Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

This manuscript by Nornes et al analyzed multiple published databases and identified a group of putative enhancers for 8 selected non-Notch arterial genes in mouse and human ECs. These enhancers were cloned and screened in fish embryos to test their effect in driving GFP reporter expression, which narrowed down a cohort of enhancers for further testing of expression activities in mouse embryonic arteries. The authors then analyzed the sequences of these enhancers, and identified binding motifs of ETS, SOX-F, FOX and MEF2 family TFs and Notch transcription regulator RBPJ commonly present in closed proximity in these arterial enhancers, suggesting interaction between these TFs in determination of arterial identity.

Major comments:

This study provides an enormous amount of bioinformatic data analysis and screening results in transgenic fish and mouse models, which led to the discovery of a group of arterial enhancers and TFs binding motifs essential in regulating arterial identity.

Other comments:

1. Choice of arterial genes is slightly biased. Acvrl1/Alk1 is not enriched in arterial ECs. Sema3G, which is highly expressed in arterial ECs, is missing. UNC5B is enriched in arterial ECs but also expressed by sprouting ECs (PMID: 38866944).
2. Exclusion of Notch genes. Although the reason for choosing non-notch genes and excluding notch genes for screening is addressed in this paper, it would be interesting to examine how the arterial enhancers identified in this study are present in the Notch genes, especially Dll4 (enriched in arterial and sprouting ECs) and Jag1 (enriched in arterial ECs).
3. SoxF family TFs. Among the 3 members of SoxF TFs, only Sox17 and Sox7 were assessed. Though not specific, Sox18 is highly expressed in the arteries. On the contrary, Sox7 is highly expressed in the vein and shows weak expression in arterial ECs (PMID: 26630461). Minor inaccuracy in Intro/paragraph 3: though sox17 is reported as indispensable for arterial specification (PMID: 24153254), losing a single SoxF factor does not seem to completely compromise the arterial program (PMID: 24153254, PMID: 26630461). A combined loss of Sox17/18, or Sox 7/17/18, seems to do the job (PMID: 26630461).
4. Fig.4 e14.5 mouse embryos. If the observation aims to assess the dorsal aorta, it would be better to use mouse embryos at mid-gestation (e9.5-10.5), when the paired DAs are formed with arterial identity but haven't been remodeled and fused as one single aorta. The morphological data in this figure would be better to show the colocalization of LacZ expression and an arterial marker (e.g. Sox17) using immulfluorescence staining instead of purely lacZ.

Significance

This novel work establishes an important foundation for future understanding of how TFs may interact to determine arterial specification.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

In this work, Nornes and collaborators have described a cohort of arterial enhancers that drive gene expression in arteries and not in veins. The paper is very well written and it is very informative. The authors have used in silico models to identified the potential artery enhancers and then used different developmental in vivo systems, zebrafish and mice, to validate their findings. Finally, the authors have explored what transcription factors may be binding the identified enhancer sequences and thus, drive arterial gene expression. I would like to congratulate the authors for this work that it has been a pleasure to read and review.

Major comments:

1. In their identification of enhancers, the authors consider a candidate every enhancer that has a putative mark in both mouse and human. Nevertheless, all the human data comes from in vitro analysis. Considering how much cell culture affects endothelial cell identity, inducing effects like EndoMT, would this have any effect on the enhancer selection? Would it be possible to search any human in vivo data? Would this allow for even stronger and more relevant sequences?
2. The human data comes from vein endothelial or microvasculature endothelial cells. Specially because some of the enhancers identified by the authors drive also vein expression, could the authors discriminate whether this is due to the identification coming from vein cells. Is there available data from HAECs? Would this not be conceptually more correct that using vein endothelial cells data? This should be at least discussed in the paper.
3. Although the authors use the mouse embryo to further validate their finding beyond the zebrafish, the expression are a bit different. While on the fish the enhancers label smaller vessels of arterial identity, in the mouse, only bigger arteries are marked. Is this defined by the time of the analysis?
4. The analysis of the enhancers is only done during development. Is the activity of these enhancers maintained through live or only important for artery vs vein determination? Is the expression of the different enhancer reporters maintained into adulthood?

Significance

This is a very well done study with potential interest for vascular biologists, in particular to those interested in the determination between artery and veins in a context of development. It advances our knowledge on the field of vascular biology as it not only proposes potential enhancers but also goes on to validation of the enhancers. Nevertheless, it is important to note that some of this enhancers have been identified from in vitro human data. In vitro culture of endothelial cells affects their cellular identity and thus, this study may have underscored many potential enhancers.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary:

Nornes et al. have generated a cohort of arterial enhancers based on in silico analysis and validation with transgenic lines in both zebrafish and mice. They utilized publicly available datasets for chromatin marks, including ATAC-seq on endothelial cells either from cell culture or isolated from mice, as well as EP300 binding, H3K27Ac, and H3K4Me1. Focusing on eight arterial-expressed genes, they identified a putative enhancer region marked by at least one enhancer feature. After validating the activity of these enhancers in zebrafish and mice, the authors assessed the regulatory pathways upstream of these genes. Using ChIP-seq and Cut&Run for key endothelial transcription factors, they discovered that binding sites for SoXF and ETS factors are shared in arterial enhancers, whereas binding sites for Notch, MEF2, and Fox are present only in the subset of identified enhancers. Together this study provides an arterial enhancer atlas that allows further characterisation of regulatory network behind endothelial cell identity.

Major comments:

The authors have assessed 15 enhancers for arterial-venous specificity, by assessing the expression in DA, ISV, cardinal and ventral veins at 2 dpf. Interestingly there is a clear difference in the expression patterns of these enhancers in the zebrafish axial vasculature, especially seen at the level of ISV. The co-localization of the enhancer expression in the endothelium was done using endothelial marks expressed in both venous and arterial EC (kdrl). To fully distinguish if the expression is venous or arterial endothelial compartment colocalization with Tg expressed in arterial (flt1) or venous (lyve1) EC would be informative. In addition, it is striking that cxc4+135 drives the expression in nearly every ISV as cxcl12+269 only every other. Similarly, not all the enhancers are enriched in the DA to the same level. Is there biological significance to this? could authors discuss these results further? The pattern of expression of the unc5b-identified enhancer is also striking, does this reflect the known roles of unc5b in the vascular formation? The final part of the paper focuses on defining the presence of "deeply conserved" transcription factor binding sites (TFSB), defined as TFBS that are as conserved as the enhancer sequence surrounding them. In literature, the term 'deep conservation' refers to evolutionary conservation (genomic sequence preservation) in a wide range of species. Therefore, the additional classification presented by the authors based on the surrounding sequence is not clear. As, the KLF motifs in the Ece1in1, which is conserved between mouse and human, are defined as "deeply conserved". However, the FLK motif in the following enhancer, Flk1in10 (one line below), gets classified as non-deeply conserved, despite also being conserved between mouse and human. Thus, in the current form, there is a contradiction in the way the authors use the term 'deeply conserved' and the accepted meaning of this term. To avoid confusion, it would be important to revise this nomenclature.

Minor:

Details on how the corresponding non-coding regions between mice and humans were established are missing, what alignment tool was used?

Not sufficient details are provided for the re-analysis of siRNA data. E.g., which clustering method was used? How the clusters were assigned to cell identities?

Details about the first HOMER analysis (in the assessment of transcription factor motifs and binding patterns at arterial enhancers) seem to be missing from the methods section.

Pg 12: "For ETS, 23/23 arterial enhancers contained at least one conserved motif (all "deeply" conserved to the same depth as the surrounding enhancer, see S7)". Is it S8, where conservation is indicated?

Figure 1 and 2 for non-zebrafish readers it would be useful to indicate in Figures 1 and 2 the non EC expression that can be observed in the embryos.

Table S1: Please, indicate in the legend what the asterisk in the H DNAseI column stands for

Figure S8: The phrasing "conserved to animal" in Figure S8 is misleading. There is no difference in something being conserved to tenrec or manatee, as both are Afrotherians. Hence, the data show that both Efnb2-141 and Ephb4-2 were present in the common ancestor of Afrotherians and humans, namely the ancestor of all placentals. Instead, it would be good to indicate the phylogenetic group for which the presence of the enhancer can be inferred (in this case, Placentalia).

Significance

To date, a systematic approach to identifying the regulatory networks driving endothelial cell identity is missing. This study provides important datasets and validation of enhancers involved in arterial gene expression and the associated transcription factors. Although this is only the tip of the iceberg, this work represents a significant milestone in the systematic understanding of how arterial gene expression is regulated. Overall, this study offers a powerful resource for understanding arterial gene regulation and conducting genome-wide studies of arterial enhancers.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Answers to reviewers

---

Reviewer #1

Sagia et al. present a manuscript using A. nidulans as model to study different transport routes of membrane proteins from the ER to the plasma membrane. They showed in earlier work that apparently at least two different transport routes exist, one involving the classical ER-ERES-ERGIC-Golgi route, one bypassing the Golgi. Unpolarized membrane proteins use the former, apically sorted membrane proteins the latter route. The study here confirms their earlier findings, uses a better model (co-expression of representatives for both routes in the same cell) and provides additional mechanistic insights about the roles of rabs, SNARES and other important proteins of the secretory pathway. The study is thoroughly done, figures are of high quality, data and methods well described and adequately replicated.

Thank you for your positive comments

I do have, however, a number of comments that could help to improve the manuscript.

-I suggest using the term polarized or apical rather than polar. Polar alone to me refers more to physico-chemical properties like water-solubility.

Amended in most parts of the revised text.

-introduction and discussion: I don’t think the literature about unconventional secretion bypassing the Golgi is complete, for example studies about TMED10 like Zhang, M. et al. Cell 181, 637-652 e615 (2020) or Zhang et al. Elife 4 (2015) are missing, there might be others. Is UapA a leader-less cargo that could be inserted via TMED10 translocation?

Thank you for letting us know, we have missed these articles. More references on UPS are now added, including the Zhang et all publications. UapA, as all transporters, is a multispan transmembrane protein with no leader peptide. In fact, we have checked the role of p24 family proteins (homologous to TMED10) in UapA trafficking. The knock-out of key p24 proteins does not affect UapA sorting to the PM (please consider this as confidential unpublished results)

-Fig. 1C. Can these intracellular structures be characterized in more detail?

As explained briefly to the handling editor above, and following the reviewer’s suggestion, we performed new experiments to better characterize the identity of the cargo-labeled fluorescent puncta. To do so, we used co-expression of a standard ERES marker, Sec16, in cells expressing either UapA or SynA, tagged with different fluorescent tags. More specifically, we constructed and analyzed strains co-expressing UapA-GFP/Sec16-mCherry or GFP-SynA/mCherry-Sec16 in the sec31ts genetic background, which allows synchronization and better analysis of ER exit, as described in our text. The new findings appear as Figure 5C __in the revised manuscript. Notice that sec16-mCherry introduced in the native sec16 locus by standard knock-in reverse genetics of A. nidulans (see Materials and methods) does not affect Aspergillus growth or secretion. Experiments depicted in __5C show that both cargoes, UapA and SynA, co-localize significantly (PCC ≈ 0.6), with Sec16, suggesting that most of these puncta are indeed ERES structures. Given that the puncta marked with UapA or SynA are clearly distinct (see Figures 1C,2A, 3A, 5B), this new experiment strongly suggests that there are indeed two distinct ERES, one populated mostly by UapA and the other by SynA. Notice, as we already outline in our response to the editor above, a three-colored approach using Sec16-BFP (or Sec13-BFP) for showing directly the existence of these two populations of cargo-specific ERES in the same cell failed as the BFP signal was problematic for colocalization studies.

Where is the Golgi localized in A. nidulans, is it decentralized like in yeast?

Yes, as in S. cerevisiae, A. nidulans Golgi cisternae are individually scattered throughout the cytoplasm, also similarly to other filamentous fungi. Notice that in A. nidulans Golgi structures are moderately polarized (Pantazopoulou and Penalva 2009).

Is the UapA at the time points shown in Fig. 1C in some sub-PM structures? To me the distribution at or near the PM is more punctate than in the steady state image shown in 1B

The punctuate appearance of PM transporters at the periphery of fungal cells is a common theme when these do not reach high, steady-state, levels of accumulation. In fact, several transporters mark specific subdomains of the PM, more evident before achieving their steady-state levels. For example, in yeast several amino acid and nucleobase transporters mark punctuate structures that colocalize with eisosomes markers (caveolin-like PM subdomains), while the proton pump ATPase Pma1 marks distinct punctuate domains. Similarly, UapA and other solute transporters mark punctuate structures before reaching their state-state accumulation in the PM. Figure 1C shows the de novo synthesis of cargoes after 100 min of transcription, while Figure 1B depicts the steady-state localization of UapA and SynA after 4h. In the latter case, the PM is ‘saturated’ with UapA molecules and thus the fluorescent signal of distinct puncta ‘fuses’, creating continuous fluorescent labeling. Notice also that in several cases, in our work, we have also performed UapA transport assays, which provide a direct tool to test and confirm the presence of UapA in the PM (see Figures 4D or 6C).

-Fig. 3A. To me it looks like there is actually a lot of colocalization of UapA and SynA, especially at or near the PM, where there is quite some white, punctate staining. The green fluorescence is just much stronger, overlaying the violet. Can you show separate channels and explain?

We think the reviewer means Figure 2A, which compares UapA and SynA (Figure 3A compares UapA with Golgi markers). If so, we have quantitatively estimated and performed statistical analysis (PCC) which indicates that this, visually apparent colocalization, is not significant (right panel in Figure 2A). Notice also that we cannot totally exclude very minimal colocalization of UapA and SynA signals as both cargoes mark very proximal early secretory domains (i.e., ERES or ERGIC), especially in fungal cells. Anyhow, in the revised Figure 2 we also added a panel depicting separate channels, as the reviewer asks.

-Fig. 3: In my opinion the statement that UapA "is probably sorted from an early secretory compartment, ultimately bypassing the need for Golgi maturation" is too strong at that point. You say for both UapA and SynA you don’t get significant colocalization with early Golgi/ERGIC marker, then you cannot conclude that one takes the conventional route via early-late Golgi and the other does not. What you can say is that UapA is apparently not going through late Golgi.

The reviewer is in principle correct. However, significant colocalization with the late Golgi marker, as SynA shows, strongly suggests that this cargo has passed via the early Golgi compartment. The fact we failed to detect significant colocalization of any cargo tested with early Golgi/ERGIC markers (e.g., SedV) is very probably due to very rapid passage of cargoes from these compartments, which conventional widefield or confocal microscopy cannot detect. To achieve this, ultra-fast fluorescent microcopy, as Lattice Light Sheet Microscopy (LLSM), should be used. In fact, we are currently initiating these studies, which will appear in the near future elsewhere.

-Fig. 4C: UapA does not seem to accumulate in the ER in the Sec24 and 13 mutants but in punctate structures. This for me is unexpected, any explanations? Can you characterize that punctate staining?

This is an interesting observation. Notice that UapA is a large homodimeric protein (e.g., 28 transmembrane domains) that oligomerizes further upon translocation into the ER membrane. Repression of Sec24, and to a less extent of Sec13, leads to inability to exit the ER properly. Consequently, this will lead to UapA overaccumulation in the ER, which might in turn lead to ER stress and turnover, reflected in UapA aggregates. In line with this, we have previously shown that specific mutants of UapA unable to exit the ER are indeed degraded by selective autophagy (Evangelinos et al., 2016). In contrast to UapA, SynA partitions in the entire ER without forming aggregates when sec24 or sec13 are repressed. This might be due to the fact that is a single-pass, much smaller, membrane protein compared to UapA and one that is not known to form oligomers. Thus, its overaccumulation in the ER might not lead to aggregation, allowing it to diffuse laterally in the membrane of the ER. A note on this is included in the Figure legend of the revised manuscript.

-Fig. 6D: You state that BFA "has only a very modest effect on UapA translocation to the PM". To me the PM (or very near PM) staining of UapA looks very different in the PFA treated cells, more uneven/punctate. Is there an explanation for that?

Our explanation is the following. When BFA is added, conventional secretion is blocked and Golgi collapses. We believe that this might have a moderate indirect effect also on cargoes bypassing the late Golgi/TGN, as UapA (i.e., lower levels of UapA present in the PM). This is based on the fact that UapA, in addition to conventional cargoes, requires the Q-SNARE complex SsoA/Sec9 to translocate to the PM. SsoA, being a membrane protein cargo itself, also needs to traffic to the PM. Interestingly, we have previously obtained evidence suggesting that SsoA traffics to the PM by both conventional and a Golgi-bypass routes (Dimou et al 2020). Thus, UapA translocation to the PM might indeed be partially impeded or delayed due to repression of proteins, such as SsoA (and probably Sec9), needed for its final integration into the PM bilayer. Importantly, in line with an indirect effect of BFA on the levels of UapA localized in the PM, notice that, unlike SynA, UapA was never trapped in brefeldin bodies (i.e., Golgi aggregates).

Reviewer #1 (Significance):

One strength of the study is the use of a model organism, A. nidulans, not cell cultures. Also, the use of both reporters, UapA and SynA, in the same cell is an advantage over previous studies using different lines and different promotors. Limitation of the study might be that it remains unclear to what extend the basic mechanism (UapA and SynA are transported to PM in different carrier and via different routes) can be generalized to other polarized (apically?) membrane proteins versus non-polarized membrane proteins in A. nidulans and whether a similar mechanism exists in other organisms. Some of the basic findings of the study are not new but were published by the same group. However, as the authors point out, the current study uses improved assays and extends their previous studies, advancing our understanding of the mechanistics of transport in the conventional secretory pathway and novel alternative routes. The study will be of interest for basic researchers in the trafficking field. My own expertise is transport through the secretory pathway in mammalian cells, many years ago more post-Golgi, now mostly ER-Golgi and ER itself.

We thank the reviewer for his positive comments.

__Reviewer #2 __

__ __The idea that transmembrane proteins of the plasma membrane move from the ER to the Golgi and then to the cell surface is firmly entrenched, and the mechanisms and components of this secretory pathway have been extensively characterized. Secretory vesicles are often delivered from the Golgi to sites of polarized growth. This paper builds on previous work by the same group to provide evidence that in Aspergillus nidulans, some non-polarly localized plasma membrane proteins follow a very different pathway, which bypasses components of the conventional secretory machinery such as SNAREs that have been implicated in secretion as well as the exocyst. In particular, they systematically compare the trafficking of the SNARE SynA, which follows the conventional secretory pathway, with that of the purine transporter UapA, which apparently does not. The two proteins were co-expressed in the same cells using the same promoter. A variety of genetic and microscopy methods are used to support the conclusion that UapA reaches the plasma membrane by a route distinct from that followed by SynA.

In my view, the authors present a convincing case. The individual experimental results are sometimes ambiguous, but the combined results favor the conclusion that UapA follows a novel pathway to the plasma membrane. I have only a few relatively minor comments.

Thank you for your positive comments

1. In the Introduction and elsewhere: to my knowledge, there is no clear evidence that AP-1-containing clathrin-coated vesicles carry cargoes from the Golgi to the plasma membrane. On the contrary, as recently reported by Robinson (https://pubmed.ncbi.nlm.nih.gov/38578286/), AP-1-containing vesicles likely mediate retrograde traffic in the late secretory pathway.

Thank you for this comment and the relative reference. We are aware that AP-1 is likely to also mediate retrograde traffic in the late secretory pathway or/and intra-Golgi recycling, as also reported by the group of Benjamin Glick. Thus, in the revised version we added a short comment on this plus relative references. Along this line, our previous work has shown that transcriptional repression of AP-1 arrests the polar localization of several apical markers in A. nidulans and we reported that this might be due to an effect on both anterograde and retrograde trafficking. Please see “Secretory Vesicle Polar Sorting, Endosome Recycling and Cytoskeleton Organization Require the AP-1 Complex in Aspergillus nidulans”. Martzoukou O, Diallinas G, Amillis S. Genetics. 2018 Aug;209(4):1121-1138. Overall, the fact that AP-1 was found absolutely dispensable for UapA trafficking, further strengthens our conclusion that UapA bypasses the Golgi.

1. In Figure 2, is there any known significance to the presence of UapA in "cytoplasmic oscillating thread structures decorated by pearl-like foci as well as a very faint vesicular/tubular network"?

At present we cannot answer this question. In order to understand what these structures represent and answer what is their role, we will need to employ super-resolution and ultra-fast microscopy and additional markers, which we envision to do. We suspect that they might be tubular networks, but this extends beyond the present work.

1. SynA is related to S. cerevisiae Snc1/2, which are known to be present in late Golgi compartments due to repeated rounds of endocytosis to the Golgi and exocytosis to the plasma membrane. The SynA shown here to colocalize with PHosbp is probably present in a similar recycling loop rather than being en route to the plasma membrane for the first time. Therefore, the differential colocalization of UapA and SynA with PHosbp does not by itself provide "strong evidence that the two cargoes studied traffic via different routes" as stated in the text but might instead indicate that only SynA undergoes frequent endocytosis. The text should be amended accordingly.

The reviewer is in principle correct. However, given that colocalization of SynA and PHosbp occurred all over the cytoplasm of hyphae and not only at the apical region, and because we record colocalization of cargoes before their steady-state accumulation to the PM, thus at a stage where recycling must be minimal, the recorded colocalization should reflect anterograde transport rather than recycling. We added this reasoning it the revised text.

1. A missing piece of the story is a test of whether the puncta visualized for the two cargoes in Figure 5B are indeed distinct populations of COPII-containing ER exit sites. The relevant experiment would involve co-labeling of the cargoes together with a COPII marker. Three-color labeling would presumably be needed.

This point was also raised by reviewer 1 (and review 3) and thus performed new experiments to better characterize the identity of the cargo-labeled fluorescent puncta. To do so, we used co-expression of a standard ERES marker, Sec16, in cells expressing either UapA or SynA, tagged with different fluorescent tags. More specifically, we constructed and analyzed strains co-expressing UapA-GFP/Sec16-mCherry or GFP-SynA/Sec16-mCherry in the sec31ts genetic background, which allows synchronization and better analysis of ER exit, as described in our text. The new findings appear as Figure 5C __in the revised manuscript. Notice that sec16-mCherry introduced in the native sec16 locus by standard knock-in reverse genetics of A. nidulans (see Materials and methods) does not affect Aspergillus growth or secretion. Experiments depicted in __5C show that both cargoes, UapA and SynA, co-localize significantly (PCC ≈ 0.6), with Sec16, suggesting that most of these puncta are indeed ERES structures. Given that the puncta marked with UapA or SynA are clearly distinct (see Figures 1C,2A, 3A, 5B), this new experiment strongly suggests that there are indeed two distinct ERES, one populated mostly by UapA and the other by SynA. Notice, as we already outline in our response to the editor above, a three-colored approach using Sec16-BFP (or Sec13-BFP) for showing directly the existence of these two populations of cargo-specific ERES in the same cell failed as the BFP signal was problematic for colocalization studies.

Reviewer #2 (Significance):

This study provides compelling evidence that in the fungus Aspergillus nidulans, some transmembrane transporter proteins reach the plasma membrane by a pathway that bypasses much of the conventional machinery associated with the Golgi apparatus and secretory vesicles. Although previous publications pointed toward a similar conclusion, the present work tackles the problem in a more rigorous and systematic way. These findings are important for cell biologists who study membrane traffic, it remains to be determined how prevalent this type of non-canonical secretion might be in other organisms.

We thank the reviewer for his positive comments

Reviewer #3

The manuscript by Sagia et al compares the trafficking of a polarized (SynA) with a non-polarized (UapA) transmembrane protein. In agreement with previous work of the same lab, they find that UapA reaches the plasma membrane through a Golgi-bypass route, which they characterize to some extent. Overall, the data are of good quality and the story is interesting and timely. Understanding trafficking routes that bypass the Golgi is highly interesting. Nevertheless, there are several points of criticism that I have and below is a list where I combine major and minor points together:

Thank you for your positive comments

Major Comments:

1- Is it possible that the polarized phenotype of SynA is caused by selective removal, i.e. SynA is delivered to the entire plasma membrane, but endocytosed rapidly from all areas except the tip of the hyphae. This would also result in a polarized distribution.

This is in principle possible, but here this is not the case. SynA is polarized due to rapid local endocytosis and immediate recycling at the subapical region, known as the subapical collar. Please see:

Taheri-Talesh N, Horio T, Araujo-Bazán L, Dou X, Espeso EA, Peñalva MA, Osmani SA, Oakley BR. The tip growth apparatus of Aspergillus nidulans. Mol Biol Cell. 2008 Apr;19(4):1439-49. doi: 10.1091/mbc.e07-05-0464.

Hernández-González M, Bravo-Plaza I, Pinar M, de Los Ríos V, Arst HN Jr, Peñalva MA. Endocytic recycling via the TGN underlies the polarized hyphal mode of life. PLoS Genet. 2018;14(4):e1007291. Published 2018 Apr 2. doi:10.1371/journal.pgen.1007291

This applies to all apical markers; they remain polarized by continuous local recycling after the diffuse laterally to the subapical collar.

2- The authors describe the distribution of SynA and UapA in cells deficient of various COPII/ERES proteins. However, these data are not shown, and it is not clear how they were quantified. It would be important to add quantitative data here.

Quantitative data are included in Figure 4C, displaying the percentages of cells with UapA either retained in the ER or reaching the PM for each background deficient in a COPII protein. Repression of SarA and Sec31 resulted in UapA retention in the ER in all analyzed cells (100%). However, repression of Sec12, Sec24, or Sec13 had a differential effect across the cell population, with UapA reaching the PM in some cells, while remaining trapped in the ER in others. To quantify these data and determine which cargo localization pattern prevails, we measured the number of cells in each category and represented them as percentages. A similar approach was used to examine the role of Golgi proteins in the trafficking of UapA and SynA (Figure 6).

3- on page 8, the authors discuss the discrepancy regarding the role of Sec13. They offer as an explanation that the previous studies have been performed in strains that separately expressed the two cargoes. However, I am unable to see why and how this would be a valid explanation.

Given that Sec13 has a variable/partial effect on UapA, we have previously been biased towards images that showed an effect on localization, as expected, and considered that the lack of an effect might have been due to inefficient repression in a fraction of cells. In our new system, we were able to directly compare UapA to SynA and find out that while SynA was always affected under our conditions, the effect of UapA was still variable. Thus, the partial effect of Sec13 on UapA is physiologically valid and not a matter of insufficient repression in a fraction of cells. This shows the importance of our new improved system where we follow the synchronous expression of two cargoes in the same cells.

4- Why is the effect of Sec24 depletion so much stronger than of Sec12 depletion? Sec12 is the GEF for SarA, without which Sec24 should not be recruited to ERES. The explanation that low amounts of Sec12 are still present and sufficient to carry out the role of this protein. What is the evidence for that?

Sec24 is the principal receptor of cargoes responsible for their recruitment to ERES. Sec12 is the catalytic effector for SarA required for the initiation of COPII vesicle formation. The question of the reviewer is thus logical.

However, Sec12 is indeed present at extremely very low levels when expressed from its native promoter under the condition of our experiment (minimal media). This is supported by our recent proteomic analysis, performed under similar conditions, which failed to detect the Sec12 protein, unlike all other COPII components (see Dimou et al., 2021, doi; 10.3390/jof7070560), but also by cellular studies of the group of M.A. Peñalva, who failed to detect Sec12 tagged with GFP (Bravo-Plaza et al., 2019, doi: 10.1016/j.bbamcr.2019.118551). Additionally, in yeast, immune detection of Sec12 has been possible only in cells harboring sec12 on a multicopy plasmid, suggesting its low abundance in wild-type cells (Nakano et al., 1988, doi:10.1083/jcb.107.3.851).

Given that repression of sec12 transcription via the thiAp promoter still allows 68% of cells to secrete normally both SynA and UapA, while 32% of cells are blocked in the trafficking of both cargoes, suggests that in most cells either SarA can catalyze the exchange of GDP for GTP without Sec12, maybe through a cryptic guanine nucleotide exchange factor (GEF), or that very small amounts of Sec12 remaining after repression are sufficient for significant SarA activation. Whichever scenario is true, Sec12, similarly to SarA, is not critical for distinguishing Golgi-dependent from Golgi-independent routes, as both cargoes are affected similarly. In the revised text we added a not on this issue.

5- In Figure 5, it would help readers who are not so familiar with Aspergillus organelle morphology to explain the figure a bit better. This might appear trivial for experts, but anyone from outside this field is slightly lost.

In the revised manuscript we added a figure panel depicting a schematic representation of A. nidulans key secretory compartments.

6- The authors write that not seeing UapA in Golgi membranes is evidence that it does not pass through this organelle. However, when they write that SynA is never seen in cis-Golgi elements, they do not conclude that SynA bypasses the cis-Golgi.

The fact that SynA, unlike UapA, colocalized significantly with late-Golgi/TGN and follows conventional secretion in general, strongly suggests that SynA also passes from the early-Golgi. Cargo traffic through the Golgi is mediated by cisternal maturation, where an individual cisterna gradually changes its nature from an earlier to a later one, while the cargo remains inside. UapA, unlike SynA, never colocalized with any Golgi marker used and was not affected by BFA. We agree with the reviewer that we did not have direct proof for passage of UapA or SynA from the early-Golgi in the wt background, which allows for the alternative, but rather unlikely hypothesis, that none of the two cargos is sorted to the early Golgi and that SynA traffics directly to late-Golgi/TGN. Our inability to detect sorting of any cargo to the early-Golgi is seemingly due to ultra-fast passage of cargoes from very early secretory compartments, such as ERGIC/early-Golgi. In fact, we have obtained evidence of this using Lattice Light Sheet microscopy (results in progress, to appear elsewhere).

7- Figure 5C: the authors claim that the CopA and ArfA affects trafficking of UapA and SynA from ER to plasma membrane and assign CopA and ArfA as regulators for anterograde trafficking. I think this interpretation is not justified by the data. Depletion of CopA and ArfA will affect the Golgi apparatus in structure and function. The more straight-forward interpretation is that repression of the COPI machinery results in a defect in Golgi exit and therefore retention in pre-Golgi compartments (including the ER and maybe the ERGIC should it exist in Aspergillus). The same is true for BFA treatment where there are also negative effects on ER export, which are rather indirect consequences of alterations of Golgi function and integrity. Likewise, the interpretation of the papers by Weigel et al and Shomron et al is not correct. It is more likely that COPI is recruited to the growing ERES-derived tubule (or ERGIC) to recycle proteins back to the ER. This is not necessarily a proof that COPI regulates anterograde trafficking

This is a highly debatable issue which our work cannot address. However, we amended the text accordingly.

8- Figure 6: The images look like in Figure 5, yet here you don't call them ER-associated.

The two images are not alike. In Figure 5 upon activation of Sec31 (permissive temperature) we detect mostly punctual structures resembling ERES, whereas at the nonpermissive temperature we detect a membranous network typical of the ER. Upon repression of CopA we also detect punctual structures similar to ERES. In Figure 6, we mostly detect an effect on SynA. Repression of early secretory steps (SedV, GeaA) lead to collapse of SynA in the entire ER network. Repression at later stages of Golgi maturation and post-Golgi secretion (RabO, HypB, RabE, AP-1) lead to the appearance of punctual structures, most probably Golgi aggregates.

9- Figure 6D: How long was the BFA treatment. I am surprised that the pool of SynA preexisting at the plasma membrane seems to also be sensitive to BFA.

Cells were grown overnight under repressed conditions for both UapA and SynA. After 12-14h cells were shifted to derepressed conditions using fructose as carbon source. BFA was added after 90min of cargo derepression, while both cargoes were still in cytoplasmic structures so there was not preexisting SynA or UapA at the PM (see also Figure 1C). Subcellular localization of both cargoes was studied for 60min after BFA treatment.

10- This might be beyond the scope of this study, but as far as I know UapA is not N-glycosylated. Would the introduction of an N-glycosylation site shift it towards the Golgi-based route?

Thank you for this suggestion. We have performed this experiment, adding a glycosylation site on UapA, based on the glycosylation sites found in tis mammalians homologues. We did not detect any effect on UapA trafficking route or its activity. As the reviewer recognizes this goes beyond the scope of this study and thus, we did not include it the manuscript. Differential cargo glycosylation is however an important issue to be studied systemically in respect to different trafficking routes, and we envision to investigate it systematically.

Minor Comments

1- This might be just a personal preference, but I think that the term polar is misleading, because it implies something about the polarity of the amino acids. I think "polarized" might be the more common term. Anyway, this is just a minor point and just a suggestion from my side.

Amended in the revised text.

2- The paper by the Saraste lab should be mentioned and discussed (PMID: 16421253), which I think is very relevant to the current story.

We thank the reviewer for pointing out this important publication. In that case, the Rab1 GTPase defined a pathway connecting a pre-Golgi intermediate compartment with the PM in mammalians nerve cells. Thus, the Saraste lab publication is indeed along the lines of findings supporting that Golgi-independent unconventional cargo trafficking routes initiate at very early secretory compartments. Notice, however, that RabO, the A. nidulans homologue of Rab1, which in their case was essential for direct cargo sorting from the ERES/ERGIC to the PM, in or system, was dispensable for Golgi bypass. The Saraste lab article is now mentioned and discussed.

3- Having worked with ERES for over two decades, I find it strange to see it written ERes. I see no reason why ER exit sites in Aspergillus should be abbreviated differently from all other types of cells (yeast, drosophila, worms, mammals). I think that the entire acronym should be capitalized.

Amended in the text

4- When discussing the data about the partial effect of Sec13, it would be good to refer to a previous paper by the Stephens lab that showed that silencing Sec13/31 results in a defect in trafficking of collagen, but not of VSVG (PMID: 18713835).

We thank the reviewer for also pointing out the publication of the Stephens lab, now mentioned in the revised text. Noticeably, in that case silencing of both Sec13 and Sec31 has no effect on the trafficking of specific cargoes, whereas in our case Sec31 is still absolutely needed for both conventional and Golgi-independent secretion of SynA and UapA, respectively.

Reviewer #3 (Significance):

Overall, the data are of good quality and the story is interesting and timely. Understanding trafficking routes that bypass the Golgi is highly interesting. The main weakness is the lack of mechanistic understanding of the Golgi-bypass pathway. In addition, the study is limited to two proteins as representatives of polarized vs. non-polarized proteins. The main target audience for this paper are scientists working in the area of secretion and trafficking in the secretory pathway.

We thank the reviewer for his positive comments.

We are aware that the mechanistic details of Golgi bypass are missing and this is our next goal, dissecting those via various approaches genetic and biochemical approaches and employment of super resolution and ultra-fast microscopy.

__Reviewer #4 __

In this study, Sagia et al investigate the trafficking of different secretory cargo in Aspergillus nidulans under conditions that repress expression of transport factors or block stages in membrane trafficking. The primary approach is to conduct dual live-cell imaging of GFP-tagged UapA (plasma membrane localized purine transporter) and SynA (plasma membrane R-SNARE) after their simultaneous derepression to monitor trafficking routes. In germlings, both secretory proteins are detected in non-overlapping intracellular compartments and puncta after 60-90 min of derepression. After 4-6 hrs, SynA localizes to hyphal tips whereas UapA localizes to non-polar regions of the PM. Colocalization studies do not show UapA overlap with Golgi markers (SedV, PH-OSBP) during its biogenesis whereas SynA displays significant co-localization. Repression of COPII and COPI components generally block transport of both cargos to the PM and cause accumulation in ER compartments, although there are some differential effects on UapA and SynA localization. Finally, repression of other transport factors (ER-Golgi SNAREs, Golgi transport factors, and exocytic machinery) had differential effects on UapA and SynA localization over time with UapA reaching the plasma membrane in many instances and SynA accumulating in intracellular compartments.

Based on these observations, the authors conclude that UapA and SynA follow distinct trafficking routes to the plasma membrane where SynA uses a canonical SNARE-dependent secretory pathway route and UapA follows a non-canonical route that may bypass Golgi compartments. The study is extensive and supports the model that biogenesis of SynA and UapA follow distinct processes. However, there are some complexities that may limit interpretation. First, the cargo studied are targeted to the ER differently. UapA is a multispanning transmembrane protein that is likely dependent on the Sec61 translocon for co-translational membrane insertion and will involve ER chaperones and quality control machinery for its biogenesis. SynA will depend on the tail-anchored machinery (GET/TRC pathway) for insertion into the ER and is processed by cytosolic factors/chaperones. Therefore, the sites of ER insertion and the rates of biogenesis of these cargoes will be different. In addition, the repression of trafficking machinery used in this study appears to be variable and may exert partial blocks on intracellular transport stages. Regardless, the study clearly documents that SynA and UapA follow distinct biogenesis and transport processes when co-expressed in cells under experimentally controlled conditions.

Thank you for your positive comments.

To our knowledge there is no evidence suggesting that SynA translocates via a tail-anchored machinery (GET/TRC pathway) and not through the translocase. Despite this, we agree with the reviewer that translocation to the ER, as well as exit from it, might be cargo-dependent, especially when it concerns proteins with very different size, structures and oligomerization. Thus, the rate of biogenesis of UapA and SynA is probably quite different. However, this still does not dismiss our basic conclusion that the two cargoes follow distinct routes to traffic to the PM. The ‘problem’ of variable transcriptional repression of some trafficking-related proteins is solved by comparing the relative effect on the two cargoes in the same cells, and this is in fact the advantage of our new system. Importantly, notice that we took care to use conditions of repression where SynA trafficking by the conventional path was totally abolished and compared it to UapA.

1. It was not clear if the translation, ER insertion and folding of UapA and SynA are fully synchronous. Is it possible that the rate of UapA synthesis and transport to the plasma membrane is substantially faster than for SynA? The imposition of transport blocks could trap SynA and not UapA if this cargo was at later transport stages.

As already discussed above translation, ER insertion and folding of UapA and SynA might indeed by different. This might somehow affect the trafficking path followed, but this issue is beyond the scope of this work. Notice, however, that the transcription of both cargoes is kept fully repressed during establishment of repression of secretion. Only when repression and blocking of secretion is established (12-14 h germination), as verified by Western blot analysis, we derepress the transcription of UapA and SynA, expressed from the same promoter, and follow their dynamic subcellular localization. Hence, this system ensures that both cargoes start from the earliest transport stage, the ER, upon imposition of transport blocks.

1. In repressing transport factors (e.g., SarA, Sec12, Sec24, Sec13, SedV, RabE), it is clear that under thiamine repressing conditions these cells do not grow or have greatly reduced growth rates. However, it was not clear if proteins are depleted to the same extent in cells after repression for 12-14 hr or 16-22 hr. as mentioned in the methods. Indeed, in some cases depleted cells display different cargo localization patterns, for example 67% of cells show normal localization of UapA and SynA after sec12 repression and 33% show ER accumulation of both cargoes. There is differential localization of UapA and SynA in many cases where transport factors are repressed, but this could be due to partial inhibition and not complete blocks. It would be helpful to clearly indicate the time points and conditions in each of the figure legends as in points 3-5 below.

In the revised manuscript we did our best to clearly indicate the time points and conditions in each of the figure legends. Differential localization of UapA and SynA in many cases where trafficking factors are repressed is indeed an interesting outcome. Inefficient repression was dismissed based on the lack of colony growth (see relative growth tests of SarA, Sec24, Sec13, Sec31, SedV, GeaA, RabO, RabE, Ykt6, Sft1, SsoA and Sec9), but also by western blots (e.g., Sec24, Sec13, Sec31 or Sec9 shown in the present manuscript, or other trafficking proteins studied previously. Martzoukou et al., 2018; Dimou et al., 2020). Repression of Sec12 and HypB, and to lower degree AP-1, allowed formation of small and/or compact colonies, but even in these cases relative protein levels could not be detected in western blots, guaranteeing efficient repression.

1. In Fig 4A immunoblot, HA-tagged proteins are not detected after thiamine repression. Please state the time of thiamine repression used before protein extraction and blot. Is this for the same length of time as for cells shown in panel 4C? It would also be helpful to state the time of cargo derepression before capturing images in 4C. The methods section mentions 12-14 hr or 16-22 hr of growth, presumably with thiamine in the culture, and then 1-8 hr or 60 min to 4 hr of cargo derepression before imaging. Please specify.

The time of thiamine repression before protein extraction was 16-18h. The same repression time was used for experiments shown in Figures 4C and 6C (ER/COPII and Golgi/post-Golgi repression respectively). More specifically, for microscopy experiments cells were grown in the presence of glucose and thiamine for 12-14h (repressed UapA/SynA and thiAp expressed gene). After this time, cells were shifted to fructose and thiamine for 4h (derepression of UapA/SynA and repression of thiAp expressed gene). In both cases (protein extraction and microscopy experiments) the total time of thiamine repression was 16-18h.

1. For the thiA-copA and thiA-arfA repression experiments (Fig 5C), the methods section states that thiamine was not added ab initio in the culture, but after an 8 h time window without thiamine at the start of spore incubation. This is interpreted to mean that repression was for a shorter period to time than the 12-14 hr overnight growth. However, the figure legend states that De novo synthesis of cargos takes place after full repression of CopA and ArfA is achieved (>16 hr). Please clarify.

We think that the review was confused with repression of cargo synthesis (via alcAp+glucose) versus repression of trafficking proteins (via thiAp+thiamine). Please see Materials and methods. We clarify our protocol also here:

For the thiAp-copA and thiAp-arfA repression experiments addition of thiamine ab initio in the culture leads to total arrest of spore germination and germling formation. Thus, we added an 8-hour time window without thiamine to allow conidiospores to germinate until the stage of young germlings, under conditions where cargo expression via the alcAp was repressed by glucose. Subsequently, thiamine was added in the media (16-18 h) to repress CopA and ArfA, while cargo expression remained glucose-repressed. The transcriptional repression of the cargoes UapA and SynA was maintained for a longer period (24-26 h) compared to other repression experiments, but longer times of repression of cargoes do not make any difference, as full repression is achieved already at 12 h. De novo cargo trafficking was followed next day by eliciting depression, via a shift to fructose media, while still maintaining thiamine to repress CopA or ArfA.

1. In Fig 6D, BFA treatment is shown to trap SynA in Golgi aggregates while UapA still reaches the plasma membrane. Please state the time of BFA treatment before collecting these images. Do longer treatments with BFA before cargo derepression cause accumulation of UapA in intracellular compartments?

As mentioned above (response to Reviewer’s #3 comment 9) cells were grown overnight under repressed conditions for both UapA and SynA. After 12-14h cells were shifted to derepressed conditions using fructose as carbon source. BFA was added after 90min of cargo derepression, while both cargoes were still in cytoplasmic structures so there was not preexisting SynA or UapA at the PM (see also Figure 1C). We have not noticed any different effect on UapA trafficking after a max of 1h of BFA treatment.

1. A minor point, but on page 21 the methods state that "cells were shifted down to the permissive temperature (25 C), to restore the secretory block...". Suggest changing to "to reverse the secretory block..."

Modified accordingly

Reviewer #4 (Significance):

This manuscript nicely builds on a developing line of investigation in the Aspergillus nidulans model that specific plasma membrane proteins are efficiently delivered to the cell surface in a pathway that is distinct from the canonical secretory pathway. Previous work from this lab has suggested that a subpopulation of COPII carriers can bypass the Golgi for delivery of specific cargo to the plasma membrane. The current study uses dual expression of UapA-GFP and mCherry-SynA to provide further support for this model. Molecular definition of a direct ER to PM transport pathway for secretory cargo would be a significant advance to a broad audience. This study provides additional depth and support that such a pathway exists but does not define how COPII vesicles or related intermediates are transported to the PM.

Again, thank you for your positive comments.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #4

Evidence, reproducibility and clarity

In this study, Sagia et al investigate the trafficking of different secretory cargo in Aspergillus nidulans under conditions that repress expression of transport factors or block stages in membrane trafficking. The primary approach is to conduct dual live-cell imaging of GFP-tagged UapA (plasma membrane localized purine transporter) and SynA (plasma membrane R-SNARE) after their simultaneous derepression to monitor trafficking routes. In germlings, both secretory proteins are detected in non-overlapping intracellular compartments and puncta after 60-90 min of derepression. After 4-6 hrs, SynA localizes to hyphal tips whereas UapA localizes to non-polar regions of the PM. Colocalization studies do not show UapA overlap with Golgi markers (SedV, PH-OSBP) during its biogenesis whereas SynA displays significant co-localization. Repression of COPII and COPI components generally block transport of both cargos to the PM and cause accumulation in ER compartments, although there are some differential effects on UapA and SynA localization. Finally, repression of other transport factors (ER-Golgi SNAREs, Golgi transport factors, and exocytic machinery) had differential effects on UapA and SynA localization over time with UapA reaching the plasma membrane in many instances and SynA accumulating in intracellular compartments.

Based on these observations, the authors conclude that UapA and SynA follow distinct trafficking routes to the plasma membrane where SynA uses a canonical SNARE-dependent secretory pathway route and UapA follows a non-canonical route that may bypass Golgi compartments. The study is extensive and supports the model that biogenesis of SynA and UapA follow distinct processes. However, there are some complexities that may limit interpretation. First, the cargo studied are targeted to the ER differently. UapA is a multispanning transmembrane protein that is likely dependent on the Sec61 translocon for co-translational membrane insertion and will involve ER chaperones and quality control machinery for its biogenesis. SynA will depend on the tail-anchored machinery (GET/TRC pathway) for insertion into the ER and is processed by cytosolic factors/chaperones. Therefore, the sites of ER insertion and the rates of biogenesis of these cargo will be different. In addition, the repression of trafficking machinery used in this study appear to be variable and may exert partial blocks on intracellular transport stages. Regardless, the study clearly documents that SynA and UapA follow distinct biogenesis and transport processes when co-expressed in cells under experimentally controlled conditions.

1. It was not clear if the translation, ER insertion and folding of UapA and SynA are fully synchronous. Is it possible that the rate of UapA synthesis and transport to the plasma membrane is substantially faster than for SynA? The imposition of transport blocks could trap SynA and not UapA if this cargo was at later transport stages.
2. In repressing transport factors (e.g. sarA, sec12, sec24, sec13, sedV, rabE), it is clear that under thiamine repressing conditions these cells do not grow or have greatly reduced growth rates. However, it was not clear if proteins are depleted to the same extent in cells after repression for 12-14 hr or 16-22 hr as mentioned in the methods. Indeed, in some cases depleted cells display different cargo localization patterns, for example 67% of cells show normal localization of UapA and SynA after sec12 repression and 33% show ER accumulation of both cargo. There is differential localization of UapA and SynA in many cases where transport factors are repressed, but this could be due to partial inhibition and not complete blocks. It would be helpful to clearly indicate the time points and conditions in each of the figure legends as in points 3-5 below.
3. In Fig 4A immunoblot, HA-tagged proteins are not detected after thiamine repression. Please state the time of thiamine repression used before protein extraction and blot. Is this for the same length of time as for cells shown in panel 4C? It would also be helpful to state the time of cargo derepression before capturing images in 4C. The methods section mentions 12-14 hr or 16-22 hr of growth, presumably with thiamine in the culture, and then 1-8 hr or 60 min to 4 hr of cargo derepression before imaging. Please specify.
4. For the thiA-copA and thiA-arfA repression experiments (Fig 5C), the methods section states that thiamine was not added ab initio in the culture, but after an 8 h time window without thiamine at the start of spore incubation. This is interpreted to mean that repression was for a shorter period to time than the 12-14 hr overnight growth. However, the figure legend states that De novo synthesis of cargos takes place after full repression of CopA and ArfA is achieved (>16 hr). Please clarify.
5. In Fig 6D, BFA treatment is shown to trap SynA in Golgi aggregates while UapA still reaches the plasma membrane. Please state the time of BFA treatment before collecting these images. Do longer treatments with BFA before cargo derepression cause accumulation of UapA in intracellular compartments?
6. A minor point, but on page 21 the methods state that "cells were shifted down to the permissive temperature (25 C), to restore the secretory block...". Suggest changing to "to reverse the secretory block..."

Significance

This manuscript nicely builds on a developing line of investigation in the Aspergillus nidulans model that specific plasma membrane proteins are efficiently delivered to the cell surface in a pathway that is distinct from the canonical secretory pathway. Previous work from this lab has suggested that a subpopulation of COPII carriers can bypass the Golgi for delivery of specific cargo to the plasma membrane. The current study uses dual expression of UapA-GFP and mCherry-SynA to provide further support for this model.

Molecular definition of a direct ER to PM transport pathway for secretory cargo would be a significant advance to a broad audience. This study provides additional depth and support that such a pathway exists but does not define how COPII vesicles or related intermediates are transported to the PM.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

The manuscript by Sagia et al compares the trafficking of a polarized (SynA) with a non-polarized (UapA) transmembrane protein. In agreement with previous work of the same lab, they find that UapA reaches the plasma membrane through a Golgi-bypass route, which they characterize to some extent. Overall, the data are of good quality and the story is interesting and timely. Understanding trafficking routes that bypass the Golgi is highly interesting. Nevertheless, there are several points of criticism that I have and below is a list where I combine major and minor points together:

Major Comments:

1. Is it possible that the polarized phenotype of SynA is caused by selective removal, i.e. SynA is delivered to the entire plasma membrane, but endocytosed rapidly from all areas except the tip of the hyphae. This would also result in a polarized distribution.
2. The authors describe the distribution of SynA and UapA in cells deficient of various COPII/ERES proteins. However, these data are not shown, and it is not clear how they were quantified. It would be important to add quantitative data here.
3. on page 8, the authors discuss the discrepancy regarding the role of Sec13. They offer as an explanation that the previous studies have been performed in strains that separately expressed the two cargoes. However, I am unable to see why and how this would be a valid explanation.
4. Why is the effect of Sec24 depletion so much stronger than of Sec12 depletion? Sec12 is the GEF for SarA, without which Sec24 should not be recruited to ERES. The explanation that low amounts of Sec12 are still present and sufficient to carry out the role of this protein. What is the evidence for that?
5. In Figure 5, it would help readers who are not so familiar with Aspergillus organelle morphology to explain the figure a bit better. This might appear trivial for experts, but anyone from outside this field is slightly lost.
6. The authors write that not seeing UapA in Golgi membranes is evidence that it does not pass through this organelle. However, when they write that SynA is never seen in cis-Golgi elements, they do not conclude that SynA bypasses the cis-Golgi.
7. Figure 5C: the authors claim that the CopA and ArfA affects trafficking of UapA and SynA from ER to plasma membrane and assign copA and ArfA as regulators fo anterograde trafficking. I think this interpretation is not justified by the data. Depletion of CopA and ArfA will affect the Golgi apparatus in structure and function. The more straight-forward interpretation is that repression of the COPI machinery results in a defect in Golgi exit and therefore retention in pre-Golgi compartments (including the ER and maybe the ERGIC should it exist in Aspergillus). The same is true for BFA treatment where there are also negative effects on ER export, which are rather indirect consequences of alterations of Golgi function and integrity. Likewise, the interpretation of the papers by Weigel et al and Shomron et al is not correct. It is more likely that COPI is recruited to the growing ERES-derived tubule (or ERGIC) to recycle proteins back to the ER. This is not necessarily a proof that COPI regulates anterograde trafficking
8. Figure 6: The images look like in Figure 5, yet here you don't call them ER-associated.
9. Figure 6D: How long was the BFA treatment. I am surprised that the pool of SynA preexisting at the plasma membrane seems to also be sensitive to BFA.
10. This might be beyond the scope of this study, but as far as I know UapA is not N-glycosylated. Would the introduction of an N-glycosylation site shift it towards the Golgi-based route?

Minor Comments

1. This might be just a personal preference, but I think that the term polar is misleading, because it implies something about the polarity of the amino acids. I think "polarized" might be the more common term. Anyway, this is just a minor point and just a suggestion from my side.
2. The paper by the Saraste lab should be mentioned and discussed (PMID: 16421253), which I think is very relevant to the current story.
3. Having worked with ERES for over two decades, I find it strange to see it written ERes. I see no reason why ER exit sites in Aspergillus should be abbreviated differently from all other types of cells (yeast, drosophila, worms, mammals). I think that the entire acronym should be capitalized.
4. When discussing the data about the partial effect of Sec13, it would be good to refer to a previous paper by the Stephens lab that showed that silencing Sec13/31 results in a defect in trafficking of collagen, but not of VSVG (PMID: 18713835)

Significance

Overall, the data are of good quality and the story is interesting and timely. Understanding trafficking routes that bypass the Golgi is highly interesting. The main weakness is the lack of mechanistic understanding of the Golgi-bypass pathway. In addition, the study is limited to two proteins as representatives of polarized vs. non-polarized proteins. The main target audience for this paper are scientists working in the area of secretion and trafficking in the secretory pathway.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

The idea that transmembrane proteins of the plasma membrane move from the ER to the Golgi and then to the cell surface is firmly entrenched, and the mechanisms and components of this secretory pathway have been extensively characterized. Secretory vesicles are often delivered from the Golgi to sites of polarized growth. This paper builds on previous work by the same group to provide evidence that in Aspergillus nidulans, some non-polarly localized plasma membrane proteins follow a very different pathway, which bypasses components of the conventional secretory machinery such as SNAREs that have been implicated in secretion as well as the exocyst. In particular, they systematically compare the trafficking of the SNARE SynA, which follows the conventional secretory pathway, with that of the purine transporter UapA, which apparently does not. The two proteins were co-expressed in the same cells using the same promoter. A variety of genetic and microscopy methods are used to support the conclusion that UapA reaches the plasma membrane by a route distinct from that followed by SynA.

In my view, the authors present a convincing case. The individual experimental results are sometimes ambiguous, but the combined results favor the conclusion that UapA follows a novel pathway to the plasma membrane. I have only a few relatively minor comments.

1. In the Introduction and elsewhere: to my knowledge, there is no clear evidence that AP-1-containing clathrin-coated vesicles carry cargoes from the Golgi to the plasma membrane. On the contrary, as recently reported by Robinson (https://pubmed.ncbi.nlm.nih.gov/38578286/), AP-1-containing vesicles likely mediate retrograde traffic in the late secretory pathway.
2. In Figure 2, is there any known significance to the presence of UapA in "cytoplasmic oscillating thread structures decorated by pearl-like foci as well as a very faint vesicular/tubular network"?
3. SynA is related to S. cerevisiae Snc1/2, which are known to be present in late Golgi compartments due to repeated rounds of endocytosis to the Golgi and exocytosis to the plasma membrane. The SynA shown here to colocalize with PH-osbp is probably present in a similar recycling loop rather than being en route to the plasma membrane for the first time. Therefore, the differential colocalization of UapA and SynA with PH-osbp does not by itself provide "strong evidence that the two cargoes studied traffic via different routes" as stated in the text, but might instead indicate that only SynA undergoes frequent endocytosis. The text should be amended accordingly.
4. A missing piece of the story is a test of whether the puncta visualized for the two cargoes in Figure 5B are indeed distinct populations of COPII-containing ER exit sites. The relevant experiment would involve co-labeling of the cargoes together with a COPII marker. Three-color labeling would presumably be needed.

Significance

This study provides compelling evidence that in the fungus Aspergillus nidulans, some transmembrane transporter proteins reach the plasma membrane by a pathway that bypasses much of the conventional machinery associated with the Golgi apparatus and secretory vesicles. Although previous publications pointed toward a similar conclusion, the present work tackles the problem in a more rigorous and systematic way. These findings are important for cell biologists who study membrane traffic, although it remains to be determined how prevalent this type of non-canonical secretion might be in other organisms.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Sagia et al. present a manuscript using A. nidulans as model to study different transport routes of membrane proteins from the ER to the plasma membrane. They showed in earlier work that apparently at least two different transport routes exist, one involving the classical ER-ERES-ERGIC-Golgi route, one bypassing the Golgi. Unpolarized membrane proteins use the former, apically sorted membrane proteins the latter route. The study here confirms their earlier findings, uses a better model (co-expression of representatives for both routes in the same cell) and provides additional mechanistic insights about the roles of rabs, SNARES and other important proteins of the secretory pathway. The study is thoroughly done, figures are of high quality, data and methods well described and adequately replicated.

I do have, however, a number of comments that could help to improve the manuscript.

• I suggest to use the term polarized or apical rather than polar. Polar alone to me refers more to physico-chemical properties like water-solubility.
• introduction and discussion: I don´t think the literature about unconventional secretion bypassing the Golgi is complete, for example studies about TMED10 like Zhang, M. et al. Cell 181, 637-652 e615 (2020) or Zhang et al. Elife 4 (2015) are missing, there might be others. Is UapA a leader-less cargo that could be inserted via TMED10 translocation?
• Fig. 1C. Can these intracellular structures be characterized in more detail? Where is the Golgi localized in A. nidulans, is it decentralized like in yeast? Is the UapA at the time points shown in Fig. 1C in some sub-PM structures? To me the distribution at or near the PM is more punctate than in the steady state image shown in 1B.
• Fig. 3A. To me it looks like there is actually a lot of colocalization of UapA and SynA, especially at or near the PM, where there is quite some white, punctate staining. The green fluorescence is just much stronger, overlaying the violet. Can you show separate channels and explain?
• Fig. 3: In my opinion the statement that UapA "is probably sorted from an early secretory compartment, ultimately bypassing the need for Golgi maturation" is too strong at that point. You say for both UapA and SynA you don´t get significant colocalization with early Golgi/ERGIC marker, then you cannot conclude that one takes the conventional route via early-late Golgi and the other does not. What you can say is that UapA is apparently not going through late Golgi.
• Fig. 4C: UapA does not seem to accumulate in the ER in the Sec24 and 13 mutants but in punctate structures. This for me is unexpected, any explanations? Can you characterize that punctate staining?
• Fig. 6D: You state that BFA "has only a very modest effect on UaPA translocation to the PM". To me the PM (or very near PM) staining of UaPA looks very different in the PFA treated cells, more uneven/punctate. Is there an explanation for that?

Significance

One strength of the study is the use of a model organism, A. nidulans, not cell cultures. Also the use of both reporters, UapA and SynA, in the same cell is an advantage over previous studies using different lines and different promotors. Limitation of the study might be that it remains unclear to what extend the basic mechanism (UapA and SynA are transported to PM in different carrier and via different routes) can be generalized to other polarized (apically?) membrane proteins versus non-polarized membrane proteins in A. nidulans and whether a similar mechanism exists in other organisms.

Some of the basic findings of the study are not new but were published by the same group. However, as the authors point out, the current study uses improved assays and extends their previous studies, advancing our understanding of the mechanistics of transport in the conventional secretory pathway and novel alternative routes. The study will be of interest for basic researchers in the trafficking field.

My own expertise is transport through the secretory pathway in mammalian cells, many years ago more post-Golgi, now mostly ER-Golgi and ER itself.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Reviewer #1 (Evidence, reproducibility and clarity (Required)):

Summary: MIC26 is a subunit of the 'mitochondrial contact site and cristae organizing system' (MICOS) complex required for crista junction (CJ) formation and was functionally linked to diabetes and modulation of lipid metabolism. In order to understand the role of MIC26 in metabolism, the authors generated MIC26-KO HepG2 cells and investigated the pathways regulated by MIC26 under normo- and hyper-glycemic culture conditions. They employed a multi-omics approach that include transcriptomics, proteomics, targeted metabolomics, and functional assays to document the changes in mRNAs, proteins, and metabolites as a result of MIC26 deletion. Through bioinformatic analyses, they showed that the function of MIC26 is critical in various pathways regulating fatty acid synthesis, oxidation, cholesterol metabolism, and glycolysis. Interestingly, they found an entirely antagonistic effect of lipogenesis in MIC26-KO cells compared to WT cells depending on the glucose concentration of the culture media. In addition, they showed that MIC26 deletion led to a major metabolic rewiring of glutamine utilization as well as oxidative phosphorylation.

Major comments: 1) This is basically a descriptive study that document the transcriptomic, proteomic, and metabolic consequences of lacking MIC26 in normal or high glucose environment. It is data rich but insight poor. The connections between MIC26 as a subunit of MICOS complex and all those metabolic pathways are so tenuous that it is hard to see what to follow up after.

__Response: __We respectfully differ from the reviewer’s opinion that the manuscript is data rich and insight poor. Our study provides significant insights by demonstrating how MIC26, strategically residing in the mitochondrial inner membrane (IM), regulates major cellular pathways.

MIC26 operates in a dual manner:

A) Depending on the nutritional status, normoglycemia or hyperglycemia, MIC26 regulates the glycolysis, lipid, cholesterol metabolism and TCA cycle intermediates in an antagonistic manner. B) Independent of the nutritional status, it regulates glutamine and OXPHOS metabolism.

In addition, based on the suggestions by Reviewer #2, we have tested whether other proteins of MICOS (MIC27 and MIC19) present in two different sub-complexes regulate important metabolic pathways. Using the experimental results achieved (See reply to comments from Reviewer #2), we conclude that MIC26 plays a unique role as metabolic regulator in the IM and this study is therefore important to the general field of metabolism.

2) In the Results section (page 5, line 114-123), the description of the Western blot (WB) analysis appears inconsistent with several blot images of Fig. 1A, which makes the result unconvincing. The authors should select appropriate representative WB images, assuming they have them, to support their claim.

Response: Thank you for the suggestion. Firstly, we have performed additional experiments in this regard, and included the relevant quantification. Secondly, we have replaced some WBs which depict the appropriate quantification.

Further, we also modified the relevant lines in the manuscript stating that ‘Mitochondrial apolipoproteins, MIC26, MIC27, and MIC25 are increased in cells exposed to hyperglycemia’ and not only MIC26 as stated before.

Minor comments:

3) As the functional role of MIC26 in metabolism is the primary focus, the authors should present the results in the figures in the order of WT-N, MIC26KO-N, WT-H, MIC26KO-H for easier comparison.

Response: We understand the reasoning to interchange the two conditions. However, such an endeavour will involve cropping images of WBs, BN-PAGE and Clear native PAGE to better represent the corresponding quantification. This will also involve modifying all figures (data-sets and functional assays) in the whole manuscript. Overall, considering that the benefits of interchanging the order, of the WT-Hyperglycemia and MIC26 KO-Normoglycemia, are relatively minor, we have decided to stick to the original representation of the figures.

Reviewer #1 (Significance (Required)):

Mitochondria play important roles in metabolism and metabolic disorders. This study generated large amount of data relating to the role of a mitochondrial protein MIC26 in metabolism. Mutations in MIC26 have been associated with mitochondrial myopathy, lactic acidosis, and cognition defects (Beninca et al. 2021) and a lethal progeria-like condition (Peifer-Weib et al. 2023). There is also a connection between MIC26 and metabolic disorders. The results of this study will be of interest to researchers in the fields of mitochondrial diseases and metabolic disorders. My field of expertise is mitochondrial disease, proteomics, lipidomics, phospholipid biochemistry.

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

Summary

This study determines the role of MIC26, a mitochondrial component of the MICOS complex, in influencing cellular metabolic status. Using a variety of multiomic profiling techniques, as well as functional assays such as Seahorse-based respirometry, the authors propose that MIC26 regulates a variety of metabolic processes, including lipid and cholesterol homeostasis, glycolysis, fatty acid oxidation, fatty acid synthesis, TCA cycle homeostasis, glutamine metabolism, and general mitochondrial bioenergetics via OxPhos activity and supercomplex formation. The data were generated in MIC26 knockout HepG2 hepatocellular carcinoma cells grown in two nutrient conditions: high glucose (which they term "hyperglycemia") and low glucose (which they term "normoglycemia") DMEM. While the data support the authors' conclusion that loss of MIC26 causes braod metabolic changes across these conditions, the authors do not distinguish whether MIC26 knockout affects metabolism due to its canonical role in MICOS, or whether it acts as "a metabolic rheostat" to directly regulate central cellular fuel pathways, as is claimed in the title of this manuscript. Given the breadth of metabolic alterations seen in the MIC26 KO cells, it seems likely that at least a subset of these changes are indirect, rather than that MIC26 plays a direct regulatory role in the eight distinct metabolic pathways outlined above. Thus, while the data generally support the conclusion that expression of MIC26 is important for metabolic homeostasis, the mechanism(s) by which MIC26 influences cellular metabolism remains unclear and should be further addressed.

Major Comments

1) The authors infer that MIC26 influences cellular metabolism by referencing a handful of papers on MIC26 transcript differences in select metabolic models, metabolic alterations seen in a MIC26 transgenic/overexpression mouse model, or metabolic effects seen due to MIC26 tissue specific KO models. They thus hypothesize that "MIC26 has an unidentified regulatory role under nutrient-enriched conditions" (line 88). They support this observation by showing that MIC26 is increased in high glucose DMEM relative to low glucose DMEM in Figure 1. However, the authors claim, in both the figure legend title as well as the results section header, that "MIC26 is selectively increased in cells exposed to hyperglycemia" (lines 103-4), when their data demonstrate that this is not true. Figure 1B shows that MIC27, MIC26, and MIC25 increase in high glucose relative to low glucose conditions, indicating that MIC26 is not selectively increased, but rather that multiple subunits of MICOS are increased under nutrient-enriched conditions.

Response: We agree to the reviewer’s comment and have now replaced the title in the results section in the manuscript accordingly - ‘Mitochondrial apolipoproteins, MIC26, MIC27, and MIC25 are increased in cells exposed to hyperglycemia’ and not only MIC26 as stated before. Further, in order to strengthen the WB data from Fig 1A & B, we also increased the number of experiments and updated the Fig 1B and also some WBs in Fig 1A to better represent the quantification.

2) This does not suggest that MIC26 does not have an important role in maintaining metabolism under such nutrient conditions, but rather supports a model in which MICOS itself dynamically responds to altered nutrient conditions in cell culture. Interestingly, other MICOS subunits do not change in abundance (e.g., MIC19 and MIC60), which have previously been associated with a separate MICOS subcomplex than MIC26 in yeast (PMID: 33053165). These data may suggest that the nutrient responsive behavior of MIC26 may be due to its assembly within this specific MICOS subcomplex, rather than an independent "unidentified regulatory role under nutrient-enriched conditions".

To test this, the authors should repeat a subset of functional assays (perhaps the Seahorse metabolic assays) in other MICOS deletion cell lines, including one that dynamically changes in expression in a similar manner to MIC26 (e.g., MIC27 KO or MIC25 KO) and one that does not dynamically change in expression in high glucose conditions (e.g., MIC60 or MIC13). In these sets of experiments the authors will be able to distinguish three possibilities:

Model 1: if MICOS in general affects metabolic pathways, similar results will be seen for all MICOS subunit KOS.

Model 2: if the MIC26-specific subcomplex is dynamically regulated to influence cellular metabolism, KO of MIC26 and other subunits of this subcomplex should show similar results, but KO of subunits of the non-dynamic subcomplex (e.g., MIC60) would not show similar phenotypes.

Model 3: If only MIC26 KO, but not other MICOS subunits, show metabolic phenotypes, this would support a MICOS-independent role for MIC26 in influencing cellular metabolism.

Importantly, any of these models are interesting, and testing these models does not invalidate any of the phenotypes presented in the manuscript. Rather, these experiments would assist the reader in understanding the underlying mechanism by which MIC26 loss causes cellular metabolic defects. Furthermore, it is worth stating that performing all experiments with multiple other MICOS cell lines is beyond the scope of the manuscript, but testing effects in select (preferably functional) assays, such as the glycolysis stress test (Fig 3E), FAO Seahorse (Fig 3H-J) glutamine oxidation (Fig 6B), and general stress test (Fig 7E) would be appropriate. Other more defined and easily achievable experiments could also be used to support these claims (e.g., western blots probing for levels of key metabolic regulators).

__Response: __We appreciate the balanced comments of the reviewer who has carefully read and appreciated our manuscript. We also appreciate the constructive criticism of the reviewer who suggested various possible models considering the MIC26 role.

In this endeavour, we have performed the following extensive experiments:

1. Generated MIC19 KOs in HepG2 cells. We had already generated MIC27 KOs during the course of a previous publication (Lubeck et al, 2023) and used them for this publication (Fig S4). WB analyses was performed using the MIC27 KO and MIC19 KO cells.
2. Measured glycolysis function using the glycolysis stress test in MIC27 and MIC19 KOs (Fig S5).
3. Analysed lipid metabolism by imaging and extensively quantifying LD number and BODIPY fluorescence intensities in MIC27 and MIC19 KO cells (Fig S6).
4. Measured glutamine oxidation using Mito Flex fuel tests in MIC27 and MIC19 KOs (Fig S10F).
5. Analysed general mitochondrial respiration by using mito stress kit in MIC27 and MIC19 KOs (Fig S11-E-H). We provide a summary of the above results:

#

Metabolic Pathway

Experiment

MIC26____ KO

MIC27____ KO

MIC19____ KO

1

Glycolysis

Glycolysis stress kit

Glycolytic reserve increased

Glycolytic reserve unchanged

Glycolytic reserve unchanged

2A

Lipid Metabolism

LD number

General increase

No consistent increase in treatment with and without palmitate in normoglycemia and hyperglycemia

No consistent increase in treatment with and without palmitate in normoglycemia and hyperglycemia

2B

Lipid Metabolism

LD (BODIPY) intensities

Presence of antagonistic regulation of LD content

Presence of antagonistic regulation of LD content

Absence of antagonistic regulation of LD content

3

Glutamine oxidation

Mito flex fuel test

No dependency

Dependent

Dependent

4

Steady-state respiration

Mito stress test

Basal respiration increased

Basal respiration unchanged

Basal respiration unchanged

5

Steady-state respiration

Mito stress test

SRC decreased in normoglycemia

SRC unchanged

SRC unchanged

Taking the above summary, we investigated three possibilities considering the role of MIC26:

1. General role of MICOS – whether deletion of any MICOS protein leads to similar phenotype as MIC26 deletion

2. Specific role of MIC26/27/10 subcomplex – whether deletion of any other protein in the MIC26-subcomplex like MIC27 leads to similar results, accompanied by dissimilar results in KOs of any protein belonging to the other MICOS subcomplex (MI19/MIC25/MIC60) and whose protein levels were not changed upon hyperglycemia treatment (MIC19 or MIC60 KOs).

3. MICOS-independent role of MIC26. Considering the various metabolic pathways analysed, we conclude that MIC26 has a MICOS-independent role in regulating major cellular pathways.

Minor Comments 1. The multiomics data as presented in Figure 2 is difficult to interpret. This is mainly driven by the fact that there are mulitpe comparisons that should be communicated (KO v WT, normoglycemia v hyperglycemia, upregulated v. downregulated), but only select enrichment values are shown (e.g., normoglycemia upregulated in 2C and hyperglycemia downregulated in 2D). It took me a long time as a reader to understand what I was looking at because only select analyses are presented. What pathways are upregulated in hyperglycemia in MIC26 KO v. WT?

__Response: __Thank you for pointing this out. We have now represented all the four conditions regarding enrichment analysis as suggested. The antagonistic metabolic regulation is only observed in MIC26 KO cells cultured in normoglycemia (upregulated) when compared to MIC26 KOs cells cultured in hyperglycemia (downregulated). When MIC26 KOs cultured in hyperglycemia (upregulated) were compared with MIC26 KOs cultured in normoglycemia (downregulated), there were also few pathways which showed an antagonistic regulation, but not directly involved with metabolism, relating to apoptosis etc. Hence, we did not focus on these pathways in the current manuscript.

The Treemaps have been shifted to Fig S1. All four Treemaps have been represented instead of the two shown before. In the process, the previous proteomics Fig S1B and S1C depicting antagonism in different pathways have been excluded.

2. Figure 3 would be stronger if expression from all glycolytic proteins were shown instead of only a subset. If the authors are making the claim that MIC26 KO increases glycolytic flux via protein-level upregulation of glycolysis, this could be substantiated at a pathway level. These data could be included in supplemental data if they are difficult to fit into the figure.

__Response: __Thank you. We have now included the data of proteins regulating glycolysis as a new figure (Fig S3C). MIC26 KO cells cultured in normoglycemia had increased levels of aldolase (ALDOA & ALDOC), phosphoglycerate kinase (PGK1) and pyruvate kinase (PKM & PKLR) when compared to control cells. We have included this data in the manuscript.

3. In Figure 4H-J - the raw data for the Seahorse traces should be shown, and OCR should be reported in pmol/min rather than relative percentages so as to help the reader more critically evaluate the data.

__Response: __For the FAO assays, we treat the cells with palmitate or mock (BSA serving as control). The histograms (Fig 4H-J) are represented as such because we normalised the oxygen consumption after palmitate treatment with oxygen consumption of mock-treated cells. We understand the reviewer’s concern and have now included the absolute values of oxygen consumption of FAO assay in an excel sheet (Supplementary Table S5). In addition, we have also included the absolute values for mito stress test and glycolysis assays where the oxygen consumption has been normalised to WT-normoglycemia condition (Supplementary Table S5). The original oxygen consumption curves for glycolysis stress kit (Fig 3E, S5A) and mito stress kits (Fig 7E, S11E) are shown as figures.

4. The majority of the plots are shown with 4 comparisons, but statistical comparisons are often only provided for a subset of comparisons. It is unclear whether statistics were compared across all comparisons and the non-annotated comparisons are not significant, or whether those calculations were not performed. Defining this in the figure, or, better, annotating all relevant comparisons on each graph with "ns" for not significant, would assist the reader with interpreting the data.

__Response: __Thanks for pointing this out. After comparing all meaningful conditions (except WT-N to MIC26 KO-H and WT-H to MIC26 KO-N), only those that were significant were represented using asterisks. We have now mentioned this information in the respective figure legends. We avoided using ‘non-significant (ns)’ in the figure as it would make some figures very crowded as seen from some of our trials.

Reviewer #2 (Significance (Required)):

This study broadly profiles the metabolic defects associated with loss of the MICOS subunit MIC26 in hepatocellular carcinoma cells in variable nutrient conditions (e.g., high glucose and low glucose). As a reviewer with expertise in multiomic profiling of metabolic models, I found the breadth of pathways studied in this manuscript to be impressive. Furthermore, the authors use a variety of techniques, including multiomic profiling, isotopic flux analysis, and functional Seahorse assays to support their conclusions. The study provides a comprehensive analysis of metabolic changes associated with MIC26, and is thus an important advance in profiling how loss of MIC26 (or MICOS in general; see below) affects cellular metabolism in the context of dynamic nutrient changes. However, the claim that "MIC26 is a metabolic rheostat regulating central cellular fuel pathways", as is proposed in the title of the manuscript, is unsubstantiated, as the authors do not test whether loss of MIC26 specifically influences cellular metabolism independent of its role in MICOS. This paper would be significantly strengthened if a subset of functional assays across metabolic pathways were repeated with other MICOS KO cell lines to delineate whether these metabolic effects are direct or indirect.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary

This study determines the role of MIC26, a mitochondrial component of the MICOS complex, in influencing cellular metabolic status. Using a variety of multiomic profiling techniques, as well as functional assays such as Seahorse-based respirometry, the authors propose that MIC26 regulates a variety of metabolic processes, including lipid and cholesterol homeostasis, glycolysis, fatty acid oxidation, fatty acid synthesis, TCA cycle homeostasis, glutamine metabolism, and general mitochondrial bioenergetics via OxPhos activity and supercomplex formation. The data were generated in MIC26 knockout HepG2 hepatocellular carcinoma cells grown in two nutrient conditions: high glucose (which they term "hyperglycemia") and low glucose (which they term "normoglycemia") DMEM. While the data support the authors' conclusion that loss of MIC26 causes braod metabolic changes across these conditions, the authors do not distinguish whether MIC26 knockout affects metabolism due to its canonical role in MICOS, or whether it acts as "a metabolic rheostat" to directly regulate central cellular fuel pathways, as is claimed in the title of this manuscript. Given the breadth of metabolic alterations seen in the MIC26 KO cells, it seems likely that at least a subset of these changes are indirect, rather than that MIC26 plays a direct regulatory role in the eight distinct metabolic pathways outlined above. Thus, while the data generally support the conclusion that expression of MIC26 is important for metabolic homeostasis, the mechanism(s) by which MIC26 influences cellular metabolism remains unclear and should be further addressed.

Major Comments

1. The authors infer that MIC26 influences cellular metabolism by referencing a handful of papers on MIC26 transcript differences in select metabolic models, metabolic alterations seen in a MIC26 transgenic/overexpression mouse model, or metabolic effects seen due to MIC26 tissue specific KO models. They thus hypothesize that "MIC26 has an unidentified regulatory role under nutrient-enriched conditions" (line 88). They support this observation by showing that MIC26 is increased in high glucose DMEM relative to low glucose DMEM in Figure 1. However, the authors claim, in both the figure legend title as well as the results section header, that "MIC26 is selectively increased in cells exposed to hyperglycemia" (lines 103-4), when their data demonstrate that this is not true. Figure 1B shows that MIC27, MIC26, and MIC25 increase in high glucose relative to low glucose conditions, indicating that MIC26 is not selectively increased, but rather that multiple subunits of MICOS are increased under nutrient-enriched conditions. This does not suggest that MIC26 does not have an important role in maintaining metabolism under such nutrient conditions, but rather supports a model in which MICOS itself dynamically responds to altered nutrient conditions in cell culture. Interestingly, other MICOS subunits do not change in abundance (e.g., MIC19 and MIC60), which have previously been associated with a separate MICOS subcomplex than MIC26 in yeast (PMID: 33053165). These data may suggest that the nutrient responsive behavior of MIC26 may be due to its assembly within this specific MICOS subcomplex, rather than an independent "unidentified regulatory role under nutrient-enriched conditions".

To test this, the authors should repeat a subset of functional assays (perhaps the Seahorse metabolic assays) in other MICOS deletion cell lines, including one that dynamically changes in expression in a similar manner to MIC26 (e.g., MIC27 KO or MIC25 KO) and one that does not dynamically change in expression in high glucose conditions (e.g., MIC60 or MIC13). In these sets of experiments the authors will be able to distinguish three possibilities:

Model 1: if MICOS in general affects metabolic pathways, similar results will be seen for all MICOS subunit KOS.

Model 2: if the MIC26-specific subcomplex is dynamically regulated to influence cellular metabolism, KO of MIC26 and other subunits of this subcomplex should show similar results, but KO of subunits of the non-dynamic subcomplex (e.g., MIC60) would not show similar phenotypes.

Model 3: If only MIC26 KO, but not other MICOS subunits, show metabolic phenotypes, this would support a MICOS-independent role for MIC26 in influencing cellular metabolism.

Importantly, any of these models are interesting, and testing these models does not invalidate any of the phenotypes presented in the manuscript. Rather, these experiments would assist the reader in understanding the underlying mechanism by which MIC26 loss causes cellular metabolic defects. Furthermore, it is worth stating that performing all experiments with multiple other MICOS cell lines is beyond the scope of the manuscript, but testing effects in select (preferably functional) assays, such as the glycolysis stress test (Fig 3E), FAO Seahorse (Fig 3H-J) glutamine oxidation (Fig 6B), and general stress test (Fig 7E) would be appropriate. Other more defined and easily achievable experiments could also be used to support these claims (e.g., western blots probing for levels of key metabolic regulators).

Minor Comments

1. The multiomics data as presented in Figure 2 is difficult to interpret. This is mainly driven by the fact that there are mulitpe comparisons that should be communicated (KO v WT, normoglycemia v hyperglycemia, upregulated v. downregulated), but only select enrichment values are shown (e.g., normoglycemia upregulated in 2C and hyperglycemia downregulated in 2D). It took me a long time as a reader to understand what I was looking at because only select analyses are presented. What pathways are upregulated in hyperglycemia in MIC26 KO v. WT?
2. Figure 3 would be stronger if expression from all glycolytic proteins were shown instead of only a subset. If the authors are making the claim that MIC26 KO increases glycolytic flux via protein-level upregulation of glycolysis, this could be substantiated at a pathway level. These data could be included in supplemental data if they are difficult to fit into the figure.
3. In Figure 4H-J - the raw data for the Seahorse traces should be shown, and OCR should be reported in pmol/min rather than relative percentages so as to help the reader more critically evaluate the data.
4. The majority of the plots are shown with 4 comparisons, but statistical comparisons are often only provided for a subset of comparisons. It is unclear whether statistics were compared across all comparisons and the non-annotated comparisons are not significant, or whether those calculations were not performed. Defining this in the figure, or, better, annotating all relevant comparisons on each graph with "ns" for not significant, would assist the reader with interpreting the data.

Significance

This study broadly profiles the metabolic defects associated with loss of the MICOS subunit MIC26 in hepatocellular carcinoma cells in variable nutrient conditions (e.g., high glucose and low glucose). As a reviewer with expertise in multiomic profiling of metabolic models, I found the breadth of pathways studied in this manuscript to be impressive. Furthermore, the authors use a variety of techniques, including multiomic profiling, isotopic flux analysis, and functional Seahorse assays to support their conclusions. The study provides a comprehensive analysis of metabolic changes associated with MIC26, and is thus an important advance in profiling how loss of MIC26 (or MICOS in general; see below) affects cellular metabolism in the context of dynamic nutrient changes. However, the claim that "MIC26 is a metabolic rheostat regulating central cellular fuel pathways", as is proposed in the title of the manuscript, is unsubstantiated, as the authors do not test whether loss of MIC26 specifically influences cellular metabolism independent of its role in MICOS. This paper would be significantly strengthened if a subset of functional assays across metabolic pathways were repeated with other MICOS KO cell lines to delineate whether these metabolic effects are direct or indirect.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary:

MIC26 is a subunit of the 'mitochondrial contact site and cristae organizing system' (MICOS) complex required for crista junction (CJ) formation and was functionally linked to diabetes and modulation of lipid metabolism. In order to understand the role of MIC26 in metabolism, the authors generated MIC26-KO HepG2 cells and investigated the pathways regulated by MIC26 under normo- and hyper-glycemic culture conditions. They employed a multi-omics approach that include transcriptomics, proteomics, targeted metabolomics, and functional assays to document the changes in mRNAs, proteins, and metabolites as a result of MIC26 deletion. Through bioinformatic analyses, they showed that the function of MIC26 is critical in various pathways regulating fatty acid synthesis, oxidation, cholesterol metabolism, and glycolysis. Interestingly, they found an entirely antagonistic effect of lipogenesis in MIC26-KO cells compared to WT cells depending on the glucose concentration of the culture media. In addition, they showed that MIC26 deletion led to a major metabolic rewiring of glutamine utilization as well as oxidative phosphorylation.

Major comments:

This is basically a descriptive study that document the transcriptomic, proteomic, and metabolic consequences of lacking MIC26 in normal or high glucose environment. It is data rich but insight poor. The connections between MIC26 as a subunit of MICOS complex and all those metabolic pathways are so tenuous that it is hard to see what to follow up after.

In the Results section (page 5, line 114-123), the description of the Western blot (WB) analysis appears inconsistent with several blot images of Fig. 1A, which makes the result unconvincing. The authors should select appropriate representative WB images, assuming they have them, to support their claim.

Minor comments:

As the functional role of MIC26 in metabolism is the primary focus, the authors should present the results in the figures in the order of WT-N, MIC26KO-N, WT-H, MIC26KO-H for easier comparison.

Significance

Mitochondria play important roles in metabolism and metabolic disorders. This study generated large amount of data relating to the role of a mitochondrial protein MIC26 in metabolism. Mutations in MIC26 have been associated with mitochondrial myopathy, lactic acidosis, and cognition defects (Beninca et al. 2021) and a lethal progeria-like condition (Peifer-Weib et al. 2023). There is also a connection between MIC26 and metabolic disorders. The results of this study will be of interest to researchers in the fields of mitochodrial diseases and metabolic disorders.

My field of expertise is mitochondrial disease, proteomics, lipidomics, phospholipid biochemistry.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Rebuttal_ Preprint- #RC-2023-02144

First of all we would like to thank the three reviewers for their constructive and positive comments and suggestions, and the time spent in reviewing our manuscript. Their suggestions and comments had contributed to improve our manuscript. We feel the manuscript is much strengthened by this revision.

Reviewer #1 (Evidence, reproducibility and clarity (Required)):

__Summary:____ __The manuscript by Dabsan et al builds on earlier work of the Igbaria lab, who showed that ER-luminal chaperones can be refluxed into the cytosol (ERCYS) during ER stress, which constitutes a pro-survival pathway potentially used by cancer cells. In the current work, they extent these observations and a role for DNAJB12&14 in ERCYS. The work is interesting and the topic is novel and of great relevance for the proteostasis community. I have a number of technical comments:

We thank the reviewer for his/her positive comments on our manuscript.

---

__Major and minor comments: __

1- In the description of Figure 2, statistics is only show to compare untreated condition with those treated with Tg or Tm, but no comparison between condition and different proteins. As such, the statement made by the authors "...DNAJB14-silenced cells were only affected in AGR2 but not in DNAJB11 or HYOU1 cytosolic accumulation" cannot be made.

Answer: We totally agree with the reviewer#1. The aim of this figure is to show that during ER stress, a subset of ER proteins are refluxed to the cytosol. This is happening in cells expressing DNAJB12 and DNAJB14. We are not comparing the identity of the expelled proteins between DNAJB12-KD cells and DNAJB14-KD cells, This is not the scoop of this paper as such the statement was removed.

2- Figure S2C: D11 seems to increase in the cytosolic fraction after Tm and Tg treatment. However, this is not reflected in the text. The membrane fraction also increases in the DKO. Is the increase of D11 in both cytosol and membrane and indication for a transcriptional induction of this protein by Tm/Tg? Again, the authors are not reflecting on this in their text.

Answer: We performed qPCR experiments in control, DNAJB12-KD, DNAJB14-KD and in the DNAJB12/DNAJB14 double knock down cells (in both A549 and PC3 cells) to follow the mRNA levels of DNAJB11. As shown in (Figure S2F-S2N), there is no increase in the mRNA levels of DNAJB11, AGR2 or HYOU1 in the different cells in normal (unstressed conditions). Upon ER stress with tunicamycin or thapsigargin there is a little increase in the mRNA levels of HYOU1 and AGR2 but not in DNAJB11 mRNA levels. On the other hand, we also performed western blot analysis and we did not detect any difference between the different knockdown cells when we analyzed the levels of DNAJB11 compared to GAPDH. Those data are now added as (Figure S2F-S2N).

We must note that although AGR2 and HYOU1 are induced at the mRNA as a result of ER stress, the data with the overexpression of DNAJB12 and DNAJB14 are important as control experiments because when DNAJB12 is overexpressed it doesn’t inducing the ER stress (Figure S3C-S3D). In those conditions there is an increase of the cytosolic accumulation of AGR2, HYOU1 and DNAJB11 despite that there was no induction of AGR2, HYOU1 or DNAJB11 (Figure 3C and Figure 3E, Figure S3, Figure 4, and Figure S4) . Those results argue against the idea that the reflux is a result of protein induction and an increase in the total proteins levels.

3- Figure 2D: Only p21 is quantified. phospho-p53 and p53 levels are not quantified.

---

Answer: We added the quantification of phospho-p53 and the p53 levels to (Figure 2E-G). Additional blots of the P21, phosphor-p53 and p53 now added to FigureS2O.

4- Figure 2D: There appears to be a labelling error

Answer: Yes, the labelling error was corrected.

5- Are there conditions where DNAJB12 would be higher?

Answer: In some cancer types there is a higher DNAJB12, DNAJB14 and SGTA expression levels that are associated with poor prognosis and reduced survival (New Figure S6E-M). The following were added to the manuscript: “Finally, we tested the effect of DNAJB12, DNAJB14, and SGTA expression levels on the survival of cancer patients. A high copy number of DNAJB12 is an unfavorable marker in colorectal cancer and in head and neck cancer because it is associated with poor prognosis in those patients (Figure S6E). A high copy number of DNAJB12, DNAJB14, and SGTA is associated with poor prognosis in many other cancer types, including colon adenocarcinoma (COAD), acute myeloid leukemia (LAML), adrenocortical carcinoma (ACC), mesothelioma (MESO), and Pheochromocytoma and paraganglioma (PCPG) (Figure S6F-M). In uveal melanoma (UVM), a high copy number of the three tested genes, DNAJB12, DNAJB14, and SGTA, are associated with poor prognosis and poor survival (Figure S6I, S6J, and S6M). The high copy number of DNAJB12, DNAJB14, and SGTA is also associated with poor prognosis in many other cancer types but with low significant scores. More data is needed to make significant differences (TCGA database). We suggest that the high expression of DNAJB12/14 and SGTA in those cancer types may account for the poor prognosis by inducing ERCYS and inhibiting pro-apoptotic signaling, increasing cancer cells' fitness.

6- What do the authors mean by "just by mass action"?

Answer: Mass action means increasing the amount of the protein (overexpression). We corrected this in the main text to overexpression.

7- Figure 3C: Should be labelled to indicate membrane and cytosolic fraction. The AGR2 blot in the left part is not publication quality and should be replaced.

Answer: We added the labelling to indicate cytosolic and membrane fractions to Figure 3C. We re-blotted the AGR2, new blot of AGR2 was added.

8- What could be the reason for the fact that DNAJB12 is necessary and sufficient for ERCYS, while DNAJB14 is only necessary?

Answer: Because of their very high homology, we speculate that the two proteins have partial redundancy. Partial because we believe that some of the roles of DNAJB12 cannot be carried by DNAJB14 in its absence. Although they are highly homologous, we expect that they probably have different affinities in recruiting other factors that are necessary for the reflux of proteins.

We further developed around this point in the discussion and the main text.

9- Figure 5A: Is the interaction between SGTA and JB12 UPR-independent?HCS70 seems to show only background binding. The interaction of JB12 with SGTA is not convincing. A better blot is needed.

Answer: In the conditions of Figure 5A, we did not observe any induction of the UPR (Figure S3C-D). Thus, we concluded that in those condition of overexpression, DNAJB12 interacts with SGTA in UPR independent manner.

We repeated this experiment another 3 times with very high number of cells (2X15cm2 culture dishes for each condition) and instead of coimmunoprecipitating with DNAJB12 antibodies we IP-ed with FLAG-beads, the results are very clear as shown in the new Figure 5A compared to Figure S5A.

10- Figure 5B: the expression of DNAJB14 was induced by Tg50, but not by Tg25 or Tm. However, the authors have not commented on this. This should be mentioned in the text and discussed.

Answer: In most of the experiments we did not see an increase in DNAJB14 upon ER stress except in this replicate. To be sure we looked at the DNAJB14 levels upon ER stress by protein and qPCR experiment as shown in new (in the Input of Figure 5 and Figure S5) and (Figure S5H-I). We also added new IP experiments in Figure 5 and Figure S5.

11- Figure 6A: Why is a double knockdown important at all? DNAJB14 does not seem to do much at all (neither in overexpression nor with single knockdown).

Answer: the data shows that DNAJB12 can compensate for the lack of DNAJB14 while DNAJB14 can only partially compensate for some of the DNAJB12 functions. DNAJB12 could have higher affinity to recruit other factor needed for the reflux process and thus the impact of DNAJB12 is higher. In summary, neither DNAJB12 or DNAJB14 is essential in the single knockdown which means that they compensate for each other. In the overexpression experiment, it is enough to have the endogenous DNAJB14 for the DNAJB12 activity. When DNAJB14 is overexpressed at very high levels, we believe that it binds to some factors that are needed for proper DNAJB12 activity (Figure 4 showing that the WT-DNAJB14 inhibits ER-stress induced ER protein reflux when overexpressed). We believe that DNAJB14 is important because only when we knock both DNAJB12 and DNAJB14 we see an effect on the ER-protein reflux. DNAJB14 is part of a complex of DNAJB12/HSC70 and DSGTA.

(DNAJB12 is sufficient while DNAJB14 is not- please refer to point #8 above).

**Referees cross-commenting**

I agree with the comments raised by reviewer 1 about the manuscript. I also agree with the points written in this consultation session. In my opinion, the comments of reviewer 2 are phrased in a harsh tone and thus the reviewer reaches the conclusion that there are "serious" problems with this manuscript. However, I think that the authors could address many of the points of this reviewer in a matter of 3 months easily. For instance, it is easy to control for the expression levels of exogenous wild type and mutant D12 and compare it to the endogenous one (point 3). This is a very good point of this reviewer and I agree with this experiment. Likewise, it is easy to provide data about the levels of AGR2 to address the concern whether its synthesis is affected by D12 and D14 overexpression. Again, an excellent suggestion, but no reason for rejecting the story. As for not citing the literature, I think this can also easily be addressed and I am sure that this is just an oversight and no ill intention by the authors. __Overall, I am unable to see why the reviewer reaches such a negative verdict about this work. With proper revisions that might take 3 months, I think the points of all reviewers can be addressed. __

Reviewer #1 (Significance (Required)):

Significance: The strength of the work is that it provides further mechanistic insight into a novel cellular phenomenon (ERCYS). The functions for DNAJB12&14 are unprecedented and therefore of great interest for the proteostasis community. Potentially, the work is also of interest for cancer researchers, who might capitalize of the ERCYS to establish DNAJB12/14 as novel therapeutic targets. The major weaknesses are as follows: (i) the work is limited to a single cell line. To better probe the cancer relevance, the work should have used at least a panel of cell lines from one (or more) cancer entity. Ideally even data from patient derived samples would have been nice. Having said this, I also appreciate that the work is primarily in the field of cell biology and the cancer-centric work could be done by others. Certainly, the current work could inspire cancer specialists to explore the relevance of ERCYS. (ii) No physiological or pathological condition is shown where DNAJB12 is induced or depleted.

Answer: We previously showed that ERCYS is conserved in many different cell lines including A549, MCF7, GL-261, U87, HEK293T, MRC5 and others and is also conserved in murine models of GBM (GL-261 and U87 derived tumors) and human patients with GBM (Sicari et al. 2021). Here, we tested the reflux process and the IP experiments in many different cell lines including A549, MCF-7, PC3 and Trex-293 cells. We also added new fractionation experiment in DNAJB12 and DNAJB14 -depleted MCF-7, PC3 and A549 cells. We added all those data to the revised version.

We also added survival curves from the TCGA database showing that high copy number of DNAB12, DNAJB14 and SGTA are associated with poor prognosis compared to conditions where DNAJB12, DNAJB14, and SGTA are at low copy number (Figure S6E-M). Finally, we included immunofluorescent experiment to show that the interaction between the refluxed AGR2 and the cytosolic SGTA occurs in tumors collected from patients with colorectal cancer patients (Figure S5F-G) compared to non-cancerous tissue.

This study is highly significant and is relevant not only to cancer but for other pathways that may behave in similar manner. For instance, DNAJB12 and DNAJB14 are part of the mechanism that is used by non-envelope viruses to escape the ER to the cytosol. Thus, the role of those DNAJB proteins seems to be mainly in the reflux of functional (not misfolded) proteins from the ER to the cytosol. We reported earlier that the UDP-Glucose-Glucosyl Transferase 1 (UGGT1) is also expelled during ER stress. UGGT1 is important because it is redeploy to the cytosol during enterovirus A71 (EA71) infection to help viral RNA synthesis (Huang et al, 2017). This redeployment of EAA71 is similar to what happens during the reflux process because on one hand, UGGT1 exit the ER by an ER stress mediated process (Sicari et al. 2021) and it is also a functional in the cytosol as a proteins which help viral RNA synthesis ((Huang et al, 2017). All those data showing that there is more of DNAJB12, DNAJB14, DNAJC14, DNAJC30 and DNAJC18 that still needs to be explored in addition to what is published. Thus, we suggest that viruses hijacked this evolutionary conserved machinery and succeeded to use it in order to escape the ER to the cytosol in a manner that depends on all the component needed for ER protein reflux.

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

The authors present a study in which they ascribe a role for a complex containing DNAJB12/14-Hsc70-SGTA in facilitating reflux of a AGR2 from the ER to cytosol during ER-stress. This function is proposed to inhibit wt-P53 during ER-stress.

Concerns: 1. The way the manuscript is written gives the impression that this is the first study about mammalian homologs of yeast HLJ1, while there are instead multiple published papers on mammalian orthologs of HLJ1. Section 1 and Figure 1 of the results section is redundant with a collection of previously published manuscripts and reviews. The lack of proper citation and discussion of previous literature prevents the reader from evaluating the results presented here, compared to those in the literature.

Answer: We highly appreciate the reviewer’s comments. This paper is not to show that DNAJB12 and DNAJB14 are the orthologues of HLJ-1 but rather to show that DNAJB12 and DNAJB14 are part of a mechanism that we recently discovered and called ERCYS that cause proteins to be refluxed out of the ER. A mechanism that is regulated in by HLJ-1 in yeast. ERCYS is an adaptive and pro-survival mechanism that results in increased chemoresistance and survival in cancer cells. The papers that reviewer #2 refer to are the ones that report DNAJB12 can replace some of the ER-Associated Degradation (ERAD) functions of HLJ-1 in degradation of membranal proteins such as CFTR. These two mechanism are totally different and the role of the yeast HLJ-1 in degradation of CFTR is not needed for ERCYS. This is because we previously showed that the role of the yeast HLJ-1 and probably its orthologues in ERCYS is independent of their activity in ERAD(Igbaria et al. 2019). Surprisingly, the role of HLJ-1 in refluxing the ER proteins is not only independent of the reported ERAD-functions of HLJ1 and the mammalian DNAJBs but rather proceeds more rigorously when the ERAD is crippled (Igbaria et al. 2019). This role of DNAJBs is unique in cancer cells and is responsible in regulating the activity of p53 during the treatment of DNA damage agents.

In our current manuscript we show by similarity, functionality, and topological orientation, that DNAJB12 and DNJB14 may be part of a well conserved mechanism to reflux proteins from the ER to the cytosol. A mechanism that is independent of DNAJB12/14’s reported activity in ERAD(Grove et al. 2011; Yamamoto et al. 2010; Youker et al. 2004). In addition, DNAJB12 and DNAJB14 facilitate the escape of non-envelope viruses from the ER to the cytosol in similar way to the reflux process(Goodwin et al. 2011; Igbaria et al. 2019; Sicari et al. 2021). All those data show that HLJ-1 reported function may be only the beginning of our understanding on the role that those orthologues carry and that are different from what is known about their ERAD function.

Action: We added the references to the main text and discussed the differences between the reported DNAJB12 and HLJ-1 functions to the function of DNAJB12, DNAJB14 and the other DNAJ proteins in the reflux process. We also developed around this in the discussion.

The conditions used to study DNAJB12 and DNAJ14 function in AGR2 reflux from the ER do not appear to be of physiological relevance. As seen below they involve two transfections and treatment with two cytotoxic drugs over a period of 42 hours. The assay for ERCY is accumulation of lumenal ER proteins in a cytosolic fraction. Yet, there is no data or controls that describe the path taken by AGR2 from the ER to cytosol. It seems like pleotropic damage to the ER due the experimental conditions and accompanying cell death could account for the reported results?

Transfection of cells with siRNA for DNAJB12 or DNAJB14 with a subsequent 24-hour growth period.

Transfection of cells with a p53-lucifease reporter.

Treatment of cells with etoposide for 2-hours to inhibit DNA synthesis and induce p53. D. Treatment of cells for 16 hours with tunicamycin to inhibit addition of N-linked glycans to secretory proteins and cause ER-stress.

Subcellular fractionation to determine the localization of AGR2, DNAJB11, and HYOU1

KD of DNAJB12 or DNAJB14 have modest if any impact on AGR2 accumulation in the cytosol. There is an effect of the double KD of DNAJB12 or DNAJB14 on AGR2 accumulation in the cytosol. Yet there are no western blots showing AGR2 levels in the different cells, so it is possible that AGR2 is not synthesized in cells lacking DNAJB12 and DNAKB14. The lack of controls showing the impact of single and double KD or DNAJB12 and DNAJB14 on cell viability and ER-homeostasis make it difficult to interpret the result presented. How many control versus siRNA KD cells survive the protocol used in these assays?

---

Answer: Despite the long protocol we see differences between the control cells and the DNAJB-silenced cells in terms of the quantity of the refluxed proteins to the cytosol. The luciferase construct was used to assess the activity of p53 so the step of the second transfection was used only in experiments were we assayed the p53-luciferase activity. The rest of the experiments especially those where we tested the levels of p53 and P21 levels, were performed with one transfection. Moreover, all the experiments with the subcellular protein fractionation were performed after one transfection without the second transfection of the p53-Luciferase reporter. Finally, the protocol of the subcellular protein fractionation requires first to trypsinize the cells to lift them up from the plates, at the time of the experiment the cells were almost at 70-80% confluency and in the right morphology under the microscope.

Here, we performed XTT assay and Caspase-3 assay to asses cell death at the end of the experiment and before the fractionation assay. We did not observe any differences at this stage between the different cell lines (Figure-RV1 for reviewers Only). This can be explained by the fact that we use low concentrations of Tm and Tg for short time of 16 hour after the pulse of etoposide.

Finally, the claim that and ER-membrane damage result in a mix between the ER and cytosolic components is not true for the following reasons: (1) In case of mixing we would expect that GAPDH levels in the membrane fraction will be increased and that we do not see, and (2) we used our previously described transmembrane-eroGFP (TM-eroGFP) that harbors a transmembrane domain and is attached to the ER membrane facing the ER lumen. The TM-eroGFP was found to be oxidized in all conditions tested. Those data argue against a rupture of the ER membrane which can results in a mix of the highly reducing cytosolic environment with the highly oxidizing ER environment by the passage of the tripeptide GSH from the cytosol to the ER. All those data argue against (1) cell death, and (2) rupture of the ER membrane. Figure RV1 Reviewers Only.

Moreover, as it is shown in Figure S2, AGR2 is found in the membrane fraction in all the four different knock downs, thus it is synthesized in all of them. Moreover, we assayed the mRNA levels of AGR2 in all the knockdowns and we so that they are at the same levels in all the 4 different conditions and still AGR2 mRNA levels increase upon ER stress in all of the 4 knockdown cells in different backgrounds (Figure S2F-N).

In Figure 3 the authors overexpress WT-D12 and H139Q-D12 and examine induction of the p53-reporter. There are no western blots showing the expression levels of WT-D12 and H139Q-D12 relative to endogenous DNAJB12. HLJ1 stands for high-copy lethal DnaJ1 as overexpression of HLJ1 kills yeast. The authors present no controls showing that WT-D12 and H139-D12 are not expressed at toxic levels, so the data presented is difficult to evaluate.

Answer: The expression levels of the overexpression of DNAJB12 and DNAJB14 were present in the initial submission of the manuscript as Figure S3A and S3B. The data showing the relationship between the expression degree and the viability were also included in the initial submission as Figure S3C (Now S3H).

There is no mechanistic data used to help explain the putative role DNAJB12 and DNAJB14 in ERCY? In Figure 4, why does H139Q JB12 prevent accumulation of AGR2 in the cytosol? There are no westerns showing the level to which DNAJB12 and DNAJB14 are overexpressed.

---

Answer: The data showing the levels of DNAJB12 compared to the endogenous were present in the initial submission as Figure S3A and S3B.

We suggest a mechanism by which DNAJB12 and DNAJB14 interact (Figure 5 and Figure S5) and oligomerize to expel those proteins in similar way to expelling non-envelope viruses to the cytosol. Thus, when expressing the mutant DNAJB12 H139Q may indicate that the J-domain dead-mutant can still be part of the complex but affects the J-domain activity in this oligomer and thus inhibit ER-protein reflux. In other words, we showed that the H139Q exhibits a dominant negative effect when overexpressed. Moreover, here we added another IP experiment in the D12/D14-DKD cells to show that in the absence of DNAJB12 and DNAJB14, SGTA cannot bind the ER-lumenal proteins because they are not refluxed (Figure 5 and Figure S5). Those data indicate that in order for SGTA bind the refluxed proteins they have to go through the DNAJB12 and DNAJB14 and their absence this interaction does not occur. This explanation was also present in the discussion of the initial submission.

Mechanistically, we show that AGR2 interacts with DNAJB12/14 which are necessary for its reflux. This mechanism involves the functionality of cytosolic HSP70 chaperones and their cochaperones (SGTA) proteins that are recruited by DNAJB12 and 14. This mechanism is conserved from yeast to mammals. Moreover, by using the alpha-fold prediction tools, we found that AGR2 is predicted to interact with SGTA in the cytosol by the interaction between the cysteines of SGTA and AGR2 in a redox-dependent manner.

**Referees cross-commenting**

__ __ I appreciate the comments of the other reviewers. I agree that the authors could revise the manuscript. Yet, based on my concerns about the physiological significance of the process under study and lack of scholarship in the original draft, I would not agree to review a revised version of the paper.

Answer: Regards the physiological relevance, we showed in our previous study (Sicari et al. 2021) how relevant is ERCYS in human patients of GBM and murine model of GBM. ERCYS is conserved from yeast to human and is constitutively active in GL-261 GBM model, U87 GBM model and human patients with GBM (Sicari et al. 2021). Here, extended that to other tumors and showed that DNAJB12, DNAJB14 and SGTA high levels are associated with poor prognosis in many cancer types (Figure S6). We also show some data from to show the relevance and added data showing the interaction of SGTA with AGR2 in CRC samples obtained from human patients compared to healthy tissue (Figure S5). This study is highly significant and is relevant not only to cancer but for other pathways that may behave in similar manner. For instance, DNAJB12 and DNAJB14 are part of the mechanism that is used by non-envelope viruses to escape the ER to the cytosol. Thus, the role of those DNAJB proteins seems to be mainly in the reflux of functional (not misfolded) proteins from the ER to the cytosol. We reported earlier that the UDP-Glucose-Glucosyl Transferase 1 (UGGT1) is also expelled during ER stress. UGGT1 is important because it is redeploy to the cytosol during enterovirus A71 (EA71) infection to help viral RNA synthesis (Huang et al, 2017). This redeployment of EAA71 is similar to what happens during the reflux process because on one hand, UGGT1 exit the ER by an ER stress mediated process (Sicari et al. 2021) and it is also a functional in the cytosol as a proteins which help viral RNA synthesis ((Huang et al, 2017). All those data showing that there is more of DNAJB12, DNAJB14, DNAJC14, DNAJC30 and DNAJC18 that still needs to be explored in addition to what is published. We suggest that viruses hijacked this evolutionary conserved machinery and succeeded to use it in order to escape.

We appreciate the time spent to review our paper and we are sorry that the reviewer reached such verdict that is also not understood by the other reviewers. Most of the points raised by reviewer 2 were already addressed and explained in the initial submission, anyways we appreciate the time and the comments of reviewer #2 on our manuscript.

Reviewer #2 (Significance (Required)):

Overall, there are serious concerns about the writing of this paper as it gives the impression that it is the first study on higher eukaryotic and mammalian homologs of yeast HLJ1. The reader is not given the ability to compare the presented data to related published work. There are also serious concerns about the quality of the data presented and the physiological significance of the process under study. In its present form, this work does not appear suitable for publication.

Answer: Again we thank reviewer #2 for giving us the opportunity to explain how significant is this manuscript especially for people who are less expert in this field. The significance of this paper (1) showing a the unique role of DNAJB12 and DNAJB14 in the molecular mechanism of the reflux process in mammalian cells (not their role in ERAD), (2) showing the implication of other cytosolic chaperones in the process including HSC70 and SGTA (3), our alpha-fold prediction show that this process may be redox dependent that implicate the cysteines of SGTA in extracting the ER proteins, (4) overexpression of the WT DNAJB12 is sufficient to drive this process, (5) mutation in the HPD motif prevent the reflux process probably by preventing the binding to the cytosolic chaperones, and (6) we need both DNAJB12 and DNAJB14 in order to make the interaction between the refluxed ER-proteins and the cytosolic chaperones occur.

In Summary, this study is highly significant in terms of physiology, we previously reported that ERCYS is conserved in mammalian cells and is constitutively active in human and murine tumors (Sicari et al. 2021). Moreover, DNAJB12 and DNAJB14 are part of the mechanism that is used by non-envelope viruses to escape the ER to the cytosol in a mechanism that is similar to reflux process (Goodwin et al. 2011; Goodwin et al. 2014). Thus, the role of those DNAJB proteins seems to be mainly in the reflux of functional proteins from the ER to the cytosol, viruses used this evolutionary conserved machinery and succeeded to use in order to escape. This paper does not deal with the functional orthologues of the HLJ-1 in ERAD but rather suggesting a mechanism by which soluble proteins exit the ER to the cytosol.

__Reviewer #3 (Evidence, reproducibility and clarity (Required)):____ __

Summary: Reflux of ER based proteins to the cytosol during ER stress inhibits wt-p53. This is a pro-survival mechanism during ER stress, but as ER stress is high in many cancers, it also promotes survival of cancer cells. Using A549 cells, Dabsan et al. demonstrate that this mechanism is conserved from yeast to mammalian cells, and identify DNAJB12 and DNAJB14 as putative mammalian orthologues of yeast HLJ1.

This paper shows that DNAJB12 and 14 are likely orthologues of HLJ1 based on their sequences, and their behaviour. The paper develops the pathway of ER-stress > protein reflux > cytosolic interactions > inhibition of p53. The authors demonstrate this nicely using knock downs of DNAJB12 and/or 14 that partially blocks protein reflux and p53 inhibition. Overexpression of WT DNAJB12, but not the J-domain inactive mutant, blocks etoposide-induced p53 activation (this is not replicated with DNAJB14) and ER-resident protein reflux. The authors then show that DNAJB12/14 interact with refluxed ER-resident proteins and cytosolic SGTA, which importantly, they show interacts with the ER-resident proteins AGR2, PRDX4 and DNAJB11. Finally, the authors show that inducing ER stress in cancer cell lines can increase proliferation (lost by etoposide treatment), and that this is partially dependent on DNAJB12/14.

This is a very interesting paper that describes a nice mechanism linking ER-stress to inhibition of p53 and thus survival in the face of ER-stress, which is a double edged sword regarding normal v cancerous cells. The data is normally good, but the conclusions drawn oversimplify the data that can be quite complex. The paper opens a lot of questions that the authors may want to develop in more detail (non-experimentally) to work on these areas in the future, or alternatively to develop experimentally and develop the observations further. There are only a few experimental comments that I make that I think should be done to publish this paper, to increase robustness of the work already here, the rest are optional for developing the paper further.

We thank the reviewer for his/her positive comments His/her comments contributed to make our manuscript stronger.

__Major comments:____ __

1. Number of experimental repeats must be mentioned in the figure legends. Figures and annotations need to be aligned properly

__Answer____: __All experiments were repeated at least 3 times. We added the number of repeats on each figure in the figures legends

Results section 2:

No intro to the proteins you've looked at for relocalization. Would be useful to have some info on why you chose AGR2. Apart from them being ER-localized, do they all share another common characteristic? Does ability to inhibit p53 vary in potency?

Answer: We previously showed that AGR2 is refluxed from the ER to the cytosol to bind and inhibit wt-p53 (Sicari et al. 2021). Here, we used AGR2 because, (1) we know that AGR2 is refluxed from the ER to the cytosol, and (2) we know which novel functions it gains in the cytosol so we are able to measure and provide a physiological significance of those novel functions when the levels of DNAJB12 and DNAJB14 are altered. Moreover, we used DNAJB11 (41 kDa) and HYOU1 (150 kDa) proteins to show that alteration in DNAJB12 or DNAJB14 prevent the reflux small, medium and large sized proteins. We added a sentence in the discussion stating that DNAJB12/14 are responsible for the reflux of ER-resident proteins independently of their size. We also added in the result section that we are looking at proteins of different sizes and activities.

---

What are the roles of DNAJB12/14 if overexpression can induce reflux? Does it allow increased binding of an already cytosolic protein, causing an overall increase in an interaction that then causes inhibition of p53? What are your suggested mechanisms?

Answer: Previously it was reported that over-expression of DNAJB12 and DNAJB14 tend to form membranous structures within cell nuclei, which was designate as DJANGOS for DNAJ-associated nuclear globular structures(Goodwin et al. 2014). Because those structures which contain both DNAJB12 and DNAJB14 also form on the ER membrane (Goodwin et al. 2014), we speculate that during stress DNAJB12/14 overexpression may facilitate ERCYS. Interestingly, those structures contain Hsc70 and markers of the ER lumen, the nuclear and ER and nuclear membranes (Goodwin et al. 2014).

The discussion was edited accordingly to further strengthen and clarify this point

Fig3: A+B show overexpression of individual DNAJs but not combined. As you go on to discuss the effect of the combination on AGR2 reflux, it would be useful to include this experimentally here.

Answer: This is a great idea, we tried to do it for long time. Unfortunately when we used cells overexpress DNAJB12 under the doxycycline promoter and transfect with DNAJB14 plasmid expressing DNAJB14 under the CMV promoter, most of the cells float within 24 hours compared to cells transfected with the empty vector alone or with DNAJB14-H136Q. We also did overexpression of DNAJB14 in cells with DNAJB12 conditional expression and also were lethal in Trex293T cells and A549-cells.

Fig 3C: Subfractionation of cells shows AGR2 in the cytosol of A549 cells. The quality of the data is good but the bands are very high on the blot. For publication is it possible to show this band more centralized so that we are sure that we are not missing bands cut off in the empty and H139Q lanes?

Also, you have some nice immunofluorescence in the 2021 EMBO reports paper, is it possible to show this by IF too? It is not essential for the story, but it would enrich the figure and support the biochemistry nicely. Also it is notable that the membrane fraction of the refluxed proteins doesn't appear to have a decrease in parallel (especially for AGR2). Is this because the % of the refluxed protein is very small? Is there a transcriptional increase of any of them (the treatments are 12+24 h so it would be enough time)? This could be a nice opportunity to discuss the amount of protein that is refluxed, whether this response is a huge emptying of the ER or more like a gentle release, and also the potency of the gain of function and effect on p53 vs the amount of protein refluxed. This latter part isn't essential but it would be a nice element to expand upon.

Answer: We re-blotted the AGR2 again, new blot of AGR2 was added. More blots also are added in Figure S2, the text is edited accordingly.

In new Figure S5 we added immunofluorescence experiment from tumors and non-tumors tissues obtained from Colorectal cancer (CRC) patients showing that the interaction between SGTA and the refluxed AGR2 also occurs in more physiological settings. It is also to emphasize that the suggested mechanism that implicates SGTA is also valid in CRC tumors.

We performed qPCR experiments in control, DNAJB12-KD, DNAJB14-KD and in the DNAJB12/DNAJB14 double knock down cells (in both A549 and PC3 cells) to follow the mRNA levels of DNAJB11. As shown in the Figure S2F-N, there is no increase in the mRNA levels of DNAJB11, AGR2 or HYOU1 in the different cells in normal (unstressed conditions). Upon ER stress with tunicamycin or thapsigargin there is a little increase in the mRNA levels of HYOU1 and AGR2 but in DNAJB11 mRNA levels. On the other hand, we also performed western blot analysis and we did not detect any difference between the different knockdown cells when we analyzed the levels of DNAJB11 compared to GAPDH. Those data are now added to Figure S2F-N. We must note that in AGR2 and HYOU1 are induced at the mRNA as a result of ER stress. The data with the overexpression of DNAJB12 and DNAJB14 are important control experiment where we show a reflux when DNAJB12 is overexpressed without inducing the ER stress (Figure 3, Figure 4, and Figure S3). In those conditions no induction of AGR2, HYOU1 or DNAJB11 were observed. Those results argue against the reflux as a result of protein induction and the increase in the proteins levels.

The overall protein levels in steady state are function of how much proteins are made, degraded and probably secreted outside the cell. We do see in Figure S2 under ER stress there are some differences in the levels of the mRNA, moreover, from our work in yeast we showed that the expelled proteins have very long half-life in the cytosol (Igbaria et al. 2019). Because it is difficult to assay how many of the mRNA is translated and how much of it is stable/degraded and the stability of the cytosolic fraction vs the ER, it is hard to interpret on the stability and the levels of the proteins.

Those data are now added to the manuscript, the text is edited accordingly.

You still mention DNAJB12 and 14 as orthologues, even though DNAJB14 has no effect on p53 activity when overexpressed. Do you think that this piece of data diminishes this statement?

Answer: The fact that DNAJB12 and DNAJB14 are highly homologous and that only the double knockdown has a great effect on the reflux process may indicate that they are redundant. Moreover, because only DNAJB12 is sufficient may indicate that some of DNAJB12 function cannot be carried by DNAJB14. In one hand they share common activities as shown in the double knock down and on the other hand DNAJB12 has a unique function that may not be compensated by DNAJB14 when overexpressed.

__ __ Fig 3D/F: Overexpression of DNAJB14 induces reflux of DNAJB11 at 24h, what does this suggest? Does this indicate having the same role as DNAJB12 but less potently? What's your hypothesis?

Answer: ERCYS is new and interesting phenomenon and the redistribution of proteins to the cytosol has been documented lately by many groups. Despite that we still do not know what is the specificity of DNAJB12 and DNAJB14 to the refluxed proteins. DNAJB11 is glycosylated protein and now we are testing whether other glycosylated proteins prefer the DNAJB14 pathway or not. This data is beyond the scope of this paper

"This suggests that the two proteins may have different functions when overexpressed, despite their overlapping and redundant functions" What does it suggest about their dependence on each other? If overexpression of WT DNAJB12 inhibits Tg induced reflux, is it also blocking the ability of DNAJB14 to permit flux?

Answer: We hypothesize that it is all about the stichometry and the ratios between proteins. When we overexpress DNAJB14 (the one that is not sufficient to cause reflux it may hijack common components and factor by non-specifically binding to them. Those factors may be needed for DNAJB12 to function properly (Like the dominant negative effect of the DNAJB12-HPD mutant for instance). On the other hand, DNAJB12 may have higher affinity for some cytosolic partner and thus can do the job when overexpressed. Here, we deal with the DNAJB12/DNAJB14 as essential components of the reflux process, yet we need to identify the interactome of each of the proteins during stress and the role of the other DNAJ proteins that also share some of the topological and structural similarity to DNAJB12, DNAJB14 and HLJ-1 (DNAJC30, DNAJC14, and DNAJC18). We edited the text accordingly and integrated this in the discussion.

__ __ Fig 4: PDI shown in blots but not commented on in text. Then included in the schematics. Please comment in the text.

Answer: We commented PDI in the text.

Fig 4F: Although the quantifications of the blots look fine, the blot shown does not convincingly demonstrate this data for AGR2. The other proteins look fine, but again it could be useful to see the individual means for each experiment, or the full gels for all replicates in a supplementary figure.

Answer: the other two repeats are in Figure S4

__ __Results section 3

Fig 5A, As there is obviously a difference between DNAJB12/14 it would be useful to do the pulldown with DNAJB14 too. Re. HSC70 binding to DNAJB12 and 14, the abstract states that DNAJB12/14 bind HSC70 and SGTA through their cytosolic J domains. Fig 5 shows pulldowns of DNAJB12 with an increased binding of SGTA in FLAG-DNAJB12 induced conditions, but the HSC70 band does not seem to be enriched in any of the conditions, including after DNAJB12 induction. This doesn't support the statement that DNAJB12 binds HSC70. In fact, in the absence of a good negative control, this would suggest that the HSC70 band seen is not specific. There is also no data to show that DNAJB14 binds HSC70. I recommend including a negative condition (ie beads only) and the data for DNAJB14 pulldown.

Answer: In Figure 5A we used the Flp-In T-REx-293 cells as it is easier to control and to tune up and down the expression levels of DNAJB12 and DNAJB14. According to new Figure S5A, DNAJB12 binds at the basal levels to HSC70 all the time. It was also surprising for us not to see the differences in the overexpression and we relate that to the fact that all the HSC70 are saturated with DNAJB12. In order to better assay that we repeated the IP in Figure 5A but instead of the IP with DNAJB12, we IP-ed with FLAG antibodies to selectively IP the transfected DNAJB12. As shown in the new Fig 5A, the increase of DNAJB12-FLAG is accompanied with an increase in the binding of HSC70.

We further tested the interaction between DNAJB12, DNAJB14 and HSC70 during ER stress in cancer cells. In those cells we found that DNAJB12 and DNAJB14 bind to HSC70 and they recruit SGTA upon stress. We also tested the binding between DNAJB12 and DNAJB14, in unstressed conditions, there was a basal binding between both, this interaction was stronger during ER stress. Those data are now added to Figure 5 and Figure S5 and the discussion was edited accordingly.

The binding of DNAJB12 to SGTA under stress conditions in Fig5B looks much more convincing than SGTA to DNAJB12 in Fig 5A. Bands in all blots need to be quantified from 3 independent experiments, and repeated if not already n=3. If this is solely a technical difference, please explain in the text.

The conclusions drawn from this interaction data are important and shold be elaborated upon to support th claims made in the paper. The authors may also chose to expand the pulldowns to demonstrate their claims made on olidomerisation of DNAJB12 and 14 here. It is also clear that the interaction data of the SGTA with ER-resident proteins AGR2, PRDX4 and DNAJB11 is strong. The authors may want to draw on this in their hypotheses of the mechanism. I would imagine a complex such as DNAJB14/DNAJB12 - SGTA - AGR2/PRDX4/DNAJB11 would be logical. Have any experiments been performed to prove if complexes like this would form?

Answer: In Figure 5A we used the Flp-In T-REx-293 cells as it is easier to control and to tune up and down the expression levels of DNAJB12 and DNAJB14. T-REx-293 are highly sensitive to ER stress, they do not die (as we did not observe apoptosis markers to be elevated) but they float and can regrow after the stress is gone. In Figure 5B we are using ER stress without the need to express DNAJB12 in A549 cell line. In order to further verify those data, we repeated the IP in another cell line as well to confirm the data in 5B. We also repeated the IP in 5A with anti-FLAG antibody to improve the IP and to specifically map he interaction with the overexpressed FLAG-DNAJB12 (discussed above). All experiments were done in triplicates and added to Figure 5 and Figure S5.

We agree with the reviewer on the complex between the refluxed proteins and SGTA. We believed that SGTA may form a complex with other refluxed ER-proteins but we were unable to see an interaction between AGR2-DNAJB11 in the cytosolic fraction or between AGR2-PRDX4 in the conditions tested in the cytosolic fraction. We could not do this in the whole cell lysate because those proteins bind each other in the ER. Finally, our structural prediction using Alpha-fold suggests that the interaction between SGTA and the refluxed AGR2 (and probably others) is redox depending and that it requires disulfide bridge between cysteine 81 on AGR2 and cysteine 153 on SGTA. Thus, we hypothesize that SGTA binds one refluxed protein at the time.

We repeated the figure with improvement: (1) using more cells in order to increase the amount of IP-ed proteins and to overcome the problem of the faint bands, (2) performing the IP with the FLAG antibodies instead of the DNAJB12 endogenous antibodies.

Fig 5B: It is clear that DNAJB12 interacts with SGTA. The authors state that DNAJB14 also interacts with SGTA under normal and stress conditions, but the band in 25/50 Tg is very feint. Why would there be stronger binding at the 2 extremes than during low stress induction? In the input, there is a much higher expression of DNAJB14 in 50 Tg. What does this say about the interaction? Is there an effect of ER stress on DNAJB14 expression? A negative control should be included to show any background binding, such as a "beads only" control

__Answer: __DNAJB14 does not change with ER stress as shown in the Ips (Input) and in the qPCR experiment in Figure S5I. We added beads only control, we also added new Ips to assess the binding between DNAJB14 and DNAJB12, and between DNAJB14-SGTA. All the new Ips and controls now added as Figure 5 and Figure S5.

Fig 5C data is sound, although a negative control should be included.

Answer: Negative control was added in Figure S5.

__Results section 4____ __

Fig 6A-B: Given that there is the complexity of overexpression v KD of DNAJB12 v 14 causing similar effects on p53 actvity (Fig 2 v 3), it would be interesting to see whether the effect of overexpression mirrors the results in Fig 6A. Is it known what SGTA overexpression does (optional)?

Answer: In the overexpression system, cells overexpressing DNAJB12 start to die between 24-48 hours as shown in Figure S3C. Thus, it is difficult to assay the proliferation of these cells in those conditions. On the other hand, overexpression of Myc-tagged SGTA in A549 cells, MCF7 or T-ReX293 did not show any reflux of ER-proteins to the cytosol and it didn’t show any significant changes in the proliferation index (Figure Reviewers only RV2).

Fig 6D: resolution very low

Answer: Figure 6D was changed

__ __ Fig 6C-D: There is an interesting difference though between the proposed cytosolic actions of the refluxed proteins. You show that AGR2, PRDX4 and DNAJB11 all bind to SGTA in stress conditions, but in the schematics you show: DNAJB11 binding to HSC70 through SGTA (not shown in the paper), then also PDIA1, PDIA3 binding to SGTA and AGR2 binding to SGTA. What role does SGTA have in these varied reactions? Sometimes it is depicted as an intermediate, sometimes a lone binder, what is its role as a binder? It should be clarified which interactions are demonstrated in the paper (or before) and which are hypothesized in a graphical way (eg. for hypotheses dotted outlines or no solid fill etc). The schematics also suggest that DNAJB14 binding to HSC70 and SGTA is inducible in stress conditions, as is PDIA3, which is not shown in the paper. Discussion "In cancer cells, DNAJB12 and DNAJB14 oligomerize and recruit cytosolic chaperones and cochaperones (HSC70 and SGTA) to reflux AGR2 and other ER-resident proteins and to inhibit wt-p53 and probably different proapoptotic signaling pathways (Figure 5, and Figure 6C-6D)." You havent shown oligomerisation between DNAJB12/14. Modify the text to make it clear that it is a hypothesis.

Answer: We removed “oligomerize” from the text and added that it as a hypothesis. Figure (C-D) also were changed to be compatible with the text.

Minor comments:

__ __ It would be useful to have page or line numbers to help with document navigation, please include them. Typos and inconsistency in how some proteins are named throughout the manuscript

Answer: Page numbers and line numbers are added. Typos are corrected

Title: Include reference to reflux. Suggest: "chaperone complexes (?proteins) reflux from the ER to cytosol..." I presume it would be more likely that the proteins go separately rather than in complex. Do you have any ideas on the size range of proteins that can undergo this process?

Answer: this is true, proteins may cross the ER membrane separately and then be in a complex with cytosolic chaperones. The title is changed accordingly. As discussed earlier, the protein we chose were of different sizes to show that they are refluxed independently of their size. Moreover, our previous work showed that the proteins that were refluxed are of different sizes. Most importantly UGGT1 (around 180 Kda) which is reported to deploy to the cytosol upon viral infection (Huang et al. 2017; Sicari et al. 2020). In this study we used AGR2 (around 19 Kda) and HYOU1 (150Kda).

ERCY in abstract, ERCYS in intro. There are typos throughout, could be a formatting problem, please check

Answer: Checked and corrected

What about the selection of refluxed proteins? Is this only a certain category of proteins? Could it be anything? Have you looked at other cargo / ER resident proteins?

__ ____Answer: __in our previous study by (Sicari, Pineau et al. 2020) we looked at many other proteins especially glycoproteins from the ER. In (Sicari, Pineau et al. 2020) we used mass spectrometry in order to identify new refluxed proteins and we found 26 new glycoprotein that are refluxed from cells treated with ER stressor and from human tissues obtained from GBM patients (Sicari, Pineau et al. 2020).

We previously showed that AGR2 is refluxed from the ER to the cytosol to bind and inhibit p53 (Sicari, Pineau et al. 2020). Here, we selected AGR2 because we know that (1) it is refluxed, and (2) we know which novel functions it acquires in the cytosol so we are able to measure and provide a physiological significance of those novel functions when the levels of DNAJB12 and DNAJB14 are altered. Moreover, we selected DNAJB11 (41 kDa) and HYOU1 (150 kDa) proteins to show that alteration in DNAJB12 or DNAJB14 prevent the reflux small, medium and large protein (independently of their size). We also showed earlier by mass spectrometry analysis that the refluxed proteins range from small to very large proteins such as UGGT1, thus we believe that soluble ER-proteins can be substrates of ERCYS independently of their size. In the discussion, we added a note that the reflux by the cytosolic and ER chaperones operates on different proteins independently of their size.

"Their role in ERCYS and cells' fate determination depends..." Suggest change to "Their role in ERCYS and determination of cell fate..."

Answer: changed and corrected

I think that the final sentence of the intro could be made stronger and more concise. There's a repeat of ER and cytosol. Instead could you comment on the reflux permitting new interactions between proteins otherwise spatially separated, then the effect on wt-p53 etc.

Answer: The sentence was rephrased as suggested to “ In this study, we found that HLJ1 is conserved through evolution and that mammalian cells have five putative functionality orthologs of the yeast HLJ1. Those five DNAJ- proteins (DNAJB12, DNAJB14, DNAJC14, DNAJC18, and DNAJC30) reside within the ER membrane with a J-domain facing the cytosol (Piette et al. 2021; Malinverni et al. 2023). Among those, we found that DNAJB12 and DNAJB14, which are strongly related to the yeast HLJ1 (Grove et al. 2011; Yamamoto et al. 2010), are essential and sufficient for determining cells' fate during ER stress by regulating ERCYS. Their role in ERCYS and determining cells' fate depends on their HPD motif in the J-domain. Downregulation of DNAJB12 and DNAJB14 increases cell toxicity and wt-p53 activity during etoposide treatment. Mechanistically, DNAJB12 and DNAJB14 interact and recruit cytosolic chaperones (HSC70/SGTA) to promote ERCYS. This later interaction is conserved in human tumors including colorectal cancer.

In summary, we propose a novel mechanism by which ER-soluble proteins are refluxed from the ER to the cytosol, permitting new inhibitory interactions between spatially separated proteins. This mechanism depends on cytosolic and ER chaperones and cochaperones, namely DNAJB12, DNAJB14, SGTA, and HSC70. As a result, the refluxed proteins gain new functions to inhibit the activity of wt-p53 in cancer cells. “

__Figure legends: __

In some cases the authors state the number of replicates, but this should be stated for all experiments. If experiments don't already include 3 independent repeats, this should be done. Check text for typos, correct letter capitalisation, spaces and random bold text (some of this could be from incompatability when saving as PDF)

Answer: all experiments were repeated at least three times. The number of repeats is now indicated in the figure legends of each experiment. Typos and capitalization is corrected as well.

Fig2E: scrambled not scramble siRNA

Answer: corrected

Fig 3: "to expel" is a term not used in the rest of the paper for reflux. Useful to remain consistent with terminology where possible

Answer: Rephrased and corrected

Results section 1:

"Protein alignment of the yeast HLJ1p showed high amino acids similarity to the mammalian..."

Answer: Rephrased to “ Comparing the amino acid sequences revealed significant similarity between the yeast protein HLJ1p and the mammalian proteins DNAJB12 and DNAJB14”

__ __ Fig 1C: state in legend which organism this is from (presumably human)

Answer: in Figure 1C legends it is stated that: “ the HPD motif within the J-domain is conserved in HLJ-1 and its putative human orthologs DNAJB12, DNAJB14, DNAJC14, DNAJC18, and DNAJC30.”

Results Section 2

"Test the two strongest hits DNAJB12/14" Add reference to previous paper showing this

Answer: the references were added.

__ __ "In the WT and J-protein-silenced A549 cells, there were no differences in the cytosolic enrichment of the three ER resident proteins AGR2, DNAJB11, and HYOU1 in normal and unstressed conditions (Figure 2A-C and Figure S2C)." I think that this is an oversimplification, and in your following discussion, you show this it's more subtle than this.

Answer: We expanded on this both in the discussion and the results section.

__ __ The text here isn't so clear: normal and unstressed conditions? Do you mean stressed? Please be careful in your phrases: "DNAJB12-silenced cells are slightly affected in AGR2 and DNAJB11 cytosolic accumulation but not HYOU1." This is the wrong way around. DNAJB12 silencing effects AGR2, not that AGR2 effects the cells (which is how you have written it). This also occurs agan in the next para:

Answer: Normal cells are non-cancer cells. Unstressed conditions= without ER stress. The sentence was rephrased to: In the absence of ER stress, the cytosolic levels of the three ER-resident proteins (AGR2, DNAJB11, and HYOU1) were similar in wild-type and J-protein-silenced A549 cells.

"During stress, DNAJB12/DNAJB14 double knockdown was highly affected in the cytosolic..." I think you mean it highly affected the cytosolic accumulation, not that it was affected by the cytosolic accumulation. Please change in the text

Answer: the sentence is now rephrased to” During stress, double knockdown of DNAJB12 and DNAJB14 highly affected the cytosolic accumulation of all three tested proteins”

__ __ "DNAJB12 and DNAJB14 are strong hits of the yeast HLJ1" Not clear, I presume you mean they are likely orthologues? Top candidates for being closest orthologues?

Answer: this is correct, the sentence is rephrased and corrected

__ __ Fig 2D: typos in WB labelling? I think Tm should be - - +, not - + +as it is now (if it's not a typo, you need more controls, eto alone.

Answer: the labeling is now corrected

Fig 2D-E-F typos for DKD? D12/D12 or D12/14?

Answer: This is correct, thank you for pointing this out. The labeling in corrected

__ __ "We assayed the phosphorylation state of wt- p53 and p21 protein expression levels (a downstream target of p53 signaling) during etoposide treatment." What are the results of this? Explain what Fig 2D-E shows, then build on this with the +Tm results. Results should be explained didactically to be clear.

Answer: The paragraph was edited and we explained the results: In these conditions, we saw an increase in the phosphorylation of wt-p53 in the control cells and in cells knocked-down with DNAJB12, DNAJB14 or both. This phosphorylation increased the protein levels of p21 as well (Figure 2D-G). Tm addition to cells treated with etoposide resulted in a reduction in wt-p53 phosphorylation, and as a consequence, the p21 protein levels were also decreased (Figure 2D-G and Figure S2O). Cells lacking DNAJB12 or DNAJB14 have partial protection in wt-p53 phosphorylation and p21 protein levels. Silencing both proteins in A549 and MCF7 cells rescued wt-p53 phosphorylation and p21 levels (Figure 2D-G and Figure S2D). Moreover, similar results were obtained when we assayed the transcriptional activity of wt-p53 in cells transfected with a luciferase reporter under the p53-DNA binding site (Figure 2H). These data confirm that DNAJB12 and DNAJB14 are involved in ER protein reflux and the inhibition of wt-p53 activity during ER stress.

---

"(Figure 2D- E). Cells lacking DNAJB12 and or DNAJB14 have partial protection in wt-p53 phosphorylation and p21 protein levels."

Answer: This sentence is now removed

You comment on p53 phosphorylation, but you haven't quantified this. This should be done, normalized to p53 levels, if you want to draw these conclusions, especially as total p53 varies between condition. Does Eto increase p53 txn? Does Tm alone increase p53 activity/phospho-p53? These are shown in the Sicari EMBO reports paper in 2021, you should briefly reference those.

Answer: The blots are now quantified and new blot is added to Figure S2D. The Paragraph was edited and referenced to our previous paper (Sicari et al. 2021). “We then wanted to examine whether the gain of function of AGR2 and the inhibition of wt-p53 depends on the activity of DNAJB12 and DNJAB14. We assayed the phosphorylation state of wt-p53 and p21 protein expression levels (a downstream target of wt-p53 signaling) during etoposide treatment. In these conditions, there was an increase in the phosphorylation of wt-p53 in the control cells and in cells knocked down with DNAJB12, DNAJB14, or both. This phosphorylation also increases protein levels of p21 (Figure 2D-G and Figure S2O). Tm addition to cells treated with etoposide resulted in a reduction in wt-p53 phosphorylation, and as a consequence, the p21 protein levels were also decreased (Figure 2D-G and Figure S2O). Silencing DNAJB12 and DNAJB14 in A549 and MCF-7 cells rescued wt-p53 phosphorylation and p21 levels (Figure 2D-G and Figure S2O). Moreover, similar results were obtained when we assayed the transcriptional activity of wt-p53 in cells transfected with a luciferase reporter under the p53-DNA binding site (Figure 2H). In the latter experiment, etoposide treatment increased the luciferase activity in all the cells tested. Adding ER stress to those cells decreased the luciferase activity except in cells silenced with DNAJB12 and DNAJB14.

These data confirm that DNAJB12 and DNAJB14 are involved in the reflux of ER proteins in general and AGR2 in particular. Inhibition of DNAJB12 and DNAJB14 prevented the inhibitory interaction between AGR2 and wt-p53 and thus rescued wt-p53 phosphorylation and its transcriptional activity as a consequence. “

Fig3A: overexpression of DNAJB12 decreases Eto induced p53 but not at steady state. Is this because at steady state the activity is already basal? Or is there another reason?

Answer: yes, at steady state the activity is already basal

Switch Figs S3D and S3C as they are not referred to in order. Also Fig S3C: vary colour (or add pattern) on bars more between conditions

Answer: The Figures now are called by their order in the new version. Colors are now added to Figure S3C.

Need to define HLJ1 at first mention

Answer: defined as” HLJ1 - High copy Lethal J-protein -an ER-resident tail-anchored HSP40 cochaperone.

Results section 3

HSC70 cochaperone (SGTA) defined twice

Answer: the second one was removed

"These data are important because SGTA and the ER-resident proteins (PRDX4, AGR2, and DNAJB11) are known to be expressed in different compartments, and the interaction occurs only when those ER-resident proteins localize to the cytosol." Is there a reference for this?

Answer: Peroxireoxin 4 is the only peroxerodin that is expressed in the ER. AGR2 and DNAJB11 are also ER luminal proteins that are known to be solely expressed in the ER. SGTA is part of the cytosolic quality control system and is expressed in the cytosol. The references are added in the main text.

Results section 4

"by almost two folds"

Answer: corrected

Fig 6A: It seems strange that the difference between purple and blue bars in scrambled, and D14-KD are very significant but D12-KD is only significant. Why is this? The error bars don't look that different. It would be interesting to see the individual means for each different replicate.

Answer: Thank you for pointing this, the two asterixis were aligned in the middle as one during figure alignments. In D14 the purple one has a lower error bar thus this changes the significance when compared to the blue while in D12-KD, the error bars in the eto treatment and the eto-Tm both are slightly higher. Graphs of the three different replicates are now added in Figure S6. Each one of the three biological replicates was repeated in three different technical repeats (averaged in the graphs).

Figures: Fig 6A: Scale bars not well placed. Annotation on final set should be D12/D14 DKD?

Answer: both were Corrected

__Discussion __47. The authors mention that they want to use DNAJB12/4-HSC70/SGTA axis to impair cancer cell fitness: What effect would this have though in a non cancer model? Would this be a viable approach Although it is obviously early days, which approach would the authors see as potentially favorable?

---

Answer: In our previous study we used an approach to target AGR2 in the cytosol because the reflux of AGR2 occurs only in cancer cells and not in normal cells. In that study we targeted AGR2 with scFv that targets AGR2 and is expressed in the cytosol, in this case it will target AGR2 in the cytosol which only occurs in cancer. Here, we suggest to target the interaction between the refluxed proteins and their new partners in the cytosol or to target the mechanism that causes their reflx to the cytosol by inhibiting for instance the interaction between SGTA and DNAJB proteins.

---

__ __ Second para: Should be "Here we present evidences"

Answer: we replaced with “Here we present evidences”

"DNAJB12 overexpression was also sufficient to promote ERCYS by refluxing AGR2 and inhibit wt-p53 signaling in cells treated with etoposide" Suggest:

Answer: DNAJB12 overexpression is also sufficient to promote ERCYS by refluxing AGR2 and inhibit wt-p53 signaling in cancer cells treated with etoposide (Figure 3). This suggests that it is enough to increase the levels of DNAJB12 without inducing the unfolded protein response in order to activate ERCYS. Moreover, the downregulation of DNAJB12 and DNAJB14 rescued the inhibition of wt-p53 during ER stress (Figure 2). Thus, wt-p53 inhibition is independent of the UPR activation but depends on the inhibitory interaction of AGR2 with wt-p53 in the cytosol.

.

DNAJB12 overexpression was also sufficient to promote ERCYS by increasing reflux of AGR2 and inhibition of wt-p53 signaling in cells treated with etoposide

Answer: This sentence is repeated twice and was removed

"Moreover, DNAJB12 was sufficient to promote this phenomenon and cause ER protein reflux by mass action without causing ER stress (Figure 3, Figure 4, and Figure S3)." You dont look at induction of ER stress here, please change the text or explain in more depth with refs if suitable

Answer: In the initial submission and in the revised version we assayed the activation of the UPR by looking at the levels of spliced Xbp1 and Bip in the different conditions when DNAJB12 and DNAJB14 are overexpressed (Figure S3C and S3D). Our data show that although DNAJB12 overexpression induces ERCYS, there was no UPR activation.

The mention of viruses is sparse in this paper. If it is a main theory, put it more centrally to the concept, and explain in more detail. As it is, its appearance in the final sentence is out of context.

Answer: DNAJB12 and DNAJB14 were reported to facilitate the escape of non-envelope viruses from the endoplasmic reticulum to the cytosol. The mechanism of non-envelope penetration is highly similar to the reflux of proteins from the ER to the cytosol. Interestingly, this mechanism takes place when the DNAJB12 and DNAJB14 form a complex with chaperones from both the ER and the cytosol including HSC70, SGTA and BiP (Walczak et al. 2014; Goodwin et al. 2011; Goodwin et al. 2014)..

Moreover, the UGGT1 that was independently found in our previous mass spectrometry analysis of the digitonin fraction obtained from HEK293T cells treated with the ER stressor thapsigargin and from isolated human GBM tumors (Sicari et al. 2020), is known to deploy to the cytosol upon viral infection (Huang et al. 2017; Sicari et al. 2020). We therefore hypothesized that the same machinary that is known to allow viruses to escape the ER to penetrate the cytosol may play an important role in the reflux of ER proteins to the cytosol.

Because ER protein reflux and the penetration of viruses from the ER to the cytosol behave similarly, we speculate that viruses hijacked an evolutionary conserved machinery -ER protein reflux- to penetrate to the cytosol. This is key because it was also reported that during the process of nonenveloped viruses penetration, large, intact and glycosylated viral particles are able to penetrate the ER membrane on their way to the cytosol (Inoue and Tsai 2011).

Action: we developed the discussion around this point and clarified it better because we believe it central to show that viruses hijacked this conserved mechanism.

**Referees cross-commenting**

I agree with the comments from Reviewer 1.

Reviewer 2 also is correct in many ways, but I think that they have somewhat overlooked the relevance of the ER-stress element and treatments. The authors do need to reference past papers more to give a full story, as this includes the groups own papers, I don't think that it is an ethical problem but rather an oversight in the writing. Regarding reviewer 2's concerns about overexpression levels and cell death, the authors do use an inducible cell line and show the levels of DNAJB12 induced (could CRISPR also be considered?). This could be used to further address reviewer 2's concerns. It would also be useful to see data on cell death in the conditions used in the paper. Re concerns about ER integrity, this could be addressed by using IF (or EM) to show a secondary ER marker that remains ER-localised, and this would also be of interest regarding my comment on which categories of proteins can undergo reflux. If everything is relocalised, then reviewer 2's point would be validated.

Reviewer #3 (Significance (Required)):

Significance

General assessment: This paper robustly shows that the yeast system of ER to cytosol reflux of ER-resident proteins is conserved in mammalian cells, and it describes clearly the link between ER stress, protein reflux and inhibition of p53 in mammalian cells. The authors have the tools to delve deeper into this mechanism and robustly explore this pathway, however the mechanistic elements - where not instantly clear from the results - have been over interpreted somewhat The results have been oversimplified in their explanations and some points and complexities of the study need to be addressed further to make the most of them - these are often some of the more interesting concepts of the paper, for example the differences in DNAJB12/14 and how the proteins orchestrate in the cytosol to play their cytosol-specific effects. I think that many points can be addressed in the text, by the authors being clear and concise with their reporting, while other experiments would turn this paper from an observational one, into a very interesting mechanistic one.

Advance: This paper is based on previous nice papers from the group. It is a nice progressions from yeast, to basic mechanism, to physiological model. But as mentioned, without a strong mechanistic improvement, the paper would remain observatory.

Audience: This paper is interesting to cell biologists (homeostasis, quality control and trafficking) as well as cancer cell biologists (fitness of cancer cells and homeostasis) and it is a very interesting demonstration of a process that is a double edged sword, depending on the environment of the cells.

My expertise: cell biology, trafficking, ER homeostasis

Answer: We would like to thank the reviewer for his/her positive feedback on our manuscript. All the comments of the three reviewers are now addressed and the manuscript has been strengthen. We put more emphasis on the mechanistic aspect with more Ips and knockdowns. We also added data to show that it is physiologically relevant. We hope that after that the revised version addressed all the concerns raised by the reviewers.

Goodwin, E. C., A. Lipovsky, T. Inoue, T. G. Magaldi, A. P. Edwards, K. E. Van Goor, A. W. Paton, J. C. Paton, W. J. Atwood, B. Tsai, and D. DiMaio. 2011. 'BiP and multiple DNAJ molecular chaperones in the endoplasmic reticulum are required for efficient simian virus 40 infection', MBio, 2: e00101-11.

Goodwin, E. C., N. Motamedi, A. Lipovsky, R. Fernandez-Busnadiego, and D. DiMaio. 2014. 'Expression of DNAJB12 or DNAJB14 causes coordinate invasion of the nucleus by membranes associated with a novel nuclear pore structure', PLoS One, 9: e94322.

Grove, D. E., C. Y. Fan, H. Y. Ren, and D. M. Cyr. 2011. 'The endoplasmic reticulum-associated Hsp40 DNAJB12 and Hsc70 cooperate to facilitate RMA1 E3-dependent degradation of nascent CFTRDeltaF508', Mol Biol Cell, 22: 301-14.

Huang, P. N., J. R. Jheng, J. J. Arnold, J. R. Wang, C. E. Cameron, and S. R. Shih. 2017. 'UGGT1 enhances enterovirus 71 pathogenicity by promoting viral RNA synthesis and viral replication', PLoS Pathog, 13: e1006375.

Igbaria, A., P. I. Merksamer, A. Trusina, F. Tilahun, J. R. Johnson, O. Brandman, N. J. Krogan, J. S. Weissman, and F. R. Papa. 2019. 'Chaperone-mediated reflux of secretory proteins to the cytosol during endoplasmic reticulum stress', Proc Natl Acad Sci U S A, 116: 11291-98.

Inoue, T., and B. Tsai. 2011. 'A large and intact viral particle penetrates the endoplasmic reticulum membrane to reach the cytosol', PLoS Pathog, 7: e1002037.

Malinverni, D., S. Zamuner, M. E. Rebeaud, A. Barducci, N. B. Nillegoda, and P. De Los Rios. 2023. 'Data-driven large-scale genomic analysis reveals an intricate phylogenetic and functional landscape in J-domain proteins', Proc Natl Acad Sci U S A, 120: e2218217120.

Piette, B. L., N. Alerasool, Z. Y. Lin, J. Lacoste, M. H. Y. Lam, W. W. Qian, S. Tran, B. Larsen, E. Campos, J. Peng, A. C. Gingras, and M. Taipale. 2021. 'Comprehensive interactome profiling of the human Hsp70 network highlights functional differentiation of J domains', Mol Cell, 81: 2549-65 e8.

Sicari, D., F. G. Centonze, R. Pineau, P. J. Le Reste, L. Negroni, S. Chat, M. A. Mohtar, D. Thomas, R. Gillet, T. Hupp, E. Chevet, and A. Igbaria. 2021. 'Reflux of Endoplasmic Reticulum proteins to the cytosol inactivates tumor suppressors', EMBO Rep: e51412.

Sicari, Daria, Raphael Pineau, Pierre-Jean Le Reste, Luc Negroni, Sophie Chat, Aiman Mohtar, Daniel Thomas, Reynald Gillet, Ted Hupp, Eric Chevet, and Aeid Igbaria. 2020. 'Reflux of Endoplasmic Reticulum proteins to the cytosol yields inactivation of tumor suppressors', bioRxiv.

Walczak, C. P., M. S. Ravindran, T. Inoue, and B. Tsai. 2014. 'A cytosolic chaperone complexes with dynamic membrane J-proteins and mobilizes a nonenveloped virus out of the endoplasmic reticulum', PLoS Pathog, 10: e1004007.

Yamamoto, Y. H., T. Kimura, S. Momohara, M. Takeuchi, T. Tani, Y. Kimata, H. Kadokura, and K. Kohno. 2010. 'A novel ER J-protein DNAJB12 accelerates ER-associated degradation of membrane proteins including CFTR', Cell Struct Funct, 35: 107-16.

Youker, R. T., P. Walsh, T. Beilharz, T. Lithgow, and J. L. Brodsky. 2004. 'Distinct roles for the Hsp40 and Hsp90 molecular chaperones during cystic fibrosis transmembrane conductance regulator degradation in yeast', Mol Biol Cell, 15: 4787-97.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Summary:

Reflux of ER based proteins to the cytosol during ER stress inhibits wt-p53. This is a pro-survival mechanism during ER stress, but as ER stress is high in many cancers, it also promotes survival of cancer cells. Using A549 cells, Dabsan et al. demonstrate that this mechanism is conserved from yeast to mammalian cells, and identify DNAJB12 and DNAJB14 as putative mammalian orthologues of yeast HLJ1.

This paper shows that DNAJB12 and 14 are likely orthologues of HLJ1 based on their sequences, and their behaviour. The paper develops the pathway of ER-stress > protein reflux > cytosolic interactions > inhibition of p53. The authors demonstrate this nicely using knock downs of DNAJB12 and/or 14 that partially blocks protein reflux and p53 inhibition. Overexpression of WT DNAJB12, but not the J-domain inactive mutant, blocks etoposide-induced p53 activation (this is not replicated with DNAJB14) and ER-resident protein reflux. The authors then show that DNAJB12/14 interact with refluxed ER-resident proteins and cytosolic SGTA, which importantly, they show interacts with the ER-resident proteins AGR2, PRDX4 and DNAJB11. Finally, the authors show that inducing ER stress in cancer cell lines can increase proliferation (lost by etoposide treatment), and that this is partially dependent on DNAJB12/14.

This is a very interesting paper that describes a nice mechanism linking ER-stress to inhibition of p53 and thus survival in the face of ER-stress, which is a double edged sword regarding normal v cancerous cells. The data is normally good, but the conclusions drawn oversimplify the data that can be quite complex. The paper opens a lot of questions that the authors may want to develop in more detail (non-experimentally) to work on these areas in the future, or alternatively to develop experimentally and develop the observations further. There are only a few experimental comments that I make that I think should be done to publish this paper, to increase robustness of the work already here, the rest are optional for developing the paper further.

Major comments:

1. Number of experimental repeats must be mentioned in the figure legends. Figures and annotations need to be aligned properly

Results section 2: 2. No intro to the proteins you've looked at for relocalisation. Would be useful to have some info on why you chose AGR2. Apart from them being ER-localised, do they all share another common characteristic? Does ability to inhibit p53 vary in potency? 3. What are the roles of DNAJB12/14 if overexpression can induce reflux? Does it allow increased binding of an already cytosolic protein, causing an overall increase in an interaction that then causes inhibition of p53? What are your suggested mechanisms? 4. Fig3: A+B show overexpression of individual DNAJs but not combined. As you go on to discuss the effect of the combination on AGR2 reflux, it would be useful to include this experimentally here. 5. Fig 3C: Subfractionation of cells shows AGR2 in the cytosol of A549 cells. The quality of the data is good but the bands are very high on the blot. For publication is it possible to show this band more centralized so that we are sure that we are not missing bands cut off in the empty and H139Q lanes? Also, you have some nice immunofluorescence in the 2021 EMBO reports paper, is it possible to show this by IF too? It is not essential for the story, but it would enrich the figure and support the biochemistry nicely. Also it is notable that the membrane fraction of the refluxed proteins doesn't appear to have a decrease in parallel (especially for AGR2). Is this because the % of the refluxed protein is very small? Is there a transcriptional increase of any of them (the treatments are 12+24 h so it would be enough time)? This could be a nice opportunity to discuss the amount of protein that is refluxed, whether this response is a huge emptying of the ER or more like a gentle release, and also the potency of the gain of function and effect on p53 vs the amount of protein refluxed. This latter part isn't essential but it would be a nice element to expand upon. 6. You still mention DNAJB12 and 14 as orthologues, even though DNAJB14 has no effect on p53 activity when overexpressed. Do you think that this piece of data diminishes this statement? 7. Fig 3D/F: Overexpression of DNAJB14 induces reflux of DNAJB11 at 24h, what does this suggest? Does this indicate having the same role as DNAJB12 but less potently? What's your hypothesis? 8. "This suggests that the two proteins may have different functions when overexpressed, despite their overlapping and redundant functions" What does it suggest about their dependence on each other? If overexpression of WT DNAJB12 inhibits Tg induced reflux, is it also blocking the ability of DNAJB14 to permit flux? 9. Fig 4: PDI shown in blots but not commented on in text. Then included in the schematics. Please comment in the text. 10. Fig 4F: Although the quantifications of the blots look fine, the blot shown does not convincingly demonstrate this data for AGR2. The other proteins look fine, but again it could be useful to see the individual means for each experiment, or the full gels for all replicates in a supplementary figure. Results section 3 11. Fig 5A, As there is obviously a difference between DNAJB12/14 it would be useful to do the pulldown with DNAJB14 too. Re. HSC70 binding to DNAJB12 and 14, the abstract states that DNAJB12/14 bind HSC70 and SGTA through their cytosolic J domains. Fig 5 shows pulldowns of DNAJB12 with an increased binding of SGTA in FLAG-DNAJB12 induced conditions, but the HSC70 band does not seem to be enriched in any of the conditions, including after DNAJB12 induction. This doesn't support the statement that DNAJB12 binds HSC70. In fact, in the absence of a good negative control, this would suggest that the HSC70 band seen is not specific. There is also no data to show that DNAJB14 binds HSC70. I recommend including a negative condition (ie beads only) and the data for DNAJB14 pulldown. 12. The binding of DNAJB12 to SGTA under stress conditions in Fig5B looks much more convincing than SGTA to DNAJB12 in Fig 5A. Bands in all blots need to be quantified from 3 independent experiments, and repeated if not already n=3. If this is solely a technical difference, please explain in the text. The conclusions drawn from this interaction data are important and shold be elaborated upon to support th claims made in the paper. The authors may also chose to expand the pulldowns to demonstrate their claims made on olidomerisation of DNAJB12 and 14 here. It is also clear that the interaction data of the SGTA with ER-resident proteins AGR2, PRDX4 and DNAJB11 is strong. The authors may want to draw on this in their hypotheses of the mechanism. I would imagine a complex such as DNAJB14/DNAJB12 - SGTA - AGR2/PRDX4/DNAJB11 would be logical. Have any experiments been performed to prove if complexes like this would form? 13. Fig 5B: It is clear that DNAJB12 interacts with SGTA. The authors state that DNAJB14 also interacts with SGTA under normal and stress conditions, but the band in 25/50 Tg is very feint. Why would there be stronger binding at the 2 extremes than during low stress induction? In the input, there is a much higher expression of DNAJB14 in 50 Tg. What does this say about the interaction? Is there an effect of ER stress on DNAJB14 expression? A negative control should be included to show any background binding, such as a "beads only" control. 14. Fig 5C data is sound, although a negative control should be included. Results section 4 15. Fig 6A-B: Given that there is the complexity of overexpression v KD of DNAJB12 v 14 causing similar effects on p53 actvity (Fig 2 v 3), it would be interesting to see whether the effect of overexpression mirrors the results in Fig 6A. Is it known what SGTA overexpression does (optional)? 16. Fig 6D: resolution very low 17. Fig 6C-D: There is an interesting difference though between the proposed cytosolic actions of the refluxed proteins. You show that AGR2, PRDX4 and DNAJB11 all bind to SGTA in stress conditions, but in the schematics you show: DNAJB11 binding to HSC70 through SGTA (not shown in the paper), then also PDIA1, PDIA3 binding to SGTA and AGR2 binding to SGTA. What role does SGTA have in these varied reactions? Sometimes it is depicted as an intermediate, sometimes a lone binder, what is its role as a binder? It should be clarified which interactions are demonstrated in the paper (or before) and which are hypothesized in a graphical way (eg. for hypotheses dotted outlines or no solid fill etc). The schematics also suggest that DNAJB14 binding to HSC70 and SGTA is inducible in stress conditions, as is PDIA3, which is not shown in the paper. Discussion "In cancer cells, DNAJB12 and DNAJB14 oligomerize and recruit cytosolic chaperones and cochaperones (HSC70 and SGTA) to reflux AGR2 and other ER-resident proteins and to inhibit wt-p53 and probably different proapoptotic signaling pathways (Figure 5, and Figure 6C-6D)." You havent shown oligomerisation between DNAJB12/14. Modify the text to make it clear that it is a hypothesis. Minor comments: 18. It would be useful to have page or line numbers to help with document navigation, please include them. Typos and inconsistency in how some proteins are named throughout the manuscript 19. Title: Include reference to reflux. Suggest: "chaperone complexes (?proteins) reflux from the ER to cytosol..." I presume it would be more likely that the proteins go separately rather than in complex. Do you have any ideas on the size range of proteins that can undergo this process? 20. ERCY in abstract, ERCYS in intro. There are typos throughout, could be a formatting problem, please check 21. What about the selection of refluxed proteins? Is this only a certain category of proteins? Could it be anything? Have you looked at other cargo / ER resident proteins? 22. "Their role in ERCYS and cells' fate determination depends..." Suggest change to "Their role in ERCYS and determination of cell fate..." 23. I think that the final sentence of the intro could be made stronger and more concise. There's a repeat of ER and cytosol. Instead could you comment on the reflux permitting new interactions between proteins otherwise spatially separated, then the effect on wt-p53 etc.

Figure legends:

1. In some cases the authors state the number of replicates, but this should be stated for all experiments. If experiments don't already include 3 independent repeats, this should be done. Check text for typos, correct letter capitalisation, spaces and random bold text (some of this could be from incompatability when saving as PDF)
2. Fig2E: scrambled not scramble siRNA
3. Fig 3: "to expel" is a term not used in the rest of the paper for reflux. Useful to remain consistent with terminology where possible

Results section 1:

1. "Protein alignment of the yeast HLJ1p showed high amino acids similarity to the mammalian..."
2. Fig 1C: state in legend which organism this is from (presumably human) Results Section 2
3. "Test the two strongest hits DNAJB12/14" Add reference to previous paper showing this
4. "In the WT and J-protein-silenced A549 cells, there were no differences in the cytosolic enrichment of the three ER resident proteins AGR2, DNAJB11, and HYOU1 in normal and unstressed conditions (Figure 2A-C and Figure S2C)." I think that this is an oversimplification, and in your following discussion, you show this it's more subtle than this.
5. The text here isn't so clear: normal and unstressed conditions? Do you mean stressed? Please be careful in your phrases: "DNAJB12-silenced cells are slightly affected in AGR2 and DNAJB11 cytosolic accumulation but not HYOU1." This is the wrong way around. DNAJB12 silencing effects AGR2, not that AGR2 effects the cells (which is how you have written it). This also occurs agan in the next para:
6. "During stress, DNAJB12/DNAJB14 double knockdown was highly affected in the cytosolic..." I think you mean it highly affected the cytosolic accumulation, not that it was affected by the cytosolic accumulation. Please change in the text
7. "DNAJB12 and DNAJB14 are strong hits of the yeast HLJ1" Not clear, I presume you mean they are likely orthologues? Top candidates for being closest orthologues?
8. Fig 2D: typos in WB labelling? I think Tm should be - - +, not - + +as it is now (if it's not a typo, you need more controls, eto alone.
9. Fig 2D-E-F typos for DKD? D12/D12 or D12/14?
10. "We assayed the phosphorylation state of wt- p53 and p21 protein expression levels (a downstream target of p53 signaling) during etoposide treatment." What are the results of this? Explain what Fig 2D-E shows, then build on this with the +Tm results. Results should be explained didactically to be clear.
11. "(Figure 2D- E). Cells lacking DNAJB12 and or DNAJB14 have partial protection in wt-p53 phosphorylation and p21 protein levels."
12. You comment on p53 phosphorylation, but you haven't quantified this. This should be done, normalized to p53 levels, if you want to draw these conclusions, especially as total p53 varies between condition. Does Eto increase p53 txn? Does Tm alone increase p53 activity/phospho-p53? These are shown in the Sicari EMBO reports paper in 2021, you should briefly reference those.
13. Fig3A: overexpression of DNAJB12 decreases Eto induced p53 but not at steady state. Is this because at steady state the activity is already basal? Or is there another reason?
14. Switch Figs S3D and S3C as they are not referred to in order. Also Fig S3C: vary colour (or add pattern) on bars more between conditions
15. Need to define HLJ1 at first mention Results section 3
16. HSC70 cochaperone (SGTA) defined twice
17. "These data are important because SGTA and the ER-resident proteins (PRDX4, AGR2, and DNAJB11) are known to be expressed in different compartments, and the interaction occurs only when those ER-resident proteins localize to the cytosol." Is there a reference for this? Results section 4
18. "by almost two folds"
19. Fig 6A: It seems strange that the difference between purple and blue bars in scrambled, and D14-KD are very significant but D12-KD is only significant. Why is this? The error bars don't look that different. It would be interesting to see the individual means for each different replicate.
20. Figures: Fig 6A: Scale bars not well placed. Annotation on final set should be D12/D14 DKD? Discussion
21. The authors mention that they want to use DNAJB12/4-HSC70/SGTA axis to impair cancer cell fitness: What effect would this have though in a non cancer model? Would this be a viable approach? Although it is obviously early days, which approach would the authors see as potentially favourable?
22. Second para: Should be "Here we present evidences"
23. "DNAJB12 overexpression was also sufficient to promote ERCYS by refluxing AGR2 and inhibit wt-p53 signaling in cells treated with etoposide" Suggest:
24. DNAJB12 overexpression was also sufficient to promote ERCYS by increasing reflux of AGR2 and inhibition of wt-p53 signaling in cells treated with etoposide
25. "Moreover, DNAJB12 was sufficient to promote this phenomenon and cause ER protein reflux by mass action without causing ER stress (Figure 3, Figure 4, and Figure S3)." You dont look at induction of ER stress here, please change the text or explain in more depth with refs if suitable
26. The mention of viruses is sparse in this paper. If it is a main theory, put it more centrally to the concept, and explain in more detail. As it is, its appearance in the final sentence is out of context.

Referees cross-commenting

I agree with the comments from Reviewer 1. Reviewer 2 also is correct in many ways, but I think that they have somewhat overlooked the relevance of the ER-stress element and treatments. The authors do need to reference past papers more to give a full story, as this includes the groups own papers, I don't think that it is an ethical problem but rather an oversight in the writing. Regarding reviewer 2's concerns about overexpression levels and cell death, the authors do use an inducible cell line and show the levels of DNAJB12 induced (could CRISPR also be considered?). This could be used to further address reviewer 2's concerns. It would also be useful to see data on cell death in the conditions used in the paper. Re concerns about ER integrity, this could be addressed by using IF (or EM) to show a secondary ER marker that remains ER-localised, and this would also be of interest regarding my comment on which categories of proteins can undergo reflux. If everything is relocalised, then reviewer 2's point would be validated.

Significance

General assessment: This paper robustly shows that the yeast system of ER to cytosol reflux of ER-resident proteins is conserved in mammalian cells, and it describes clearly the link between ER stress, protein reflux and inhibition of p53 in mammalian cells. The authors have the tools to delve deeper into this mechanism and robustly explore this pathway, however the mechanistic elements - where not instantly clear from the results - have been over interpreted somewhat. The results have been oversimplified in their explanations and some points and complexities of the study need to be addressed further to make the most of them - these are often some of the more interesting concepts of the paper, for example the differences in DNAJB12/14 and how the proteins orchestrate in the cytosol to play their cytosol-specific effects. I think that many points can be addressed in the text, by the authors being clear and concise with their reporting, while other experiments would turn this paper from an observational one, into a very interesting mechanistic one.

Advance: This paper is based on previous nice papers from the group. It is a nice progressions from yeast, to basic mechanism, to physiological model. But as mentioned, without a strong mechanistic improvement, the paper would remain observatory.

Audience: This paper is interesting to cell biologists (homeostasis, quality control and trafficking) as well as cancer cell biologists (fitness of cancer cells and homeostasis) and it is a very interesting demonstration of a process that is a double edged sword, depending on the environment of the cells.

My expertise: cell biology, trafficking, ER homeostasis

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

The authors present a study in which they ascribe a role for a complex containing DNAJB12/14-Hsc70-SGTA in facilitating reflux of a AGR2 from the ER to cytosol during ER-stress. This function is proposed to inhibit wt-P53 during ER-stress.

Concerns:

1. The way the manuscript is written gives the impression that this is the first study about mammalian homologs of yeast HLJ1, while there are instead multiple published papers on mammalian orthologs of HLJ1. Section 1 and Figure 1 of the results section is redundant with a collection of previously published manuscripts and reviews. The lack of proper citation and discussion of previous literature prevents the reader from evaluating the results presented here, compared to those in the literature.
2. The conditions used to study DNAJB12 and DNAJ14 function in AGR2 reflux from the ER do not appear to be of physiological relevance. As seen below they involve two transfections and treatment with two cytotoxic drugs over a period of 42 hours. The assay for ERCY is accumulation of lumenal ER proteins in a cytosolic fraction. Yet, there is no data or controls that describe the path taken by AGR2 from the ER to cytosol. It seems like pleotropic damage to the ER due the experimental conditions and accompanying cell death could account for the reported results?

A. Transfection of cells with siRNA for DNAJB12 or DNAJB14 with a subsequent 24-hour growth period.

B. Transfection of cells with a p53-lucifease reporter.

C. Treatment of cells with etoposide for 2-hours to inhibit DNA synthesis and induce p53.

D. Treatment of cells for 16 hours with tunicamycin to inhibit addition of N-linked glycans to secretory proteins and cause ER-stress.

E. Subcellular fractionation to determine the localization of AGR2, DNAJB11, and HYOU1

KD of DNAJB12 or DNAJB14 have modest if any impact on AGR2 accumulation in the cytosol. There is an effect of the double KD of DNAJB12 or DNAJB14 on AGR2 accumulation in the cytosol. Yet there are no western blots showing AGR2 levels in the different cells, so it is possible that AGR2 is not synthesized in cells lacking DNAJB12 and DNAKB14. The lack of controls showing the impact of single and double KD or DNAJB12 and DNAJB14 on cell viability and ER-homeostasis make it difficult to interpret the result presented. How many control versus siRNA KD cells survive the protocol used in these assays? 3. In Figure 3 the authors overexpress WT-D12 and H139Q-D12 and examine induction of the p53-reporter. There are no western blots showing the expression levels of WT-D12 and H139Q-D12 relative to endogenous DNAJB12. HLJ1 stands for high-copy lethal DnaJ1 as overexpression of HLJ1 kills yeast. The authors present no controls showing that WT-D12 and H139-D12 are not expressed at toxic levels, so the data presented is difficult to evaluate. 4. There is no mechanistic data used to help explain the putative role DNAJB12 and DNAJB14 in ERCY? In Figure 4, why does H139Q JB12 prevent accumulation of AGR2 in the cytosol? There are no westerns showing the level to which DNAJB12 and DNAJB14 are overexpressed.

Referees cross-commenting

I appreciate the comments of the other reviewers. I agree that the authors could revise the manuscript. Yet, based on my concerns about the physiological significance of the process under study and lack of scholarship in the original draft, I would not agree to review a revised version of the paper.

Significance

Overall, there are serious concerns about the writing of this paper as it gives the impression that it is the first study on higher eukaryotic and mammalian homologs of yeast HLJ1. The reader is not given the ability to compare the presented data to related published work. There are also serious concerns about the quality of the data presented and the physiological significance of the process under study. In its present form, this work does not appear suitable for publication

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary:

The manuscript by Dabsan et al builds on earlier work of the Igbaria lab, who showed that ER-luminal chaperones can be refluxed into the cytosol (ERCYS) during ER stress, which constitutes a pro-survival pathway potentially used by cancer cells. In the current work, they extent these observations and a role for DNAJB12&14 in ERCYS. The work is interesting and the topic is novel and of great relevance for the proteostasis community. I have a number of technical comments:

Major and minor comments:

1. In the description of Figure 2, statistics is only show to compare untreated condition with those treated with Tg or Tm, but no comparison between condition and different proteins. As such, the statement made by the authors "...DNAJB14-silenced cells were only affected in AGR2 but not in DNAJB11 or HYOU1 cytosolic accumulation" cannot be made.
2. Figure S2C: D11 seems to increase in the cytosolic fraction after Tm and Tg treatment. However, this is not reflected in the text. The membrane fraction also increases in the DKO. Is the increase of D11 in both cytosol and membrane and indication for a transcriptional induction of this protein by Tm/Tg? Again, the authors are not reflecting on this in their text.
3. Figure 2D: Only p21 is quantified. phospho-p53 and p53 levels are not quantified.
4. Figure 2D: There appears to be a labelling error
5. Are there conditions where DNAJB12 would be higher?
6. What do the authors mean by "just by mass action"?
7. Figure 3C: Should be labelled to indicate membrane and cytosolic fraction. The AGR2 blot in the left part is not publication quality and should be replaced.
8. What could be the reason for the fact that DNAJB12 is necessary and sufficient for ERCYS, while DNAJB14 is only necessary?
9. Figure 5A: Is the interaction between SGTA and JB12 UPR-independent?HCS70 seems to show only background binding. The interaction of JB12 with SGTA is not convincing. A better blot is needed.
10. Figure 5B: the expression of DNAJB14 was induced by Tg50, but not by Tg25 or Tm. However, the authors have not commented on this. This should be mentioned in the text and discussed.
11. Figure 6A: Why is a double knockdown important at all? DNAJB14 does not seem to do much at all (neither in overexpression nor with single knockdown).

Referees cross-commenting

I agree with the comments raised by reviewer 1 about the manuscript. I also agree with the points written in this consultation session. In my opinion, the comments of reviewer 2 are phrased in a harsh tone and thus the reviewer reaches the conclusion that there are "serious" problems with this manuscript. However, I think that the authors could address many of the points of this reviewer in a matter of 3 months easily. For instance, it is easy to control for the expression levels of exogenous wild type and mutant D12 and compare it to the endogenous one (point 3). This is a very good point of this reviewer and I agree with this experiment. Likewise, it is easy to provide data about the levels of AGR2 to address the concern whether its synthesis is affected by D12 and D14 overexpression. Again, an excellent suggestion, but no reason for rejecting the story. As for not citing the literature, I think this can also easily be addressed and I am sure that this is just an oversight and no ill intention by the authors. Overall, I am unable to see why the reviewer reaches such a negative verdict about this work. With proper revisions that might take 3 months, I think the points of all reviewers can be addressed.

Significance

The strength of the work is that it provides further mechanistic insight into a novel cellular phenomenon (ERCYS). The functions for DNAJB12&14 are unprecedented and therefore of great interest for the proteostasis community. Potentially, the work is also of interest for cancer researchers, who might capitalize of the ERCYS to establish DNAJB12/14 as novel therapeutic targets.

The major weaknesses are as follows:

• (i) the work is limited to a single cell line. To better probe the cancer relevance, the work should have used at least a panel of cell lines from one (or more) cancer entity. Ideally even data from patient derived samples would have been nice. Having said this, I also appreciate that the work is primarily in the field of cell biology and the cancer-centric work could be done by others. Certainly, the current work could inspire cancer specialists to explore the relevance of ERCYS.
• (ii) No physiological or pathological condition is shown where DNAJB12 is induced or depleted.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Manuscript number: RC-2024-02491

Corresponding author(s): Gilbert, Vassart

1. General Statements [optional]

We thank referees 1 and 2 for their in-depth analysis of our manuscript. They see interest in our study, with questions to be answered. Referee 3 is essentially negative, considering that there is nothing new ("novel finding is missing"). We respectfully disagree with him/her, comforted by the opinion of referee 2 that "the authors developed a protocol to reproducibly generate fetal-like spheroids from adult tissue is an important advancement in the field and ... the manuscript should attract a significant amount of attention in the intestinal field" and we provide evidence in our answers that he/she did not read the manuscript with the same attention as referees 1 and 2 (see in particular answer to his/her question 5).

Here is a summary of the main reason why we consider that our study represents valuable new information in the field of intestinal regeneration.

It is based on the serendipitous observation that dissociation of adult intestinal tissue by collagenase generates stably replatable spheroids upon culture in matrigel. Surprisingly and contrary to canonical EDTA-generated intestinal organoids and fetal spheroids, these spheroids are not traced in Rosa26Tomato mice harboring a VilCre transgene, despite expressing robustly endogenous Villin. Our interpretation is that adult intestinal spheroids originate from a cell lineage, distinct from the main developmental intestinal lineage, in which the VilCre transgene is unexpectedly not expressed, probaly due to the absence of cis regulatory sequences required for expression in this lineage.

Adult spheroid transcriptome shares a gene signature with the YAP/TAZ signature commonly expressed in models of intestinal regeneration. This led us to look for VilCre negative crypts in the regenerating intestine of Lgr5/DTR mice in which Lgr5-positive stem cells have been ablated by diphtheria toxin. Numerous VilCre negative clones were observed, identifying a novel lineage of stem cells implicated in intestinal regeneration.

FACS purification and scRNAseq analysis of the rare VilCre negative cells present at homeostasis identified a population of cells with characteristics of quiescent stem cells.

In sum, we believe that our study demonstrates the existence of a hitherto undescribed stem cell lineage involved in intestinal regeneration. It points to the existence of a hierarchical model of intestinal regeneration in addition to the well-accepted plasticity model.

2. Description of the planned revisions

See section 3 below.

3. Description of the revisions that have already been incorporated in the transferred manuscript

Here is a point-by-point reply to the queries of the three referees, with indication of the revisions introduced in the manuscript.

Reviewer #1 (Evidence, reproducibility and clarity (Required)):

• *In this manuscript, Marefati et al report an Lgr5-independent lineage in the regenerating intestine using in vitro organoids and in vivo injury-coupled lineage tracing model. In organoids, collagenase/dispase dissociated resulted in "immortal spheroids" that maintain a cystic and undifferentiated phenotype in the absence of standard growth factors (Rspondin/Noggin/EGF). Bulk RNAseq of spheroids demonstrates downregulation of classical CBC signatures and upregulation of fetal spheroid, mesenchymal, inflammation and regenerative signatures. In mice, Villin-Cre lineage tracing revealed some Villin- negative progenies that lack reporter tracing throughout crypt-villus ribbons after injury.

*The authors proposed that there is Lgr5-independent population support the regenerative response upon CBC depletion. A major caveat of this study is the identification of this population is based on absence of VilCre expression. *

We respectfully disagree. It is precisely this characteristic that makes the interest of our study. Whereas mosaicism of transgene expression is widespread and usually of little significance, our study shows that the rare VilCre-negative cells in the intestinal epithelium are not randomly showing this phenotype: they give specifically birth to what we call adult spheroids and regenerating crypts, which cannot be due to chance. The absence of VilCre expression allows tracing these cells from the zygote stage of the various VilCre/Ros26 reporter mice. We have modified our text to emphasize this point.

*It is surprising that there is no characterisation of Lgr5 expression throughout the manuscript whilst claiming of a Lgr5- independent lineage. *

We understand the perplexity of the referee not to see direct Lgr5 expression data in our manuscript, given our title. However, our point is that it is the cells at the origin of adult spheroids and the regenerating crypts we have identified that are Lgr5-negative, not the spheroids or the regenerated crypts themselves. Those are downstream offspring that may, and indeed have, gained some Lgr5 expression (e.g. figure 3F). We believe that our data showing that VilCre-negative spheroids are not traced in Lgr5-CreERT2/Rosa reporter mice convincingly demonstrate absence of Lgr5 expression in the cells at the origin of adult spheroids (figure 4G). We think that this experiment is better evidence than attempts to show absence of two markers (Tom and Lgr5) in the rare "white" cells present in the epithelium. Regarding the Lgr5 status of cells at the origin of the regenerating "white" crypts that we have identified, the early appearance of these crypts following ablation of CBC (i.e. Lgr5+ve) cells is a strong argument that they originate from Lgr5-negative cells. Regarding the scRNAseq experiment, Lgr5 transcripts are notoriously low and difficult to measure reliably in CBCs (Haber et al 2017). However, blowing up the pertinent regions of the merged UMAP allows showing some Lgr5 transcripts in clusters 5,6 and none in cluster 1 of figure 8GH. Given the very low level of detection, we had chosen not to include these data in the manuscript, but we hope they may help answer the point of the referee (see portion of UMAP below, with Olfm4 as a control, together with the corresponding violin plot). Several markers that gave significant signals in the CBC cluster (Smoc2, Axin2, Slc12a2) were virtually undetectable in the Olfm4-low /Tom-negative cluster of our scRNAseq data (figure 8I) supporting our conclusion.

Although the research question is potentially interesting, the concept of epithelial reprogramming upon injury is well documented in the field. The data generated in this manuscript also seem to be preliminary and lack of detailed characterisation. Below are specific comments.

We do not question the existence of epithelial reprogramming upon injury. We believe our data show, in addition to this well demonstrated phenomenon, the existence of rare cells traced by absence of VilCre expression that are at the origin of a developmental cell lineage distinct from Lgr5+ stem cells and also implicated in regeneration.

• Expression of Lgr5 should be properly characterised throughout the manuscript in both organoid models and injury-induced regeneration in vivo.
• *

See above for a detailed answer to this point.

• An important question is the origin of these "Lgr5-independent" adult spheroids. They look and appear like fetal organoids, which could be induced by injury (e.g. upon collagenase/dispase dissociation). Have the authors tried to culture fetal spheroids in BCM over extensive period of time? Do they behave the same? This would be a great way to directly compare the collagenase/dispase-derived organoids with fetal origin. * *Fetal spheroids require ENR for survival and die in BCM. We have chosen to illustrate this point in Fig2A by showing that, contrary to adult spheroid, they die even when only Rspondin is missing.

• Fig 1C, Why is the replating spheroid culture time different between mesenchymal cells and conditioned medium? We took the earliest time showing convincingly the return to the organoid phenotype. This timing difference does not modify the conclusion that EDTA organoids becoming spheroid-like when exposed to factors originating from mesenchymal cells revert to the organoid phenotype when returned to ENR medium without mesenchymal influence.

• *It is unclear how the bulk RNA-seq data in Fig. 3 were compared. How long were the adult organoids and spheroids cultured for (how many passages)? Were they culture in the same condition of were they in ENR vs BCM? * Both EDTA organoids and spheroids displaying a stable phenotype were used in this experiment. Organoids were collected at passage 4, day 5; spheroids were collected at passage passage 9 day 3.

As stated in the legend to the figure: "...to allow pertinent comparison spheroids and organoids were cultured in the same ENR-containing medium...".

These are important information to consider when interpreting the results. For instance, are Ptgs1 & Ptgs2 expression in adult spheroids the same in ENR vs BCM? Are the gene signatures (regenerative, fetal and YAP) changed in adult spheroids culturing in ENR vs BCM?

We did compare bulk RNAseq of EDTA organoids to ENR-cultured spheroids, short term (passage 6, day 6) BCM-cultured spheroids and long term BCM-cultured (passage 26, day 6) spheroids. To avoid overloading the manuscript these data were not shown in the original manuscript. In summary the BCM-cultured spheroids display a similar phenotype as those cultured in ENR, but with further de-differentiation. See in revision plan folder the results for PTGS, some differentiation markers and fetal regenerative markers including YAP induced genes.

We have included a brief description of these data in the new version of the manuscript and added an additional supplementary file (Suppl table 2) presenting the whole data set.

• It is stated: "In agreement with their aptitude to grow indefinitely, adult spheroids express a set of upregulated genes overlapping significantly with an "adult tissue stem cell module" [159/721 genes; q value 2.11 e-94) (Fig.S2F)].". What is the definition of "indefinitely"? Are they referring to the Fig 1B where spheroid were passaged to P10? The authors should avoid the term "indefinitely" but use a more specific time scale, e.g. passages, months etc.

We agree that the term indefinitely should be avoided, as it is vague. We have introduced the maximum number of passages during which we have maintained the stable spheroid phenotype (26 passages). Also worth noting, the spheroids could be frozen and cultured repeatedly over many months.

SuppFig 3D: Row Z-Score is missing the "e" in Score.

Corrected

• Fig 4E: Figure legend says QNRQ instead of CNRQ. Corrected

• Fig 4G: The brightfield image of adult spheroids 5 days after 3x TAM injections doesn't look like a spheroid. It seems to be differentiating. True, the choice was not the best as the spheroids started to darken. When further replated, however, the offspring of these spheroids showing a clear phenotype remain negative 30 days after tamoxifen administration as shown on the figure. We are sorry, but for reasons explained in section 4 below, we cannot redo the experiment to get a better picture.

• Fig 4: Most mouse model data are missing the number of mice & their respective age used for organoid isolation. We have introduced these data in the legend.

• *Fig 4A-D, H-G: How was fluorescent signal of organoids quantified? *

The settings of fluo imaging or time of LacZ staining were the same for organoids and spheroid pictures. This has been added to the material and methods of the figure and an example is shown below for Rosa26Tomato.

*How many images? * 2 per animal per condition.

*Were there equal numbers of organoids? *

No, see number of total elements counted added to the figure

This all needs to be included in methods/figure legends.

We have introduced additional pertinent information in the material and methods section.

• Figure 4B-D, G-H: Which culturing conditions were used for adult spheroids? Original method or sandwich method? These data were obtained with the original protocol

• Fig 6D-E: Please add the timepoint after DT administration these samples are from. It is not listed in text or figure legend. These samples were those obtained from mice sacrificed at the end of the 5 day period as indicated in panel A. This has been emphasized in the legend of the figure.

• SuppFig 6D: again timepoint is missing. In this experiment all samples were untreated as indicated. This has been emphasized in the legend of the figure.

• SuppFig 6: How were the crypts of these mice (DT WT & DT HE) isolated? Was this via EDTA? This was RNA extracted from total uncultured EDTA-released material (crypts). This has been emphasized in the legend of the figure.

Also, what is the timepoint for isolation for these samples? Even if untreated, the timepoint adds context to the data. Please add more context to describing these different experiments, either in the figure legends or methods section.

All these experiments were from 2 month old animals. We have indicated this in the legend of the figure.

• SuppFig 6E: The quality of the heatmap resolution is too poor to read gene names. We have improved the resolution of the figure and hope the name of the genes are readable now.

• 5-7, are the regenerating crypt-villus units fully differentiated or are they maintained in the developmental state? Immunostaining of markers for stem cells (Lgr5), differentiated lineages (Alpi, Muc2, Lyz, ChgA etc.) and fetal state (Sca1, Trop2 etc) should be analysed in those "white" unrecombined crypt-villus units. The differentiation phenotype is shown by the clear presence of morphologically-identified Paneth and Goblet cells. We agree that specific immunostainings could have been performed to further explore this point. Regarding the fetal state, Clu expression was shown during the regeneration period (see figure 7D,E).

Unfortunately, for reasons explained in section 4 below, we are not in a position to perform these additional experiments.

• The following text needs clarification: "The kinetics of appearance of newly formed un-recombined ("white") crypts was studied after a single pulse of DT (Fig.7A). This demonstrated an increase at 48 hours, with further increase at day 10 and stable maintenance at day 30. The presence of newly formed white crypts one month after toxin administration indicates that the VilCre-negative lineage is developmentally stable and does not turn on the transgene during differentiation of the various epithelial lineages occurring after regeneration (Fig.7B).

*Comment: The "newly formed" is an overstatement, the data doesn't conclude that those are "new" crypts. *

Except if we do not understand the point, we think we can write that a fraction of "white" crypts must be "newly formed", since they are in excess of those present in untreated animals at the same time point.

*The end of the sentence states that these "white" crypts form developmentally stable lineages, thus these white crypts at day 30 could originate from the initial injury. *

As stated above, we consider that crypts found in excess of those present in untreated animals result from the initial injury.

*There was no characterisation of the various epitheial lineages. Are they fully differentiated? *

See above the point related to Paneth cells and Goblet cells.

Is Lgr5 expressed one month after toxin administration? Can the VilCre neg lineage give rise to CBCs?

We have tried hard to show presence or absence of Lgr5 in white crypts at the various times following DT administration. We tried double RFP / Lgr5-RNA scope labeling and double GFP/RFP immunolabeling. Unfortunately, we could not get these methods to produce convincing specific labeling of CBCs in homeostatic crypts, which explains why we could not reach a conclusion regarding the white crypts.

However, there is an indirect indication that "chronic" white crypts (i.e. those caused by DTR expression in CBC, plus those observed 30 days after DT administration) do not express Lgr5. Indeed, acute regeneration indicated by Clu expression at day 5 in Fig.7C is lower in white crypts than in red ones strongly suggesting that white crypts preexisting DT administration (the "chronic ones) do not express Lgr5DTR.

The relationship between white crypt generation and appearance of Clu-positive revival cells (Ayyaz et al., 2019) was then explored. In agreement with others and similar to what happens in the irradiation model, (Ayyaz et al., 2019; Yuan et al., 2023) Clu-positive cells were rare in crypts of untreated mice and their number transiently increased forty-eight hours after a single pulse of DT, and more so after three pulses of DT (Fig.7C,D).

Comment: Comparing 1 pulse at day 2 vs 3 pulses at day 5 makes the data hard to interpret. How is the Clu ISH level for 1 pulse at day 5? Are they equivalent?

After a single pulse of of DT, Clu is only transiently increased. As shown by Ayyaz et al it is back to the starting point at day 5 (supplementary figure 4 of Ayyaz et al).

Clu-positive cells were less frequently observed in white crypts (see "Total" versus "White" in Fig.7C). This fits with the hypothesis that Clu expression marks acutely regenerating crypts and that a proportion of the white crypts are chronically regenerating due to DTR expression in CBCs."

*Comment: I believe the authors suggested that the discrepancy of less Clu expression in white crypts is due to the ectopic expression of DTR in CBCs causing low grade injury without DT administration. This means that some white crypts could have been formed before the administration of DT, and thus are on a different regenerative timeline compared to the white crypts formed from DT administration. *

Yes, this is our interpretation. We have clarified it in the text.

Is there any proof of the chronic regeneration? Immunostaining of chronic regenerative markers such as Sca1, Anxa1 or Yap1 nuclear localization would support the claim. It'd be important to show only the white crypts, but not the RFP+ ones, show regenerative markers.

We think that the steady state higher number of white crypts in untreated Lgr5-DTR animals, compared to wild type siblings indicates chronical low-grade regeneration, which is supported by the RNAseq data (Suppl fig6). It must be noted, however, that this phenotype is mild compared to the well described fetal-like regeneration phenotype described in most injury models. Since these white crypts were made at undetermined earlier stages, the great majority of them are not expected to show markers of acute regeneration like Clu, Sca1....

Fig 7D-E: What are the timepoints of harvest for HE-WT-HE 1 pulse DT mice and HE- HE-HE PBS injected mice?

We have added this information in the figure.

• *Fig 8-9: Regarding the CBC-like Olfm4 low population, what is the status of Lgr5? This should be shown in the figure since the argument is that this is an Lgr5-independent lineage. * See response to the second point.

And what about the regenerative, Yap, mesenchymal and inflammatory signatures? Are they enriched in the white crypts similar to the in vitro spheroids?

In a portion of white crypts, those we believe are newly formed after CBC ablation (see above), there is a transient increase in Clu, which may be considered a marker of Yap activation. In the CBC-like Olfm4 low cells, as seen by scRNAseq, there is nothing like an actively regenerating phenotype. This is expected, since these cells are coming from homeostatic untreated VilCre/Rosa26Tom animals and are supposed to be quiescent "awaiting to be activated".

Reviewer #1 (Significance (Required)):

Strengths: The study employed a range of in vitro and in vivo models to test the hypothesis.

• *

*Limitations: Unfortunately, the models chosen did not provide sufficient evidence to draw the conclusions. Injury induced reprogramming, both in vivo and in vitro, has been well documented in the field. The new message here is to show that such reprogrammed state is continuous rather than transient; instead of regenerating Lgr5+ stem cells, it can continue to differentiate to all cell lineages in Lgr5-independent manner.

*

We respectfully disagree with this analysis of our results. What we show is not "that such reprogrammed state is continuous rather than transient; instead of regenerating Lgr5+ stem cells, it can continue to differentiate to all cell lineages in Lgr5-independent manner", but that a quiescent stem cell line, not previously identified, is activated to regenerate a portion of crypts following CBC ablation. These cells are not reprogrammed, they correspond to a developmental lineage waiting to be activated and keep their VilCre-negative state at least of 30 days. We believe that their "by default tracing" (VilCre negative from the zygote stage) is as strong an evidence for the existence of such a lineage as positive lineage tracing would be. The increase in crypts originating from this lineage after CBC ablation indicates that it is implicated in regeneration. We do not question the well-demonstrated plasticity-associated reprogramming taking place during regeneration; we simply suggest that this would coexist with the involvement of the quiescent VilCre-negative lineage we have identified.

*However, through the manuscript, there was no immunostaining of Lgr5 and other differentiation markers. The conclusion is an overstatement without solid proof. * We have provided the best answer we could to this point in our answer to the second question of the referee hereabove.

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

In this manuscript, the Marefati et al. developed a novel approach to generate spheroids from adult intestinal epithelium using a collagenase/dispase based protocol. Adult spheroids were found to be distinct from classic budding-type organoids normally generated from EDTA based release of the crypt epithelium. Transcriptional profiling indicated that adult spheroids were undifferentiated and similar to regenerating crypts or fetal spheroids. To identify the cell of origin that generates adult spheroids, the authors labelled epithelial cells with VilCreERT-LSL-Tom, VilCre-LSL-GFP and Lgr5CreERT- LSLTom mice. From these experiments the authors conclude that that spheroids are only generated from Vil-Cre negative and Lgr5 negative cells. Next the authors deleted the anti- apoptotic gene Mcl1 using Vil-CreERT mice. This led to a strong apoptotic response throughout the crypt epithelium and tissues processed from knockout mice readily generated spheroids, and in vivo, replenishment of the gut epithelium was mediated by unrecombined cells. In a second model, CBCs were ablated using Lgr5DTR mice and VilCre negative cells were found again to contribute to regeneration of the crypt epithelium. Finally based on the absence of Vil-Cre reporter activity, the authors were able to sort out and perform scRNAseq to profile VilCre negative cells. These cells were found to be quiescent, express the stem cell marker Olfm4 and were also abundant in ribosomal gene expression.

• *

The fact that the authors developed a protocol to reproducibly generate fetal-like spheroids from adult tissue is an important advancement in the field. Previous reports have shown that treatment with various small molecule inhibitors can revert budding organoids into a spheroid morphology, but this manuscript demonstrates that spheroids can also be generated from otherwise untreated cells. This new methodology will provide new tools to dissect the molecular determinants of fetal/regenerative cells in the gut. Based on this, the manuscript should attract a significant amount of attention in the intestinal field.

• *

As pointed out by the authors themselves the study has important limitations that diminish enthusiasm. The primary issue relates to the inability of the team to identify markers of VilCre neg cells other than the fact that these cells are Olfm4+ and quiescent. Nonetheless, for the reasons stated above the manuscript should reach the target audience within the research community, if the authors can address the specific points below related to issues with methodology as well as defining more precisely the characteristics and growth requirements of adult spheroid cultures.

Thank you for this positive analysis of our study.

Major comments

The main conclusion of the study is that Vil-Cre neg cells are rare quiescent Olfm4+ crypt cells. If this is the case, then standard EDTA treatment should release these cells as well. Consequently, spheroids should also emerge from isolated crypts grown in the absence of ENR. If this is not the case how do the authors explain this?

We have tried hard to generate spheroids by culturing EDTA organoids in medium lacking ENR and by treating EDTA organoids with collagenase/dispase, without success. Therefore, we are left with the conclusion that spheroid-generating cells must be more tightly attached to the matrix than those released by EDTA, and that it is their release from this attachment by collagenase that triggers a regeneration-like phenotype. This hypothesis is supported by several models of regeneration in other tissues as indicated in our references (Gilbert et al., 2010; Machado et al., 2021; Montarras et al., 2005).

From the text the authors appear to suggest that growth of adult spheroids is dependent initially on "material" released by collagenase/dispase treatment. An obvious candidate would be mesenchymal cells, which are known to secrete factors such as Wnts and PGE2 that drive spheroid morphology. To test this, the authors should treat spheroid cultures with Porcupine and/or PGE2 inhibitors.

We followed similar reasoning, considering that spheroids express strongly Ptgs1 ,2 (Figure 3A). We thought their phenotype might be maintained by autocrine prostaglandin action. We tested aspirin, a Ptgs inhibitor, which was without effect on the spheroid phenotype. Besides, we explored a wide variety of conditions to test whether they would affect the spheroid phenotype [Aspirin-see above, cAMP agonists/antagonists, YapTaz inhibitors (verteporfin and CA3), valproic acid, Notch inhibitors (DAPT, DBZ, LY511455), all-trans retinoic acid, NFkB inhibitors (TCPA, BMS), TGFbeta inhibitor (SB431542)]. As these results were negative, we did not include them in the manuscript.

• If these inhibitors block growth then this would suggest that either stromal cells or autocrine signalling involving these pathways is important. Overall, more in-depth analysis of the growth requirements of adult spheroids is required.*

Figure 1d indicates that adult spheroids can be propagated for at least 10 passages. The abstract mentions they are "immortal". The text itself does not address this issue. More precise information as to how long spheroids can be propagated is required. If these cultures can be propagated for 10 passages or more it becomes important to determine what nutrients/mitogens in the basal media are driving growth? Alternatively, what is the evidence that spheroid cultures are completely devoid of mesenchymal cells. The text only mentions that "Upon replating, these spheroids could be stably cultured free of mesenchymal cells (Fig.1B)". No validation is shown to support this.

We agree that "immortal" is not a good way to characterize our spheroids, as also pointed out by referee nr 1. We have changed that in the text, indicating the maximal number of replating we tested was 26 and replacing immortal by stably replatable. Of note, the spheroids could frozen/thawed and recultured many times.

Related to the question whether mesenchymal cells could still contaminate the spheroid cultures, we can provide the following answers:

• No fibroblasts could be seen in replated cultures and multiple spheroids could be repeatedly propagated from a single starting spheroid.
• The bulk RNAseq experiment comparing organoids to ENR or BCM cultured spheroids show, despite expression of several mesenchymal markers (see matrisome in Fig3), absence of significant expression of Pdgfra (see in revision plan folder for CP20Millions results from the raw data of new suppl table 2, with Clu, Tacstd2 and Alpi shown as controls).
• Regarding the nutrients/mitogens in the medium driving spheroid growth, we did not explore the point further than showing that they grow in basal medium (i.e. advanced DMEM), given that the presence of Matrigel makes it difficult to pinpoint what is really needed. In Figure 2, the authors describe the growth requirements for adult spheroids and indicate that spheroids grown in ENR or EN became dark and shrink. The representative images showing this are clear, but this analysis should be quantified.

Added to the manuscript.

In SF3, the gene expression profile of organoids from the sandwich method only partially overlaps with that of organoids from the old protocol. What are the gene expression differences between the 2 culture systems? Secondly, the sandwich method appears to sustain growth of Tom+ spheroids based on RNAseq and the IF images. This suggest that Vil-Cre negative cells are not necessarily the only source of adult spheroids and thus this experiment seems to indicate that any cell may be converted to grow as a spheroid under the right conditions. These points should be addressed.

Looking back to our data in order to answer the point raised by the referee, we realized that we had inadvertently-compared organoids to ENR-cultured spheroids generated by the first protocol to BCM-cultured spheroids generated by the sandwich method. We have corrected this error in a new version of suppl fig3. This shows increased correspondence between genes up- or downregulated in the spheroids obtained in the two protocols (from 49/48% to 57/57% (Venn diagram on the new figure). We agree that, even after this correction, the spheroids obtained with the two protocols present sizeable differences in their transcriptome. However, considering the very different way these spheroids were obtained and cultured initially, we do not believe this to be unexpected. The important point in our opinion is that the core of the up- and down-regulated genes typical of the de-differentiation phenotype of adult spheroids is very similar, as shown in the heatmap (which was made with the correct samples!). Also, a key observation is that that both kind of spheroids survive and can be replated in basal medium. As already stated, this characteristic is only seen rare cases [spheroids obtained from rare FACS-purified cells (Smith et al 2018) or helminth-infected intestinal tissue (Nusse et al.2018)]. Together with the observation that the majority of them is not traced by VilCre constitutes what we consider the halmark of the spheroids described in our study. As shown in figure 4E (old protocol) and Suppl Fig.3 (sandwich protocol) both red and white spheroids were extremely low in VilCre expression. As stated in the text, the fact that some spheroids are nevertheless red is most probably related to the extreme sensitivity of the Rosa26Tom marker to recombination (Liu et al., 2013), but this does not mean that there are two phenotypically different kind of spheroids. It means that the arbitrary threshold of Rosa26Tom recombination introduces an artificial subdivision of spheroids with no phenotypical significance.

Regarding the point made by the referee that "that any cell may be converted to grow as a spheroid under the right conditions", we agree and have shown with others that organoids acquire indeed a spheroid phenotype when cultured for instance in fibroblasts-conditioned medium (see suppl fig1B and (Lahar et al., 2011; Roulis et al., 2020) quoted in the manuscript). However, these spheroids cannot be propagated in basal medium, and revert to an organoid phenotype when put back in ENR (Suppl fig1B).

*In Figure 4, the authors conclude that spheroids do not originate from Lgr5 cell derived clones even after 30days post Tam induction. Does this suggest that in vivo and under homeostatic conditions VilCre neg cells are derived from a distinct stem cell pool or are themselves a quiescent stem cell. Given the rarity of VilCre neg cells, the latter seems unlikely.

*

Despite their rarity, we believe VilCre-negative cells observed under homeostatic conditions are themselves quiescent stem cells. Actually, if they were derived from a larger stem cell pool, this pool should also be VilCre-negative. And we do not see such larger number of VilCre-neg cells under homeostatic conditions.

The problem with the original assertion is that Lgr5-CreERT mice are mosaic and therefore not all Lgr5+ cells are labelled in this model. "White" spheroids may thus derive from cells that in turn derive from these unlabelled Lgr5 cells.

We had considered the possibility that mosaicism [very low for VilCre (Madison et al., 2002); in the 40-50% range for Lgr5CreERT2 (Barker & Clevers. Curr Protoc Stem Cell Biol. 2010 Chapter 5)] could explain our data. We think, however that we can exclude this possibility on the basis that spheroids do not conform to the expected ratio of unrecombined cells, given the observed level of mosaicism. Indeed, for VilCre, a few percent, at most, of unrecombined cells in the epithelium translates into almost 100% unrecombined spheroids. For Lgr5CreERT2 mice, the mosaicism level is in the range of 40%, which is what we observe for EDTA organoids (Figure 4G), while spheroids were in their vast majority unrecombined.

We have included a discussion about the possible role of mosaicism in the new version.

ATACseq experiments were briefly mentioned in the manuscript but unfortunately little information was extracted from this experiment. What does this experiment reveal about the chromatin landscape of adult spheroids relative to normal organoids?

We only performed this experiment to search for an explanation to the paradoxical absence of expression of the VilCre transgene in spheroids, despite robust expression of endogenous villin (Suppl Fig.4). We chose to show the chromatin landscape of a gene equally expressed in both organoids and spheroids (Krt19), a gene specifically expressed in spheroids (Tacstd2) and the endogenous Villin gene also expressed in both. We believe that the observation of a clear difference in pattern of the chromatin accessibility around the endogenous villin gene in organoids and spheroids provides an explanation to the observed results. The cis regulatory sequences needed for expression of the endogenous villin gene seem to be different in organoids and spheroids, which may explain why the regulatory sequences present in the transgene (only 12.4kb) might not allow expression of the transgene in spheroids. We have added a sentence in the manuscript clarifying this point. Missing is obviously the chromatin landscape around the VilCre transgene, but this is beyond reach in such kind of experiments.

Reviewer #2 (Significance (Required)):

The fact that the authors developed a protocol to reproducibly generate fetal-like spheroids from adult tissue is an important advancement in the field. Previous reports have shown that treatment with various small molecule inhibitors can revert budding organoids into a spheroid morphology, but this manuscript demonstrates that spheroids can also be generated from otherwise untreated cells. This new methodology will provide new tools to dissect the molecular determinants of fetal/regenerative cells in the gut. Based on this, the manuscript should attract a significant amount of attention in the intestinal field.

Reviewer #3 (Evidence, reproducibility and clarity (Required)): CR-2024-02491

An Lgr5-independent developmental lineage is involved in mouse intestinal regeneration

Marefati et al.

Homeostatic maintenance of the intestinal epithelium has long been thought to rely upon Wnt signaling responsive Lgr5-expressing stem cells that reside at the crypt base.

However, myriad reported mechanisms or populations have been reported to underlie epithelial regeneration after injury. Many groups have reported that reacquisition of a fetal- link intestinal phenotype is an import part of the regenerative response, however the originating cell type has not been definitively identified. Herein, the authors demonstrate that cells from adult homeostatic intestine can generate immortal spheroids that resemble fetal spheroids and are derived independent of Lgr5+ intestinal stem cells (ISCs). The authors then draw the conclusion that this indicates that a hierarchical stem cell model applies to regeneration of the intestinal epithelium, in addition to the plasticity model.

• *

Comments:

1. Please indicate what species is used for studies in Fig 1.

All experiments were performed in Mus musculus.

Please clarify if Figure 2 studies utilize Matrigel or not.

Yes

RNA-seq analyses of adult intestinal generated spheroids lack the granularity of single cell analyses and thus it is unclear if this is a homogeneous population or if the population has diversity across it (i.e., enteroids/organoids have a high level of diversity). Many of the conclusions from the RNA-seq study are broad and generalized-for example Fig 3F indicates that markers of the +4 ISC populations (Bmi1, tert, lrig1, hopx) were all expressed similarly in adult spheroids as compared to adult organoids. However, while this may be true in the bulk-RNA-seq analyses, clearly scRNA-seq would provide a better foundation to make this statement, as enteroids/organoids are comprised of heterogeneous subpopulations. . .and it might indicate that these +4 markers have only very low expression in the spheroids. Based upon these concerns, misconclusions are likely to be drawn.

We agree and it would be certainly worthwhile to perform scRNAseq of adult spheroid populations. This would certainly be worth doing in future studies to explore the possible heterogeneity of adult spheroids. We nevertheless believe that our scRNAseq performed on homeostatic intestinal tissue from VilCre/Rosa26Tom mice identify Olfm4-low VilCre-neg cells that are likely at the origin of adult spheroids and display a quite homogenous phenotype.

*The language around Figure 4 results is confusing. Please define "white" and "red". It might be simpler to designate recombined versus not recombined lineage.

*

We have clarified this in the figure.

The hypothesis that collagenase/dispase solution acts as a proxy for injury is not demonstrated and backed by data. Thus, it is difficult to make the conclusion that this approach could represent a "stable avatar" of intestinal regenerating cells. It is clear that subpopulations of crypt-based cells generate spheroids in culture without collagenase/dispase (see the cited reference Smith et al, 2018).

* *Smith et al demonstrate clearly the possibility to obtain spheroids with properties probably similar to ours from EDTA derived intestinal crypt cells. However they need to prepurify them by FACS. Besides, Nusse et al describe spheroids similar to ours after infection of the intestine by helminths (Nusse et al. 2018). In our case, and for most labs preparing enteroids with the EDTA protocol, the result is close to 100% organoids. Even if we treat EDTA organoids with collagenase, we do not obtain spheroids. This brought us to the conclusion that spheroid-generating cells must be more tightly attached to the matrix than CBCs and that it is their release from the matrix that activates the spheroid regeneration-like phenotype. This hypothesis is supported by several models of regeneration in other tissues as indicated in our references (Gilbert et al., 2010; Machado et al., 2021; Montarras et al., 2005)

A study based on the absence of recombination in a VilCre lineage tracing scenario is not well-established to be strong experimental approach, as there are many reasons why recombination may not cells may not be lineage marked. In order to use this system as the authors intend, they first need to demonstrate that villin is not expressed in the discrete cell population that they are targeting. For the presented observational studies, this would be difficult to do. While they do demonstrate differences in chromatin accessibility between cells from organoids versus spheroids (fig s4), some of these differences could merely be due to the bulk analytical nature of the study and the lack of comparing stem cell populations from spheroids to stem cell populations from organoids-since the spheroids are likely homogenous versus the organoids that only have a small fraction of stem cells-and thus represent a mix of stem cell and differentiated cell populations. The authors do not demonstrate that villin protein expression varies in these cells.

If it were found that villin is not expressed in their "novel" population, then one would expect that the downstream use of villin-based recombination would demonstrate the same recombination potential (i.e., Mcl1 would not be recombined). Both recombination studies in Fig 6 are difficult to interpret, and thus it is not clear if these studies support the stated conclusions. Quantification of number of crypts that are negative should be reported as a percentage of recombined crypts.

We are sorry but there seems to be a complete misunderstanding of our data regarding the point raised by the referee. The important point of our initial observation is that despite robust expression of villin in spheroids, the VilCre transgene is not expressed (see figure 4E). This in our opinion makes absence of VilCre expression (or of Rosa marker recombination) a trustful marker of a new developmental lineage. All the data in figure 4 constitute an answer.

*The reasoning about heterogeneity of cell type in organoids versus probable homogeneity of spheroids is well taken. However, as the endogenous villin gene is expressed in all cells of both organoids and spheroids, it is highly significant that only spheroids do not express the transgene. *

We performed the ATACseq experiment to search for an explanation to the paradoxical absence of expression of the VilCre transgene in spheroids, despite robust expression of endogenous villin (Suppl Fig.4). We chose to show the chromatin landscape of a gene equally expressed in both organoids and spheroids (Krt19), a gene specifically expressed in spheroids (Tacstd2) and the endogenous Villin gene also expressed in both. We believe that the observation of a clear difference in pattern of the chromatin accessibility around the endogenous villin gene in organoids and spheroids provides an explanation to the observed results. The cis regulatory sequences needed for expression of the endogenous villin gene seem to be different in organoids and spheroids, which may explain why the regulatory sequences present in the transgene (only 12.4kb) might not allow expression of the transgene in spheroids. We have added a sentence in the manuscript clarifying this point. Missing is obviously the chromatin landscape around the VilCre transgene, but this is beyond reach in such kind of experiments.

*Figure 8 indicates that the cell population identified by scRNA-seq may be quiescent. Companion IF or IHC should be conducted to confirm this finding, as well as other conclusions from the informatics conducted.

*

We agree that additional experiments could be performed to support this point. We are unfortunately not in a position to perform these experiments (see section 4 below).

Clearly the data is intriguing, however, the conclusion is strong and is an over interpretation of the presented data. There are a number of validation or extension data that would enhance the overall interpretation of the study: 1. validation of scRNA-seq or bulk RNA-seq concepts by protein staining of intestinal tissues in the damage model will serve as a secondary observation. 2. identification of the ISC that they are defining is critical and important. There is already the notion that this cell type exists and it has been shown with various different markers. 3. expand the analyses of the fetal-like expression profiling to injured intestines to demonstrate that the lineage negative cells indeed express fetal-like proteins. 4. expand the discussion of the Clu+ cell type. Is this cell the previously described revival cell? If so, how does this body of work provide unique aspects to the field?

We agree that all these suggested experiments could be performed and would be of interest. However, we consider that they would not modify the main message of our study and would only constitute an expansion of the present work. As already stated, we are not in the position to perform them (see section 4).

*There is some level of conflicting data, with the stem population being proliferative in culture stimulated by the stromal cells, but quiescent in vivo and also based upon scRNA- seq data in Fig 9.

*

We do not see any conflict in our observation regarding this point. The observation that cells that are quiescent in vivo become proliferative when subjected to culture (with or without addition of stromal cells) is routinely made in a multitude of cell culture systems. In particular, it has been shown that intestinal tissue dissociation activates the Yap/Taz pathway, resulting in proliferation (Yu et al. Hippo Pathway Regulation of Gastrointestinal Tissues. Annual Review of Physiology, 2015 Volume 77, 201-227).

Many of the findings have been previously reported: Population that grows as spheroids (Figure 2), Population that is Wnt independent (Figure 2), Lgr5 independent regenerative growth of the intestine (figure 3F, Figure 4), Clu+ ISCs drive regeneration (Figure 7).

Whereas these individual findings have indeed been reported, it was in a different context. We strongly disagree with the underlying suggestion that our study would not bring new information. We have identified here a developmental lineage involved in intestinal regeneration that has not been described up to now.

Minor comments:

1. • The statement that spheroids must originate from collagenase/dispase digested material might be an overstatement. As spheroids generation from EDTA treated intestines have been previously reported (Smith et al, 2018). * See answer to point 4 above. *Overall while the study includes an extensive amount of work and different approaches, a foundationally supported novel finding is missing. Many of the statements have already been demonstrated by others in the fields. In addition, one of the most intriguing aspects of the study is that the stromal population impacts this stem cell population, however, interactions and factors stimulating the crosstalk are not addressed.

*

Reviewer #3 (Significance (Required)):

Overal while the study includes an extensive amount of work and different approaches, a foundationally supported novel finding is missing. Many of the statements have already been demonstrated by others in the fields. In addition, one of the most intriguing aspects of the study is that the stromal population impacts this stem cell population, however, interactions and factors stimulating the crosstalk are not addressed.

We can only disagree.

4. Description of analyses that authors prefer not to carry out

• *

We have answered most questions raised by the referees by explaining our view, by clarifying individual points and, in several cases, by providing additional information that was not included in the original manuscript.

In a limited number of cases when additional experiments were suggested, we were unfortunately obliged to write that we are not in a position to perform them. This is because my lab is closing after more than fifty years of uninterrupted activity. There will unfortunately be nobody to perform additional experiments.

Nevertheless, as written by referees 1 and 2, we believe that the revised manuscript, as it stands, contains data that will be of interest to the people in the field and may be the bases for future developments. We hope editors will find interest in publishing it.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

RC-2024-02491

An Lgr5-independent developmental lineage is involved in mouse intestinal regeneration Marefati et al.

Homeostatic maintenance of the intestinal epithelium has long been thought to rely upon Wnt signaling responsive Lgr5-expressing stem cells that reside at the crypt base. However, myriad reported mechanisms or populations have been reported to underlie epithelial regeneration after injury. Many groups have reported that reacquisition of a fetal-link intestinal phenotype is an import part of the regenerative response, however the originating cell type has not been definitively identified. Herein, the authors demonstrate that cells from adult homeostatic intestine can generate immortal spheroids that resemble fetal spheroids and are derived independent of Lgr5+ intestinal stem cells (ISCs). The authors then draw the conclusion that this indicates that a hierarchical stem cell model applies to regeneration of the intestinal epithelium, in addition to the plasticity model.

Comments:

1. Please indicate what species is used for studies in Fig 1.
2. Please clarify if Figure 2 studies utilize Matrigel or not.
3. RNA-seq analyses of adult intestinal generated spheroids lack the granularity of single cell analyses and thus it is unclear if this is a homogeneous population or if the population has diversity across it (i.e., enteroids/organoids have a high level of diversity). Many of the conclusions from the RNA-seq study are broad and generalized-for example Fig 3F indicates that markers of the +4 ISC populations (Bmi1, tert, lrig1, hopx) were all expressed similarly in adult spheroids as compared to adult organoids. However, while this may be true in the bulk-RNA-seq analyses, clearly scRNA-seq would provide a better foundation to make this statement, as enteroids/organoids are comprised of heterogeneous subpopulations. . .and it might indicate that these +4 markers have only very low expression in the spheroids. Based upon these concerns, misconclusions are likely to be drawn.
4. The language around Figure 4 results is confusing. Please define "white" and "red". It might be simpler to designate recombined versus not recombined lineage.
5. The hypothesis that collagenase/dispase solution acts as a proxy for injury is not demonstrated and backed by data. Thus, it is difficult to make the conclusion that this approach could represent a "stable avatar" of intestinal regenerating cells. It is clear that subpopulations of crypt-based cells generate spheroids in culture without collagenase/dispase (see the cited reference Smith et al, 2018).
6. A study based on the absence of recombination in a VilCre lineage tracing scenario is not well-established to be strong experimental approach, as there are many reasons why recombination may not cells may not be lineage marked. In order to use this system as the authors intend, they first need to demonstrate that villin is not expressed in the discrete cell population that they are targeting. For the presented observational studies, this would be difficult to do. While they do demonstrate differences in chromatin accessibility between cells from organoids versus spheroids (fig s4), some of these differences could merely be due to the bulk analytical nature of the study and the lack of comparing stem cell populations from spheroids to stem cell populations from organoids-since the spheroids are likely homogenous versus the organoids that only have a small fraction of stem cells-and thus represent a mix of stem cell and differentiated cell populations. The authors do not demonstrate that villin protein expression varies in these cells. If it were found that villin is not expressed in their "novel" population, then one would expect that the downstream use of villin-based recombination would demonstrate the same recombination potential (i.e., Mcl1 would not be recombined). Both recombination studies in Fig 6 are difficult to interpret, and thus it is not clear if these studies support the stated conclusions. Quantification of number of crypts that are negative should be reported as a percentage of recombined crypts.
7. Figure 8 indicates that the cell population identified by scRNA-seq may be quiescent. Companion IF or IHC should be conducted to confirm this finding, as well as other conclusions from the informatics conducted.
8. Clearly the data is intriguing, however, the conclusion is strong and is an over interpretation of the presented data. There are a number of validation or extension data that would enhance the overall interpretation of the study:
   • a. validation of scRNA-seq or bulk RNA-seq concepts by protein staining of intestinal tissues in the damage model will serve as a secondary observation.
   • b. identification of the ISC that they are defining is critical and important. There is already the notion that this cell type exists and it has been shown with various different markers.
   • c. expand the analyses of the fetal-like expression profiling to injured intestines to demonstrate that the lineage negative cells indeed express fetal-like proteins.
   • d. expand the discussion of the Clu+ cell type. Is this cell the previously described revival cell? If so, how does this body of work provide unique aspects to the field?
9. There is some level of conflicting data, with the stem population being proliferative in culture stimulated by the stromal cells, but quiescent in vivo and also based upon scRNA-seq data in Fig 9.
10. Many of the findings have been previously reported: Population that grows as spheroids (Figure 2), Population that is Wnt independent (Figure 2), Lgr5 independent regenerative growth of the intestine (figure 3F, Figure 4), Clu+ ISCs drive regeneration (Figure 7).

Minor comments:

1. The statement that spheroids must originate from collagenase/dispase digested material might be an overstatement. As spheroids generation from EDTA treated intestines have been previously reported (Smith et al, 2018).

Overall while the study includes an extensive amount of work and different approaches, a foundationally supported novel finding is missing. Many of the statements have already been demonstrated by others in the fields. In addition, one of the most intriguing aspects of the study is that the stromal population impacts this stem cell population, however, interactions and factors stimulating the crosstalk are not addressed.

Significance

Overall while the study includes an extensive amount of work and different approaches, a foundationally supported novel finding is missing. Many of the statements have already been demonstrated by others in the fields. In addition, one of the most intriguing aspects of the study is that the stromal population impacts this stem cell population, however, interactions and factors stimulating the crosstalk are not addressed.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

In this manuscript, the Marefati et al. developed a novel approach to generate spheroids from adult intestinal epithelium using a collagenase/dispase based protocol. Adult spheroids were found to be distinct from classic budding-type organoids normally generated from EDTA based release of the crypt epithelium. Transcriptional profiling indicated that adult spheroids were undifferentiated and similar to regenerating crypts or fetal spheroids. To identify the cell of origin that generates adult spheroids, the authors labelled epithelial cells with VilCreERT-LSL-Tom, VilCre-LSL-GFP and Lgr5CreERT-LSLTom mice. From these experiments the authors conclude that that spheroids are only generated from Vil-Cre negative and Lgr5 negative cells. Next the authors deleted the anti-apoptotic gene Mcl1 using Vil-CreERT mice. This led to a strong apoptotic response throughout the crypt epithelium and tissues processed from knockout mice readily generated spheroids, and in vivo, replenishment of the gut epithelium was mediated by unrecombined cells. In a second model, CBCs were ablated using Lgr5DTR mice and VilCre negative cells were found again to contribute to regeneration of the crypt epithelium. Finally based on the absence of Vil-Cre reporter activity, the authors were able to sort out and perform scRNAseq to profile VilCre negative cells. These cells were found to be quiescent, express the stem cell marker Olfm4 and were also abundant in ribosomal gene expression.

The fact that the authors developed a protocol to reproducibly generate fetal-like spheroids from adult tissue is an important advancement in the field. Previous reports have shown that treatment with various small molecule inhibitors can revert budding organoids into a spheroid morphology, but this manuscript demonstrates that spheroids can also be generated from otherwise untreated cells. This new methodology will provide new tools to dissect the molecular determinants of fetal/regenerative cells in the gut. Based on this, the manuscript should attract a significant amount of attention in the intestinal field.

As pointed out by the authors themselves the study has important limitations that diminish enthusiasm. The primary issue relates to the inability of the team to identify markers of VilCre neg cells other than the fact that these cells are Olfm4+ and quiescent. Nonetheless, for the reasons stated above the manuscript should reach the target audience within the research community, if the authors can address the specific points below related to issues with methodology as well as defining more precisely the characteristics and growth requirements of adult spheroid cultures.

Major comments

The main conclusion of the study is that Vil-Cre neg cells are rare quiescent Olfm4+ crypt cells. If this is the case, then standard EDTA treatment should release these cells as well. Consequently, spheroids should also emerge from isolated crypts grown in the absence of ENR. If this is not the case how do the authors explain this?

From the text the authors appear to suggest that growth of adult spheroids is dependent initially on "material" released by collagenase/dispase treatment. An obvious candidate would be mesenchymal cells, which are known to secrete factors such as Wnts and PGE2 that drive spheroid morphology. To test this, the authors should treat spheroid cultures with Porcupine and/or PGE2 inhibitors. If these inhibitors block growth then this would suggest that either stromal cells or autocrine signalling involving these pathways is important. Overall, more in-depth analysis of the growth requirements of adult spheroids is required.

Figure 1d indicates that adult spheroids can be propagated for at least 10 passages. The abstract mentions they are "immortal". The text itself does not address this issue. More precise information as to how long spheroids can be propagated is required. If these cultures can be propagated for 10 passages or more it becomes important to determine what nutrients/mitogens in the basal media are driving growth? Alternatively, what is the evidence that spheroid cultures are completely devoid of mesenchymal cells. The text only mentions that "Upon replating, these spheroids could be stably cultured free of mesenchymal cells (Fig.1B)". No validation is shown to support this.

In Figure 2, the authors describe the growth requirements for adult spheroids and indicate that spheroids grown in ENR or EN became dark and shrink. The representative images showing this are clear, but this analysis should be quantified.

In SF3, the gene expression profile of organoids from the sandwich method only partially overlaps with that of organoids from the old protocol. What are the gene expression differences between the 2 culture systems? Secondly, the sandwich method appears to sustain growth of Tom+ spheroids based on RNAseq and the IF images. This suggest that Vil-Cre negative cells are not necessarily the only source of adult spheroids and thus this experiment seems to indicate that any cell may be converted to grow as a spheroid under the right conditions. These points should be addressed.

In Figure 4, the authors conclude that spheroids do not originate from Lgr5 cell derived clones even after 30days post Tam induction. Does this suggest that in vivo and under homeostatic conditions VilCre neg cells are derived from a distinct stem cell pool or are themselves a quiescent stem cell. Given the rarity of VilCre neg cells, the latter seems unlikely. The problem with the original assertion is that Lgr5-CreERT mice are mosaic and therefore not all Lgr5+ cells are labelled in this model. "White" spheroids may thus derive from cells that in turn derive from these unlabelled Lgr5 cells.

ATACseq experiments were briefly mentioned in the manuscript but unfortunately little information was extracted from this experiment. What does this experiment reveal about the chromatin landscape of adult spheroids relative to normal organoids?

Significance

The fact that the authors developed a protocol to reproducibly generate fetal-like spheroids from adult tissue is an important advancement in the field. Previous reports have shown that treatment with various small molecule inhibitors can revert budding organoids into a spheroid morphology, but this manuscript demonstrates that spheroids can also be generated from otherwise untreated cells. This new methodology will provide new tools to dissect the molecular determinants of fetal/regenerative cells in the gut. Based on this, the manuscript should attract a significant amount of attention in the intestinal field.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

In this manuscript, Marefati et al report an Lgr5-independent lineage in the regenerating intestine using in vitro organoids and in vivo injury-coupled lineage tracing model. In organoids, collagenase/dispase dissociated resulted in "immortal spheroids" that maintain a cystic and undifferentiated phenotype in the absence of standard growth factors (Rspondin/Noggin/EGF). Bulk RNAseq of spheroids demonstrates downregulation of classical CBC signatures and upregulation of fetal spheroid, mesenchymal, inflammation and regenerative signatures. In mice, Villin-Cre lineage tracing revealed some Villin-negative progenies that lack reporter tracing throughout crypt-villus ribbons after injury. The authors proposed that there is Lgr5-independent population support the regenerative response upon CBC depletion. A major caveat of this study is the identification of this population is based on absence of VilCre expression. It is surprising that there is no characterisation of Lgr5 expression throughout the manuscript whilst claiming of a Lgr5-independent lineage. Although the research question is potentially interesting, the concept of epithelial reprogramming upon injury is well documented in the field. The data generated in this manuscript also seem to be preliminary and lack of detailed characterisation. Below are specific comments.

• Expression of Lgr5 should be properly characterised throughout the manuscript in both organoid models and injury-induced regeneration in vivo.
• An important question is the origin of these "Lgr5-independent" adult spheroids. They look and appear like fetal organoids, which could be induced by injury (e.g. upon collagenase/dispase dissociation). Have the authors tried to culture fetal spheroids in BCM over extensive period of time? Do they behave the same? This would be a great way to directly compare the collagenase/dispase-derived organoids with fetal origin.
• Fig 1C, Why is the replating spheroid culture time different between mesenchymal cells and conditioned medium?
• It is unclear how the bulk RNA-seq data in Fig. 3 were compared. How long were the adult organoids and spheroids cultured for (how many passages)? Were they culture in the same condition of were they in ENR vs BCM? These are important information to consider when interpreting the results. For instance, are Ptgs1 & Ptgs2 expression in adult spheroids the same in ENR vs BCM? Are the gene signatures (regenerative, fetal and YAP) changed in adult spheroids culturing in ENR vs BCM?
• It is stated: "In agreement with their aptitude to grow indefinitely, adult spheroids express a set of upregulated genes overlapping significantly with an "adult tissue stem cell module" [159/721 genes; q value 2.11 e-94) (Fig.S2F)].". What is the definition of "indefinitely"? Are they referring to the Fig 1B where spheroid were passaged to P10? The authors should avoid the term "indefinitely" but use a more specific time scale, e.g. passages, months etc.
• SuppFig 3D: Row Z-Score is missing the "e" in Score.
• Fig 4E: Figure legend says QNRQ instead of CNRQ.
• Fig 4G: The brightfield image of adult spheroids 5 days after 3x TAM injections doesn't look like a spheroid. It seems to be differentiating.
• Fig 4: Most mouse model data are missing the number of mice & their respective age used for organoid isolation.
• Fig 4A-D, H-G: How was fluorescent signal of organoids quantified? How many images? Were there equal numbers of organoids? This all needs to be included in methods/figure legends.
• Figure 4B-D, G-H: Which culturing conditions were used for adult spheroids? Original method or sandwich method?
• Fig 6D-E: Please add the timepoint after DT administration these samples are from. It is not listed in text or figure legend.
• SuppFig 6D: again timepoint is missing.
• SuppFig 6: How were the crypts of these mice (DT WT & DT HE) isolated? Was this via EDTA? Also, what is the timepoint for isolation for these samples? Even if untreated, the timepoint adds context to the data. Please add more context to describing these different experiments, either in the figure legends or methods section.
• SuppFig 6E: The quality of the heatmap resolution is too poor to read gene names.
• Fig.5-7, are the regenerating crypt-villus units fully differentiated or are they maintained in the developmental state? Immunostaining of markers for stem cells (Lgr5), differentiated lineages (Alpi, Muc2, Lyz, ChgA etc.) and fetal state (Sca1, Trop2 etc) should be analysed in those "white" unrecombined crypt-villus units.
• The following text needs clarification:

"The kinetics of appearance of newly formed un-recombined ("white") crypts was studied after a single pulse of DT (Fig.7A). This demonstrated an increase at 48 hours, with further increase at day 10 and stable maintenance at day 30. The presence of newly formed white crypts one month after toxin administration indicates that the VilCre-negative lineage is developmentally stable and does not turn on the transgene during differentiation of the various epithelial lineages occurring after regeneration (Fig.7B). Comment: The "newly formed" is an overstatement, the data doesn't conclude that those are "new" crypts. The end of the sentence states that these "white" crypts form developmentally stable lineages, thus these white crypts at day 30 could originate from the initial injury. There was no characterisation of the various epitheial lineages. Are they fully differentiated? Is Lgr5 expressed one month after toxin administration? Can the VilCre neg lineage give rise to CBCs?

The relationship between white crypt generation and appearance of Clu-positive revival cells (Ayyaz et al., 2019) was then explored. In agreement with others and similar to what happens in the irradiation model, (Ayyaz et al., 2019; Yuan et al., 2023) Clu-positive cells were rare in crypts of untreated mice and their number transiently increased forty-eight hours after a single pulse of DT, and more so after three pulses of DT (Fig.7C,D). Comment: Comparing 1 pulse at day 2 vs 3 pulses at day 5 makes the data hard to interpret. How is the Clu ISH level for 1 pulse at day 5? Are they equivalent?

Clu-positive cells were less frequently observed in white crypts (see "Total" versus "White" in Fig.7C). This fits with the hypothesis that Clu expression marks acutely regenerating crypts and that a proportion of the white crypts are chronically regenerating due to DTR expression in CBCs." Comment: I believe the authors suggested that the discrepancy of less Clu expression in white crypts is due to the ectopic expression of DTR in CBCs causing low grade injury without DT administration. This means that some white crypts could have been formed before the administration of DT, and thus are on a different regenerative timeline compared to the white crypts formed from DT administration. Is there any proof of the chronic regeneration? Immunostaining of chronic regenerative markers such as Sca1, Anxa1 or Yap1 nuclear localization would support the claim. It'd be important to show only the white crypts, but not the RFP+ ones, show regenerative markers. - Fig 7D-E: What are the timepoints of harvest for HE-WT-HE 1 pulse DT mice and HE-HE-HE PBS injected mice? - Fig 8-9: Regarding the CBC-like Olfm4 low population, what is the status of Lgr5? This should be shown in the figure since the argument is that this is an Lgr5-independent lineage. And what about the regenerative, Yap, mesenchymal and inflammatory signatures? Are they enriched in the white crypts similar to the in vitro spheroids?

Significance

Strengths: The study employed a range of in vitro and in vivo models to test the hypothesis.

Limitations: Unfortunately, the models chosen did not provide sufficient evidence to draw the conclusions. Injury induced reprogramming, both in vivo and in vitro, has been well documented in the field. The new message here is to show that such reprogrammed state is continuous rather than transient; instead of regenerating Lgr5+ stem cells, it can continue to differentiate to all cell lineages in Lgr5-independent manner. However, through the manuscript, there was no immunostaining of Lgr5 and other differentiation markers. The conclusion is an overstatement without solid proof.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Manuscript number: RC-2023-02306

Corresponding author(s): John, Yates

[Please use this template only if the submitted manuscript should be considered by the affiliate journal as a full revision in response to the points raised by the reviewers.

• *

If you wish to submit a preliminary revision with a revision plan, please use our "Revision Plan" template. It is important to use the appropriate template to clearly inform the editors of your intentions.]

1. General Statements [optional]

We greatly appreciate the reviewers taking time from their busy scientific careers to evaluate our manuscript. We were elated to read all the positive comments, such as “the conclusions are well-supported and convincing”, “should contribute to a more nuanced understanding of SCZ pathogenesis”; “The potential implications for drug development underscore the broader significance of the study in advancing our knowledge of neurobiology and its relevance to neurological disorders like schizophrenia”, and “The study is informative, and has great potential to enrich the specific literature of this field”. We also found the constructive criticism very helpful for improving our manuscript. We performed additional experiments and bioinformatic analyses, as requested. We modified the manuscript to answer the reviewers’ questions. Due to its complexity, it is difficult to describe the different and sometimes conflicting hypotheses of SCZ pathogenesis in a single manuscript. This complexity is reflected in the conflicting requests from the reviewers. One reviewer requested we investigate and highlight the role of non-neuronal cells in SCZ while another reviewer suggested we did not focus enough on synaptic proteins. We believe we have achieved a balance to represent the intricacy of SCZ biology and the different opinions of the reviewers.

Thanks again.

2. Point-by-point description of the revisions

This section is mandatory. *Please insert a point-by-point reply describing the revisions that were already carried out and included in the transferred manuscript. *

• *

Reviewer #1 (Evidence, reproducibility and clarity (Required)):

Summary: Provide a short summary of the findings and key conclusions (including methodology and model system(s) where appropriate). In this manuscript, McClatchy and colleagues used a conventional approach combining immunoprecipitation (IP) of endogenous target proteins (baits) followed by liquid chromatography mass spectrometry (MS) analysis of the co-immunoprecipitating proteins to map protein-protein interaction (PPI). This interaction network is centered around baits that had been annotated as susceptibility factors for schizophrenia (SCZ). A variety of previous studies have identified thousands of such SCZ susceptibility factors. Mostly based on the availability of antibodies, 8 bait proteins were selected in this study. The authors reasoned that immunoprecipitating endogenous proteins from tissues using specific antibodies was a more accurate view of physiological conditions than epitope tagging followed by affinity purification (AP) from cells in culture. The model system from which proteins were extracted was the hippocampus dissected from mice that had been treated or not by phencyclidine (PCP), a drug that has been shown to induce SCZ symptoms in humans and animals. By comparing the proteins identified and quantified from the PCP-treated samples against control IPs and/or saline-injected mouse controls, a large number of PPI were deemed statistically significant. Most of these potential interactors were not present in PPI databases (BioGRID), most likely because such databases are populated with large-scale APMS datasets from cell cultures, with very few studies using brain tissue. Strikingly, many of the co-immunoprecipitated proteins were also known as SCZ susceptibility factors, which lend weight to the hypothesis that these factors form a large protein interaction network, localized at the synapses.

Major comments: - Are the key conclusions convincing? Overall, the conclusions drawn from the experimental design, data analysis, and corroboration with existing literature are well-supported and convincing. When selecting the SCZ susceptibility factors, the authors clearly state their goal, the databases used for gene selection, and the rationale for choosing proteins with synaptic localization. The inclusion of evidence from genetic studies and previous publications strengthens the credibility of the selected genes. The methodology used to establish the novel SCZ PPI network is mostly well-described (see minor comments below). The use of an 15N internal standard also adds rigor to the quantitation of PPI. The GO enrichment analysis provides valuable insights into the biological functions and cellular components associated with the SCZ PPI network. The annotation of identified proteins using the SynGo synaptic database and the distribution of annotated synaptic proteins among different baits further support the biological relevance of this PPI network. The cross-referencing of the PPI network with published genetic studies on SCZ susceptibility genes adds robustness to the findings. Specifically, the observation that 68% of protein interactors have evidence of being potential SCZ risk factors is a strong corroboration of the prevailing hypothesis in the field. Finally, the significant changes induced by PCP that were identified for all baits except Syt1, along with the comparison of altered proteins with SAINT-identified PPI, add depth to the understanding of PCP modulation.

- Should the authors qualify some of their claims as preliminary or speculative, or remove them altogether? No, but note that APMS/IPMS has been around for more than a decade (Introduction page 3).

We agree and did not mean to imply that IP-MS is new technology. We tried to convey that IP-MS is not new technology, but the number of IP-MS studies employed to study the PPI of endogenous proteins in brain tissue is a small percentage of all the published PPI MS studies.

We added the following to the Conclusions to clarify this point: “Although IP-LC-MS technology has been employed for more than a decade, quantitation of proteins using this strategy in mammalian tissue is scarce in the literature.””

- Would additional experiments be essential to support the claims of the paper? Request additional experiments only where necessary for the paper as it is, and do not ask authors to open new lines of experimentation. One piece of data that is missing are Western blots using the 8 selected antibodies against the proteins extracted from their experimental samples to validate the antibodies recognize 1 protein of the expected size from these tissue extracts.

We took your suggestion and performed immunoblots with our 8 IP antibodies using the starting material (i.e. rat brain hippocampus). All antibodies recognized a single band of the approximate molecular weight of the target except for the Gsk3b, which produced a doublet instead of a single band. This image is similar to what has been observed with the phosphorylation of Gsk3b(Krishnankutty, Kimura et al. 2017, Vainio, Taponen et al. 2021). To provide evidence that the additional band observed for Gsk3b is the phosphorylated target protein, we searched our Gsk3b IP dataset for a differential phosphorylation (i.e. 79.9663) on S,T, or Y. Even though we did not perform phosphorylation enrichment, we identified S389 as abundantly phosphorylated in all Sal and PCP samples consistent with our immunoblot. Images of these immunoblots are now Supplementary Figure 1.

• Are the suggested experiments realistic in terms of time and resources? It would help if you could add an estimated cost and time investment for substantial experiments. Running SDS-PAGE and Western blotting should be straightforward and cheap.

- Are the data and the methods presented in such a way that they can be reproduced? Yes

- Are the experiments adequately replicated and statistical analysis adequate? Yes

Minor comments: - Specific experimental issues that are easily addressable. The rationale for the short duration between PCP injection and animal sacrifice is only explained in the discussion section (page 17). The fact that this short treatment of less than 30 min should prevent any change in transcription or translation should be introduced earlier (in the experimental procedures).

We agree this is an important aspect of the study and that it suggests that the effect of PCP is independent of changes in transcription and translation as stated in the Discussion.

We added the following to the Introduction:

“PCP was administered for less than 30min., which precluded any changes in transcription or translation and allowed us to focus on PPI.*” *

Note that the duration is written as 26 min on page 4 and 25 min on page 9. Please reconcile these numbers*. *

We have corrected this typo. It was 26min.<br /> Is there any biological significance for this SCZ study that the mice were maintained on a reverse day-night cycle?

Rats are nocturnal animals, i.e. active at night and sleep during the day. In this study, rats were housed on a reverse day-night cycle so that assessment of the response to PCP could be evaluated during their active phase. This is not specific SCZ research and is the routine protocol for behavioral testing in the Powell laboratory. It is not clear from reading Experimental Procedures/Bioinformatic Analysis section (page 6) if normalized N14/N15 protein ratios measured in the bait-IPs and control-IPs were used for the SAINT analysis? Or did the authors used label-free quantitation with spectral counts?

We apologize for not making the methods clearer. In the results, it is stated that the N14 identifications are used in the SAINT analysis, and we state in the Discussion that SAINT uses spectral counts. We modified the Experimental Procedures/Bioinformatic Analysis section (page 6) to state: The input for SAINT was only the 14N identifications.

*- Are prior studies referenced appropriately? Yes

• Are the text and figures clear and accurate? *Fig1C: The workflow is a little too simple, the authors might want to add more details.

We revised Fig1C with more details as suggested.

FigS1C: Please add x-axis title (spectral counts) directly to the figure.

“Spectral counts” was added to the x-axis. FigS1C is now FigS2C ,with the addition of the immunoblots you suggested. Fig2B-D: The color scale bar should have number values to denote lower and upper limits in % (as opposed to "lowest" and "highest"). Numerical values were added to replace the upper and lower limits. - Do you have suggestions that would help the authors improve the presentation of their data and conclusions? No * *

Reviewer #1 (Significance (Required)):

• Describe the nature and significance of the advance (e.g. conceptual, technical, clinical) for the field. In this study, the authors have drastically expanded the protein interaction landscape around 8 known SCZ susceptibility factors by using a conventional IPMS approach. Performing the IPs on protein extracted from hippocampus dissected from mice treated with phencyclidine to model SCZ increases the biological significance of such lists of proteins. Furthermore, the co-immunoprecipitation of many other SCZ susceptibility factors along with the 8 selected baits supports the hypothesis that these proteins of varied functions are part of large interaction networks. Overall, the integration of experimental data with in silico networks, along with the quantification of PPI changes in response to PCP, should contribute to a more nuanced understanding of SCZ pathogenesis. The potential implications for drug development underscore the broader significance of the study in advancing our knowledge of neurobiology and its relevance to neurological disorders like schizophrenia.

• Place the work in the context of the existing literature (provide references, where appropriate). Overall, this study contributes to the existing literature by providing experimental data on in vivo PPI networks related to SCZ risk factors. Not only do the authors validate 124 known interactions but also they identify many novel PPI, due to a gap in the existing literature regarding the comprehensive mapping of PPI directly from tissue extracts, especially brain tissue. The authors advocate for more IPMS studies in mammalian tissues to generate robust tissue-specific in silico networks, which agrees with the growing understanding of the importance of tissue-specific networks for identifying disease mechanisms and potential drug targets. Furthermore, the SCZ PPI network reported here is enriched in proteins previously associated with SCZ, which aligns with the existing literature emphasizing the involvement of certain proteins and pathways in the pathogenesis of SCZ [References: 78-85]. The authors also investigate the response of the SCZ network to PCP treatment, hence providing insights into the potential effects of post-translational modifications, protein trafficking, and PPI alterations in a model of schizophrenia, which adds to existing knowledge about the impact of PCP on the molecular processes associated with SCZ [References: 88, 89, 92].

• State what audience might be interested in and influenced by the reported findings. Overall, the findings reported in this manuscript have implications for both basic research in molecular biology and potential translational applications in the development of targeted therapies for neurological disorders, particularly schizophrenia. The study delves into in vivo protein-protein interaction (PPI) networks related to genes implicated in schizophrenia (SCZ) risk factors. Researchers in neuroscience, molecular biology, and psychiatry would find the information valuable for understanding the molecular basis of SCZ. The study highlights the potential for identifying disease "hubs" that could be drug targets. Pharmacologists and drug developers interested in targeting protein complexes for drug development, especially in the context of neurological disorders, may find the study relevant.

• Define your field of expertise with a few keywords to help the authors contextualize your point of view. Indicate if there are any parts of the paper that you do not have sufficient expertise to evaluate. Technical Expertise | biochemistry, liquid chromatography mass spectrometry, proteomics, computational biology, protein engineering, protein interaction networks, post-translational modifications, protein crosslinking, proximity labeling, limited proteolysis, thermal shift assay, label-free and isotope-labeled quantitation. Biological Applications | human transcriptional complexes, apicomplexan parasites, viruses, nuclear envelope, ubiquitin ligases, non-model organisms.

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

Summary: McClatchy, Powell and Yates aimed at identifying a protein interactome associated to schizophrenia. For that, they treated rats (N14 and N15) with PCP, which disturbs gutamatergic transmission, as a model for the disease and co-immunoprecipitated hippocampi proteins, which were further analyzed by standard LC-MS.

The study is new, considering not much has been done in this direction in the field of schizophrenia. This justifies its publication. On the other hand, a major flaw of the is the lack of information on the level of interaction of the so called protein interactome. Meaning, we cannot distinguish, as the study was performed, which proteins are directly interacting with the targets of interest from proteins which are interacting with targets´ interactors. The different shells of interaction are crucial information in protein interactomics.

Major: most of I am pointing below must be at least discussed or better presented in the paper, as It may not be solvable considering how the study has been conducted.

1) The study fails in defining the level of interaction of the protein interactome with the considered targets. This has been shortly mentioned in the discussion, but must be more explicit to readers, for instance, in the abstract, introduction and in the methods sections. We agree this is crucial information that is absent from our dataset. As we explained in the Discussion, we cannot distinguish between PPI that are direct interactors with the target protein and PPI that reside in a multi-protein complex that includes the protein (i.e. indirect). This is an inherent problem with any IP-MS study. We amended the Introduction to highlight the ambiguity of the interaction data produced by the IP-MS approach, as you suggested.

Text added to the Introduction:

“Regardless of whether Ab or tagged proteins are employed to identify PPI from a biological sample, it cannot be determined if the identified interactor binds directly to the target protein or reside in a complex of proteins that includes the target protein (i.e. indirect).”

Since this important information is routinely missing from IP-MS studies, we decided to try to determine the level of interaction by using the artificial intelligence algorithm AlphaFold3(AF3). We believe it is not yet optimized for PPI, but AF3 is a big leap forward in the field of structural biology. For example, we observed AF3 did not predict high confident structures for our large membrane target proteins and was unable to validate known direct PPI of these targets. In addition, analyzing data with AF3 is currently not automated or streamlined so with ~1600 PPI identified in our dataset, we chose to look at one target protein, Ppp1ca. AF3 identified many known direct binding proteins in our Ppp1ca PPI dataset, which gives high confidence to the novel PPI predicted to be direct interactors. The AF3 data is encompassed in an additional Figure 6.

The following was added to the Results Section:

“A disadvantage of IP-MS studies is that it cannot distinguish between a PPI that binds directly to the target protein, and a PPI in which the interactor and target protein reside the same multiprotein complex (i.e. indirect). We sought to predict which PPI may be directly interacting with its target protein by using the artificial intelligence algorithm AlphaFold3(AF3) (Abramson, Adler et al. 2024). First, we analyzed the predicted AF3 structure of the targets using the pTM score and the fraction of each structure calculated to be disordered (Figure 6A and Supplementary Table7). Our reasoning was that if targets have a poorly resolved structures, it will be difficult to screen them for direct PPI. A pTM score >0.5 suggests that the structure may be correct (the highest confidence score is 1). Undefined or disordered regions hinder the accuracy of the prediction. All targets possessed a pTM score > 0.5 except Syt1. The disordered fraction negatively correlated with the pTM score, as expected. Gsk3b, Ppp1ca, and Map2k1 had the highest pTM scores and were also the smallest of our target proteins (Figure 6B). Ppp1ca had the most confident structure (i.e. pTM 0.9) and the smallest disordered fraction (i.e. 0.07). Next, we determined the AF3 prediction of previously reported direct interactions of the targets. We used the iPTM score to determine interaction confidence. An iPTM score >0.8 is considered a highly confident direct interaction, whereas 0.8. These eight PPI have all previously been reported to form a direct interaction with Ppp1ca, except Phactr3 (Zhang, Zhang et al. 1998, Terrak, Kerff et al. 2004, Hurley, Yang et al. 2007, Marsh, Dancheck et al. 2010, Ragusa, Dancheck et al. 2010, Ferrar, Chamousset et al. 2012, Choy, Srivastava et al. 2024, Xu, Sadleir et al. 2024)*. Phactr3 is structurally similar to, but less studied than, the reported direct interactor Phactr1. These interactors are all inhibitors of PP1 except Ppp1r9b which targets Ppp1ca to specific subcellular compartments. Nine PPI were assigned a score The following has been added to the Discussion:

“Our SCZ PPI network consists of two types of PPI: direct physical interactions and “co-complex” or indirect interactions. Typically, the nature of the interaction can be distinguished in IP-MS studies. We decided to employ the new AF3 algorithm to screen the PPI of Ppp1ca to provide evidence for direct interactors. We chose to examine the PPI assigned to Ppp1ca, because its structure was the most confident among our target proteins and AF3 correctly predicted a known direct interactor with high confidence. Ppp1ca is a catalytic subunit of the phosphatase PP1, which is required to associate with regulatory subunits to create holoenzymes (Li, Wilmanns et al. 2013). Eighteen PPI were predicted to be directly interacting with Ppp1ca using a 0.6 or higher iPTM filter. This filter may be too conservative and generate false negatives, because another study employed a 0.3 filter followed by additional interrogation to screen for direct PPI (Weeratunga, Gormal et al. 2024). Forty-four percent of these predictions were confirmed by previous publications. Most of the validated direct interactions are inhibitors of the phosphatase, but one, Ppp1r9b (aka spinophilin), is known to target Ppp1ca to dendrite spines to enhance its activity to specific substrates (Allen, Ouimet et al. 1997, Salek, Claeboe et al. 2023). This high correlation with the literature provides substantial confidence in the novel PPI predicted to be direct Ppp1ca interactors. The AF3 screen predicted that NDRG2 directly interacts with Ppp1ca. This protein is known to regulate many phosphorylation dependent signaling pathways by directly interacting with other phosphatases including Pp1ma and PP2A (Feng, Zhou et al. 2022, Lee, Lim et al. 2022). Actin binding protein Capza1 was also predicted to directly interact with Ppp1ca and Ppp1ca interacts with actin and its binding proteins to maintain optimal localization for efficient activity to specific substrates (Foley, Ward et al. 2023). Hsp1e is a heat shock protein predicted to directly interact with Ppp1ca. Although there is no direct connection to Ppp1ca, other heat shock proteins have been reported to regulate Ppp1ca (Mivechi, Trainor et al. 1993, Flores-Delgado, Liu et al. 2007, Qian, Vafiadaki et al. 2011). We also observed that many of these direct PPI were altered with PCP treatment. One direct interactor, Ppp1r1b (aka DARPP-32), is phosphorylated at Thr34 by PKA in the brain upon PCP treatment. This phosphorylation event converts Ppp1rb to a potent inhibitor of Ppp1ca(Svenningsson, Tzavara et al. 2003). Importantly, manipulation of Thr34 attenuated the behavioral effects of PCP. Consistent with this report, Ppp1r1b-Ppp1ca interaction was only observed with PCP in our study. Further investigation is needed to determine if our novel direct interactors regulate the PCP phenotype. We conclude that AF3 can provide important structural insights into the nature of PPI obtained from large scale IP-MS studies.”

2) Considering the protein extraction protocol, it is fair to mention that only the most soluble proteins are being considered here. I am bringing this up since the importance of membrane receptors is clear in the studied context. This is an interesting point. It has been predicted that transmembrane proteins constitute 25-30% of the proteome(Dobson, Remenyi et al. 2015). Thus, we would predict our dataset will have more soluble proteins than membrane proteins. Half of our target proteins were transmembrane proteins, so in designing the protocol for this study we ensured that these membrane proteins could be significantly enriched compared to the control IPs (Supplementary Figure 2C). In addition, compared to soluble proteins, membrane proteins are notoriously difficult to identify by bottom-up proteomics (Savas, Stein et al. 2011). We decided to investigate how many of our protein interactors were transmembrane proteins. Using Uniprot, 199 (20%) of our protein interactors were determined to have a transmembrane domain. Therefore, this data does not support the statement that only the most soluble proteins are being considered in our study. We added this percentage of transmembrane proteins in our network to the text of the Results section.

3) It is not clear from the methods description if antibodies from all 8 targets were all together in one Co-IP or have been incubated separately in 8 different hippocampi samples. It seems the first, given how results have been presented. If so, this maximizes the major issue raised above (in 1). We apologize for not clearly describing our experimental design. All the targets were immunoprecipitated separately and analyzed separately on the mass spectrometer. With all the biological replicates and two conditions (i.e. Saline and PCP), we performed 48 individual, separate IPs. There were an additional 48 individual, separate IPs run in parallel that were the control IPs.

We modified the schematic of our experimental design in Figure 1C to clarify that the 8 targets IPs were analyzed separately. In addition, we modified the Results to read:

“In total, 96 (48 bait and 48 control) IPs were performed, and each was analyzed separately by LC-MS analysis.”

4) Definitely, results here are not representing a "SCZ PPI network". PCP-treated animals, as any other animal model, are rather limited models to schizophrenia. As a complex multifactorial disease, synaptic deficits, which is the focus of this study, can no longer be considered "the pivot" of the disease. Synaptic dysfunction is only one among many other factors associated to schizophrenia.

We do agree that synaptic dysfunction is only one factor associated with SCZ and we will discuss this more in our response to your next comment.

We understand the limitations of PCP as an animal model of SCZ. It is quite difficult to model a specific human complex multifactorial neurological disease in rodents and we would contend that there is no single universal SCZ model that everyone agrees with. We addressed this by adding the following to the Introduction:

“Since many SCZ symptoms are uniquely human, this is no single animal model that truly replicates all the complex human SCZ phenotypes(Winship, Dursun et al. 2019). In this respect, all SCZ animal models can be considered limited.* “ *

We respectfully disagree, however, with the term SCZ PPI network. This study is focused on SCZ by choosing proteins implicated in SCZ, quantitating how the PPI changes in a SCZ model, and discussing how our findings are relevant to SCZ pathogenesis. So, it seems logical to call our dataset a SCZ PPI network. We do concede that without further experimentation we do not know if these PPI play a causal role in SCZ. Furthermore, our novel PPI may involve biological pathways unrelated to SCZ and that have relevance to other biological conditions.

We added the following statement to the Discussion to address this comment:

“Even though our network was constructed in the context of SCZ, our dataset has relevance to other neurological diseases where our targets have been implicated in the pathogenesis.”

5) Authors should look for protein interactions that might be happening also in glial cells. They are not the majority in hippocampus, but are present in the type of tissue analyzed here. Thus, some of the interactions observed might be more abundantly present in those cells. Maybe enriching using bioinformatics tools the PPI network to different cell types.

As mentioned above, we agree that synaptic dysfunction is just one of the hypotheses of SCZ pathogenesis and emerging evidence suggests that dysfunction in astrocytes and microglia are factors. Since these non-neuronal cells can regulate synapses, these hypotheses are not mutually exclusively and suggests that at the cellular level SCZ etiology involves multiple cell types.

We addressed your query by comparing our PPI network to an RNA-seq analysis of different cell types in the rodent brain(Zhang, Chen et al. 2014). First, we analyzed our target proteins, and found that they were expressed in all cell types to varying degrees except Syngap which was not in the RNA-seq database. This data is now represented in Figure 3E. We then determined the RNA abundance distribution of all the protein interactors, which is represented in Figure 3D as a heatmap. From a bird’s eye view, it suggests that some PPI exist in non-neuronal cells. Next, we determine how many of our protein interactors were enriched in one cell type, which is shown in Figure 3F. We defined an enriched protein as having >50% of the RNA signal in one cell type. We identified 175 proteins that were enriched in one cell type compared to the entire RNA-seq dataset which had 4008 enriched proteins. In the entire RNA-seq dataset, 24% of the enriched proteins were in neurons whereas 47% of our protein interactors were enriched in neurons. This is consistent with the enrichment of synaptic proteins in our network. There was also an increased percentage of astrocytes (19%) and oligodendrocytes (6%) in our network compared to the entire database (i.e. astrocytes-11% and oligodendrocytes-4%). In other cell types, such as microglia, there was less protein enrichment in our network compared to the database. We have amended this cell type analysis to our manuscript and concluded that a portion of our PPI network may occur in non-neuronal cells. We also created a supplementary table of our network with its associated RNA-seq data.

Text added to the Results:

“Non-synaptic proteins represented 59% of our network suggesting that some PPI may occur in non-neuronal cells. To investigate this possibility, we annotated our network with a transcriptome rodent brain database of eight cell types(Zhang, Chen et al. 2014). All the targets were detected in all cell types but there was obvious enrichment in specific cell types for some targets (Figure 3E). Syngap1 was not in the database. We also observed a large variation of cellular distributions for the interactors (Figure 3D). Next, we sought to determine how many interactors are enriched in a particular cell type by defining cell enrichment as a protein having >50% RNA signal in one cell type. We identified 175 protein interactors enriched in one cell type, whereas the entire database had 4008 proteins enriched (Figure 3F). Consistent with our synaptic enrichment, 47% of the enriched protein interactors were in neurons whereas only 24% of the enriched protein in the entire database were in neurons. We also observed an increase in protein interactors enriched in astrocytes compared to the database. Overall, this analysis provides evidence that our identified PPI may occur in non-neuronal cells.”

Text added to the Discussion:

“The exact etiology of SCZ, however, remains unclear and synaptic dysfunction is only one hypothesis (Misir and Akay 2023). There is evidence for the involvement of non-neuronal cell types, including endothelial cells, astrocytes, and microglia(Tarasov, Svistunov et al. 2019, Rodrigues-Neves, Ambrosio et al. 2022, Stanca, Rossetti et al. 2024). Although we observed an enrichment of synaptic proteins in our SCZ network, we provided evidence that a portion of our network may occur in non-neuronal cells. Since non-neuronal cells can regulate synapses(Vilalta and Brown 2018, Bauminger and Gaisler-Salomon 2022), synaptic dysfunction and perturbations in non-neuron cells in SCZ etiology are not mutually exclusive. Our data corresponds with emerging evidence that pathogenesis is multifaceted, involving dysfunction in multiple cell types. “

Minor: 1) in the abstract, it is not clear if 90% of the PPI are novel to brain tissue in general or specifically schizophrenia. We apologize for the confusing sentence. 90% are novel meaning the PPI have not been reported in any study. We changed the abstract to read:

“Over 90% of the PPI have not been previously reported.”

2) authors refer to LC-MS-based proteomics as "MS" all across the text. Who am I to say this to Yates et al, but I think it is rather simplified use "Mass Spectrometry Analysis", when this is a typical LC-MS type of analysis We agree with you. We have replaced MS analysis with LC-MS analysis in the manuscript.

3) Several references used to construct the hypothesis of the paper are rather outdated: several from 10-15 years ago. It would be interesting to provide to the reader up to date references, given the rapid pace science has been progressing. We agree many of the references are 10-15 years old. Many of the hypotheses and biological mechanisms we discussed can be supported by too many studies to cite them all, due to space. If we could, we would. We also agree that there are many more recent studies that have confirmed and added more details to the original discovery or hypothesis cited. We cite the first study to support our conclusions because it deserves the most credit.

4) "UniProt rat database". Please, state the version and if reviewed or unreviewed.

This information was added to the Methods section. UniProt reviewed rat database with isoforms 03-25-2014.

Reviewer #2 (Significance (Required)):

The study is informative, and has great potential to enrich the specific literature of this field. But should tone down some arguments, given the experimental limitations of the PPI network (as described above) and should state PCP-treated rats as a limited model to schizophrenia.

Reviewer #3 (Evidence, reproducibility and clarity (Required)):

Summary

It is now widely accepted that schizophrenia is polygenic disorder in which a large fraction of the genetic risk is in variants affecting the expression of synaptic proteins. Moreover, it is known that these synaptic proteins are found in multiprotein complexes and that many proteins encoded by schizophrenia risk genes interact directly or indirectly in these complexes. It is also known that some drugs including phencyclidine, which binds to NMDA receptors and to Dopamine D2 receptors (not mentioned by the authors) can induce schizophreniform psychosis. The authors have set out to advance on this position by performing proteomic mass spectrometry studies on proteins identified as encoded by schizophrenia risk genes. They target 8 proteins for immunoprecipitation from rat brain and identify coisolated proteins and perform various network analyses. In the most interesting part of the paper they ask if PCP-treatment altered protein interactions and report various changes.

Major comments:

1. Choice of target proteins. It was not until the first paragraph of the results section that the authors first name the 8 synaptic proteins that have chosen to study. This information should be in the abstract.

This information was added to the abstract as requested.

The authors then use figure 1A and 1B as evidence that these 8 "baits" are schizophrenia-relevant proteins. Figure 1A does not provide any evidence at all and Figure 1B is about as weak a line of evidence imaginable - a histogram of the number of papers that have the search term "schizophrenia" and the protein name. I tried this search for Grin2B and almost immediately found papers that reported no association between Grin2B and schizophrenia (e.g. PMID: 33237434). Figure 1B should be scrapped.

The purpose of Figure 1A was not to demonstrate that there is evidence that our proteins are involved in SCZ. The purpose of this figure is to show that these proteins are diverse in function and structure (blue = membrane proteins; yellow = soluble proteins), and that there are published studies reporting physical and functional interactions between these 8 proteins. This suggests that a more extensive network may exist.

We agree that Figure 1B does not specifically describe how each protein is related to SCZ but demonstrates how many papers investigating their connection to SCZ have been published. We understand how by itself, this can be considered weak. We still think it is important to show that multiple laboratories have published papers connecting these proteins to SCZ. Instead of scrapping this figure, we have moved it to the Supplementary Figure 2A.

We read PMID: 33237434 and interpret their findings quite differently than you. This report examined whether one single nucleotide mutation (SNV) in Grin2b is associated with the cognitive dysfunction in SCZ but did not examine if this mutation is associated with the other major SCZ phenotypes (i.e. psychotic and emotional). Specifically, the study selected 117 “patients in whom cognitive dysfunctions are present despite effective antipsychotic treatment of other schizophrenia symptoms.” The study concluded that Grin2B SNV was not associated with this subset of patients but concluded that they need to search for other NMDAR variants and study their association with SCZ. We would argue that the only reason this group performed these experiments was the well-known association between Grin2b and SCZ. Many studies have found SNVs in Grin2B that are associated with SCZ, but there are conflicting reports. It is unclear if the discrepancies are connected to different cohorts, complexity of SCZ phenotype, or small sample sizes. Regardless of Grin2B mutations significantly associated with SCZ, there are several lines of evidence that Grin2B is involved in SCZ. Most importantly, Grin2b is a component of the NMDAR, which is a key player to the SCZ hypo-glutamate hypothesis and the receptor that binds PCP. By immunoprecipitating Grin2b, we are analyzing the PPI network of NMDAR, which is arguably the most studied complex in SCZ research.

The remaining part of paragraph 1 of the results does not provide an adequate, let alone systematic, justification for the use of the 8 baits. It would be appropriate to construct a table with the 8 proteins and cite relevant papers and identify the basis for why they are implicated in schizophrenia (is it a direct mutation or some other evidence?). What makes these 8 proteins better than many others that are cited as synaptic schizophrenia relevant proteins?

We apologize for not clearly and thoroughly describing the reasons for choosing our baits. As stated in the first paragraph of the Results, we chose the proteins that had evidence of being a SCZ risk factor in SCZ databases that included a plethora of human genomic studies. This criterion by itself results in ~5000 genes. To further narrow our candidates, we chose targets that were synaptic and were observed to have phosphorylation changes in response to PCP in an SCZ animal model. Since protein-protein interactions (PPI) are often dependent on phosphorylation, we believe this is an important criterion for quantitation of PPI in response to PCP. These requirements still resulted in a list of hundreds of proteins. So, what makes these better than any other SCZ relevant protein? As stated in the manuscript, the major limiting criterion was identifying commercial antibodies that can efficiently immunoprecipitate their target in brain tissue. Since there are many reports associating our targets with SCZ, we directed the reader to SCZ databases that compile large genomic association studies. We understand, however, the request for more specific information regarding the biological connection between these proteins and SCZ. We took your suggestion and constructed a table with our 8 targets, and it is now Figure 1A. In this table, we selected references to indicate if the target has reported changes in expression and/or activity in SCZ samples (i.e. human and animal model) or genetic association with SCZ in human studies.

The methods of protein extraction are particularly concerning. The postsynaptic density of excitatory synapses (which contains several of the target proteins in this study) has been notoriously difficult to solubilise unless one uses high pH (9) and harsh detergent extraction (1% deoxycholate). The authors use pH 7 and weak detergent conditions, which are likely to be inefficient for solubilising at least several of the target proteins. Nowhere do the authors report how much of the total of their target protein is being solubilised. Indeed, there are no figures showing biochemical conditions at all. What if only a small percentage of the target protein is being immunoprecipitated - what does this mean for the interaction data? How do we know if the fraction being immunoprecipitated is from the synapse? (why did they not use synaptosomes).

How do we know if the fraction being immunoprecipitated is from the synapse? (why did they not use synaptosomes). The absence of this kind of data undermines the reader's confidence in the findings.

We apologize for not clearly explaining our experimental design We were not interested in identifying the PPI of the PSD. All these proteins have been localized to the synapse, but they are also localized to other neuronal compartments and non-neuronal cell types. Synaptic dysfunction is one hypothesis of SCZ pathogenesis, but there is evidence of other cell types, including astrocytes, microglia, and oligodendrocytes(Kerns, Vong et al. 2010, Ma, Abazyan et al. 2013, Goudriaan, de Leeuw et al. 2014, Park, Noh et al. 2020). For these reasons, we chose an unbiased approach to identifying PPI.

The Results have been amended to read: “All the targets are localized to the synapse, but also localized to non-synaptic compartments and expressed in non-neuronal cells. Thus, since there is also evidence for non-synaptic perturbations contributing to SCZ pathogenesis, we chose to perform an unbiased analysis in unfractionated brain tissue (Tarasov, Svistunov et al. 2019, Rodrigues-Neves, Ambrosio et al. 2022, Stanca, Rossetti et al. 2024). “

Why do we choose a specific solubilization strategy? Harsh detergents can disrupt PPI and prevent efficient enrichment of the target by disrupting the target-antibody interaction(Pankow, Bamberger et al. 2015). To identify protein interactions, mild detergent conditions are typically employed in PPI studies. We used a combination of “weak” detergents (i.e. 0.5% NP-40, 0.5% Triton, and 0.01% Deoxycholate) to help prevent non-specific PPI, but still allowing efficient enrichment of the target proteins. We do agree that with our conditions the targets were not completely solubilized. It is a balancing act to find the correct conditions for IP-MS analysis. Since we are unable to immunoprecipitate all the target protein, we did not identify all the PPI for each target, and we did not make this claim. Importantly, we did identify known interactions for all our targets. Our mild detergent protocol is similar to other PPI studies and our results validates results reported in previous studies. It is more important to significantly enrich the target protein over control than to achieve complete solubilization (Supplementary Figure 2D). This allows us to use control IPs to successfully employ the SAINT algorithm to determine which proteins are confident PPI using a 5% FDR.

How do we know protein are being immunoprecipitated from the synapse? As we show in Figures 2B and 3A, multiple proteins are annotated to the synapse with different databases, Gene ontology and SynGO. Well-known synaptic PPI were also observed, such as Grin2B-Dlg4(i.e. PSD-95), providing further evidence for proteins being immunoprecipitated for the synapses. Besides validating over a hundred published PPI interactions, we also identified many reciprocal interactions between the target datasets demonstrating the reproducibility of our protocol. Thus, we respectfully disagree with you and assert that our PPI network is very confident.

The immunoprecipitation protocol is unusual in that the homogenates were incubated overnight (twice), which is a very long period compared to most published protocols. This is a concern because spurious protein interactions could form during this long incubation.

There are many different immunoprecipitation protocols in the literature. The IP conditions depend upon the target protein and the antibody employed. Specifically, the abundance of the target and the affinity of the antibody to the target will dictate the IP conditions. We routinely perform overnight incubation for our IP-MS studies(Pankow, Bamberger et al. 2016, McClatchy, Yu et al. 2018). In our experience with brain tissue, this results in the highest enrichment of the target protein and the best reproducibility between biological replicates compared to IP protocols with shorter incubation times. Many other laboratories use overnight incubations(Lin and Lai 2017, Iqbal, Akins et al. 2018, Lagundzin, Krieger et al. 2022), so we do not consider our protocol unusual. We do find that IPs with tagged proteins in cell culture are more amenable to short incubation times. We have no evidence that overnight incubation causes spurious protein interactions nor could find any in the literature. Non-specific interactions are a concern with IP-MS experiments regardless of the incubation time. We took multiple steps to reduce the non-specific PPI from affecting our dataset. The first overnight incubation was incubating the brain lysate with agarose beads linked to IgGs to preclear the lysate from “sticky” non-specific interactors binding to IgGs and the beads. In addition, control IPs with IgG crosslinked to beads were incubated with brain lysate in parallel to each target IP. We computationally compared the non-specific control IPs with the target IPs using the SAINT algorithm to generate a confident list of PPI with a stringent 5% FDR. Therefore, our pipeline is specifically designed to prevent spurious PPI.

In the section "Biological interpretation of scz PPI network". Surprisingly the authors found that synaptic proteins that are exclusively postsynaptic (Grin2B, SynGAP) or exclusively presynaptic (Syt1) show very high percentages of their interacting proteins are from the synaptic compartments where the target protein is not expressed. The authors offer no explanation for this paradox. One explanation for this could be that spurious PPIs have formed in the protein extraction/immunoprecipitation protocol. These findings need validation by biochemical fractionation of synapses into pre and post synaptic fractions and immunohistochemistry to demonstrate the subsynaptic localisation of the proteins. Grin2b is traditionally described as exclusively post-synaptic, but there is evidence for other localizations, including presynaptic(Berretta and Jones 1996, Sjostrom, Turrigiano et al. 2003, Bouvier, Larsen et al. 2018) and expression in astrocytes(Serrano, Robitaille et al. 2008, Lee, Ting et al. 2010, Lalo, Koh et al. 2021, Kim, Choi et al. 2024). Syngap has been localized to non-synaptic sites and glia expression in addition to its heavily studied role at the post synapse(Moon, Sakagami et al. 2008, Araki, Zeng et al. 2015, Birtele, Del Dosso et al. 2023). Syt1 is commonly used as a presynaptic marker, but along with other proteins previously reported to be exclusively presynaptic (such as SNAP-25), it has been localized to the postsynapse (Selak, Paternain et al. 2009, Tomasoni, Repetto et al. 2013, Hussain, Egbenya et al. 2017, Madrigal, Portales et al. 2019, Sumi and Harada 2023). Similarly, SynGo database assigns both post-synaptic and pre-synaptic localizations to Grin2b as stated in the manuscript. Thus, our data is not paradoxical, but supports the emerging evidence against the canonical exclusivity of the pre- and post-synaptic compartments. Determining subsynaptic localization of a protein is a huge undertaking and requires expertise we do not possess. This is why we relied on synaptic databases and the literature for our interpretation of our data, as other publications have done.

We added the following to the Discussion to address this issue:

“Using the SynGo database, 418 proteins (i.e. 41% of our network) were identified as synaptic proteins consistent with the targets having a synaptic localization. Defining the synaptic proteome is inherently difficult because the synapse is an “open organelle”, and many synaptic proteins also have non-synaptic localizations and are expressed in non-neuronal cells. We further attempted to define our synaptic PPI by differentiating between pre- and post- synaptic compartments via SynGo. Half of our targets were annotated to both compartments and all targets had PPI that were annotated to both. This data supports the emerging evidence against the canonical localization exclusivity of the pre and post synapse(Bouvier, Larsen et al. 2018, Madrigal, Portales et al. 2019).”

My concerns about spurious interactions are raised again because the authors say that 92% of their interactions are novel (I note that they authors have not compared their interaction data of the NMDA receptor with published datasets from Dr Seth Grant's laboratory). BioGrid itself is good but not enough for comparison, maybe at this point it worth taking String, which accumulates several sources of PPIs, just select the direct PPIs.

Since the MS-IP experiments in our study have never been performed before, we are not surprised by the extent of novel data we produced. As described above, we took many steps to prevent spurious PPI from entering our final dataset, including the use of detergents, preclearing and stringent bioinformatic filtering. Our entire dataset is very large, so the 8% of PPI that we replicated from other studies represents 124 interactions. We believe this to be an impressive number which correlates to the confidence of our data. Providing more confidence, we identified many reciprocal PPI where shared protein interactors between target proteins were identified in both target protein datasets.

    The PPI described for our targets in BioGrid encompassed 713 publications.  Two of the BioGrid datasets that were compared to our Grin2b PPI data were from the laboratory of Seth Grant.  Arbuckle et al (2010) is a low-throughout paper that describes a Grin2b and DLG4 PPI (that we also identified) and Husi et al (__2000__) is a seminal paper using high-throughput LC-MS to identify PPI in the PSD of mouse brain.  There were many differences between Husi et al and our pipeline.  Husi et al employed the C-terminal Grin2b peptide to pull down interactors from the PSD fraction whereas we employed Grin2b antibody to enrich Grin2b and its interactors from unfractionated brain tissue.  Despite these differences, our studies found 8 proteins in common.

We took your suggestion and compared our data to String which includes direct PPI and functional PPI. Our input was the high confidence PPI identified by SAINT with 5% FDR as with the BioGrid comparison. The PPI network for each target protein had a more significant enrichment (p We think the problem you suggest with SynGO is more of an inherent problem with characterizing the synaptic proteome. The synaptic proteome is difficult to define since it is an “open organelle” with proteins transporting in and out. In addition, most synaptic proteins, such as mitochondrial and translational proteins, also have non-synaptic localizations. It is not possible to isolate a contaminant-free “pure” synaptic preparation by biochemical fractionation. Recently, SynGO was used in a meta-analysis of previously published PSD datasets(Kaizuka, Hirouchi et al. 2024). Kaizuka et al. found 123 proteins identified in 20 PSD datasets. SynGo annotated proteins with post-synaptic localization from this list. To a lesser extent they also identified presynaptic localizations, but it is unclear if the presynaptic proteins are novel localizations. Kaizuka et al. continued the investigation and identified a novel PSD protein, thus demonstrating that our knowledge of pre- and post- synaptic proteomes is incomplete.

Minor comments

1. A number of papers have reported protein interactions of native NMDA receptor complexes and their associated proteins isolated from rodent brain and are neither referenced in this paper. It would be relevant to compare these published datasets with the Grin2B IP datasets.

We employed BioGrid as a reference of reported PPI for each of our target proteins. For Grin2B, the PPI came from 142 different publications. For eight target proteins, we decided *BioGrid * was the best resource for determining the novelty of our PPI because it is routinely used for large-scale unbiased PPI analysis. To determine the novelty of our network, we compared our PPI network to 713 publications via BioGrid. We are unsure whether the papers you are referring to are included in the BioGrid database. To make it easier for readers with similar queries, we added an additional supplementary table (TableS4) including all the publications (i.e. PMID numbers) included in BioGrid comparison for each target protein.

We amended the Results with the following sentence, so the readers realized the extensiveness of the Biogrid comparison analysis:

“There were 713 publications in BioGrid that describe at least one interaction with one of our targets (Supplementary Table4).”

The use of the term "bait" in purification experiments typically refers to a protein and not an antibody. I suggest removing the word bait to avoid ambiguity and simply use the word target. We took your suggestion and used “target” instead of “bait” to avoid ambiguity.

26 mins of treatment gives completely different set of PPIs between PCP and saline which is very interesting, so both networks should be included in Supplementary. Also, it would be useful to have a list of modulated (phosphorylated in their case, but also ubiquitinated etc) proteins, which is not presented. Table S1 lists the PPI for each target, and we designated whether the interactors were for Sal, PCP, or both. Phosphorylated and ubiquitinated proteins are very hard to reproducibly identify without an additional enrichment step. Since we did not perform this enrichment step, we did not search for these modifications and do not have any modified proteins to report.

As they say their final network is composed of "direct physical and "co-complex" interactors and they cannot distinguish between them. This is particularly bad for the postsynapse, where all the PSD components can be co-IP-ed in different combinations. It can explain the Figure 5C, where most of the proteins have FDR = 1, which means they do not reproduce. Figure 5C represents the intersection of 15N quantification and SAINT analysis. The x-axis is the FDR reported for SAINT analysis, and the y-axis is the significant proteins from the N15 analysis. This figure demonstrates that some proteins that were significantly different with PCP via N15 quantification also were annotated as PPI by SAINT (i.e. 5%. As stated in the Discussion, we concluded that the SAINT analysis and N15 quantitation are complementary in identifying PPI and that the quantification of a biological perturbation may aid the identification of PPI. Figure 5C is not related to whether our PPI are direct physical or "co-complex" interactors. Distinguishing between direct physical and co-complex interactors is an inherent problem for all IP studies. Since another reviewer also highlighted this deficit in our manuscript, we decided to analyze our PPI dataset with the artificial intelligence algorithm AlphaFold 3(AF3). The AF3 data is encompassed in Figure 6.

The following AF3 data was added to the Results Section:

“A disadvantage of IP-MS studies is that it cannot distinguish between a PPI that binds directly to the target protein, and a PPI in which the interactor and target protein reside in the same multiprotein complex (i.e. indirect). We sought to predict which PPI may be directly interacting with its target protein by using the artificial intelligence algorithm AlphaFold3(AF3) (Abramson, Adler et al. 2024). First, we analyzed the predicted AF3 structure of the targets using the pTM score, and determined the fraction of each structure that was calculated to be disordered (Figure 6A and Supplementary Table7). Our reasoning was that if our targets have a poorly resolved structures then it will be difficult to screen for direct PPI. A pTM score >0.5 suggests that the structure may be correct, with the highest confidence equaling 1. Undefined or disordered regions hinder the accuracy of the prediction, and all our targets possessed a pTM score > 0.5 except Syt1. The fraction of disordered negatively correlated with the pTM score, as expected. Gsk3b, Ppp1ca, and Map2k1 were the target proteins with the highest pTM scores and were also the smallest of our targets (Figure 6B). Ppp1ca had the most confident structure (i.e. pTM 0.9) and the least fraction disordered (i.e. 0.07). Next, we determined the AF3 prediction of previously reported direct interactions of the targets. We used the iPTM score to determine an interaction confidence. An iPTM score >0.8 is a highly confident direct interaction, whereas 0.8. These eight PPI have all previously been reported to form a direct interaction with Ppp1ca, except Phactr3 (Zhang, Zhang et al. 1998, Terrak, Kerff et al. 2004, Hurley, Yang et al. 2007, Marsh, Dancheck et al. 2010, Ragusa, Dancheck et al. 2010, Ferrar, Chamousset et al. 2012, Choy, Srivastava et al. 2024, Xu, Sadleir et al. 2024)*. Phactr3 is structurally similar to, but less studied than, the reported direct interactor, Phactr1. These interactors are all inhibitors of PP1 except for Ppp1r9b which targets Ppp1ca to specific subcellular compartments. Nine PPI were assigned a score The following AF3 interpretation was added to the Discussion:

“Our SCZ PPI network consists of two types of PPI: direct physical interactions and “co-complex” or indirect interactions. Typically, the nature of the interaction cannot be distinguished in IP-MS studies. We decided to employ the new AF3 algorithm to screen the PPI of Ppp1ca to provide evidence for direct interactors. We chose to examine the PPI assigned to Ppp1ca, because its structure was the most confident among our target proteins and AF3 correctly predicted a known direct interactor with high confidence. Ppp1ca is a catalytic subunit of the phosphatase PP1, which is required to associate with regulatory subunits to create holoenzymes (Li, Wilmanns et al. 2013). Eighteen PPI were predicted to be directly interacting with Ppp1ca using a 0.6 or higher iPTM filter. This filter may be too conservative and may generate false negatives, because another study employed a 0.3 filter followed by additional interrogation to screen for direct PPI (Weeratunga, Gormal et al. 2024). Forty-four percent of these predictions were confirmed by previous publications. Most of these validated direct interactions are inhibitors of the phosphatase, but one, Ppp1r9b (aka spinophilin), is known to target Ppp1ca to dendritic spines (Allen, Ouimet et al. 1997, Salek, Claeboe et al. 2023). This high correlation with the literature provides substantial confidence to the novel PPI predicted to be direct Ppp1ca interactors. The AF3 screen predicted that NDRG2 directly interacts with Ppp1ca. This protein is known to regulate many phosphorylation dependent signaling pathways by directly interacting with other phosphatases including Pp1ma and PP2A (Feng, Zhou et al. 2022, Lee, Lim et al. 2022). Actin binding protein Capza1 was also predicted to directly interact with Ppp1ca and Ppp1ca interacts with actin and its binding proteins to maintain optimal localization for efficient activity to specific substrates (Foley, Ward et al. 2023). Hsp1e is a heat shock protein predicted to directly interact with Ppp1ca. Although there is no direct connection to Ppp1ca, other heat shock proteins have been reported to regulate Ppp1ca (Mivechi, Trainor et al. 1993, Flores-Delgado, Liu et al. 2007, Qian, Vafiadaki et al. 2011). We also observed that many of the direct PPI were altered with PCP treatment. One direct interactor, Ppp1r1b (aka DARPP-32), is phosphorylated at Thr34 by PKA in the brain upon PCP treatment. This phosphorylation event converts Ppp1rb to a potent inhibitor of Ppp1ca(Svenningsson, Tzavara et al. 2003). Importantly, the manipulation of Thr34 attenuated the behavioral effects of PCP. Consistent with this report, Ppp1r1b-Ppp1ca interaction was only observed with PCP in our study. Further investigation is needed to determine if our novel direct interactors regulate the PCP phenotype. We conclude that AF3 can provide important structural insights into the nature of PPI obtained from large scale IP-MS studies.”

The way PPI data is reported can be improved so that I does not have to be extracted from Table 1 and 2. It would be good if they provide just two columns PPI list, with names or IDs, plus PSP/saline/both conditions in third column, for ease of comparison with other sources and building the graph. They can add it as another spreadsheet to Table 2. We generated this table (TableS2) as you requested.

Is Figure 2 built for Sal or PCP conditions? as they have only 23% interactions in common (Figure 4A) the Figure 2 should be pretty different for two conditions. Are the 1007 interactors combined from SAL and PCP?

Figure 2 contains ALL the unique PPI for each target regardless of Sal or PCP conditions. The 1007 protein interactors shown in Figure 2Awhere Sal and PCP were combined to generate a non-redundant list of proteins for each target.

We amended the Results to make this clearer:

“When the PCP and SAL datasets were combined, there were 1007 unique proteins.”

This sentence was added to Figure 2A:

“For this comparison, Sal and PCP PPI were combined into a unique PPI list for each target.”

Figure 1F is mentioned but no figure is shown. We apologize for this oversight, and we have corrected the manuscript. 8. Overall the paper could be edited and made more concise, especially the introduction and discussion. We extensively edited the manuscript to be more concise.

Reviewer #3 (Significance (Required)):

General assessment

Proteomic mass spectrometry of immunoprecipitated complexes from synapses has been extensively studied since Husi et al (2000) first study of NMDA receptor and AMPA receptor complexes. Since then, a wide variety of methods have been employed to purify synaptic protein complexes including peptide affinity, tandem-affinity purification of endogenous proteins tagged with FLAG and Histine-affinity tags amongst other methods. Purification of protein complexes and the postsynaptic density from the postsynaptic terminal of mammalian excitatory synapses have been crucial for establishing that schizophrenia is a polygenic disorder affecting synapses (e.g. Fernandez et al, 2009; Kirov et al, 2012; Purcell et al, 2014, Fromer et al, 2014 etc). Network analyses of the postsynaptic proteome have described networks of schizophrenia interacting proteins (e.g. Pocklington et al, 2006; Fernandez et al, 2009) and other neuropsychiatric disorders.

Hundreds of synaptic protein complexes have been identified (Frank et al, 2016), but very few have been characterised using proteomic mass spectrometry. This paper has chosen 8 protein targets for such analysis and identified many proteins that a putative interactors of the target protein. At this level the current manuscript does not represent a conceptual advance and the value of the data lies in its utility as a resource that may be used in future studies.

The findings from the 8 target proteins from normal adult rat brain were used for a secondary study that describes the effects that PCP has on the interaction networks. Interestingly, this work shows that 26 minutes of drug treatment leads to considerable changes in the interactomes of the target proteins. These descriptive data could be used in future studies to understand the cell biological mechanisms that mediate these rapid changes in the proteome. PCP and drugs that interact with NMDA receptors are known to induce changes in synaptic proteome phosphorylation including modifications in protein-protein interaction sites, which may explain the PCP effects.

The study would benefit from validation of experimental protocols for solubilisation and immunoprecipitation and validation of described interactions using orthogonal biochemical or localisation experiments.

Audience Specialists in synapse proteins and mechanisms of schizophrenia.

Expertise

The reviewers' expertise is in molecular biology of synapses including synapse proteomics, protein interaction and network analysis, and genetics of schizophrenia and other brain disorders.

Abramson, J., J. Adler, J. Dunger, R. Evans, T. Green, A. Pritzel, O. Ronneberger, L. Willmore, A. J. Ballard, J. Bambrick, S. W. Bodenstein, D. A. Evans, C. C. Hung, M. O'Neill, D. Reiman, K. Tunyasuvunakool, Z. Wu, A. Zemgulyte, E. Arvaniti, C. Beattie, O. Bertolli, A. Bridgland, A. Cherepanov, M. Congreve, A. I. Cowen-Rivers, A. Cowie, M. Figurnov, F. B. Fuchs, H. Gladman, R. Jain, Y. A. Khan, C. M. R. Low, K. Perlin, A. Potapenko, P. Savy, S. Singh, A. Stecula, A. Thillaisundaram, C. Tong, S. Yakneen, E. D. Zhong, M. Zielinski, A. Zidek, V. Bapst, P. Kohli, M. Jaderberg, D. Hassabis and J. M. Jumper (2024). "Accurate structure prediction of biomolecular interactions with AlphaFold 3." Nature 630(8016): 493-500.

Allen, P. B., C. C. Ouimet and P. Greengard (1997). "Spinophilin, a novel protein phosphatase 1 binding protein localized to dendritic spines." Proc Natl Acad Sci U S A 94(18): 9956-9961.

Anschuetz, A., K. Schwab, C. R. Harrington, C. M. Wischik and G. Riedel (2024). "A Meta-Analysis on Presynaptic Changes in Alzheimer's Disease." J Alzheimers Dis 97(1): 145-162.

Araki, Y., M. Zeng, M. Zhang and R. L. Huganir (2015). "Rapid dispersion of SynGAP from synaptic spines triggers AMPA receptor insertion and spine enlargement during LTP." Neuron 85(1): 173-189.

Bauminger, H. and I. Gaisler-Salomon (2022). "Beyond NMDA Receptors: Homeostasis at the Glutamate Tripartite Synapse and Its Contributions to Cognitive Dysfunction in Schizophrenia." Int J Mol Sci 23(15).

Berretta, N. and R. S. Jones (1996). "Tonic facilitation of glutamate release by presynaptic N-methyl-D-aspartate autoreceptors in the entorhinal cortex." Neuroscience 75(2): 339-344.

Birtele, M., A. Del Dosso, T. Xu, T. Nguyen, B. Wilkinson, N. Hosseini, S. Nguyen, J. P. Urenda, G. Knight, C. Rojas, I. Flores, A. Atamian, R. Moore, R. Sharma, P. Pirrotte, R. S. Ashton, E. J. Huang, G. Rumbaugh, M. P. Coba and G. Quadrato (2023). "Non-synaptic function of the autism spectrum disorder-associated gene SYNGAP1 in cortical neurogenesis." Nat Neurosci 26(12): 2090-2103.

Bouvier, G., R. S. Larsen, A. Rodriguez-Moreno, O. Paulsen and P. J. Sjostrom (2018). "Towards resolving the presynaptic NMDA receptor debate." Curr Opin Neurobiol 51: 1-7.

Choy, M. S., G. Srivastava, L. C. Robinson, K. Tatchell, R. Page and W. Peti (2024). "The SDS22:PP1:I3 complex: SDS22 binding to PP1 loosens the active site metal to prime metal exchange." J Biol Chem 300(1): 105515.

Dobson, L., I. Remenyi and G. E. Tusnady (2015). "The human transmembrane proteome." Biol Direct 10: 31.

Feng, D., J. Zhou, H. Liu, X. Wu, F. Li, J. Zhao, Y. Zhang, L. Wang, M. Chao, Q. Wang, H. Qin, S. Ge, Q. Liu, J. Zhang and Y. Qu (2022). "Astrocytic NDRG2-PPM1A interaction exacerbates blood-brain barrier disruption after subarachnoid hemorrhage." Sci Adv 8(39): eabq2423.

Ferrar, T., D. Chamousset, V. De Wever, M. Nimick, J. Andersen, L. Trinkle-Mulcahy and G. B. Moorhead (2012). "Taperin (c9orf75), a mutated gene in nonsyndromic deafness, encodes a vertebrate specific, nuclear localized protein phosphatase one alpha (PP1alpha) docking protein." Biol Open 1(2): 128-139.

Flores-Delgado, G., C. W. Liu, R. Sposto and N. Berndt (2007). "A limited screen for protein interactions reveals new roles for protein phosphatase 1 in cell cycle control and apoptosis." J Proteome Res 6(3): 1165-1175.

Foley, K., N. Ward, H. Hou, A. Mayer, C. McKee and H. Xia (2023). "Regulation of PP1 interaction with I-2, neurabin, and F-actin." Mol Cell Neurosci 124: 103796.

Goudriaan, A., C. de Leeuw, S. Ripke, C. M. Hultman, P. Sklar, P. F. Sullivan, A. B. Smit, D. Posthuma and M. H. Verheijen (2014). "Specific glial functions contribute to schizophrenia susceptibility." Schizophr Bull 40(4): 925-935.

Hemmings, H. C., Jr., P. Greengard, H. Y. Tung and P. Cohen (1984). "DARPP-32, a dopamine-regulated neuronal phosphoprotein, is a potent inhibitor of protein phosphatase-1." Nature 310(5977): 503-505.

Hurley, T. D., J. Yang, L. Zhang, K. D. Goodwin, Q. Zou, M. Cortese, A. K. Dunker and A. A. DePaoli-Roach (2007). "Structural basis for regulation of protein phosphatase 1 by inhibitor-2." J Biol Chem 282(39): 28874-28883.

Hussain, S., D. L. Egbenya, Y. C. Lai, Z. J. Dosa, J. B. Sorensen, A. E. Anderson and S. Davanger (2017). "The calcium sensor synaptotagmin 1 is expressed and regulated in hippocampal postsynaptic spines." Hippocampus 27(11): 1168-1177.

Iqbal, H., D. R. Akins and M. R. Kenedy (2018). "Co-immunoprecipitation for Identifying Protein-Protein Interactions in Borrelia burgdorferi." Methods Mol Biol 1690: 47-55.

Kaizuka, T., T. Hirouchi, T. Saneyoshi, T. Shirafuji, M. O. Collins, S. G. N. Grant, Y. Hayashi and T. Takumi (2024). "FAM81A is a postsynaptic protein that regulates the condensation of postsynaptic proteins via liquid-liquid phase separation." PLoS Biol 22(3): e3002006.

Kaizuka, T., T. Suzuki, N. Kishi, K. Tamada, M. W. Kilimann, T. Ueyama, M. Watanabe, T. Shimogori, H. Okano, N. Dohmae and T. Takumi (2024). "Remodeling of the postsynaptic proteome in male mice and marmosets during synapse development." Nat Commun 15(1): 2496.

Kerns, D., G. S. Vong, K. Barley, S. Dracheva, P. Katsel, P. Casaccia, V. Haroutunian and W. Byne (2010). "Gene expression abnormalities and oligodendrocyte deficits in the internal capsule in schizophrenia." Schizophr Res 120(1-3): 150-158.

Kim, H., S. Choi, E. Lee, W. Koh and C. J. Lee (2024). "Tonic NMDAR Currents in the Brain: Regulation and Cognitive Functions." Biol Psychiatry.

Koopmans, F., P. van Nierop, M. Andres-Alonso, A. Byrnes, T. Cijsouw, M. P. Coba, L. N. Cornelisse, R. J. Farrell, H. L. Goldschmidt, D. P. Howrigan, N. K. Hussain, C. Imig, A. P. H. de Jong, H. Jung, M. Kohansalnodehi, B. Kramarz, N. Lipstein, R. C. Lovering, H. MacGillavry, V. Mariano, H. Mi, M. Ninov, D. Osumi-Sutherland, R. Pielot, K. H. Smalla, H. Tang, K. Tashman, R. F. G. Toonen, C. Verpelli, R. Reig-Viader, K. Watanabe, J. van Weering, T. Achsel, G. Ashrafi, N. Asi, T. C. Brown, P. De Camilli, M. Feuermann, R. E. Foulger, P. Gaudet, A. Joglekar, A. Kanellopoulos, R. Malenka, R. A. Nicoll, C. Pulido, J. de Juan-Sanz, M. Sheng, T. C. Sudhof, H. U. Tilgner, C. Bagni, A. Bayes, T. Biederer, N. Brose, J. J. E. Chua, D. C. Dieterich, E. D. Gundelfinger, C. Hoogenraad, R. L. Huganir, R. Jahn, P. S. Kaeser, E. Kim, M. R. Kreutz, P. S. McPherson, B. M. Neale, V. O'Connor, D. Posthuma, T. A. Ryan, C. Sala, G. Feng, S. E. Hyman, P. D. Thomas, A. B. Smit and M. Verhage (2019). "SynGO: An Evidence-Based, Expert-Curated Knowledge Base for the Synapse." Neuron 103(2): 217-234 e214.

Krishnankutty, A., T. Kimura, T. Saito, K. Aoyagi, A. Asada, S. I. Takahashi, K. Ando, M. Ohara-Imaizumi, K. Ishiguro and S. I. Hisanaga (2017). "In vivo regulation of glycogen synthase kinase 3beta activity in neurons and brains." Sci Rep 7(1): 8602.

Lagundzin, D., K. L. Krieger, H. C. Law and N. T. Woods (2022). "An optimized co-immunoprecipitation protocol for the analysis of endogenous protein-protein interactions in cell lines using mass spectrometry." STAR Protoc 3(1): 101234.

Lalo, U., W. Koh, C. J. Lee and Y. Pankratov (2021). "The tripartite glutamatergic synapse." Neuropharmacology 199: 108758.

Lee, B. H., F. Schwager, P. Meraldi and M. Gotta (2018). "p37/UBXN2B regulates spindle orientation by limiting cortical NuMA recruitment via PP1/Repo-Man." J Cell Biol 217(2): 483-493.

Lee, K. W., S. Lim and K. D. Kim (2022). "The Function of N-Myc Downstream-Regulated Gene 2 (NDRG2) as a Negative Regulator in Tumor Cell Metastasis." Int J Mol Sci 23(16).

Lee, M. C., K. K. Ting, S. Adams, B. J. Brew, R. Chung and G. J. Guillemin (2010). "Characterisation of the expression of NMDA receptors in human astrocytes." PLoS One 5(11): e14123.

Li, X., M. Wilmanns, J. Thornton and M. Kohn (2013). "Elucidating human phosphatase-substrate networks." Sci Signal 6(275): rs10.

Lin, J. S. and E. M. Lai (2017). "Protein-Protein Interactions: Co-Immunoprecipitation." Methods Mol Biol 1615: 211-219.

Ma, T. M., S. Abazyan, B. Abazyan, J. Nomura, C. Yang, S. Seshadri, A. Sawa, S. H. Snyder and M. V. Pletnikov (2013). "Pathogenic disruption of DISC1-serine racemase binding elicits schizophrenia-like behavior via D-serine depletion." Mol Psychiatry 18(5): 557-567.

Madrigal, M. P., A. Portales, M. P. SanJuan and S. Jurado (2019). "Postsynaptic SNARE Proteins: Role in Synaptic Transmission and Plasticity." Neuroscience 420: 12-21.

Marsh, J. A., B. Dancheck, M. J. Ragusa, M. Allaire, J. D. Forman-Kay and W. Peti (2010). "Structural diversity in free and bound states of intrinsically disordered protein phosphatase 1 regulators." Structure 18(9): 1094-1103.

McClatchy, D. B., N. K. Yu, S. Martinez-Bartolome, R. Patel, A. R. Pelletier, M. Lavalle-Adam, S. B. Powell, M. Roberto and J. R. Yates (2018). "Structural Analysis of Hippocampal Kinase Signal Transduction." ACS Chem Neurosci 9(12): 3072-3085.

Misir, E. and G. G. Akay (2023). "Synaptic dysfunction in schizophrenia." Synapse 77(5): e22276.

Mivechi, N. F., L. D. Trainor and G. M. Hahn (1993). "Purified mammalian HSP-70 KDA activates phosphoprotein phosphatases in vitro." Biochem Biophys Res Commun 192(2): 954-963.

Moon, I. S., H. Sakagami, J. Nakayama and T. Suzuki (2008). "Differential distribution of synGAP alpha1 and synGAP beta isoforms in rat neurons." Brain Res 1241: 62-75.

Pankow, S., C. Bamberger, D. Calzolari, A. Bamberger and J. R. Yates, 3rd (2016). "Deep interactome profiling of membrane proteins by co-interacting protein identification technology." Nat Protoc 11(12): 2515-2528.

Pankow, S., C. Bamberger, D. Calzolari, S. Martinez-Bartolome, M. Lavallee-Adam, W. E. Balch and J. R. Yates, 3rd (2015). "∆F508 CFTR interactome remodelling promotes rescue of cystic fibrosis." Nature 528(7583): 510-516.

Park, G. H., H. Noh, Z. Shao, P. Ni, Y. Qin, D. Liu, C. P. Beaudreault, J. S. Park, C. P. Abani, J. M. Park, D. T. Le, S. Z. Gonzalez, Y. Guan, B. M. Cohen, D. L. McPhie, J. T. Coyle, T. A. Lanz, H. S. Xi, C. Yin, W. Huang, H. Y. Kim and S. Chung (2020). "Activated microglia cause metabolic disruptions in developmental cortical interneurons that persist in interneurons from individuals with schizophrenia." Nat Neurosci 23(11): 1352-1364.

Partiot, E., A. Hirschler, S. Colomb, W. Lutz, T. Claeys, F. Delalande, M. S. Deffieu, Y. Bare, J. R. E. Roels, B. Gorda, J. Bons, D. Callon, L. Andreoletti, M. Labrousse, F. M. J. Jacobs, V. Rigau, B. Charlot, L. Martens, C. Carapito, G. Ganesh and R. Gaudin (2024). "Brain exposure to SARS-CoV-2 virions perturbs synaptic homeostasis." Nat Microbiol.

Qian, J., E. Vafiadaki, S. M. Florea, V. P. Singh, W. Song, C. K. Lam, Y. Wang, Q. Yuan, T. J. Pritchard, W. Cai, K. Haghighi, P. Rodriguez, H. S. Wang, D. Sanoudou, G. C. Fan and E. G. Kranias (2011). "Small heat shock protein 20 interacts with protein phosphatase-1 and enhances sarcoplasmic reticulum calcium cycling." Circ Res 108(12): 1429-1438.

Ragusa, M. J., B. Dancheck, D. A. Critton, A. C. Nairn, R. Page and W. Peti (2010). "Spinophilin directs protein phosphatase 1 specificity by blocking substrate binding sites." Nat Struct Mol Biol 17(4): 459-464.

Rodrigues-Neves, A. C., A. F. Ambrosio and C. A. Gomes (2022). "Microglia sequelae: brain signature of innate immunity in schizophrenia." Transl Psychiatry 12(1): 493.

Salek, A. B., E. T. Claeboe, R. Bansal, N. F. Berbari and A. J. Baucum, 2nd (2023). "Spinophilin-dependent regulation of GluN2B-containing NMDAR-dependent calcium influx, GluN2B surface expression, and cleaved caspase expression." Synapse 77(3): e22264.

Savas, J. N., B. D. Stein, C. C. Wu and J. R. Yates, 3rd (2011). "Mass spectrometry accelerates membrane protein analysis." Trends Biochem Sci 36(7): 388-396.

Selak, S., A. V. Paternain, M. I. Aller, E. Pico, R. Rivera and J. Lerma (2009). "A role for SNAP25 in internalization of kainate receptors and synaptic plasticity." Neuron 63(3): 357-371.

Serrano, A., R. Robitaille and J. C. Lacaille (2008). "Differential NMDA-dependent activation of glial cells in mouse hippocampus." Glia 56(15): 1648-1663.

Sjostrom, P. J., G. G. Turrigiano and S. B. Nelson (2003). "Neocortical LTD via coincident activation of presynaptic NMDA and cannabinoid receptors." Neuron 39(4): 641-654.

Stanca, S., M. Rossetti, L. Bokulic Panichi and P. Bongioanni (2024). "The Cellular Dysfunction of the Brain-Blood Barrier from Endothelial Cells to Astrocytes: The Pathway towards Neurotransmitter Impairment in Schizophrenia." Int J Mol Sci 25(2).

Sumi, T. and K. Harada (2023). "Muscarinic acetylcholine receptor-dependent and NMDA receptor-dependent LTP and LTD share the common AMPAR trafficking pathway." iScience 26(3): 106133.

Svenningsson, P., E. T. Tzavara, R. Carruthers, I. Rachleff, S. Wattler, M. Nehls, D. L. McKinzie, A. A. Fienberg, G. G. Nomikos and P. Greengard (2003). "Diverse psychotomimetics act through a common signaling pathway." Science 302(5649): 1412-1415.

Tarasov, V. V., A. A. Svistunov, V. N. Chubarev, S. S. Sologova, P. Mukhortova, D. Levushkin, S. G. Somasundaram, C. E. Kirkland, S. O. Bachurin and G. Aliev (2019). "Alterations of Astrocytes in the Context of Schizophrenic Dementia." Front Pharmacol 10: 1612.

Terrak, M., F. Kerff, K. Langsetmo, T. Tao and R. Dominguez (2004). "Structural basis of protein phosphatase 1 regulation." Nature 429(6993): 780-784.

Tokizane, K., C. S. Brace and S. I. Imai (2024). "DMH(Ppp1r17) neurons regulate aging and lifespan in mice through hypothalamic-adipose inter-tissue communication." Cell Metab 36(2): 377-392 e311.

Tomasoni, R., D. Repetto, R. Morini, C. Elia, F. Gardoni, M. Di Luca, E. Turco, P. Defilippi and M. Matteoli (2013). "SNAP-25 regulates spine formation through postsynaptic binding to p140Cap." Nat Commun 4: 2136.

Vainio, L., S. Taponen, S. M. Kinnunen, E. Halmetoja, Z. Szabo, T. Alakoski, J. Ulvila, J. Junttila, P. Lakkisto, J. Magga and R. Kerkela (2021). "GSK3beta Serine 389 Phosphorylation Modulates Cardiomyocyte Hypertrophy and Ischemic Injury." Int J Mol Sci 22(24).

van Oostrum, M., T. M. Blok, S. L. Giandomenico, S. Tom Dieck, G. Tushev, N. Furst, J. D. Langer and E. M. Schuman (2023). "The proteomic landscape of synaptic diversity across brain regions and cell types." Cell 186(24): 5411-5427 e5423.

Vilalta, A. and G. C. Brown (2018). "Neurophagy, the phagocytosis of live neurons and synapses by glia, contributes to brain development and disease." FEBS J 285(19): 3566-3575.

Weeratunga, S., R. S. Gormal, M. Liu, D. Eldershaw, E. K. Livingstone, A. Malapaka, T. P. Wallis, A. T. Bademosi, A. Jiang, M. D. Healy, F. A. Meunier and B. M. Collins (2024). "Interrogation and validation of the interactome of neuronal Munc18-interacting Mint proteins with AlphaFold2." J Biol Chem 300(1): 105541.

Winship, I. R., S. M. Dursun, G. B. Baker, P. A. Balista, L. Kandratavicius, J. P. Maia-de-Oliveira, J. Hallak and J. G. Howland (2019). "An Overview of Animal Models Related to Schizophrenia." Can J Psychiatry 64(1): 5-17.

Xu, Z., L. Sadleir, H. Goel, X. Jiao, Y. Niu, Z. Zhou, G. de Valles-Ibanez, G. Poke, M. Hildebrand, N. Lieffering, J. Qin and Z. Yang (2024). "Genotype and phenotype correlation of PHACTR1-related neurological disorders." J Med Genet 61(6): 536-542.

Zhang, J., L. Zhang, S. Zhao and E. Y. Lee (1998). "Identification and characterization of the human HCG V gene product as a novel inhibitor of protein phosphatase-1." Biochemistry 37(47): 16728-16734.

Zhang, Y., K. Chen, S. A. Sloan, M. L. Bennett, A. R. Scholze, S. O'Keeffe, H. P. Phatnani, P. Guarnieri, C. Caneda, N. Ruderisch, S. Deng, S. A. Liddelow, C. Zhang, R. Daneman, T. Maniatis, B. A. Barres and J. Q. Wu (2014). "An RNA-sequencing transcriptome and splicing database of glia, neurons, and vascular cells of the cerebral cortex." J Neurosci 34(36): 11929-11947.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Summary

It is now widely accepted that schizophrenia is polygenic disorder in which a large fraction of the genetic risk is in variants affecting the expression of synaptic proteins. Moreover, it is known that these synaptic proteins are found in multiprotein complexes and that many proteins encoded by schizophrenia risk genes interact directly or indirectly in these complexes. It is also known that some drugs including phencyclidine, which binds to NMDA receptors and to Dopamine D2 receptors (not mentioned by the authors) can induce schizophreniform psychosis. The authors have set out to advance on this position by performing proteomic mass spectrometry studies on proteins identified as encoded by schizophrenia risk genes. They target 8 proteins for immunoprecipitation from rat brain and identify coisolated proteins and perform various network analyses. In the most interesting part of the paper they ask if PCP-treatment altered protein interactions and report various changes.

Major comments:

1. Choice of target proteins. It was not until the first paragraph of the results section that the authors first name the 8 synaptic proteins that have chosen to study. This information should be in the abstract. The authors then use figure 1A and 1B as evidence that these 8 "baits" are schizophrenia-relevant proteins. Figure 1A does not provide any evidence at all and Figure 1B is about as weak a line of evidence imaginable - a histogram of the number of papers that have the search term "schizophrenia" and the protein name. I tried this search for Grin2B and almost immediately found papers that reported no association between Grin2B and schizophrenia (e.g. PMID: 33237434). Figure 1B should be scrapped. The remaining part of paragraph 1 of the results does not provide an adequate, let alone systematic, justification for the use of the 8 baits. It would be appropriate to construct a table with the 8 proteins and cite relevant papers and identify the basis for why they are implicated in schizophrenia (is it a direct mutation or some other evidence?). What makes these 8 proteins better than many others that are cited as synaptic schizophrenia relevant proteins?
2. The methods of protein extraction are particularly concerning. The postsynaptic density of excitatory synapses (which contains several of the target proteins in this study) has been notoriously difficult to solubilise unless one uses high pH (9) and harsh detergent extraction (1% deoxycholate). The authors use pH 7 and weak detergent conditions, which are likely to be inefficient for solubilising at least several of the target proteins. Nowhere do the authors report how much of the total of their target protein is being solubilised. Indeed, there are no figures showing biochemical conditions at all. What if only a small percentage of the target protein is being immunoprecipitated - what does this mean for the interaction data? How do we know if the fraction being immunoprecipitated is from the synapse? (why did they not use synaptosomes). The absence of this kind of data undermines the reader's confidence in the findings.
3. The immunoprecipitation protocol is unusual in that the homogenates were incubated overnight (twice), which is a very long period compared to most published protocols. This is a concern because spurious protein interactions could form during this long incubation.
4. In the section "Biological interpretation of scz PPI network". Surprisingly the authors found that synaptic proteins that are exclusively postsynaptic (Grin2B, SynGAP) or exclusively presynaptic (Syt1) show very high percentages of their interacting proteins are from the synaptic compartments where the target protein is not expressed. The authors offer no explanation for this paradox. One explanation for this could be that spurious PPIs have formed in the protein extraction/immunoprecipitation protocol. These findings need validation by biochemical fractionation of synapses into pre and post synaptic fractions and immunohistochemistry to demonstrate the subsynaptic localisation of the proteins.
5. My concerns about spurious interactions are raised again because the authors say that 92% of their interactions are novel (I note that they authors have not compared their interaction data of the NMDA receptor with published datasets from Dr Seth Grant's laboratory). BioGrid itself is good but not enough for comparison, maybe at this point it worth taking String, which accumulates several sources of PPIs, just select the direct PPIs.
6. A major concern is that they use SynGO as a reference database, and even test the enrichment against it. SynGO is about ~ 2000 genes in size and was built around the presynaptic datasets, so it is biased and incomplete in terms of the whole synapse. This may be one of the reasons there is the strangely high percentage of presynaptic proteins interacting with postsynaptic proteins as noted above.

Minor comments

1. A number of papers have reported protein interactions of native NMDA receptor complexes and their associated proteins isolated from rodent brain and are neither referenced in this paper. It would be relevant to compare these published datasets with the Grin2B IP datasets.
2. The use of the term "bait" in purification experiments typically refers to a protein and not an antibody. I suggest removing the word bait to avoid ambiguity and simply use the word target.
3. 26 mins of treatment gives completely different set of PPIs between PCP and saline which is very interesting, so both networks should be included in Supplementary. Also, it would be useful to have a list of modulated (phosphorylated in their case, but also ubiquitinated etc) proteins, which is not presented.
4. As they say their final network is composed of "direct physical and "co-complex" interactors and they cannot distinguish between them. This is particularly bad for the postsynapse, where all the PSD components can be co-IP-ed in different combinations. It can explain the Figure 5C, where most of the proteins have FDR = 1, which means they do not reproduce.
5. The way PPI data is reported can be improved so that I does not have to be extracted from Table 1 and 2. It would be good if they provide just two columns PPI list, with names or IDs, plus PSP/saline/both conditions in third column, for ease of comparison with other sources and building the graph. They can add it as another spreadsheet to Table 2.
6. Is Figure 2 built for Sal or PCP conditions? as they have only 23% interactions in common (Figure 4A) the Figure 2 should be pretty different for two conditions. Are the 1007 interactors combined from SAL and PCP?
7. Figure 1F is mentioned but no figure is shown.
8. Overall the paper could be edited and made more concise, especially the introduction and discussion.

Significance

General assessment

Proteomic mass spectrometry of immunoprecipitated complexes from synapses has been extensively studied since Husi et al (2000) first study of NMDA receptor and AMPA receptor complexes. Since then, a wide variety of methods have been employed to purify synaptic protein complexes including peptide affinity, tandem-affinity purification of endogenous proteins tagged with FLAG and Histine-affinity tags amongst other methods. Purification of protein complexes and the postsynaptic density from the postsynaptic terminal of mammalian excitatory synapses have been crucial for establishing that schizophrenia is a polygenic disorder affecting synapses (e.g. Fernandez et al, 2009; Kirov et al, 2012; Purcell et al, 2014, Fromer et al, 2014 etc). Network analyses of the postsynaptic proteome have described networks of schizophrenia interacting proteins (e.g. Pocklington et al, 2006; Fernandez et al, 2009) and other neuropsychiatric disorders.

Hundreds of synaptic protein complexes have been identified (Frank et al, 2016), but very few have been characterised using proteomic mass spectrometry. This paper has chosen 8 protein targets for such analysis and identified many proteins that a putative interactors of the target protein. At this level the current manuscript does not represent a conceptual advance and the value of the data lies in its utility as a resource that may be used in future studies.

The findings from the 8 target proteins from normal adult rat brain were used for a secondary study that describes the effects that PCP has on the interaction networks. Interestingly, this work shows that 26 minutes of drug treatment leads to considerable changes in the interactomes of the target proteins. These descriptive data could be used in future studies to understand the cell biological mechanisms that mediate these rapid changes in the proteome. PCP and drugs that interact with NMDA receptors are known to induce changes in synaptic proteome phosphorylation including modifications in protein-protein interaction sites, which may explain the PCP effects.

The study would benefit from validation of experimental protocols for solubilisation and immunoprecipitation and validation of described interactions using orthogonal biochemical or localisation experiments.

Audience

Specialists in synapse proteins and mechanisms of schizophrenia.

Expertise

The reviewers' expertise is in molecular biology of synapses including synapse proteomics, protein interaction and network analysis, and genetics of schizophrenia and other brain disorders.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary: McClatchy, Powell and Yates aimed at identifying a protein interactome associated to schizophrenia. For that, they treated rats (N14 and N15) with PCP, which disturbs gutamatergic transmission, as a model for the disease and co-immunoprecipitated hippocampi proteins, which were further analyzed by standard LC-MS.

The study is new, considering not much has been done in this direction in the field of schizophrenia. This justifies its publication. On the other hand, a major flaw of the is the lack of information on the level of interaction of the so called protein interactome. Meaning, we cannot distinguish, as the study was performed, which proteins are directly interacting with the targets of interest from proteins which are interacting with targets´ interactors. The different shells of interaction are crucial information in protein interactomics.

Major: most of I am pointing below must be at least discussed or better presented in the paper, as It may not be solvable considering how the study has been conducted.

1. The study fails in defining the level of interaction of the protein interactome with the considered targets. This has been shortly mentioned in the discussion, but must be more explicit to readers, for instance, in the abstract, introduction and in the methods sections.
2. Considering the protein extraction protocol, it is fair to mention that only the most soluble proteins are being considered here. I am bringing this up since the importance of membrane receptors is clear in the studied context.
3. It is not clear from the methods description if antibodies from all 8 targets were all together in one Co-IP or have been incubated separately in 8 different hippocampi samples. It seems the first, given how results have been presented. If so, this maximizes the major issue raised above (in 1).
4. Definitely, results here are not representing a "SCZ PPI network". PCP-treated animals, as any other animal model, are rather limited models to schizophrenia. As a complex multifactorial disease, synaptic deficits, which is the focus of this study, can no longer be considered "the pivot" of the disease. Synaptic dysfunction is only one among many other factors associated to schizophrenia.
5. Authors should look for protein interactions that might be happening also in glial cells. They are not the majority in hippocampus, but are present in the type of tissue analyzed here. Thus, some of the interactions observed might be more abundantly present in those cells. Maybe enriching using bioinformatics tools the PPI network to different cell types.

Minor:

1. in the abstract, it is not clear if 90% of the PPI are novel to brain tissue in general or specifically schizophrenia.
2. authors refer to LC-MS-based proteomics as "MS" all across the text. Who am I to say this to Yates et al, but I think it is rather simplified use "Mass Spectrometry Analysis", when this is a typical LC-MS type of analysis
3. Several references used to construct the hypothesis of the paper are rather outdated: several from 10-15 years ago. It would be interesting to provide to the reader up to date references, given the rapid pace science has been progressing.
4. "UniProt rat database". Please, state the version and if reviewed or unreviewed.

Significance

The study is informative, and has great potential to enrich the specific literature of this field. But should tone down some arguments, given the experimental limitations of the PPI network (as described above) and should state PCP-treated rats as a limited model to schizophrenia.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary:

Provide a short summary of the findings and key conclusions (including methodology and model system(s) where appropriate).

In this manuscript, McClatchy and colleagues used a conventional approach combining immunoprecipitation (IP) of endogenous target proteins (baits) followed by liquid chromatography mass spectrometry (MS) analysis of the co-immunoprecipitating proteins to map protein-protein interaction (PPI). This interaction network is centered around baits that had been annotated as susceptibility factors for schizophrenia (SCZ). A variety of previous studies have identified thousands of such SCZ susceptibility factors. Mostly based on the availability of antibodies, 8 bait proteins were selected in this study. The authors reasoned that immunoprecipitating endogenous proteins from tissues using specific antibodies was a more accurate view of physiological conditions than epitope tagging followed by affinity purification (AP) from cells in culture. The model system from which proteins were extracted was the hippocampus dissected from mice that had been treated or not by phencyclidine (PCP), a drug that has been shown to induce SCZ symptoms in humans and animals. By comparing the proteins identified and quantified from the PCP-treated samples against control IPs and/or saline-injected mouse controls, a large number of PPI were deemed statistically significant. Most of these potential interactors were not present in PPI databases (BioGRID), most likely because such databases are populated with large-scale APMS datasets from cell cultures, with very few studies using brain tissue. Strikingly, many of the co-immunoprecipitated proteins were also known as SCZ susceptibility factors, which lend weight to the hypothesis that these factors form a large protein interaction network, localized at the synapses.

Major comments:

• Are the key conclusions convincing?

Overall, the conclusions drawn from the experimental design, data analysis, and corroboration with existing literature are well-supported and convincing. When selecting the SCZ susceptibility factors, the authors clearly state their goal, the databases used for gene selection, and the rationale for choosing proteins with synaptic localization. The inclusion of evidence from genetic studies and previous publications strengthens the credibility of the selected genes. The methodology used to establish the novel SCZ PPI network is mostly well-described (see minor comments below). The use of an 15N internal standard also adds rigor to the quantitation of PPI. The GO enrichment analysis provides valuable insights into the biological functions and cellular components associated with the SCZ PPI network. The annotation of identified proteins using the SynGo synaptic database and the distribution of annotated synaptic proteins among different baits further support the biological relevance of this PPI network. The cross-referencing of the PPI network with published genetic studies on SCZ susceptibility genes adds robustness to the findings. Specifically, the observation that 68% of protein interactors have evidence of being potential SCZ risk factors is a strong corroboration of the prevailing hypothesis in the field. Finally, the significant changes induced by PCP that were identified for all baits except Syt1, along with the comparison of altered proteins with SAINT-identified PPI, add depth to the understanding of PCP modulation. - Should the authors qualify some of their claims as preliminary or speculative, or remove them altogether?

No, but note that APMS/IPMS has been around for more than a decade (Introduction page 3). - Would additional experiments be essential to support the claims of the paper? Request additional experiments only where necessary for the paper as it is, and do not ask authors to open new lines of experimentation.

One piece of data that is missing are Western blots using the 8 selected antibodies against the proteins extracted from their experimental samples to validate the antibodies recognize 1 protein of the expected size from these tissue extracts. - Are the suggested experiments realistic in terms of time and resources? It would help if you could add an estimated cost and time investment for substantial experiments.

Running SDS-PAGE and Western blotting should be straightforward and cheap. - Are the data and the methods presented in such a way that they can be reproduced?

Yes - Are the experiments adequately replicated and statistical analysis adequate?

Yes

Minor comments:

• Specific experimental issues that are easily addressable.

The rationale for the short duration between PCP injection and animal sacrifice is only explained in the discussion section (page 17). The fact that this short treatment of less than 30 min should prevent any change in transcription or translation should be introduced earlier (in the experimental procedures). Note that the duration is written as 26 min on page 4 and 25 min on page 9. Please reconcile these numbers. Is there any biological significance for this SCZ study that the mice were maintained on a reverse day-night cycle? It is not clear from reading Experimental Procedures/Bioinformatic Analysis section (page 6) if normalized N14/N15 protein ratios measured in the bait-IPs and control-IPs were used for the SAINT analysis? Or did the authors used label-free quantitation with spectral counts? - Are prior studies referenced appropriately?

Yes - Are the text and figures clear and accurate?

Fig1C: The workflow is a little too simple, the authors might want to add more details. FigS1C: Please add x-axis title (spectral counts) directly to the figure. Fig2B-D: The color scale bar should have number values to denote lower and upper limits in % (as opposed to "lowest" and "highest"). - Do you have suggestions that would help the authors improve the presentation of their data and conclusions?

No

Significance

• Describe the nature and significance of the advance (e.g. conceptual, technical, clinical) for the field.

In this study, the authors have drastically expanded the protein interaction landscape around 8 known SCZ susceptibility factors by using a conventional IPMS approach. Performing the IPs on protein extracted from hippocampus dissected from mice treated with phencyclidine to model SCZ increases the biological significance of such lists of proteins. Furthermore, the co-immunoprecipitation of many other SCZ susceptibility factors along with the 8 selected baits supports the hypothesis that these proteins of varied functions are part of large interaction networks. Overall, the integration of experimental data with in silico networks, along with the quantification of PPI changes in response to PCP, should contribute to a more nuanced understanding of SCZ pathogenesis. The potential implications for drug development underscore the broader significance of the study in advancing our knowledge of neurobiology and its relevance to neurological disorders like schizophrenia. - Place the work in the context of the existing literature (provide references, where appropriate).

Overall, this study contributes to the existing literature by providing experimental data on in vivo PPI networks related to SCZ risk factors. Not only do the authors validate 124 known interactions but also they identify many novel PPI, due to a gap in the existing literature regarding the comprehensive mapping of PPI directly from tissue extracts, especially brain tissue. The authors advocate for more IPMS studies in mammalian tissues to generate robust tissue-specific in silico networks, which agrees with the growing understanding of the importance of tissue-specific networks for identifying disease mechanisms and potential drug targets.

Furthermore, the SCZ PPI network reported here is enriched in proteins previously associated with SCZ, which aligns with the existing literature emphasizing the involvement of certain proteins and pathways in the pathogenesis of SCZ [References: 78-85]. The authors also investigate the response of the SCZ network to PCP treatment, hence providing insights into the potential effects of post-translational modifications, protein trafficking, and PPI alterations in a model of schizophrenia, which adds to existing knowledge about the impact of PCP on the molecular processes associated with SCZ [References: 88, 89, 92]. - State what audience might be interested in and influenced by the reported findings.

Overall, the findings reported in this manuscript have implications for both basic research in molecular biology and potential translational applications in the development of targeted therapies for neurological disorders, particularly schizophrenia. The study delves into in vivo protein-protein interaction (PPI) networks related to genes implicated in schizophrenia (SCZ) risk factors. Researchers in neuroscience, molecular biology, and psychiatry would find the information valuable for understanding the molecular basis of SCZ. The study highlights the potential for identifying disease "hubs" that could be drug targets. Pharmacologists and drug developers interested in targeting protein complexes for drug development, especially in the context of neurological disorders, may find the study relevant. - Define your field of expertise with a few keywords to help the authors contextualize your point of view. Indicate if there are any parts of the paper that you do not have sufficient expertise to evaluate.

Technical Expertise | biochemistry, liquid chromatography mass spectrometry, proteomics, computational biology, protein engineering, protein interaction networks, post-translational modifications, protein crosslinking, proximity labeling, limited proteolysis, thermal shift assay, label-free and isotope-labeled quantitation. Biological Applications | human transcriptional complexes, apicomplexan parasites, viruses, nuclear envelope, ubiquitin ligases, non-model organisms.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

The authors do not wish to provide a response at this time.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

The manuscript by Balachandra and Amodeo presents Bellymount-Pulsed Tracking as a technique for continuous long-term imaging of Drosophila oogenesis. This approach modifies the existing Bellymount technique by exposing restrained female flies to pulses of CO2 anesthesia in combination with image acquisition. Flies that survived the restraint were kept alive for many hours by addition of a liquid diet in the restraint apparatus. This allowed for imaging and tracking of egg chamber development over longer time periods than capable with ex vivo culturing methods. However, the authors did report a 40% mortality rate and decreased fecundity compared to unrestrained flies. Using this method the authors were able to image and measure the growth rate of developing egg chambers in living flies, and capture events like vitellogenesis which relies on the interactions of multiple organ systems.

This technique is a notable contribution to the fly community, as it could be useful for studying processes that require interactions between multiple tissues and organs, as well as for long-term imaging of other internal structures in the adult fly. The significance is somewhat reduced due to the relatively high mortality rate and the decreased fecundity and egg chamber growth rate reported. However, the authors should be commended for their diligence in documenting the limitations of the procedure, as this now provides a strong jumping off point to improve the technique if it becomes widely adopted by the fly community. Overall, the experiments appear to have been carefully performed and the manuscript is clearly written. However, there are several issues that should be addressed prior to publication.

Major concerns

1. The movies of egg chamber development are challenging to interpret. They could be improved by the addition of timestamps and other annotations. Having multiple example movies of the same process would also be valuable. It could be helpful to potential users of this technique to show the process the authors used for identifying the same egg chamber between such long time points.
2. Figure 4 - Given that the Bellymount PT technique slows oogenesis and reduces egg chamber growth in vitellogenic stages (Figure 3E), it is possible that Bellymount PT slows yolk protein uptake. It would be important to establish a baseline for how much to expect yolk protein levels to change across stages to compare to measurements obtained with Bellymount PT. It would be a relatively simple experiment to show the change in yolk protein uptake across stages in fixed samples. This could also be performed for His2Av dynamics during nurse cell dumping.
3. Movie 11 - The authors propose that Bellymount-PT can be used to visualize the process of border cell migration. However, there is no obvious movement of the cluster relative to the nurse cell nuclei over the course of the 3 hour long movie. The authors should either show a better movie of border cell migration, or remove this claim from the manuscript.
4. Movie 13 - The authors claim that they see egg chamber rotation continue in stage 9 and 10 egg chambers. This movie is not convincing. There is also very strong evidence in the literature that egg chamber rotation ends at stage 8. Chen et al., Cell Reports, 2017 showed using a method that tracks follicle cell migration in vivo that rotational migration ends during stage 8. The only movement of follicle cells after stage 8 is due to the epithelial reorganization that occurs during the posterior movement of the follicle cells as the stretch cells flatten. Additionally, after stage 8 follicle cells lose their circumferentially oriented actin protrusions that drive rotation. This claim should be removed from the manuscript.

Minor comments

1. Line 104 - The authors mention that CO2 affects fertility in flies. They should also reference Sustar et al., Genetics, 2023 and Zimmerman and Berg, PLoS One, 2024 for wider ranging effects of CO2 on oogenesis.
2. Line 244 - Although it is true that the original paper describing egg chamber rotation reported that it starts at 5, subsequent studies from multiple labs have confirmed that it begins much earlier. First shown by Cetera et al., Nature Communications, 2014 but later confirmed by Bilder, Dahmann, and Mirouse labs. Chen et al., Cell Reports, 2016 has even published a movie of an egg chamber initiating rotation as it buds from the germarium.
3. Figures of egg chambers are generally oriented anterior on the left and posterior on the right. Reorienting all the figures would be challenging, so the recommendation is to be clear in the figure legends the orientation of the images. This is important given they are shown in different orientations in Figure 1 than throughout the rest of the paper, and also will be helpful for readers who may not be familiar with the structure of the ovary/egg chambers.
4. Figure 1B and Methods line 334 - Should "Rely" be "Relay"?
5. Figure 1E - Oocyte nuclei are missing from the diagrams of stage 7, 13 and 14 egg chambers. Also, "G" looks like a figure panel label, could just say Germarium
6. Figure 3F-H - "Stagee" should be "Stage"
7. Figure 4B - Why is the fluorescence for egg chamber #6 so much higher than the others? It makes the slopes of the other samples hard to see.
8. Figure 4D,E,G - For clarity, the labeled boxes should be the same color as the lines on the associated graphs. In line 790 "Note the steady increase of H2Av in all three regions as it exits the nurse cell nuclei" - this is not actually shown without the nurse cell nuclei average intensity being on the graph as well.
9. Line 787 - "Note the flow of H2Av" - "flow" is not actually shown in these static images. Consider a more precise description.

Referee Cross-commenting

The other reviewers make several excellent points. We personally feel that it is beyond the scope of this initial report to ask the authors to show that they can see all aspects of oogenesis with this technique. If the method becomes widely adopted by the oogenesis community, individual researchers can optimize it to suit the exact process they want to study. If the authors want to claim they can see a particular process, it needs to be well documented and convincing. For example, we agree that the movies that claim to show egg chamber rotation (both during established stages and later) and border cell migration need to be improved or the claims need to be removed. However, we feel that the authors have documented enough other interesting processes to make the study worthy of publication. Likewise, asking the authors to determine the minimal time window that can be used for imaging could take months of open-ended work and is something that could be better tackled by subsequent users depending on the requirements of the biological process they want to study. It seems better to get the work out into the public sooner rather than later so that improvements can be crowd sourced.

Finally, although Flp-out clones were used for cell tracking in the original Belly mount paper, this technique will be less effective during the first half of oogenesis when the egg chamber is rotating, as the clone is likely to rotate into and out of sight between imaging time points.

Significance

This technique is a notable contribution to the fly community, as it could be useful for studying processes that require interactions between multiple tissues and organs, as well as for long-term imaging of other internal structures in the adult fly. The significance is somewhat reduced due to the relatively high mortality rate and the decreased fecundity and egg chamber growth rate reported. However, the authors should be commended for their diligence in documenting the limitations of the procedure, as this now provides a strong jumping off point to improve the technique if it becomes widely adopted by the fly community. Overall, the experiments appear to have been carefully performed and the manuscript is clearly written. However, there are several issues that should be addressed prior to publication.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary:

The authors describe an improvement of the Bellymount imaging method for internal tissues of the fly's abdomen. They are able to increase the total duration of the imaging by introducing pulsed anesthesia. This allows the immobilized flies to take up food in between the imaging; this increases survival rate and allows for longer total imaging times. The authors illustrate the technique by tracking the development of egg chambers.

Major Points

• The Bellymount PT method results in decreased fecundity, which might affect the processes (oogenesis) the authors looked at. Indeed, the authors conclude that "oogenesis is not completely stalled under the Bellymount-PT protocol" (line 140). The authors do provide some data indicating that egg chambers develop (Fig. 2G,H; Fig. 3F,H), in particular a stage 10 egg chamber proceeding to a stage where dorsal appendages seem to form. However, for early stage egg chambers this is less convincing. The egg chambers show an increase in (cross-sectional) area, however, what is the evidence that they also mature? For example, during egg chamber maturation, the ratio of oocyte/nurse cell volume changes, follicle cells re-arrange, etc. The authors should test whether any of these characteristics can be observed in egg chambers imaged using Bellymount PT. This may include the imaging of egg chambers in which both nuclei and plasma membranes are visualized.
• A potential advantage of the Bellymount PT method is the ability to follow the dynamics of processes. A current drawback, however, is the rather low temporal resolution as the fly needs to wake up between single images. The authors should provide an estimate for the minimal possible cycle time and should test whether flies imaged at 10 minutes interval show lower survival/fecundity than flies imaged at 2 hours interval.
• The authors claim that they can track on a cellular level (based on nuclei), but it is unclear how accurate the tracking is. Especially cell tracking over very long times might be challenging here, as the time delay between two time points is big. The authors should test the accuracy of their tracking, potentially by creating Flip-out clones and using them as a control.
• The authors show that they can visualize cell membranes (Moesin-GFP, Fig. 2C). Tracking cells over time based on their membranes would greatly widen the applicability of the method as it would enable to analyze the complex cellular dynamics during egg chamber maturation. The authors should test whether cells can be tracked over time (e.g. using Moesin-GFP) using their technique.
• Movie 11. The authors claim that they can capture border cell migration. However, it is unclear whether the border cells actually migrate towards posterior. The authors should track and quantitatively analyze the migration path of the border cells in their movies.
• Movie 12. The authors claim that they can observe egg chamber rotation. However, it is unclear whether the egg chambers actually rotate. The authors should track cells and quantify the angular velocity of movement.

Minor Points

• Please move the labels of the scale bars to the legends.
• The figures (especially 2 and 3) would benefit from a clearer structuring. Moving part of them to supplementary figures would also help.
• "stage" typo in figure 3

Significance

The authors describe here an improvement of an existing technique. The advantage of the improved technique is the longer imaging time, which potentially allows users to track cells/organelles/proteins over time. However, tracking requires the user to connect single time points with each other, which is somewhat unclear at this time. Moreover, the potential applicability (and significance) of the technique would be widened if visualization and tracking of cell membranes/organelles/vesicles would be possible. With these further optimizations, the technique would add a useful tool to the Drosophila community.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary

The Drosophila ovary is an established model system for many aspects of development and cell biology. In vitro culture of live ovaries has provided valuable insight, yet these methods do not accurately mimic oogenesis in vivo for some stages. Here the authors develop a new method that allows for sustained imaging of ovaries in intact flies, maintaining normal physiology.

The method provides a valuable addition to the field. Processes such as growth, cell migration, egg chamber rotation, yolk uptake and nurse cell dumping can be observed in the intact fly. Time lapse and 3D reconstruction provide valuable tools. While the detail/resolution of the images is not as good as ex vivo or fixed samples, the ability to maintain normal development and homeostasis provides a novel advantage. The figures and movies are well-presented and sufficient detail is provided in the methods.

Major comments

1. Why do the authors think that growth is slowed? The imaging process or the trapping/anesthesia of the fly? For example, if the frequency of imaging was varied, it could reveal whether it was the actual imaging that affected development. Did the length of time the fly had been in the trap make a difference? The sentence on lines 190-191 is not clear.
2. In Movie 6, the nurse cell nuclear shape does not look normal - more ovoid than round. Perhaps some settings are off in the 3D reconstruction.
3. Movie 11 - why do the border cells seem stalled?
4. There is no discussion of the earliest stages of oogenesis. Is it possible to see egg chambers forming from the germarium?

Minor comments

1. It would be helpful to mention if the egg chambers stay in similar locations or move around - is it challenging to locate the same egg chamber after 2 hours?
2. Are any egg chambers degenerating? This could indicate stress in the fly.
3. In Figure 4D, release of HisAV into the cytoplasm is described. Similar release of nuclear proteins was described by Cooley et al. 1992 so this paper could be cited.
4. At 321 minutes in Figure 4D, a large nucleus is apparent in the oocyte. Is this an oocyte nucleus or evidence for nurse cell translocation to the oocyte as described in Ali-Murthy et al. 2021?

Significance

The technique provides a significant advance to the field, extending the time period currently possible to image ovaries through the Belly Mount method. It will immediately benefit researchers working on the ovary but could be extended to many other tissues in the fly abdomen such as the gut and tumor models.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Reviewer #1

__Evidence, reproducibility and clarity __

The work by Przanowska et al., sought to understand the role of ORC2 in murine development and further wanted to discover its role in liver endo-reduplication. The overall methods used is sufficient enough to address its role but is not very conclusive based on their overall results and data provided as elaborated in below comments.

Major Comments:

1. The major issue of the paper is how well is ORC2 depleted in perinatal liver (Fig. 2C) and is not very clear from the data as all the western blots are at very low exposure levels and bands are very weak (still weak bands seen). There are good antibodies of ORC2 which can be used for IHC staining and can be used to address the extent of ORC2 depletion.

We have now shown that ORC2 protein is significantly decreased in the hepatocytes of the Orc2 KO and DKO livers (New Fig. 2C and 6D). The decrease is consistent, with 4-5 mice examined, and all showing the depletion. We have been unable to do immunohistochemistry on tissue sections of the mouse livers with the anti-ORC antibodies we have tried, and this could be a reflection of the low level of the proteins. On hepatocytes in culture we have obtained faint signal with the anti-ORC2 antibody in WT cells, and this is clearly absent in 100% of the hepatocytes. See Fig. R1 below.

__Reviewer Fig R1: __

---

A) Immunofluorescence of hepatocytes in culture from livers of WT and two DKO mice.

B) Quantitation of A) from counting 70-100 cells from each specimen.

However, the calculations in the methods and the discussion are very compelling that at least the last 6-9 cell divisions in normal development start with 2n nuclei in the livers at baseline (Fig. 3B-G and 6I).

Why in Fig 2C, the M2 mice is showing an equivalent level of ORC2 protein compared to mice M1 with NO CRE expression (compare lane1 and lane5). So, the results are based on one mouse which I do not think is significant enough to come to the conclusion. The authors need to add more data from different mice for statistical significance. Please use IHC to show the depletion of ORC2 protein in the liver sections.

We had used total liver and had pointed out that residual ORC2 protein will be seen from stromal cells (endothelia, blood vessels and blood cells). We have therefore removed the figure which measured ORC2 levels in total liver and have now shown that when hepatocytes are isolated from five animals there was a massive depletion of ORC2 in all five animals (new Fig. 3C).

As nicely demonstrated in the previous paper by Okano-Uchida et al., 2018 that ORC1 depletion in the liver shows an DNA ploidy effect from 6-week onwards. The authors need to demonstrate in this paper also when the 16N phenotype is observed starting from week1 to 12 months.

Based on the results from our previous paper (Okano-Uchida et al., 2018) we decided to measure 16N phenotype at 6 weeks of age. The endoreduplication occurs at a stage when ORC2 protein is undetectable during normal development or during regeneration.

In the double knockout experiments (ORC1 and ORC2) the authors are not even bothered to demonstrate that how much are both the proteins are actually depleted from the cells, so on the results obtained from these mice experiments are not conclusive or explanatory.

We have performed immunoblotting of isolated hepatocytes and immunohistochemistry of livers for ORC1 and ORC2. Our data shows that both proteins are depleted in all four mice tested (New Fig. 6D).

Minor points:

1. Why are scale bars missing in right panel of Fig. 2G, Fig. 6D Supp Fig. 2B KO studies. The authors need to confirm that that all the large nuclei have NO or less significant ORC2 protein through IHC H&E staining.

The scale bars are missing from the right panels to avoid redundancy. We have added “Both panels are at the same scale.” in the figure legend, according to https://doi.org/10.1371/journal.pbio.3001161.

1. Please explain why is EYFP in Fig. 5G is cytoplasmic compared to Fig 4C (nuclear). We consistently see this variability and it was there in our previous results (Okano-Uchida et al., 2018), where EYFP was cytoplasmic in tissues, but was nuclear (and some cytoplasmic) in hepatocytes in culture.

We do not know the reason for this difference but consistently see this difference. We now say in the text: “We did not explore why the EYFP protein is mostly nuclear in hepatocytes in culture (Fig. 4C) and mostly cytoplasmic in hepatocytes in the liver tissue (Fig. 5G, 7G), but speculate that differences in signaling pathways or fixation techniques between the two conditions contribute to this difference.”

Are authors using the same genotype of Alb-Cre mice as shown by Okano-Uchida et al., 2018 as I do not find the reference of Schuler et. al., 2004 (PMID:15282742).

We have been using two independent Alb-Cre animals. This is now described in the Methods.

---

Significance

The article is exactly based on their previous published paper but instead of ORC1, they were interested in dissecting the role of ORC2. Although they have discussed that CDC6 may be involved in replacing ORC1 KO mice to rescue the extensive DNA replication in endoreduplication, but instead of going to hunt the role of CDC6 in endoreduplication they checked the effect of ORC2 which actually lower the overall impact of the paper.

We studied ORC2 conditional KO mice in a similar manner to the previously published ORC1 conditional KO in order to ensure (1) that the lack of effect in the Orc1 KO was not because ORC1 can theoretically be substituted for by CDC6 and (2) to establish the double KO of Orc1 and Orc2. To the best of our knowledge this is the first description of removal of two subunits of ORC complex at once in a mouse model. Moreover, in the light of rising recognition of sex as biological variable, we report sex-dependent effects which are very intriguing.

We have not attempted knocking out CDC6 to uncover novel mechanisms of DNA replication, because we first needed to make sure that the mice can truly endo-reduplicate without two of the six subunits of ORC. Note that our published results in cancer cell lines (Shibata, 2016) show that CDC6 is still essential in the ORC KO cell lines, so a future experiment will likely reveal that CDC6 is still essential for endoreduplication in the ORC KO mice in vivo.

Reviewer #2

__Evidence, reproducibility and clarity __

It has been reported that in the absence of ORC1, liver cells can still endoreduplicate and it has been speculated that this might occur if CDC6 can replace, at least partially, the function of ORC1. Here, authors evaluate if this is also true in the absence of ORC2 and found that ORC2 is required for cell proliferation in mouse hepatocytes but not for endoreduplication. This is also the case after combining the conditional mutations of ORC1 and ORC2. They propose that a mechanism must exist to load sufficient MCM2-7 to support DNA replication in the absence of these two ORC subunits. Some of the conclusions need further experimental support. The rationale for testing the requirement of ORC2, with or without ORC1, for endoreduplication is valid. However, a key point is that the endoreduplication level seems to be higher in the absence of ORC2 or both ORC1 and ORC2, and this is not properly addressed. Also, mechanistic details on how this could be triggered are absent from this study. As indicated below almost every figure in this manuscript contains weak points (see below).

We now discuss the following: “One possible explanation of the greater endoreduplication in both our papers is that mitosis may be arrested earlier in development by G2 DNA damage checkpoints activated by incomplete licensing and replication of the genome in the absence of ORC. As a result, endoreduplication cycles could begin earlier in development resulting in greater endoreduplication.”

Major 1. Fig 1G, needs a detailed comment and justification.

We have added the following to the text: “The proliferation rate of the MEF were measured by MTT assays. Even in the Orc2+/+ MEF, the infection with adeno-Cre decreased proliferation a little (the orange line compared to the blue line in Fig. 1G). However, for Orc2f/f MEF infection with adeno-Cre impairs proliferation even further (yellow line compared to black line in Fig. 1G)..

Note that Adeno-Cre has been reported to be toxic for cell proliferation (citations 1, 2, 3), and so we included Adeno-Cre expression in ORC2+/+ (WT) as a background control.

Citation:

1. Pfeifer A, Brandon EP, Kootstra N, Gage FH, Verma IM: Delivery of the Cre recombinase by a self deleting lentiviral vector: Efficient gene targeting in vivo. Proc Natl Acad Sci USA. 2001, 98: 11450-11455. 10.1073/pnas.201415498.
2. Loonstra A, Vooijs M, Beverloo HB, Allak BA, Drunen EV, Kanaar R, Berns A, Jonkers J: Growth inhibition and DNA damage induced by Cre recombinase in mammalian cells. Proc Natl Acad Sci USA. 2001, 98: 9209-9214. 10.1073/pnas.161269798.
3. Schmidt EE, Taylor DS, Prigge JR, Barnet S, Capecchi R: Illegitimate Cre-dependent chromosome rearrangements in transgenic mouse spermatids. Proc Natl Acad Sci USA. 2000, 97: 13702-13707. 10.1073/pnas.240471297.
4. Fig 2D-F. Is this conclusion applicable to other endoreplicating tissues? Have authors consider to analyze body weight and liver weight measurements after normalization with similar data from a non-affected organ? The conditional KO was performed specifically in the liver. ORC is intact in other tissues in these animals. As a future direction our lab plans to study cardiac-specific conditional KO of ORC subunits to test whether other endo-reduplicating tissues can also synthesize DNA in the absence of ORC subunits.

Fig 3 shows inconsistent results or results that lack proper justification in the text. The 2C peak is missing in Fig 3E (yellow line, positive control). However, 2n nuclei appear in Fig 3F-H. Also, the blue and yellow peaks do not coincide in the flow cytometry profiles, in particular for 8C and 16C.

There was an error in the plotting of the former Fig. 3E. The information is better presented in the former Fig. 3F-H (now Fig. 3E-G) and so have removed the former Fig. 3E from the paper.

Fig 4. Shorter EdU pulses could be more informative of the actual amount of S-phase cells. Thus, the use of a 2h EdU pulse needs a clear justification.

The half-life of EDU incorporation differs slightly between in vivo and in vitro conditions. In vivo, slower cell proliferation requires a longer time, approximately 4 hours. However, in vitro, liver cells grow faster, and a 2-hour EDU pulse with 20 µM is sufficient for detection compared to a 3-hour pulse with 10 µM BrdU (Okano-Uchida et al., 2018). Several publications also use a 2-hour EDU incubation time (https://doi.org/10.1098/rsob.150172).

Fig 5. EYFP is cytoplasmic, in contrast with results shown in Fig 4C

We consistently see this variability and it was there in our previous results (Okano-Uchida et al., 2018), where EYFP was cytoplasmic in tissues, but was nuclear (and some cytoplasmic) in hepatocytes in culture.

We do not know the reason for this difference but consistently see this difference. We now say in the text: “We did not explore why the EYFP protein is mostly nuclear in hepatocytes in culture (Fig. 4C) and mostly cytoplasmic in hepatocytes in the liver tissue (Fig. 5G, 7G), but speculate that differences in signaling pathways or fixation techniques between the two conditions contribute to this difference.”

Fig 6. Results obtained with the double mutant are poorly described.

We have split the figure into two figures (New Fig. 6 and 7) edited the results section to ensure that they are easily comprehended by the readers. We have also included Westerns from hepatocyte cell lysates of four DKO mice to show that ORC1 and ORC2 proteins are reproducible decreased (New Fig. 6D).

What are the level of other pre-RC components in the mutants used in this study. This could be easily evaluated by Western blotting

Despite the technical difficulty of not having antibodies that recognize all the mouse initiation proteins, we have now measured mouse ORC1, ORC2, ORC3, ORC5, ORC6, CDC6 and the MCM2 and MCM3 subunits of MCM2-7. The results do not show a consistent decrease or increase of any of these proteins in individual mice of the two genotypes, Orc2-/- or DKO (New Fig. 2D and 6E)

How do authors justify their claim that a very limited amount of ORC are sufficient to load a vast excess of MCM2-7 hexamers?

The rationale is stated in the introduction from data from cancer cell lines: “Given that WT cells have about 150,000 molecules of ORC2, even if this truncated protein is functional ORC2, ~150 molecules of the protein would be expected to load MCM2-7 double hexamers on at least 50,000 origins of replication. Experimentally, we show in Shibata, 2020 (Fig. 7C), that although ORC subunits are undetectable on Westerns, MCM2-7 association with the chromatin is unchanged. By the way, we do not say “vast excess” of MCM2-7, just sufficient MCM2-7 to fire 50,000 origins.

Minor 1. The titles of the Results section could be more informative of the main conclusion rather than simply descriptive

We updated our Results titles to be more informative.

The Discussion is too long

We have shortened the discussion by removing our calculations to the Results section and abbreviating some of the discussion on endoreduplication. However we had to insert new items brough forth by the reviewers. Due to the controversy of this topic in our field, we had to include extensive discussion of current literature and put our results in their proper context.

Significance

The topic is relevant and the hypothesis tested is reasonable, although the conceptual advance is limited (see also below). The major limitation is the absence of mechanistic details addressing the occurrence of extra endoreduplication cycles (compared to controls) in the ORC1 and ORC2 mutants.

Reviewer #3

__Evidence, reproducibility and clarity: __

The origin recognition complex (ORC) is an essential loading factor for the replicative Mcm2-7 helicase complex. Despite ORC's critical role in DNA replication, there have been instances where the loss of specific ORC subunits has still seemingly supported DNA replication in cancer cells, endocycling hepatocytes, and Drosophila polyploid cells. Critically, all tested ORC subunits are essential for development and proliferation in normal cells. This presents a challenge, as conditional knockouts need to be generated, and a skeptic can always claim that there were limiting but sufficient ORC levels for helicase loading and replication in polyploid or transformed cells. That being said, the authors have consistently pushed the system to demonstrate replication in the absence or extreme depletion of ORC subunits.

Here, the authors generate conditional ORC2 mutants to counter a potential argument with prior conditional ORC1 mutants that Cdc6 may substitute for ORC1 function based on homology. They also generate a double ORC1 and ORC2 mutant, which is still capable of DNA replication in polyploid hepatocytes. While this manuscript provides significantly more support for the ability of select cells to replicate in the absence or near absence of select ORC subunits, it does not shed light on a potential mechanism. While a mechanistic understanding of how these cells proliferate in the absence or extreme depletion of ORC subunits is outside the scope of the current manuscript, it would have been beneficial to see more functional analyses to help guide the field. For example, is there a delay or impairment in Mcm2-7 loading in G1 (FACs-based loading assay from the Cook Lab (Matson et al., eLife. 2017)) in primary hepatocytes with the ORC2 conditional deletion? Is copy number maintained as cells increase polyploidy in the absence of ORC subunits, or are some regions of the genome more sensitive to ORC depletion (CGH arrays or sequencing of the flow-sorted polyploid cells)?

We thank the reviewer for recognizing the main point of these experiments: to dispel the argument that CDC6 can substitute for ORC1 in the six-subunit ORC (although no one has demonstrated this, the argument is made on the basis of close sequence homology between CDC6 and ORC1). The second point, also appreciated by the reviewer is to show that it is possible to find cells that replicate in the absence or near absence of two ORC subunits.

The mechanistic questions raised are important, and we will address them here:

Is there a delay or impairment of MCM2-7 loading in G1? The hepatocytes in culture are fragile and not immortalized and thus, this issue can be much more easily addressed in the cancer cell lines we have made that are missing several ORC subunits and will do that in a later paper. Note however, the surprising lack of change in MCM2-7 association in cell lines where both ORC2 and ORC5 are deleted (Shibata, 2020, Fig. 7C).

Are some regions of the genome more sensitive to ORC deletion during the polyploidization? We could not find any paper where people have investigated whether the whole genome is uniformly polyploidized in livers. In other words, the baseline conditions in WT livers have not been established. We therefore have postponed experiments to answer this question for a later paper. Note that in unpublished data from mapping SNS-seq origins in WT and ORC deletion cell lines there does not appear to be selective firing of certain origins over others in the deletion cell lines.

Additional points: I didn't understand how the numbers were derived in Table 2. Was there really a 20-fold decrease in nuclear density for female ORC1 and ORC2 double-deletion hepatocytes? The differences in Figure S2 are dramatic, but not 20-fold dramatic.

We measure the relative nuclear density by counting the number of plump nuclei (hepatocytes) per field as described for Fig. 5F and 7F now in the Methods section. The reviewer is correct in that we overestimated the decrease of nuclear density in the female DKO mice by two-fold. The revised calculations suggest that 6 cell divisions occur in the female DKO mice after the ORC proteins have decreased to at least __Significance: __

The strengths of this manuscript are the mouse genetics and the generation of conditional alleles of Orc2 and the rigorous assessment of phenotypes resulting from limiting amounts of specific ORC subunits. It also builds on prior work with ORC1 to rule out Cdc6 complementing the loss of ORC1. The weakness is that it is a very hard task to resolve the fundamental question of how much ORC is enough for replication in cancer cells or hepatocytes. Clearly, there is a marked reduction in specific ORC subunits that is sufficient to impact replication during development and in fibroblasts, but the devil's advocate can always claim limiting levels of ORC remaining in these specialized cells. The significance of the work is that the authors keep improving their conditional alleles (and combining them), thus making it harder and harder (but not impossible) to invoke limiting but sufficient levels of ORC. At this point, the investigators and the field are well-positioned to attempt future functional CRISPR screens to identify other factors that may modulate the response to the loss of ORC subunits. This work will be of interest to the DNA replication, polyploidy, and genome stability communities.

We thank the reviewer for getting the important point of this paper: “making it harder and harder (but not impossible) to invoke limiting but sufficient levels of ORC….” In other words, either ORC is completely dispensable for loading MCM2-7 in certain cancer cell lines and hepatocytes or it is highly catalytic and one molecule of ORC can load a few hundred MCM2-7 doublets so that most origins in the genome are licensed and capable of firing. We are trying the CRISPR screens in cancer cell lines that the reviewer envisages

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

The origin recognition complex (ORC) is an essential loading factor for the replicative Mcm2-7 helicase complex. Despite ORC's critical role in DNA replication, there have been instances where the loss of specific ORC subunits has still seemingly supported DNA replication in cancer cells, endocycling hepatocytes, and Drosophila polyploid cells. Critically, all tested ORC subunits are essential for development and proliferation in normal cells. This presents a challenge, as conditional knockouts need to be generated, and a skeptic can always claim that there were limiting but sufficient ORC levels for helicase loading and replication in polyploid or transformed cells. That being said, the authors have consistently pushed the system to demonstrate replication in the absence or extreme depletion of ORC subunits.

Here, the authors generate conditional ORC2 mutants to counter a potential argument with prior conditional ORC1 mutants that Cdc6 may substitute for ORC1 function based on homology. They also generate a double ORC1 and ORC2 mutant, which is still capable of DNA replication in polyploid hepatocytes. While this manuscript provides significantly more support for the ability of select cells to replicate in the absence or near absence of select ORC subunits, it does not shed light on a potential mechanism. While a mechanistic understanding of how these cells proliferate in the absence or extreme depletion of ORC subunits is outside the scope of the current manuscript, it would have been beneficial to see more functional analyses to help guide the field. For example, is there a delay or impairment in Mcm2-7 loading in G1 (FACs-based loading assay from the Cook Lab (Matson et al., eLife. 2017)) in primary hepatocytes with the ORC2 conditional deletion? Is copy number maintained as cells increase polyploidy in the absence of ORC subunits, or are some regions of the genome more sensitive to ORC depletion (CGH arrays or sequencing of the flow-sorted polyploid cells)?

Additional points: I didn't understand how the numbers were derived in Table 2. Was there really a 20-fold decrease in nuclear density for female ORC1 and ORC2 double-deletion hepatocytes? The differences in Figure S2 are dramatic, but not 20-fold dramatic.

Significance

The strengths of this manuscript are the mouse genetics and the generation of conditional alleles of Orc2 and the rigorous assessment of phenotypes resulting from limiting amounts of specific ORC subunits. It also builds on prior work with ORC1 to rule out Cdc6 complementing the loss of ORC1. The weakness is that it is a very hard task to resolve the fundamental question of how much ORC is enough for replication in cancer cells or hepatocytes. Clearly, there is a marked reduction in specific ORC subunits that is sufficient to impact replication during development and in fibroblasts, but the devil's advocate can always claim limiting levels of ORC remaining in these specialized cells. The significance of the work is that the authors keep improving their conditional alleles (and combining them), thus making it harder and harder (but not impossible) to invoke limiting but sufficient levels of ORC. At this point, the investigators and the field are well-positioned to attempt future functional CRISPR screens to identify other factors that may modulate the response to the loss of ORC subunits. This work will be of interest to the DNA replication, polyploidy, and genome stability communities.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

It has been reported that in the absence of ORC1, liver cells can still endoreduplicate and it has been speculated that this might occur if CDC6 can replace, at least partially, the function of ORC1. Here, authors evaluate if this is also true in the absence of ORC2 and found that ORC2 is required for cell proliferation in mouse hepatocytes but not for endoreduplication. This is also the case after combining the conditional mutations of ORC1 and ORC2. They propose that a mechanism must exist to load sufficient MCM2-7 to support DNA replication in the absence of these two ORC subunits.

Some of the conclusions need further experimental support. The rationale for testing the requirement of ORC2, with or without ORC1, for endoreduplication is valid. However, a key point is that the endoreduplication level seems to be higher in the absence of ORC2 or both ORC1 and ORC2, and this is not properly addressed. Also, mechanistic details on how this could be triggered are absent from this study. As indicated below almost every figure in this manuscript contains weak points (see below).

Major

1. Fig 1G, needs a detailed comment and justification.
2. Fig 2D-F. Is this conclusion applicable to other endoreplicating tissues? Have authors consider to analyze body weight and liver weight measurements after normalization with similar data from a non-affected organ?
3. Fig 3 shows inconsistent results or results that lack proper justification in the text. The 2C peak is missing in Fig 3E (yellow line, positive control). However, 2n nuclei appear in Fig 3F-H. Also, the blue and yellow peaks do not coincide in the flow cytometry profiles, in particular for 8C and 16C.
4. Fig 4. Shorter EdU pulses could be more informative of the actual amount of S-phase cells. Thus, the use of a 2h EdU pulse needs a clear justification.
5. Fig 5. EYFP is cytoplasmic, in contrast with results shown in Fig 4C
6. Fig 6. Results obtained with the double mutant are poorly described.
7. What are the level of other pre-RC components in the mutants used in this study. This could be easily evaluated by Western blotting
8. How do authors justify their claim that a very limited amount of ORC are sufficient to load a vast excess of MCM2-7 hexamers?

Minor

1. The titles of the Results section could be more informative of the main conclusion rather than simply descriptive
2. The Discussion is too long

Significance

The topic is relevant and the hypothesis tested is reasonable, although the conceptual advance is limited (see also below). The major limitation is the absence of mechanistic details addressing the occurrence of extra endoreduplication cycles (compared to controls) in the ORC1 and ORC2 mutants

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

The work by Przanowska et al., sought to understand the role of ORC2 in murine development and further wanted to discover its role in liver endo-reduplication. The overall methods used is sufficient enough to address its role but is not very conclusive based on their overall results and data provided as elaborated in below comments.

Major Comments:

1. The major issue of the paper is how well is ORC2 depleted in perinatal liver (Fig. 2C) and is not very clear from the data as all the western blots are at very low exposure levels and bands are very weak (still weak bands seen). There are good antibodies of ORC2 which can be used for IHC staining and can be used to address the extent of ORC2 depletion.
2. Why in Fig 2C, the M2 mice is showing an equivalent level of ORC2 protein compared to mice M1 with NO CRE expression (compare lane1 and lane5). So, the results are based on one mouse which I do not think is significant enough to come to the conclusion. The authors need to add more data from different mice for statistical significance. Please use IHC to show the depletion of ORC2 protein in the liver sections.
3. As nicely demonstrated in the previous paper by Okano-Uchida et al., 2018 that ORC1 depletion in the liver shows an DNA ploidy effect from 6-week onwards. The authors need to demonstrate in this paper also when the 16N phenotype is observed starting from week1 to 12 months.
4. In the double knockout experiments (ORC1 and ORC2) the authors are not even bothered to demonstrate that how much are both the proteins are actually depleted from the cells, so on the results obtained from these mice experiments are not conclusive or explanatory.

Minor points:

1. Why are scale bars missing in right panel of Fig. 2G, Fig. 6D Supp Fig. 2B KO studies. The authors need to confirm that that all the large nuclei have NO or less significant ORC2 protein through IHC H&E staining.
2. Please explain why is EYFP in Fig. 5G is cytoplasmic compared to Fig 4C (nuclear).
3. Are authors using the same genotype of Alb-Cre mice as shown by Okano-Uchida et al., 2018 as I do not find the reference of Schuler et. al., 2004 (PMID:15282742).

Significance

The article is exactly based on their previous published paper but instead of ORC1, they were interested in dissecting the role of ORC2. Although they have discussed that CDC6 may be involved in replacing ORC1 KO mice to rescue the extensive DNA replication in endoreduplication, but instead of going to hunt the role of CDC6 in endoreduplication they checked the effect of ORC2 which actually lower the overall impact of the paper.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Reviewer 1

R1 Cell profiling is an emerging field with many applications in academia and industry. Finding better representations for heterogeneous cell populations is important and timely. However, unless convinced otherwise after a rebuttal/revision, the contribution of this paper, in our opinion, is mostly conceptual, but in its current form - not yet practical. This manuscript combined two concepts that were previously reported in the context of cell profiling, weakly supervised representations. Our expertise is in computational biology, and specifically applications of machine learning in microscopy.

In our revised manuscript, we have aimed to better clarify the practical contributions of our work by demonstrating the effectiveness of the proposed concepts on real-world datasets. We hope that these revisions and our detailed responses address your concerns and highlight the potential impact of our approach.

R1.1a. CytoSummaryNet is evaluated in comparison to aggregate-average profiling, although previous work has already reported representations that capture heterogeneity and self-supervision independently. To argue that both components of contrastive learning and sets representations are contributing to MoA prediction we believe that a separate evaluation for each component is required. Specifically, the authors can benchmark their previous work to directly evaluate a simpler population representation (PMID: 31064985, ref #13) - we are aware that the authors report a 20% improvement, but this was reported on a separate dataset. The authors can also compare to contrastive learning-based representations that rely on the aggregate (average) profile to assess and quantify the contribution of the sets representation.

We agree that evaluating the individual contributions of the contrastive learning framework and single-cell data usage is important for understanding CytoSummaryNet's performance gains.

To assess the impact of the contrastive formulation independently, we applied CytoSummaryNet to averaged profiles from the cpg0004 dataset. This isolated the effect of contrastive learning by eliminating single-cell heterogeneity. The experiment yielded a 32% relative improvement in mechanism of action retrieval, compared to the 68% gain achieved with single-cell data. These findings suggest that while the contrastive formulation contributes significantly to CytoSummaryNet's performance, leveraging single-cell information is crucial for maximizing its effectiveness. We have added a discussion of this experiment to the Results section:

“We conducted an experiment to determine whether the improvements in mechanism of action retrieval were due solely to CytoSummaryNet's contrastive formulation or also influenced by the incorporation of single-cell data. We applied the CytoSummaryNet framework to pre-processed average profiles from the 10 μM dose point data of Batch 1 (cpg0004 dataset). This approach isolated the effect of the contrastive architecture by eliminating single-cell data variability. We adjusted the experimental setup by reducing the learning rate by a factor of 100, acknowledging the reduced task complexity. All other parameters remained as described in earlier experiments.

This method yielded a less pronounced but still substantial improvement in mechanism of action retrieval, with an increase of 0.010 (32% enhancement - Table 1). However, this improvement was not as high as when the model processed single-cell level data (68% as noted above). These findings suggest that while CytoSummaryNet's contrastive formulation contributes to performance improvements, the integration of single-cell data plays a critical role in maximizing the efficacy of mechanism of action retrieval.”

We don't believe comparing with PMID: 31064985 is useful: while the study showcased the usefulness of modeling heterogeneity using second-order statistics, its methodology is limited in scalability due to the computational burden of computing pairwise similarities for all perturbations, particularly in large datasets. Additionally, the study's reliance on similarity network fusion, while expedient, introduces complexity and inefficiency. We contend that this comparison does not align with our objective of testing the effectiveness of heterogeneity in isolation, as it primarily focuses on capturing second and first-order information. Thus, we do not consider this study a suitable baseline for comparison.

R1.1b. The evaluation metric of mAP improvement in percentage is misleading, because a tiny improvement for a MoA prediction can lead to huge improvement in percentage, while a much larger improvement in MoA prediction can lead to a small improvement in percentage. For example, in Fig. 4, MEK inhibitor mAP improvement of ~0.35 is measured as ~50% improvement, while a much smaller mAP improvement can have the same effect near the origins (i.e., very poor MoA prediction).

We agree that relying solely on percentage improvements can be misleading, especially when small absolute changes result in large percentage differences.

However, we would like to clarify two key points regarding our reporting of percentage improvements:

• We calculate the percentage improvement by first computing the average mAP across all compounds for both CytoSummaryNet and average profiling, and then comparing these averages. This approach is less susceptible to the influence of outlier improvements compared to calculating the average of individual compound percentage improvements.
• We report percentage improvements alongside their corresponding absolute improvements. For example, the mAP improvement for Stain4 (test set) is reported as 0.052 (60%). To further clarify this point, we have updated the caption of Table 1 to explicitly state how the percentage improvements are calculated:

“The improvements are calculated as mAP(CytoSummaryNet)-mAP(average profiling). The percentage improvements are calculated as (mAP(CytoSummaryNet)-mAP(average profiling))/mAP(average profiling).”

R1.1b. (Subjective) visual assessment of this figure does not show a convincing contribution of CytoSummaryNet representations of the average profiling on the test set (3.33 uM). This issue might also be relevant for the task of replicate retrieval. All in all, the mAP improvement reported in Table 1 and throughout the manuscript (including the Abstract), is not a proper evaluation metric for CytoSummaryNet contribution. We suggest reporting the following evaluations:

1. Visualizing the results of cpg0001 (Figs. 1-3) similarly to cpg0004 (Fig. 4), i.e., plotting the matched mAP for CytoSummaryNet vs. average profile.

2. In Table 1, we suggest referring to the change in the number of predictable MoAs (MoAs that pass a mAP threshold) rather than the improvement in percentages. Another option is showing a graph of the predictability, with the X axis representing a threshold and Y-axis showing the number of MoAs passing it. For example see (PMID: 36344834, Fig. 2B) and (PMID: 37031208, Fig. 2A), both papers included contributions from the corresponding author of this manuscript.

Regarding the suggestion to visualize the results for compound group cpg0001 similarly to cpg0004, unfortunately, this is not feasible due to the differences in data splitting between the two datasets. In cpg0001, an MoA might have one compound in the training set and another in the test or validation set. Reporting a single value per MoA would require combining these splits, which could be misleading as it would conflate performance across different data subsets.

However, we appreciate the suggestion to represent the number of predictable MoAs that surpass a certain mAP threshold, as it provides another intuitive measure of performance. To address this, we have created a graph that visualizes the predictability of MoAs across various thresholds, similar to the examples provided in the referenced papers (PMID: 36344834, Figure 2B and PMID: 37031208, Figure 2A). This graph, with the x-axis depicting the threshold and the y-axis showing the number of MoAs meeting the criterion, has been added to Supplementary Material K.

R1.1c.i. "a subset of 18 compounds were designated as validation compounds" - 5 cross-validations of 18 compounds can make the evaluation complete. This can also enhance statistical power in figures 1-3.

We appreciate your suggestion and acknowledge the potential benefits of employing cross-validation, particularly in enhancing statistical power. While we understand the merit of cross-validation for evaluating model performance and generalization to unseen data, we believe the results as presented already highlight the generalization characterics of our methods.

Specifically, (the new) Figure 3 demonstrates the model's improvement over average profiling in both training and validation plates, supporting its ability to generalize to unseen compounds (but not to unseen plates).

While cross-validation could potentially enhance our analysis, retraining five new models solely for different validation set results may not substantially alter our conclusions, given the observed trends in Suppl Figure A1 and (the new) Figure 4, both of which show results across multiple stain sets (but a single train-test-validation split).

---

R1.1c.ii. Clarify if the MoA results for cpg0001 are drawn from compounds from both the training and the validation datasets. If so, describe how the results differ between the sets in text and graphs.

We confirm that the Mechanism of Action (MoA) retrieval results for cpg0001 are derived from all available compounds. It's important to note that the training and validation dataset split for the replicate retrieval task is different from the MoA prediction task. For replicate retrieval, we train using all available compounds and validate on a held-out set (see Figure 2). For MoA prediction, we train using the replicate retrieval task as the objective on all available compounds but validate using MoA retrieval, which is a distinct task. We have added a brief clarification in the main text to highlight the distinction between these tasks and how validation is performed for each:

“We next addressed a more challenging task: predicting the mechanism of action class for each compound at the individual well level, rather than simply matching replicates of the exact same compound (Figure 5). It's also important to note that mechanism of action matching is a downstream task on which CytoSummaryNet is not explicitly trained. Consequently, improvements observed on the training and validation plates are more meaningful in this context, unlike in the previous task where only improvements on the test plate were meaningful. For similar reasons, we calculate the mechanism of action retrieval performance on all available compounds, combining both the training and validation sets. This approach is acceptable because we calculate the score on so-called "sister compounds" only—that is, different compounds that have the same mechanism of action annotation. This ensures there is no overlap between the mechanism of action retrieval task and the training task, maintaining the integrity of our evaluation. ”

R1.1c.iii. "Mechanism of action retrieval is evaluated by quantifying a profile's ability to retrieve the profile of other compounds with the same annotated mechanism of action.". It was unclear to us if the evaluation of mAP for MoA identification can include finding replicates of the same compound. That is, whether finding a close replicate of the same compound would be included in the AP calculation. This would provide CytoSummaryNet with an inherent advantage as this is the task it is trained to do. We assume that this was not the case (and thus should be more clearly articulated), but if it was - results need to be re-evaluated excluding same-compound replicates.

The evaluation excludes replicate wells of the same compound and only considers wells of other compounds with the same MoA. This methodology ensures that the model's performance on the MoA prediction task is not inflated by its ability to find replicates of the same compound, which is the objective of the replicate retrieval task. Please see the explanation we have added to the main text in our response to R1.1c.ii. Additionally, we have updated the Methods section to clearly describe this evaluation procedure:

“Mechanism of action retrieval is evaluated by quantifying a profile’s ability to retrieve the profile of different compounds with the same annotated mechanism of action.”

---

---

__R1.2a. __The description of Stain2-5 was not clear for us at first (and second) read. The information is there, but more details will greatly enhance the reader's ability to follow. One suggestion is explicitly stating that these "stains" partitioning was already defined in ref 26. Another suggestion is laying out explicitly a concrete example on the differences between two of these stains. We believe highlighting the differences between stains will strengthen the claim of the paper, emphasizing the difficulty of generalizing to the out-of-distribution stain.

We appreciate your feedback on the clarity of the Stain2-5 dataset descriptions; we certainly struggled to balance detail and concepts in describing these. We have made the following changes:

• Explicitly mentioned that the partitioning of the Stain experiments was defined in https://pubmed.ncbi.nlm.nih.gov/37344608/: “The partitioning of the Stain experiments have been defined and explained previously [21].”
• Moved an improved version of (now) Figure 2 from the Methods section to the main text to help visually explain how the stratification is done early on.
• Added a new section in the Experimental Setup: Diversity of stain sets, which includes a concrete example highlighting the differences between Stain2, and Stain5 to emphasize the diversity in experimental setups within the same dataset: “Stain2-5 comprise a series of experiments which were conducted sequentially to optimize the experimental conditions for image-based cell profiling. These experiments gradually converged on the most optimal set of conditions; however, within each experiment, there were significant variations in the assay across plates. To illustrate the diversity in experimental setups within the dataset, we will highlight the differences between Stain2 and Stain5.

Stain2 encompasses a wide range of nine different experimental protocols, employing various imaging techniques such as Widefield and Confocal microscopy, as well as specialized conditions like multiplane imaging and specific stains like MitoTracker Orange. This subset also includes plates acquired with strong pixel binning instead of default imaging and plates with varying concentrations of dyes like Hoechst. As a result, Stain2 exhibits greater variance in the experimental conditions across different plates compared to Stain5.

In contrast, Stain5, the last experiment in the series, follows a more systematic approach, consistently using either confocal or default imaging across three well-defined conditions. Each condition in Stain5 utilizes a lower cell density of 1,000 cells per well compared to Stain2's 2,500 cells per well. Being the final experiment in the series, Stain5 had the least variance in experimental conditions.

For training the models, we typically select the data containing the most variance to capture the broadest range of experimental variation. Therefore, we chose Stain2-4 for training, as they represented the majority of the data and captured the most experimental variation. We reserved Stain5 for testing to evaluate the model's ability to generalize to new experimental conditions with less variance.

All StainX experiments were acquired in different passes, which may introduce additional batch effects.”

These changes aim to provide a clearer understanding of the dataset's complexity and the challenges associated with generalizing to out-of-distribution data.

R1.2b. What does each data point in Figures 1-3 represent? Is it the average mAP for the 18 validation compounds, using different seeds for model training? Why not visualize the data similarly to Fig. 4 so the improvement per compound can be clearly seen?

The data points in (the new) Figures 3,4,5 represent the average mAP for each plate, calculated by first computing the mAP for each compound and then averaging across compounds to obtain the average mAP per plate. We have updated the figure captions to clarify this:

"... (each data point is the average mAP of a plate) ..."

While visualizing the mAP per compound, similar to (the new) Figure 6 for cpg0004, could provide insights into compound-level improvements, it would require creating numerous additional figures or one complex figure to adequately represent all the stratifications we are analyzing (plate, compound, Stain subset). By averaging the data per plate across different stratifications, we aim to provide a clearer and more comprehensible overview of the trends and improvements while allowing us to draw conclusions about generalization.

Please note: this comment is related to the comment R1.1b (Subjective)

R1.2.c [On the topic of enhancing clarity and readability:] Justification and interpretation of the evaluation metrics.

Please refer to our response to comment R1.1b, where we have addressed your concerns regarding the justification and interpretation of the evaluation metrics.

R1.2d. Explicitly mentioning the number of MoAs for each datasets and statistics of number of compounds per MoA (e.g., average\median, min, max).

We have added the following to the Experimental Setup: Data section:

“A subset of the data was used for evaluating the mechanism of action retrieval task, focusing exclusively on compounds that belong to the same mechanism class. The Stain plates contained 47 unique mechanisms of action, with each compound replicated four times. Four mechanisms had only a single compound; the four mechanisms (and corresponding compounds) were excluded, resulting in 43 unique mechanisms used for evaluation. In the LINCS dataset, there were 1436 different mechanisms, but only 661 were used for evaluation because the remaining had only one compound.”

R1.2e. The data split in general is not easily understood. Figure 8 is somewhat helpful, however in our view, it can be improved to enhance understanding of the different splits. Specifically, the training and validation compounds need to be embedded and highlighted within the figure.

Thank you for highlighting this. We have completely revised the figure, now Figure 2 which we hope more clearly conveys the data split strategy.

Please note: this comment is related to the comment R1.2a.

---

---

---

---

R1.3a. Why was stain 5 used for the test, rather than the other stains?

Stain2-5 were part of a series of experiments aimed at optimizing the experimental conditions for image-based cell profiling using Cell Painting. These experiments were conducted sequentially, gradually converging on the most optimal set of conditions. However, within each experiment, there were significant variations in the assay across plates, with earlier iterations (Stain2-4) having more variance in the experimental conditions compared to Stain5. As Stain5 was the last experiment in the series and consisted of only three different conditions, it had the least variance. For training the models, we typically select the data containing the most variance to capture the broadest range of experimental variation. Therefore, Stain2-4 were chosen for training, while Stain5 was reserved for testing to evaluate the model's ability to generalize to new experimental conditions with less variance.

We have now clarified this in the Experimental Setup: Diversity of stain sets section. Please see our response to comment R1.2a. for the full citation.

R1.3b How were the 18 validation compounds selected?

20% of the compounds (n=18) were randomly selected and designated as validation compounds, with the remaining compounds assigned to the training set. We have now clarified this in the Results section:

“Additionally, 20% of the compounds (n=18) were randomly selected and designated as validation compounds, with the remaining compounds assigned to the training set (Supplementary Material H).”

R1.3c. For cpg0004, no justification for the specific doses selected (10uM - train, 3.33 uM - test) for the analysis in Figure 4. Why was the data split for the two dosages? For example, why not perform 5-fold cross validation on the compounds (e.g., of the highest dose)?

We chose to use the 10 μM dose point as the training set because we expected this higher dosage to consist of stronger profiles with more variance than lower dose points, making it more suitable for training a model. We decided to use a separate test set at a different dose (3.33 μM) to assess the model's ability to generalize to new dosages. While cross-validation on the highest dose could also be informative, our approach aimed to balance the evaluation of the model's generalization capability with its ability to capture biologically relevant patterns across different dosages.

This explanation has been added to the text:

“We chose the 10 μM dose point for training because we expected this high dosage to produce stronger profiles with more variance than lower dose points, making it more suitable for model training.”

“The multiple dose points in this dataset allowed us to create a separate hold-out test set using the 3.33 μM dose point data. This approach aimed to evaluate the model's performance on data with potentially weaker profiles and less variance, providing insights into its robustness and ability to capture biologically relevant patterns across dosages. While cross-validation on the 10 μM dose could also be informative, focusing on lower dose points offers a more challenging test of the model's capacity to generalize beyond its training conditions, although we do note that all compounds’ phenotypes would likely have been present in the 10 μM training dataset, given the compounds tested are the same in both.”

R1.3d. A more detailed explanation on the logic behind using a training stain to test MoA retrieval will help readers appreciate these results. In our first read of this manuscript we did not grasp that, we did in a second read, but spoon-feeding your readers will help.

This comment is related to the rationale behind training on one task and testing on another, which is addressed in our responses to comments R1.1.cii and R1.1.ciii.

R1.4 Assessment of interpretability is always tricky. But in this case, the authors can directly confirm their interpretation that the CytoSummaryNet representation prioritizes large uncrowded cells, by explicitly selecting these cells, and using their average profile re

We progressively filtered out cells based on a quantile threshold for Cells_AreaShape features (MeanRadius, MaximumRadius, MedianRadius, and Area), which were identified as important in our interpretability analysis, and then computed average profiles using the remaining cells before determining the replicate retrieval mAP. In the exclusion experiment, we gradually left out cells as the threshold increased, while in the inclusion experiment, we progressively included larger cells from left to right.

The results show that using only the largest cells does not significantly increase the performance. Instead, it is more important to include the large cells rather than only including small cells. The mAP saturates after a threshold of around 0.4, indicating that larger cells define the profile the most, and once enough cells are included to outweigh the smaller cell features, the profile does not change significantly by including even larger cells.

These findings support our interpretation that CytoSummaryNet prioritizes large, uncrowded cells. While this approach could potentially be used as a general outlier removal strategy for cell profiling, further investigation is needed to assess its robustness and generalizability across different datasets and experimental conditions.

We have created Supplementary Material L to report these findings and we additionally highlight them in the Results:

“To further validate CytoSummaryNet's prioritization of large, uncrowded cells, we progressively filtered cells based on Cells_AreaShape features and observed the impact on replicate retrieval mAP (Supplementary Material L). The results support our interpretation and highlight the key role of larger cells in profile strength.”

__R1.5. __Placing this work in context of other weakly supervised representations. Previous papers used weakly supervised labels of proteins / experimental perturbations (e.g., compounds) to improve image-derived representations, but were not discussed in this context. These include PMID: 35879608, https://www.biorxiv.org/content/10.1101/2022.08.12.503783v2 (from the same research groups and can also be benchmarked in this context), https://pubs.rsc.org/en/content/articlelanding/2023/dd/d3dd00060e , and https://www.biorxiv.org/content/10.1101/2023.02.24.529975v1. We believe that a discussion explicitly referencing these papers in this specific context is important.

While these studies provide valuable insights into improving cell population profiles using representation learning, our work focuses specifically on the question of single-cell aggregation methods. We chose to use classical features for our comparisons because they are the current standard in the field. This approach allows us to directly assess the performance of our method in the context of the most widely used feature extraction pipeline in practice. However, we see the value in incorporating them in future work and have mentioned them in the Discussion:

“Recent studies exploring image-derived representations using self-supervised and self-supervised learning [35][36] could inspire future research on using learned embeddings instead of classical features to enhance model-aggregated profiles.”

R1.minor1. "Because the improved results could stem from prioritizing certain features over others during aggregation, we investigated each cell's importance during CytoSummaryNet aggregation by calculating a relevance score for each" - what is the relevance score? Would be helpful to provide some intuition in the Results.

We have included more explanation of the relevance score in the Results section, following the explanation of sensitivity analysis (SA) and critical point analysis (CPA):

“SA evaluates the model's predictions by analyzing the partial derivatives in a localized context, while CPA identifies the input cells with the most significant contribution to the model's output. The relevance scores of SA and CPA are min-max normalized per well and then combined by addition. The combination of the two is again min-max normalized, resulting in the SA and CPA combined relevance score (see Methods for details).”

R1.minor2. Figure 1:

1. Colors of the two methods too similar
2. The dots are too close. It will be more easily interpreted if they were further apart.
3. What do the dots stand for?
4. We recommend considering moving this figure to the supp. material (the most important part of it is the results on the test set and it appears in Fig.2).

1. We chose a lighter and darker version of the same color as a theme to simplify visualization, as this theme is used throughout (the new) Figures 3,4,5.
2. We agree; we have now redrawn the figure to fix this.
3. Each data point is the average mAP of a plate. Please see our answer for R1.2b as well.
4. We believe that (the new) Figures 3,4,5 serve distinct purposes in testing various generalization hypotheses. We have added the following text to emphasize that the first figures are specifically about generalization hypothesis testing: “We first investigated CytoSummaryNet’s capacity to generalize to out-of-distribution data: unseen compounds, unseen experimental protocols, and unseen batches. The results of these investigations are visualized in Figures 3, 4, and 5, respectively.”

R1.minor3 Figure 4: It is somewhat misleading to look at the training MoAs and validation MoAs embedded together in the same graph. We recommend showing only the test MoAs (train MoAs can move to SI).

We addressed this comment in R1.1c.ii. To reiterate briefly, there are no training, validation, or test MoAs because these are not used as labels during the training process. There is an option to split them based on training and validation compounds, which is addressed in R1.1c.ii.

---

R1.minor4 Figure 5

1. Why only Stain3? What happens if we look at Stains 2,3 and 4 together? Stain 5?

2. Should validation compounds and training compounds be analyzed separately?

3. Subfigure (d): it is expected that the data will be classified by compound labels as it is the training task, but for this to be persuasive I would like to see this separately on the training compounds first and then and more importantly on the validation compounds.

4. For subfigures (b) and (d): it appears there are not enough colors for d, which makes it partially not understandable. For example, the pink label in (d) shows a single compound which appears to represent two different MoAs. This is probably not the case, and it has two different compounds, but it cannot be inferred when they are represented by the same color.

5. For the Subfigure (e) - only 1 circle looks justified (in the top left). And for that one, is it not a case of an outlier plate that would perhaps need to be removed from analysis? Is it not good that such a plate will be identified?

We have addressed this point in the text, stating that the results are similar for Stain2 and Stain4. Stain5 represents an out-of-distribution subset because of a very different set of experimental conditions (see Experimental Setup: Diversity of stain sets). To improve clarity, we have revised the figure caption to reiterate this information:

“... Stain2 and Stain4 yielded similar results (data not shown). …”

1. For replicate retrieval, analyzing validation and training compounds separately is appropriate. However, this is not the case for MoA retrieval, as discussed in our responses to R1.1c.ii and R1.1c.i.
2. We have created the requested plot (below) but ultimately decided not to include it in the manuscript because we believe that (the new) Figures 3 and 4 are more effective for making quantitative comparative claims.

[Please see the full revision document for the figures]

Top: training compounds (validation compounds grayed out); not all compounds are listed in the legend.

*Bottom: validation compounds (training compounds grayed out). *

Left: average profiling; Right: CytoSummaryNet

1. We agree with your observation and have addressed this issue by labeling the center mass as a single class (gray) and highlighting only the outstanding pairs in color. Please refer to the updated figure and our response to R3.6 for more details.

2. In the updated figure, we have revised the figure caption to focus solely on the annotation of same mechanism of action profile clusters, as indicated by the green ellipses. The annotation of isolated plate clusters has been removed (Figures 7e and 7f) to maintain consistency and avoid potential confusion. Despite being an outlier for Stain3, the plate (BR00115134bin1) clusters with Stain4 plates (Supplementary Figure F1, green annotated square inside the yellow annotated square), indicating it is not merely a noisy outlier and can provide insights into the out-of-sample performance of our model.

R1.minor5a. Discussion: "perhaps in part due to its correction of batch effects" - is this statement based on Fig. 5F - we are not convinced.

We appreciate the reviewer's scrutiny regarding our statement about batch effect correction. Upon reevaluation, we agree that this claim was not adequately substantiated by empirical data. We quantified the batch effects using comparison mean average precision for both average profiles and CytoSummaryNet profiles, and the statistical analysis revealed no significant difference between these profiles in terms of batch effect correction. Therefore, we have removed this theoretical argument from the manuscript entirely to ensure that all claims are strongly supported by the data presented.

R1.minor5b. "Overall, these results improve upon the ~20% gains we previously observed using covariance features" - this is not the same dataset so it is hard to reach conclusions - perhaps compare performance directly on the same data?

We have now explicitly clarified this is a different dataset. Please see our response to R1.1a for why a direct comparison was not performed. The following clarification can be found in the Discussion:

“These results improve upon the ~20% gains previously observed using covariance features [13] albeit on a different dataset, and importantly, CytoSummaryNet effectively overcomes the challenge of recomputation after training, making it easier to use.”

Reviewer 2

R2.1 The authors present a well-developed and useful algorithm. The technical motivation and validation are very carefully and clearly explained, and their work is potentially useful to a varied audience.

That said, I think the authors could do a better job, especially in the figures, of putting the algorithm in context for an audience that is unfamiliar with the cell painting assay. (a) For example, a figure towards the beginning of the paper with example images might help to set the stage. (b) Similarly a schematic of the algorithm earlier in the paper would provide a graphical overview. (c) For the sake of a biologically inclined audience, I would consider labeling the images in the caption by cell type and label.

Thank you for your valuable suggestions on improving the accessibility of our figures for readers unfamiliar with the Cell Painting assay. We have made the following changes to address your comments:

1. and b. To provide visual context and a graphical overview of the algorithm, we have moved the original Figure 7 to Figure 1. This figure now includes example images that help readers new to the Cell Painting assay.
2. We have added relevant details to the example images in (the new) Figure 1

   R2.2 The interpretability results were intriguing. The authors might consider further validating these interpretations by removing weakly informative cells from the dataset and retraining. Are the cells so uninformative that the algorithm does better without them, or are they just less informative than other cells?

Please see our responses to R1.4 and R3.0

R2.3 As far as I can tell, the authors only oblique state whether the code associated with the manuscript is openly available. Posting the code is needed for reproducibility. I would provide not only a github, but a doi linked to the code, or some other permanent link.

We have now added a Code Availability and Data Availability section, clearing stating that the code and data associated with the manuscript are openly available.

R2.4 Incorporating biological heterogeneity into machine-learning driven problems is a critical research question. Replacing means/modes and such with a machine learning framework, the authors have identified a problem with potentially wide significance. The application to cell painting and related assays is of broad enough significance for many journals, However, the authors could further broaden the significance by commenting on other possible cell biology applications. What other applications might the algorithm be particularly suited for? Are there any possible roadblocks to wider use. What sorts of data has the code been tested on so far?

We have added the following paragraph to discuss the broader applicability of CytoSummaryNet:

“The architecture of CytoSummaryNet holds significant potential for broader applications beyond image-based cell profiling, accommodating tabular, permutation-invariant data and enhancing downstream task performance when applied to processed population-level profiles. Its versatility makes it valuable for any omics measurements where downstream tasks depend on measuring similarity between profiles. Future research could also explore CytoSummaryNet's applicability to genetic perturbations, expanding its utility in functional genomics.”

Reviewer 3

R3.0 The authors have done a commendable job discussing the method, demonstrating its potential to outperform current models in profiling cell-based features. The work is of considerable significance and interest to a wide field of researchers working on the understanding of cell heterogeneity's impact on various biological phenomena and practical studies in pharmacology.

One aspect that would further enhance the value of this work is an exploration of the method's separation power across different modes of action. For instance, it would be interesting to ascertain if the method's performance varies when dealing with actions that primarily affect size, those that affect marker expression, or compounds that significantly diminish cell numbers.

Thank you for encouraging comments!

We have added the following to Results: Relevance scores reveal CytoSummaryNet's preference for large, isolated cells:

“Statistical t-tests were conducted to identify the features that most effectively differentiate mechanisms of action from negative controls in average profiles, focusing on the three mechanisms of action where CytoSummaryNet demonstrates the most significant improvement and the three mechanisms where it shows the least. Consistent with our hypothesis that CytoSummaryNet emphasizes larger, more sparse cells, the important features for the CytoSummaryNet-improved mechanisms of action (Supplementary Material I1) often involve the radial distribution for the mitochondria and RNA channels. These metrics capture the fraction of those stains near the edge of the cell versus concentric rings towards the nucleus, which are more readily detectable in larger cells compared to small, rounded cells.

In contrast, the important features for mechanisms of action not improved by CytoSummaryNet (Supplementary Material I) predominantly include correlation metrics between brightfield and various fluorescent channels, capturing spatial relationships between cellular components. Some of these mechanisms of action included compounds that were not individually distinguishable from negative controls, and CytoSummaryNet did not overcome the lack of phenotype in these cases. This suggests that while CytoSummaryNet excels in identifying certain cellular features, its effectiveness is limited when dealing with mechanisms of action that do not exhibit pronounced phenotypic changes.”

We have also added supplementary material to support (I. Relevant features for CytoSummaryNet improvement).

R3.0 Another test on datasets that are not concerned with chemical compounds, but rather genetic perturbations would greatly increase the reach of the method into the functional genomics community and beyond. This additional analysis could provide valuable insights into the versatility and applicability of the proposed method.

We agree that testing the method's behavior on genetic perturbations would be interesting and could provide insights into its versatility. However, the efficacy of the methodology may vary depending on the specific properties of different genetic perturbation types.

For example, the penetrance of phenotypes may differ between genetic and chemical perturbations. In some experimental setups, a selection agent ensures that nearly all cells receive a genetic perturbation (though not all may express a phenotype due to heterogeneity or varying levels of the target protein). Other experiments may omit such an agent. Additionally, different patterns might be observed in various classes of reagents, such as overexpression, CRISPR-Cas9 knockdown (CRISPRn), CRISPR-interference (CRISPRi), and CRISPR-activation (CRISPRa).

We believe that selecting a single experiment with one of these technologies would not adequately address the question of versatility. Instead, we propose future studies that may conclusively assess the method's performance across a variety of genetic perturbation types. This would provide a more comprehensive understanding of CytoSummaryNet's applicability in functional genomics and beyond. We have update the Discussion section to reflect this:

“Future research could also explore CytoSummaryNet's applicability to genetic perturbations, expanding its utility in functional genomics.”

R3.1. The datasets were stratified based on plates and compounds. It would be beneficial to clarify the basis for data stratification applied for compounds. Was the data sampled based on structural or functional similarity of compounds? If not, what can be expected from the model if trained and validated using structurally or functionally diverse and non-diverse compounds?

Thank you for raising the important question of data stratification based on compound similarity. In our study, the data stratification was performed by randomly sampling the compounds, without considering their structural or functional similarity.

This approach may limit the generalizability of the learned representations to new structural or functional classes not captured in the training set. Consequently, the current methodology may not fully characterize the model’s performance across diverse compound structures.

In future work, it would be valuable to explore the impact of compound diversity on model performance by stratifying data based on structural or functional similarity and comparing the results to our current random stratification approach to more thoroughly characterize the learned representations.

R3.2. Is the method prioritizing a particular biological reaction of cells toward common chemical compounds, such as mitotic failure? Could this be oncology-specific, or is there more utility to it in other datasets?

Our analysis of CytoSummaryNet's performance in (the new) Figure 6 reveals a strong improvement in MoAs targeting cancer-related pathways, such as MEK, HSP, MDM, dehydrogenase, and purine antagonist inhibitors. These MoAs share a common focus on cellular proliferation, survival, and metabolic processes, which are key characteristics of cancer cells.

Given the composition of the cpg0004 dataset, which contains 1,258 unique MoAs with only 28 annotated as oncology-related, the likelihood of randomly selecting five oncology-related MoAs that show strong improvement is extremely low. This suggests that the observed prioritization is not due to chance.

Furthermore, the prioritization cannot be solely attributed to the frequency of oncology-related MoAs in the dataset. Other prevalent disease areas, such as neurology/psychiatry, infectious disease, and cardiology, do not exhibit similar improvements despite having higher MoA counts.

While these findings indicate a potential prioritization of oncology-related MoAs by CytoSummaryNet, further research is necessary to fully understand the extent and implications of this bias. Future work should involve conducting similar analyses across other disease areas and cell types to assess the method's broader utility and identify areas for refinement and application. However, given the speculative nature of these observations, we have chosen not to update the manuscript to discuss this potential bias at this time.

R3.3 Figures 1 and 2 demonstrate that the CytoSummaryNet profiles outperform average-aggregated profiles. However, the average profiling results seem more consistent when compared to CytoSummaryNet profiling. What further conditions or approaches can help improve CytoSummaryNet profiling results to be more consistent?

The observed variability in CytoSummaryNet's performance is primarily due to the intentional technical variance in our datasets, where each plate tested different staining protocol variations. It's important to note that this level of technical variance is not typical in standard cell profiling experiments. In practice, the variance across plates would be much lower. We want to emphasize that while a model capable of generalizing across diverse experimental conditions might seem ideal, it may not be as practically useful in real-world scenarios. This is because such non-uniform conditions are uncommon in typical cell profiling experiments. In normal experimental settings, where technical variance is more controlled, we expect CytoSummaryNet's performance to be more consistent.

R3.4 Can the poor performance on unseen data (in the case of stain 5) be overcome? If yes, how? If no, why not?

We believe that the poor performance on unseen data, such as Stain 5, can be overcome depending on the nature of the unseen data. As shown in Figure 4 (panel 3), the model improves upon average profiling for unseen data when the experimental conditions are similar to the training set.

The issue lies in the different experimental conditions. As explained in our response to R3.3, this could be addressed by including these experimental conditions in the training dataset. As long as CytoSummaryNet is trained (seen) and tested (unseen) on data generated under similar experimental conditions, we are confident that it will improve or perform as well as average profiling.

It's important to note that the issue of generalization to vastly different experimental conditions was considered out of scope for this paper. The main focus is to introduce a new method that improves upon average profiling and can be readily used within a consistent experimental setup.

R3.5 It needs to be mentioned how the feature data used for CytoSummaryNet profiling was normalized before training the model. What would be the impact of feature data normalization before model training? Would the model still outperform if the skewed feature data is normalized using square or log transformation before model training?

We have clarified in the manuscript that we standardized the feature data on a plate-by-plate basis to achieve zero mean and unit variance across all cells per feature within each plate. We have added the following statement to improve clarity:

“The data used to compute the average profiles and train the model were standardized at the plate-level, ensuring that all cell features across the plate had a zero mean and unit variance. The negative control wells were then removed from all plates."

We chose standardization over transformations like squaring or logging to maintain a balanced scale across features while preserving the biological and morphological information inherent in the data. While transformations can reduce skewness and are useful for data spanning several orders of magnitude, they might distort biological relevance by compressing or expanding data ranges in ways that could obscure important cellular variations.

Regarding the potential impact of square or log transformations on skewed feature data, these methods could improve the model's learning efficiency by making the feature distribution more symmetrical. However, the suitability and effectiveness of these techniques would depend on the specific data characteristics and the model architecture.

Although not explored in this study, investigating various normalization techniques could be a valuable direction for future research to assess their impact on the performance and adaptability of CytoSummaryNet across diverse datasets and experimental setups.

R3.6. In Figure 5 b and c, MoAs often seem to be represented by singular compounds and thus, the test (MoA prediction) is very similar to the training (compound ID). Given this context, a discussion about the extent this presents a circular argument supported by stats on the compound library used for training and testing would be beneficial.

Clusters in (the new) Figure 7 that contain only replicates of a single compound would not yield an improved performance on the MoA task unless they also include replicates of other compounds sharing the same MoA in close proximity. Please see our response to R1.1c.iii. for details. To improve visual clarity and avoid misinterpretation, we have recomputed the colors for (the new) Figure 7 and grayed out overlapping points.

R3.7 Can you estimate the minimum amount of supervision (fuzzy/sparse labels, often present in mislabeled compound libraries with dirty compounds and polypharmacology being present) that is needed for it to be efficiently trained?

It's important to note that the metadata used by the model is only based on identifying replicates of the same compound. Mechanism of action (MoA) annotations, which can be erroneous due to dirty compounds, polypharmacology, and incomplete information, are not used in training at all. MoA annotations are only used in our evaluation, specifically for calculating the mAP for MoA retrieval.

We have successfully trained CytoSummaryNet on 72 unique compounds with 4 replicates each. This is the current empirical minimum, but it is possible that the model could be trained effectively with even fewer compounds or replicates.

Determining the absolute minimum amount of supervision required for efficient training would require further experimentation and analysis. Factors such as data quality, feature dimensionality, and model complexity could influence the required level of supervision.

R3.minor1 Figure 5: The x-axis and y-axis tick values are too small, and image resolution/size needs to be increased.

We have made the following changes to address the concerns:

• Increased the image resolution and size to improve clarity and readability.
• Removed the x-axis and y-axis tick values, as they do not provide meaningful information in the context of UMAP visualizations. We believe these modifications enhance the visual presentation of the data and make it easier for readers to interpret the results.

R3.minor2 The methods applied to optimize hyperparameters in supplementary data need to be included.

We added the following to Supplementary Material D:

“We used the Weights & Biases (WandB) sweep suite in combination with the BOHB (Bayesian Optimization and HyperBand) algorithm for hyperparameter sweeps. The BOHB algorithm [47] combines Bayesian optimization with bandit-based strategies to efficiently find optimal hyperparameters.

Additionally Table D1 provides an overview of all tunable hyperparameters and their chosen values based on a BOHB hyperparameter optimization.”

R3.minor3 Figure 5(c, d): The names of compound 2 and Compound 5 need to be included in the labels.

These compounds were obtained from external companies and are proprietary, necessitating their anonymization in our study. This has now been added in the caption of (the new) Figure 7:

“Note that Compound2 and Compound5 are intentionally anonymized.”

R3.minor4 Table C1: Plate descriptions need to be included.

*Table C1: The training, validation, and test set stratification for Stain2, Stain3, Stain4, and Stain5. Five training, four validation, and three test plates are used for Stain2, Stain3, and Stain4. Stain5 contains six test set plates only. *

__Stain2 __

Stain3

Stain4

Stain5

Training plates

Test plates

BR00113818

BR00115128

BR00116627

BR00120532

BR00113820

BR00115125highexp

BR00116631

BR00120270

BR00112202

BR00115133highexp

BR00116625

BR00120536

BR00112197binned

BR00115131

BR00116630highexp

BR00120530

BR00112198

BR00115134

200922_015124-Vhighexp

BR00120526

Validation plates

BR00120274

BR00112197standard

BR00115129

BR00116628highexp

BR00112197repeat

BR00115133

BR00116629highexp

BR00112204

BR00115128highexp

BR00116627highexp

BR00112201

BR00115127

BR00116629

Test plates

BR00112199

BR00115134bin1

200922_044247-Vbin1

BR00113819

BR00115134multiplane

200922_015124-V

BR00113821

BR00115126highexp

BR00116633bin1

We have added a reference to the metadata file in the description of Table C1: https://github.com/carpenter-singh-lab/2023_Cimini_NatureProtocols/blob/main/JUMPExperimentMasterTable.csv

R3.minor5 Figure F1: Does the green box (stain 3) also involve training on plates from stain 4 (BR00116630highexp) and 5 (BR00120530) mentioned in Table C1? Please check the figure once again for possible errors.

We have carefully re-examined Figure F1 and Table C1 to ensure their accuracy and consistency. Upon double-checking, we can confirm that the figure is indeed correct. We intentionally omitted the training and validation plates from Figure F1 to maintain clarity and readability, as including them resulted in a cluttered and difficult-to-interpret figure.

Regarding the specific plates mentioned:

• BR00116630highexp (Stain4) is used for training, as correctly stated in Table C1. This plate is considered an outlier within the Stain4 dataset and happens to cluster with the Stain3 plates in Figure F1.
• BR00120530 (Stain5) is part of the test set only and correctly falls within the Stain5 cluster in Figure F1. To improve the clarity of the training, validation, and test split in Table C1, we have added a color scheme that visually distinguishes the different data subsets. This should make it easier for readers to understand the distribution of plates across the various splits.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Summary:

In the manuscript by Van Dijk et al., a novel deep learning technique is introduced that aims to summarize informative cells from heterogeneous populations in image-based profiling. This technique is based on a network that utilizes contrastive learning with a multiple-instance learning framework, a significant departure from existing average-based cell profiling models.

The authors have done a commendable job discussing the method, demonstrating its potential to outperform current models in profiling cell-based features. The work is of considerable significance and interest to a wide field of researchers working on the understanding of cell heterogeneity's impact on various biological phenomena and practical studies in pharmacology.

One aspect that would further enhance the value of this work is an exploration of the method's separation power across different modes of action. For instance, it would be interesting to ascertain if the method's performance varies when dealing with actions that primarily affect size, those that affect marker expression, or compounds that significantly diminish cell numbers. Another test on datasets that are not concerned with chemical compounds, but rather genetic perturbations would greatly increase the reach of the method into the functional genomics community and beyond. This additional analysis could provide valuable insights into the versatility and applicability of the proposed method. Please find my detailed comments below:

Major Comments:

1. The datasets were stratified based on plates and compounds. It would be beneficial to clarify the basis for data stratification applied for compounds. Was the data sampled based on structural or functional similarity of compounds? If not, what can be expected from the model if trained and validated using structurally or functionally diverse and non-diverse compounds?
2. Is the method prioritizing a particular biological reaction of cells toward common chemical compounds, such as mitotic failure? Could this be oncology-specific, or is there more utility to it in other datasets?
3. Figures 1 and 2 demonstrate that the CytoSummaryNet profiles outperform average-aggregated profiles. However, the average profiling results seem more consistent when compared to CytoSummaryNet profiling. What further conditions or approaches can help improve CytoSummaryNet profiling results to be more consistent?
4. Can the poor performance on unseen data (in the case of stain 5) be overcome? If yes, how? If no, why not?
5. It needs to be mentioned how the feature data used for CytoSummaryNet profiling was normalized before training the model. What would be the impact of feature data normalization before model training? Would the model still outperform if the skewed feature data is normalized using square or log transformation before model training?
6. In Figure 5 b and c, MoAs often seem to be represented by singular compounds and thus, the test (MoA prediction) is very similar to the training (compound ID). Given this context, a discussion about the extent this presents a circular argument supported by stats on the compound library used for training and testing would be beneficial.
7. Can you estimate the minimum amount of supervision (fuzzy/sparse labels, often present in mislabeled compound libraries with dirty compounds and polypharmacology being present) that is needed for it to be efficiently trained?

Minor Comments:

1. Figure 5: The x-axis and y-axis tick values are too small, and image resolution/size needs to be increased.
2. The methods applied to optimize hyperparameters in supplementary data need to be included.
3. Figure 5(c, d): The names of compound 2 and Compound 5 need to be included in the labels.
4. Table C1: Plate descriptions need to be included.
5. Figure F1: Does the green box (stain 3) also involve training on plates from stain 4 (BR00116630highexp) and 5 (BR00120530) mentioned in Table C1? Please check the figure once again for possible errors.

Significance

This work presents a significant move forward in the ways we deal with cellular heterogeneity in all single-cell assays. Though the model in its current state has trouble extrapolating to out of distribution data, I am confident that it provides a considerable step forward in the process of extracting "informative" knowledge from data in the form of optimized profiles.

The optimization is yet based on optimizing a similarity metric for group assignments, I will be interesting to see if other objectives could be more effective in developing aggregation techniques.

The work is of considerable significance and interest to a wide field of researchers working on the understanding of cell heterogeneity's impact on various biological phenomena and practical studies in pharmacology.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

The authors present a well-developed and useful algorithm. The technical motivation and validation are very carefully and clearly explained, and their work is potentially useful to a varied audience.

That said, I think the authors could do a better job, especially in the figures, of putting the algorithm in context for an audience that is unfamiliar with the cell painting assay. For example, a figure towards the beginning of the paper with example images might help to set the stage. Similarly a schematic of the algorithm earlier in the paper would provide a graphical overview. For the sake of a biologically inclined audience, I would consider labeling the images in the caption by cell type and label.

The interpretability results were intriguing. The authors might consider further validating these interpretations by removing weakly informative cells from the dataset and retraining. Are the cells so uninformative that the algorithm does better without them, or are they just less informative than other cells?

As far as I can tell, the authors only oblique state whether the code associated with the manuscript is openly available. Posting the code is needed for reproducibility. I would provide not only a github, but a doi linked to the code, or some other permanent link.

Significance

Incorporating biological heterogeneity into machine-learning driven problems is a critical research question. Replacing means/modes and such with a machine learning framework, the authors have identified a problem with potentially wide significance. The application to cell painting and related assays is of broad enough significance for many journals, However, the authors could further broaden the significance by commenting on other possible cell biology applications. What other applications might the algorithm be particularly suited for? Are there any possible roadblocks to wider use. What sorts of data has the code been tested on so far?

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary:

Provide a short summary of the findings and key conclusions (including methodology and model system(s) where appropriate).

Cell (non-genetic) heterogeneity is an important concept in cell biology, but there are currently only a few studies that try to incorporate this information to represent cell populations in the field of high-content image-based phenotypic profiling. The authors present CytoSummaryNet, a machine learning approach for representing heterogeneous cell populations, and apply it to a high-content image-based Cell Painting dataset to demonstrate superior performance in predicting a compound's mechanism of action (MoA), in relation to the average profile representation. CytoSummaryNet relies on Cell Profiler morphological features and simultaneous optimization of two components, both novel in the cell profiling field: (i) learning representations using weakly supervised contrastive learning according to the perturbation identifications (i.e., the compound), (ii) using a representation method called Deep Sets to create permutation-invariant population representations. The authors evaluate their representation on the task of replicate retrieval and of MoA retrieval using the public dataset cpg0001 (and cpg0004), and report superior performance in respect to the average-aggregated profiles for the experimental protocols and compounds seen on training (that do not generalize to out-of-distribution compounds + experimental protocols). By interpreting which cells were most important for the MoA model predictions, the authors propose that their representation prioritizes large uncrowded cells.

Major comments:

The strength of the manuscript is the new idea of combining contrastive learning and sets representations for better representation of heterogeneous cell populations. However, we are not convinced that the conclusion that this representation improves MoA prediction is fully supported by the data, for several reasons.

1. Evaluations. This is the most critical point in our review.

a. CytoSummaryNet is evaluated in comparison to aggregate-average profiling, although previous work has already reported representations that capture heterogeneity and self-supervision independently. To argue that both components of contrastive learning and sets representations are contributing to MoA prediction we believe that a separate evaluation for each component is required. Specifically, the authors can benchmark their previous work to directly evaluate a simpler population representation (PMID: 31064985, ref #13) - we are aware that the authors report a 20% improvement, but this was reported on a separate dataset. The authors can also compare to contrastive learning-based representations that rely on the aggregate (average) profile to assess and quantify the contribution of the sets representation.

b. The evaluation metric of mAP improvement in percentage is misleading, because a tiny improvement for a MoA prediction can lead to huge improvement in percentage, while a much larger improvement in MoA prediction can lead to a small improvement in percentage. For example, in Fig. 4, MEK inhibitor mAP improvement of ~0.35 is measured as ~50% improvement, while a much smaller mAP improvement can have the same effect near the origins (i.e., very poor MoA prediction). (Subjective) visual assessment of this figure does not show a convincing contribution of CytoSummaryNet representations of the average profiling on the test set (3.33 uM). This issue might also be relevant for the task of replicate retrieval. All in all, the mAP improvement reported in Table 1 and throughout the manuscript (including the Abstract), is not a proper evaluation metric for CytoSummaryNet contribution. We suggest reporting the following evaluations:

i. Visualizing the results of cpg0001 (Figs. 1-3) similarly to cpg0004 (Fig. 4), i.e., plotting the matched mAP for CytoSummaryNet vs. average profile. ii. In Table 1, we suggest referring to the change in the number of predictable MoAs (MoAs that pass a mAP threshold) rather than the improvement in percentages. Another option is showing a graph of the predictability, with the X axis representing a threshold and Y-axis showing the number of MoAs passing it. For example see (PMID: 36344834, Fig. 2B) and (PMID: 37031208, Fig. 2A), both papers included contributions from the corresponding author of this manuscript.

c. Additional evaluation-related concerns were: i. "a subset of 18 compounds were designated as validation compounds" - 5 cross-validations of 18 compounds can make the evaluation complete. This can also enhance statistical power in figures 1-3.

ii. Clarify if the MoA results for cpg0001 are drawn from compounds from both the training and the validation datasets. If so, describe how the results differ between the sets in text and graphs.

iii. "Mechanism of action retrieval is evaluated by quantifying a profile's ability to retrieve the profile of other compounds with the same annotated mechanism of action.". It was unclear to us if the evaluation of mAP for MoA identification can include finding replicates of the same compound. That is, whether finding a close replicate of the same compound would be included in the AP calculation. This would provide CytoSummaryNet with an inherent advantage as this is the task it is trained to do. We assume that this was not the case (and thus should be more clearly articulated), but if it was - results need to be re-evaluated excluding same-compound replicates. 2. Lack of clarity in the description of the data and evaluation. While the concept of constructive learning + sets representation is elegant and intuitive, we found it very hard to follow the technical aspects of data and performance evaluation, even after digging in deep into the Methods. Figuring out these important aspects required us for vast investment in time, more than the vast majority of manuscripts we reviewed in the last couple of years. It is highly recommended that the authors provide more details to make this manuscript easier to follow. Some examples include:

a. The description of Stain2-5 was not clear for us at first (and second) read. The information is there, but more details will greatly enhance the reader's ability to follow. One suggestion is explicitly stating that these "stains" partitioning was already defined in ref 26. Another suggestion is laying out explicitly a concrete example on the differences between two of these stains. We believe highlighting the differences between stains will strengthen the claim of the paper, emphasizing the difficulty of generalizing to the out-of-distribution stain.

b. What does each data point in Figures 1-3 represent? Is it the average mAP for the 18 validation compounds, using different seeds for model training? Why not visualize the data similarly to Fig. 4 so the improvement per compound can be clearly seen?

c. Justification and interpretation of the evaluation metrics.

d. Explicitly mentioning the number of MoAs for each datasets and statistics of number of compounds per MoA (e.g., average\median, min, max).

e. The data split in general is not easily understood. Figure 8 is somewhat helpful, however in our view, it can be improved to enhance understanding of the different splits. Specifically, the training and validation compounds need to be embedded and highlighted within the figure. 3. Lack of justification of design choices. There were multiple design choices that were not justified. This adds to the lack of clarity and makes it harder to evaluate the merits of the new method. For example:

a. Why was stain 5 used for the test, rather than the other stains?

b. How were the 18 validation compounds selected?

c. For cpg0004, no justification for the specific doses selected (10uM - train, 3.33 uM - test) for the analysis in Figure 4. Why was the data split for the two dosages? For example, why not perform 5-fold cross validation on the compounds (e.g., of the highest dose)?

d. A more detailed explanation on the logic behind using a training stain to test MoA retrieval will help readers appreciate these results. In our first read of this manuscript we did not grasp that, we did in a second read, but spoon-feeding your readers will help. 4. The interpretability analysis is speculative. Assessment of interpretability is always tricky. But in this case, the authors can directly confirm their interpretation that the CytoSummaryNet representation prioritizes large uncrowded cells, by explicitly selecting these cells, and using their average profile representation to demonstrate that they achieve improved results. If this works, it could be applied as a general outlier removal strategy for cell profiling.

a. "We identified the likely mechanism by which the learned CytoSummaryNet aggregates cells: the most salient cells are generally larger and more isolated from other cells, while the least salient cells appear to be smaller and more crowded, and tend to contain spots of high-intensity pixels (whether dying, debris or in some stage of cell division)." - doesn't such a mechanism should generalize to out-of-distribution data? 5. Placing this work in context of other weakly supervised representations. Previous papers used weakly supervised labels of proteins / experimental perturbations (e.g., compounds) to improve image-derived representations, but were not discussed in this context. These include PMID: 35879608, https://www.biorxiv.org/content/10.1101/2022.08.12.503783v2 (from the same research groups and can also be benchmarked in this context),https://pubs.rsc.org/en/content/articlelanding/2023/dd/d3dd00060e , and https://www.biorxiv.org/content/10.1101/2023.02.24.529975v1. We believe that a discussion explicitly referencing these papers in this specific context is important.

Minor comments:

In our opinion, evaluation of the training task using the training data (Figure 1) is not contributing to the manuscript and could be excluded. Also we feel that the subjectiveness of the UMAP analysis (Figure 5) is not contributing much and could be excluded, especially if the authors follow our suggestions regarding quantification. Of course, this is up to the authors to decide (along with most of the other suggestions below).

Suggested clarifications:

1. "Because the improved results could stem from prioritizing certain features over others during aggregation, we investigated each cell's importance during CytoSummaryNet aggregation by calculating a relevance score for each" - what is the relevance score? Would be helpful to provide some intuition in the Results.
2. Figure 1:

a. Colors of the two methods too similar

b. The dots are too close. It will be more easily interpreted if they were further apart.

c. What do the dots stand for?

d. We recommend considering moving this figure to the supp. material (the most important part of it is the results on the test set and it appears in Fig.2). 3. Figure 4: It is somewhat misleading to look at the training MoAs and validation MoAs embedded together in the same graph. We recommend showing only the test MoAs (train MoAs can move to SI). 4. Figure 5

a. Why only Stain3? What happens if we look at Stains 2,3 and 4 together? Stain 5?

b. Should validation compounds and training compounds be analyzed separately?

c. Subfigure (d): it is expected that the data will be classified by compound labels as it is the training task, but for this to be persuasive I would like to see this separately on the training compounds first and then and more importantly on the validation compounds.

d. For subfigures (b) and (d): it appears there are not enough colors for d, which makes it partially not understandable. For example, the pink label in (d) shows a single compound which appears to represent two different MoAs. This is probably not the case, and it has two different compounds, but it cannot be inferred when they are represented by the same color.

e. For the Subfigure (e) - only 1 circle looks justified (in the top left). And for that one, is it not a case of an outlier plate that would perhaps need to be removed from analysis? Is it not good that such a plate will be identified? 5. Discussion:

a. "perhaps in part due to its correction of batch effects" - is this statement based on Fig. 5F - we are not convinced.

b. "Overall, these results improve upon the ~20% gains we previously observed using covariance features" - this is not the same dataset so it is hard to reach conclusions - perhaps compare performance directly on the same data?

Significance

Cell profiling is an emerging field with many applications in academia and industry. Finding better representations for heterogeneous cell populations is important and timely. However, unless convinced otherwise after a rebuttal/revision, the contribution of this paper, in our opinion, is mostly conceptual, but in its current form - not yet practical. This manuscript combined two concepts that were previously reported in the context of cell profiling, weakly supervised representations. Our expertise is in computational biology, and specifically applications of machine learning in microscopy.

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

The authors do not wish to post a response at this time. This is because this is not the submission of the revised version, which we have not completed yet. This is a preliminary revision together with a revision plan instead.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

In this manuscript, Singh et al. demonstrate that infection of D. melanogaster flies with B. bassiana fungus induces neurodegeneration via Toll/wek/sarm signalling. It is already known that fungal infection can be associated with neurodegeneration, but the exact mechanism is unclear. The authors demonstrate that the fungus enters the brain, causes hallmark symptoms of neurodegeneration, and requires Toll-1, Wek, and Sarm in order to do so. This is an important step forward as it demonstrates specific genes in the fly immune pathways that are involved in fungus-induced neurodegeneration, which could be informative for infections in humans. Overall, the manuscript is thorough and well-written and the conclusions are broadly supported. A few mostly minor comments and questions are below, which could mostly be addressed by including additional details in the methods or discussion. The only major comments would be 1) that the control fly genotypes used in experiments were not always the most ideal controls (eg, compared WT genotypes to RNAi against a gene of interest; ideally would be RNAi against a control gene compared to RNAi against a gene of interest), and 2) negative controls of fluorescence microscopy imaging were not always included. It would be important to address these through clarification in the figures/methods, and/or discussion of the potential caveats, even though it is likely the conclusions would still hold. Notably, these comments are relatively easily addressed through edits in the text.

Major comments:

• For fluorescence imaging, were negative controls included (no infection or no gene expression etc.) for all stains (as with Figure 1H)? If so, it would help to include representative images as supplemental figures. Also, for all positive samples, was presence of the fungus noted in all samples?
• Figure 4: Here, it appears that the control fly genotypes are wildtype vs an RNAi line (similar for some other figures/assays as well, but using this one as an example). The best control would be RNAi against a control gene compared to RNAi against a gene of interest, rather than just a control WT genotype with no RNAi compared to RNAi against a gene of interest. This should be included as a caveat in the discussion since the experiments do not all account for the effect of RNAi (or other gene expression) on the phenotypes regardless of the gene.

Minor comments:

• It would help to have line numbers throughout
• Figure 1- what are the arrows in panels D-G?
• Methods: A few details are unclear:
  • Was only one fly sex used or were both used for the various assays? If both were used, were they statistically assessed for differences? Sex is only mentioned in a couple of the methods sections.
  • How old were the flies at the start of the experiments? A few experiments noted age, but it was not clear for all
  • For longevity, was the fungal culture ever replaced during the experiment?
  • For the climbing assay when the flies were initially flipped, how much time was there between flips?
• Figures 2A, 2D, 2E, 3E, & 3H: If multiple replicates or samples are represented in the data, it would help to be able to see the data points underlying these bars. If so, please add them to the graphs to see the spread of data points.
• Figure 3F- what do arrows indicate?
• It is interesting that Wek-RNAi with infection not only rescues loss caused by the infection alone, but also increases YFP cells beyond the uninfected controls (Figure 5C). The same is true with toll-1 RNAi (Figure 4C). Why might this be?
• It would be ideal if data underlying data points and full statistical models and outputs could be included through a public repository such as Dryad. This would be ideal for full assessment of statistical approaches

Very Minor comments:

• Check italicizing throughout- missed a few "Drosophila" or "B. bassiana" in main text or figures
• Looks like no space between C. and elegans in C. elegans in a few cases
• Word missing: "No effect was seen after three days exposure to B. bassiana, but seven days exposure impaired climbing"... seven days of exposure?
• Toll-1 misspelled pg 6 last paragraph

Referee Cross-Commenting

Regarding the major comments, I agree with Reviewer 1 that more thorough proof of spores entering the brain (and what proportion of exposed flies this happens to) would be beneficial. I also agree with Reviewer 2 that a rescue experiment for the climbing assay and my earlier suggestion for more controls in the microscopy could help address this concern, at least in part. Other responses or experiments may also be appropriate to address some of the major concerns- maybe additional assay(s) of brain function other than climbing?

Reviewer 1 also brought up the point that flies with advanced infection were used for the experiments- it would be helpful to know if earlier time points were ever checked for BBB damage, loss of brain cells, or presence of fungus etc. This would clarify if the same phenotypes are present in flies that die early, along with other concerns from Reviewer 1.

However, whether directly or indirectly, several later figures show loss of brain cells with infection followed by rescue with RNAi against genes of interest. This does lend support to the conclusions that fungal infection negatively impacts brain cells and the fungus requires these host genes to do so.

Other concerns Reviewer 2 and I raised about the fly genetic controls being unclear should also be addressed. What is the full genotype of the flies in each case? What is considered "+" in each case? Were these driver background strains, WT (like Oregon R), or RNAi against control genes (best controls)?

Significance

General assessment: The manuscript by Singh et al. is a thorough investigation into the fungus-host interactions in the brain, demonstrating that the common insect fungal pathogen B. bassiana requires the host genes Toll, wek, and sarm to induce negative phenotypes in the brain. The strengths are in the multi-pronged approaches that use several independent techniques (fly behavior assays, gene expression, microscopy, etc.) and multiple genes, conducted with many replicates, that all show clear and consistent trends supporting the conclusions of the authors. The weaknesses include some cases where controls are either not completely clear or not the most ideal controls. This weakness could be addressed with either edits to the text, where appropriate, or addition of supplemental figures. However, the conclusions are still broadly supported.

Advance: Although it is known that fungal infections can impair brain function, it is not fully understood how this happens. This manuscript identifies Toll-associated molecules that are required for fungus-mediated neurodegeneration, which is a critical first step to understanding the process and for future development of therapies.

Audience: This finding would be of broad interest to scientists in immunology, microbiology, neuroscience, and other areas.

Expertise of reviewer: Drosophila, fly genetics, invertebrate immunology, insect-fungal interactions

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary

The authors describe a role for Toll signaling in detrimental neuronal loss associated with B. bassiana fungal infection in Drosophila melanogaster model. They show that this effect is mediated by wek/sarm as silencing either of them prevents neuronal loss after the infection. Similar results are obtained with Toll-1 RNAi, suggesting that the response is dependent on the activation of Toll signaling by B. bassiana. The study is well executed,main conclusions are backed up by the data presented and experiments are conducted with adequate numbers of replications and individuals. Below I give some comments that I think would help in further improving the manuscript.

Major comments

As the initial experiments (including the effect on survival and climbing assay) have been performed using OR/CantonS, it would be interesting to investigate if the same is seen with a more similar background to that what is used in the genetic experiments. In addition, I'd suggest an experiment to see if the Toll (or wek/sarm) RNAi in the brain rescues the climbing defect caused by the fungal infection.

It is somewhat unclear what are the controls in the genetic experiments. For example, in Figure 2, the control is UAS-TrpA1/+. Does this mean that the UAS-TrpA1 flies have been crossed to something (like the driver background strain) or used as it is? In Figure 4, controls are ">+". Again, are MyD88>histoneYFP;tubulinGal80ts flies crossed to something (in this case, maybe the w1118 background of the KK library RNAi strains) or used as homozygous? And same for the subsequent figures. I'd ask the authors to clarify these points in the manuscript.

Could the authors please explain why they opted for MyD88-GAL4 in the experiments in Figures 5-7? What is the overall expression pattern of MyD88-GAL4? Is there a possibility that some of the effects seen could arise from the Toll/sarm/wek knockdown elsewhere in the fly? How do the flies survive the infection with Toll knockdown in MyD88-epressing cells (expressed at least in all immunogenic tissues)? A bit more explanation would clarify the situation.

Minor comments

Page 3: Full species name should be given here (Drosophila melanogaster)

A short description of the FM4-64 dye (what it stains etc) would be useful for the readers unfamiliar with it.

Page 6: Please explain shortly why TrpA1 overexpression was used to activate the neurons.

Figure 2E: What is the genotype of the flies? mtk is lacking statistics

Page 8: third row refers to Figure 2 but should be Figure 3.

Although antibody stainings are performed using "standard methods", a short overview on the process should be presented also in the current manuscript. Also, I imagine fungal spores are all over the flies retrieved from the infection chamber. I'd like to know (and this could also be described in the materials) how the flies (and the brains) were treated/washed prior to preparing brains for immunostaining and imaging?

Some typos and inconsistencies at various places. For example, at some occasions B. bassiana written without a space in between "B" and "bassiana" and not in italics (both in figures and in text); on page 5, first line: "mimicked" misspelled

Referee Cross-Commenting

As the fungal infiltration into the brain is central to the conclusions made in the manuscript, I agree that care should be taken in making this argument solid. I believe this can be achieved adding controls as reviewer #3 suggests together with additional experiment(s) verifying that Toll/wek/sarm in the brain is mediating the neuronal loss caused by the fungal infection (rescue experiments). Of note, I wonder if, similarly to mammalian macrophages, hemocytes could be responsible for delivering the fungal cells into the brain?

I agree with the reviewer #1 that the climbing defect could be because of multiple reasons other than the fungal spores in the brain causing neuronal loss (for instance flies being generally weak at this point, ). However, the authors do show convincingly that there is neuronal loss in fungal-infected flies.

Significance

Fungal infections are understudied in any research model considering the threat they pose to humans and other animals alike. Due to the high conservation of the signaling components studied here, the results provide a good basis for future research, extending to mammalian models. I think these results will be of interest to a wider audience because of the reasons stated above.

My fields of expertise are Drosophila melanogaster, innate immunity, cell-mediated immunity, blood cell homeostasis

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Singh et al. report that, after exposure to the entomopathogenic fungus Beauveria bassiana, the Drosophila adults impaired fly locomotion and died within two weeks. During which time, the authors designed experiments and showed the decline of brain cells via a Toll-1/Wek/Sarm pathway, mimicking the neurodegenerative diseases in humans in association with fungal infections. Providing that the rather solid genetic evidence was shown for the pathway in mediating fly brain cell losses, critical issues of experiment design/setup and conclusion validity were concerned.

Specific comments:

The fungus-exposed flies died within two weeks were largely typical. However, it was unclear how those flies could be uniformly contaminated with fungal spores in the infection chamber shown in Fig. 1A, by landing on fungal "carpet"? It is publicly known that entomopathogenic fungi (EPF) like B. bassiana infect insects via spore germination on cuticle and then penetration of cuticles by fungal hyphae/mycelia (e.g., Trends Microbiol. 2024. 32, 302-316).

It is typical that EPF killed and mycosed insects within 5-14 days after topical infection by immersion in or spraying spore suspensions, or dusting on the sporulated plates. Fungal spores can be ingested by insects, largely those with chewing mouthparts. However, fungal spores can barely survive the highly-alkaline foreguts. It is questionable that flies could ingest spores and spores "infiltrated" the brains.

Regarding the detection of fungal cells in fly brains, on the one hand, the authors argued that detection of fungal SPORES in fly brain THREE days post exposure (page 5) by infiltration. It would be impossible that, even fungus could breach the blood brain barrier (BBB), it might be the fungal hyphae/mycelia but not the spores. One the other hand, the authors provided the evidence of the damaged BBB SEVEN days post exposure, a few days LATER than the detection of fungal spores in brains (THREE days) post treatment mentioned above. Did "spore infiltration" (even impossible) occur before BBB damage?

The authors stated that "by day seven more than half of the flies had died" (Fig. 1B). It is questionable therefore that the "other half" of the diseased and dying insects were used for the following experiments. There would be no wonder that the climbing of these diseased and dying flies was impaired, however, which could be due to muscle damage, hemocyte number decline and reduction of energy production etc. apart from brain cell loss. The brain function of dying animals could be compromised by multiple direct or indirect factors.

Issue of Fig. 2D labelling.

Referee Cross-Commenting

I agree with that Reviewers 2 and 3 that rather solid evidence of fly brain loss was shown in this work, however, at most in association with exposure to fungal cultures (volatiles could not be excluded etc.). "Spores" entry into fly brains were suspicious or impossible. If the dying flies had been used for these neurological experiments, the reliability of conclusions would be highly concerned.

Significance

Since there are critical concerns of experiment designs/setup in this work, it is questionable that fly brain cell loss was caused by fungal entry into brains.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Manuscript number: RC-2024-02491

Corresponding author(s): Gilbert, Vassart

1. General Statements [optional]

We thank referees 1 and 2 for their in-depth analysis of our manuscript. They see interest in our study, with questions to be answered. Referee 3 is essentially negative, considering that there is nothing new ("novel finding is missing"). We respectfully disagree with him/her, comforted by the opinion of referee 2 that "the authors developed a protocol to reproducibly generate fetal-like spheroids from adult tissue is an important advancement in the field and ... the manuscript should attract a significant amount of attention in the intestinal field" and we provide evidence in our answers that he/she did not read the manuscript with the same attention as referees 1 and 2 (see in particular answer to his/her question 5).

Here is a summary of the main reason why we consider that our study represents valuable new information in the field of intestinal regeneration.

It is based on the serendipitous observation that dissociation of adult intestinal tissue by collagenase generates stably replatable spheroids upon culture in matrigel. Surprisingly and contrary to canonical EDTA-generated intestinal organoids and fetal spheroids, these spheroids are not traced in Rosa26Tomato mice harboring a VilCre transgene, despite expressing robustly endogenous Villin. Our interpretation is that adult intestinal spheroids originate from a cell lineage, distinct from the main developmental intestinal lineage, in which the VilCre transgene is unexpectedly not expressed, probaly due to the absence of cis regulatory sequences required for expression in this lineage.

Adult spheroid transcriptome shares a gene signature with the YAP/TAZ signature commonly expressed in models of intestinal regeneration. This led us to look for VilCre negative crypts in the regenerating intestine of Lgr5/DTR mice in which Lgr5-positive stem cells have been ablated by diphtheria toxin. Numerous VilCre negative clones were observed, identifying a novel lineage of stem cells implicated in intestinal regeneration.

FACS purification and scRNAseq analysis of the rare VilCre negative cells present at homeostasis identified a population of cells with characteristics of quiescent stem cells.

In sum, we believe that our study demonstrates the existence of a hitherto undescribed stem cell lineage involved in intestinal regeneration. It points to the existence of a hierarchical model of intestinal regeneration in addition to the well-accepted plasticity model.

2. Description of the planned revisions

See section 3 below.

3. Description of the revisions that have already been incorporated in the transferred manuscript

Here is a point-by-point reply to the queries of the three referees, with indication of the revisions introduced in the manuscript.

Reviewer #1 (Evidence, reproducibility and clarity (Required)):

• *In this manuscript, Marefati et al report an Lgr5-independent lineage in the regenerating intestine using in vitro organoids and in vivo injury-coupled lineage tracing model. In organoids, collagenase/dispase dissociated resulted in "immortal spheroids" that maintain a cystic and undifferentiated phenotype in the absence of standard growth factors (Rspondin/Noggin/EGF). Bulk RNAseq of spheroids demonstrates downregulation of classical CBC signatures and upregulation of fetal spheroid, mesenchymal, inflammation and regenerative signatures. In mice, Villin-Cre lineage tracing revealed some Villin- negative progenies that lack reporter tracing throughout crypt-villus ribbons after injury.

*The authors proposed that there is Lgr5-independent population support the regenerative response upon CBC depletion. A major caveat of this study is the identification of this population is based on absence of VilCre expression. *

We respectfully disagree. It is precisely this characteristic that makes the interest of our study. Whereas mosaicism of transgene expression is widespread and usually of little significance, our study shows that the rare VilCre-negative cells in the intestinal epithelium are not randomly showing this phenotype: they give specifically birth to what we call adult spheroids and regenerating crypts, which cannot be due to chance. The absence of VilCre expression allows tracing these cells from the zygote stage of the various VilCre/Ros26 reporter mice. We have modified our text to emphasize this point.

*It is surprising that there is no characterisation of Lgr5 expression throughout the manuscript whilst claiming of a Lgr5- independent lineage. *

We understand the perplexity of the referee not to see direct Lgr5 expression data in our manuscript, given our title. However, our point is that it is the cells at the origin of adult spheroids and the regenerating crypts we have identified that are Lgr5-negative, not the spheroids or the regenerated crypts themselves. Those are downstream offspring that may, and indeed have, gained some Lgr5 expression (e.g. figure 3F). We believe that our data showing that VilCre-negative spheroids are not traced in Lgr5-CreERT2/Rosa reporter mice convincingly demonstrate absence of Lgr5 expression in the cells at the origin of adult spheroids (figure 4G). We think that this experiment is better evidence than attempts to show absence of two markers (Tom and Lgr5) in the rare "white" cells present in the epithelium. Regarding the Lgr5 status of cells at the origin of the regenerating "white" crypts that we have identified, the early appearance of these crypts following ablation of CBC (i.e. Lgr5+ve) cells is a strong argument that they originate from Lgr5-negative cells. Regarding the scRNAseq experiment, Lgr5 transcripts are notoriously low and difficult to measure reliably in CBCs (Haber et al 2017). However, blowing up the pertinent regions of the merged UMAP allows showing some Lgr5 transcripts in clusters 5,6 and none in cluster 1 of figure 8GH. Given the very low level of detection, we had chosen not to include these data in the manuscript, but we hope they may help answer the point of the referee (see portion of UMAP below, with Olfm4 as a control, together with the corresponding violin plot). Several markers that gave significant signals in the CBC cluster (Smoc2, Axin2, Slc12a2) were virtually undetectable in the Olfm4-low /Tom-negative cluster of our scRNAseq data (figure 8I) supporting our conclusion.

Although the research question is potentially interesting, the concept of epithelial reprogramming upon injury is well documented in the field. The data generated in this manuscript also seem to be preliminary and lack of detailed characterisation. Below are specific comments.

We do not question the existence of epithelial reprogramming upon injury. We believe our data show, in addition to this well demonstrated phenomenon, the existence of rare cells traced by absence of VilCre expression that are at the origin of a developmental cell lineage distinct from Lgr5+ stem cells and also implicated in regeneration.

• Expression of Lgr5 should be properly characterised throughout the manuscript in both organoid models and injury-induced regeneration in vivo.
• *

See above for a detailed answer to this point.

• An important question is the origin of these "Lgr5-independent" adult spheroids. They look and appear like fetal organoids, which could be induced by injury (e.g. upon collagenase/dispase dissociation). Have the authors tried to culture fetal spheroids in BCM over extensive period of time? Do they behave the same? This would be a great way to directly compare the collagenase/dispase-derived organoids with fetal origin. * *Fetal spheroids require ENR for survival and die in BCM. We have chosen to illustrate this point in Fig2A by showing that, contrary to adult spheroid, they die even when only Rspondin is missing.

• Fig 1C, Why is the replating spheroid culture time different between mesenchymal cells and conditioned medium? We took the earliest time showing convincingly the return to the organoid phenotype. This timing difference does not modify the conclusion that EDTA organoids becoming spheroid-like when exposed to factors originating from mesenchymal cells revert to the organoid phenotype when returned to ENR medium without mesenchymal influence.

• *It is unclear how the bulk RNA-seq data in Fig. 3 were compared. How long were the adult organoids and spheroids cultured for (how many passages)? Were they culture in the same condition of were they in ENR vs BCM? * Both EDTA organoids and spheroids displaying a stable phenotype were used in this experiment. Organoids were collected at passage 4, day 5; spheroids were collected at passage passage 9 day 3.

As stated in the legend to the figure: "...to allow pertinent comparison spheroids and organoids were cultured in the same ENR-containing medium...".

These are important information to consider when interpreting the results. For instance, are Ptgs1 & Ptgs2 expression in adult spheroids the same in ENR vs BCM? Are the gene signatures (regenerative, fetal and YAP) changed in adult spheroids culturing in ENR vs BCM?

We did compare bulk RNAseq of EDTA organoids to ENR-cultured spheroids, short term (passage 6, day 6) BCM-cultured spheroids and long term BCM-cultured (passage 26, day 6) spheroids. To avoid overloading the manuscript these data were not shown in the original manuscript. In summary the BCM-cultured spheroids display a similar phenotype as those cultured in ENR, but with further de-differentiation. See in revision plan folder the results for PTGS, some differentiation markers and fetal regenerative markers including YAP induced genes.

We have included a brief description of these data in the new version of the manuscript and added an additional supplementary file (Suppl table 2) presenting the whole data set.

• It is stated: "In agreement with their aptitude to grow indefinitely, adult spheroids express a set of upregulated genes overlapping significantly with an "adult tissue stem cell module" [159/721 genes; q value 2.11 e-94) (Fig.S2F)].". What is the definition of "indefinitely"? Are they referring to the Fig 1B where spheroid were passaged to P10? The authors should avoid the term "indefinitely" but use a more specific time scale, e.g. passages, months etc.

We agree that the term indefinitely should be avoided, as it is vague. We have introduced the maximum number of passages during which we have maintained the stable spheroid phenotype (26 passages). Also worth noting, the spheroids could be frozen and cultured repeatedly over many months.

SuppFig 3D: Row Z-Score is missing the "e" in Score.

Corrected

• Fig 4E: Figure legend says QNRQ instead of CNRQ. Corrected

• Fig 4G: The brightfield image of adult spheroids 5 days after 3x TAM injections doesn't look like a spheroid. It seems to be differentiating. True, the choice was not the best as the spheroids started to darken. When further replated, however, the offspring of these spheroids showing a clear phenotype remain negative 30 days after tamoxifen administration as shown on the figure. We are sorry, but for reasons explained in section 4 below, we cannot redo the experiment to get a better picture.

• Fig 4: Most mouse model data are missing the number of mice & their respective age used for organoid isolation. We have introduced these data in the legend.

• *Fig 4A-D, H-G: How was fluorescent signal of organoids quantified? *

The settings of fluo imaging or time of LacZ staining were the same for organoids and spheroid pictures. This has been added to the material and methods of the figure and an example is shown below for Rosa26Tomato.

*How many images? * 2 per animal per condition.

*Were there equal numbers of organoids? *

No, see number of total elements counted added to the figure

This all needs to be included in methods/figure legends.

We have introduced additional pertinent information in the material and methods section.

• Figure 4B-D, G-H: Which culturing conditions were used for adult spheroids? Original method or sandwich method? These data were obtained with the original protocol

• Fig 6D-E: Please add the timepoint after DT administration these samples are from. It is not listed in text or figure legend. These samples were those obtained from mice sacrificed at the end of the 5 day period as indicated in panel A. This has been emphasized in the legend of the figure.

• SuppFig 6D: again timepoint is missing. In this experiment all samples were untreated as indicated. This has been emphasized in the legend of the figure.

• SuppFig 6: How were the crypts of these mice (DT WT & DT HE) isolated? Was this via EDTA? This was RNA extracted from total uncultured EDTA-released material (crypts). This has been emphasized in the legend of the figure.

Also, what is the timepoint for isolation for these samples? Even if untreated, the timepoint adds context to the data. Please add more context to describing these different experiments, either in the figure legends or methods section.

All these experiments were from 2 month old animals. We have indicated this in the legend of the figure.

• SuppFig 6E: The quality of the heatmap resolution is too poor to read gene names. We have improved the resolution of the figure and hope the name of the genes are readable now.

• 5-7, are the regenerating crypt-villus units fully differentiated or are they maintained in the developmental state? Immunostaining of markers for stem cells (Lgr5), differentiated lineages (Alpi, Muc2, Lyz, ChgA etc.) and fetal state (Sca1, Trop2 etc) should be analysed in those "white" unrecombined crypt-villus units. The differentiation phenotype is shown by the clear presence of morphologically-identified Paneth and Goblet cells. We agree that specific immunostainings could have been performed to further explore this point. Regarding the fetal state, Clu expression was shown during the regeneration period (see figure 7D,E).

Unfortunately, for reasons explained in section 4 below, we are not in a position to perform these additional experiments.

• The following text needs clarification: "The kinetics of appearance of newly formed un-recombined ("white") crypts was studied after a single pulse of DT (Fig.7A). This demonstrated an increase at 48 hours, with further increase at day 10 and stable maintenance at day 30. The presence of newly formed white crypts one month after toxin administration indicates that the VilCre-negative lineage is developmentally stable and does not turn on the transgene during differentiation of the various epithelial lineages occurring after regeneration (Fig.7B).

*Comment: The "newly formed" is an overstatement, the data doesn't conclude that those are "new" crypts. *

Except if we do not understand the point, we think we can write that a fraction of "white" crypts must be "newly formed", since they are in excess of those present in untreated animals at the same time point.

*The end of the sentence states that these "white" crypts form developmentally stable lineages, thus these white crypts at day 30 could originate from the initial injury. *

As stated above, we consider that crypts found in excess of those present in untreated animals result from the initial injury.

*There was no characterisation of the various epitheial lineages. Are they fully differentiated? *

See above the point related to Paneth cells and Goblet cells.

Is Lgr5 expressed one month after toxin administration? Can the VilCre neg lineage give rise to CBCs?

We have tried hard to show presence or absence of Lgr5 in white crypts at the various times following DT administration. We tried double RFP / Lgr5-RNA scope labeling and double GFP/RFP immunolabeling. Unfortunately, we could not get these methods to produce convincing specific labeling of CBCs in homeostatic crypts, which explains why we could not reach a conclusion regarding the white crypts.

However, there is an indirect indication that "chronic" white crypts (i.e. those caused by DTR expression in CBC, plus those observed 30 days after DT administration) do not express Lgr5. Indeed, acute regeneration indicated by Clu expression at day 5 in Fig.7C is lower in white crypts than in red ones strongly suggesting that white crypts preexisting DT administration (the "chronic ones) do not express Lgr5DTR.

The relationship between white crypt generation and appearance of Clu-positive revival cells (Ayyaz et al., 2019) was then explored. In agreement with others and similar to what happens in the irradiation model, (Ayyaz et al., 2019; Yuan et al., 2023) Clu-positive cells were rare in crypts of untreated mice and their number transiently increased forty-eight hours after a single pulse of DT, and more so after three pulses of DT (Fig.7C,D).

Comment: Comparing 1 pulse at day 2 vs 3 pulses at day 5 makes the data hard to interpret. How is the Clu ISH level for 1 pulse at day 5? Are they equivalent?

After a single pulse of of DT, Clu is only transiently increased. As shown by Ayyaz et al it is back to the starting point at day 5 (supplementary figure 4 of Ayyaz et al).

Clu-positive cells were less frequently observed in white crypts (see "Total" versus "White" in Fig.7C). This fits with the hypothesis that Clu expression marks acutely regenerating crypts and that a proportion of the white crypts are chronically regenerating due to DTR expression in CBCs."

*Comment: I believe the authors suggested that the discrepancy of less Clu expression in white crypts is due to the ectopic expression of DTR in CBCs causing low grade injury without DT administration. This means that some white crypts could have been formed before the administration of DT, and thus are on a different regenerative timeline compared to the white crypts formed from DT administration. *

Yes, this is our interpretation. We have clarified it in the text.

Is there any proof of the chronic regeneration? Immunostaining of chronic regenerative markers such as Sca1, Anxa1 or Yap1 nuclear localization would support the claim. It'd be important to show only the white crypts, but not the RFP+ ones, show regenerative markers.

We think that the steady state higher number of white crypts in untreated Lgr5-DTR animals, compared to wild type siblings indicates chronical low-grade regeneration, which is supported by the RNAseq data (Suppl fig6). It must be noted, however, that this phenotype is mild compared to the well described fetal-like regeneration phenotype described in most injury models. Since these white crypts were made at undetermined earlier stages, the great majority of them are not expected to show markers of acute regeneration like Clu, Sca1....

Fig 7D-E: What are the timepoints of harvest for HE-WT-HE 1 pulse DT mice and HE- HE-HE PBS injected mice?

We have added this information in the figure.

• *Fig 8-9: Regarding the CBC-like Olfm4 low population, what is the status of Lgr5? This should be shown in the figure since the argument is that this is an Lgr5-independent lineage. * See response to the second point.

And what about the regenerative, Yap, mesenchymal and inflammatory signatures? Are they enriched in the white crypts similar to the in vitro spheroids?

In a portion of white crypts, those we believe are newly formed after CBC ablation (see above), there is a transient increase in Clu, which may be considered a marker of Yap activation. In the CBC-like Olfm4 low cells, as seen by scRNAseq, there is nothing like an actively regenerating phenotype. This is expected, since these cells are coming from homeostatic untreated VilCre/Rosa26Tom animals and are supposed to be quiescent "awaiting to be activated".

Reviewer #1 (Significance (Required)):

Strengths: The study employed a range of in vitro and in vivo models to test the hypothesis.

• *

*Limitations: Unfortunately, the models chosen did not provide sufficient evidence to draw the conclusions. Injury induced reprogramming, both in vivo and in vitro, has been well documented in the field. The new message here is to show that such reprogrammed state is continuous rather than transient; instead of regenerating Lgr5+ stem cells, it can continue to differentiate to all cell lineages in Lgr5-independent manner.

*

We respectfully disagree with this analysis of our results. What we show is not "that such reprogrammed state is continuous rather than transient; instead of regenerating Lgr5+ stem cells, it can continue to differentiate to all cell lineages in Lgr5-independent manner", but that a quiescent stem cell line, not previously identified, is activated to regenerate a portion of crypts following CBC ablation. These cells are not reprogrammed, they correspond to a developmental lineage waiting to be activated and keep their VilCre-negative state at least of 30 days. We believe that their "by default tracing" (VilCre negative from the zygote stage) is as strong an evidence for the existence of such a lineage as positive lineage tracing would be. The increase in crypts originating from this lineage after CBC ablation indicates that it is implicated in regeneration. We do not question the well-demonstrated plasticity-associated reprogramming taking place during regeneration; we simply suggest that this would coexist with the involvement of the quiescent VilCre-negative lineage we have identified.

*However, through the manuscript, there was no immunostaining of Lgr5 and other differentiation markers. The conclusion is an overstatement without solid proof. * We have provided the best answer we could to this point in our answer to the second question of the referee hereabove.

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

In this manuscript, the Marefati et al. developed a novel approach to generate spheroids from adult intestinal epithelium using a collagenase/dispase based protocol. Adult spheroids were found to be distinct from classic budding-type organoids normally generated from EDTA based release of the crypt epithelium. Transcriptional profiling indicated that adult spheroids were undifferentiated and similar to regenerating crypts or fetal spheroids. To identify the cell of origin that generates adult spheroids, the authors labelled epithelial cells with VilCreERT-LSL-Tom, VilCre-LSL-GFP and Lgr5CreERT- LSLTom mice. From these experiments the authors conclude that that spheroids are only generated from Vil-Cre negative and Lgr5 negative cells. Next the authors deleted the anti- apoptotic gene Mcl1 using Vil-CreERT mice. This led to a strong apoptotic response throughout the crypt epithelium and tissues processed from knockout mice readily generated spheroids, and in vivo, replenishment of the gut epithelium was mediated by unrecombined cells. In a second model, CBCs were ablated using Lgr5DTR mice and VilCre negative cells were found again to contribute to regeneration of the crypt epithelium. Finally based on the absence of Vil-Cre reporter activity, the authors were able to sort out and perform scRNAseq to profile VilCre negative cells. These cells were found to be quiescent, express the stem cell marker Olfm4 and were also abundant in ribosomal gene expression.

• *

The fact that the authors developed a protocol to reproducibly generate fetal-like spheroids from adult tissue is an important advancement in the field. Previous reports have shown that treatment with various small molecule inhibitors can revert budding organoids into a spheroid morphology, but this manuscript demonstrates that spheroids can also be generated from otherwise untreated cells. This new methodology will provide new tools to dissect the molecular determinants of fetal/regenerative cells in the gut. Based on this, the manuscript should attract a significant amount of attention in the intestinal field.

• *

As pointed out by the authors themselves the study has important limitations that diminish enthusiasm. The primary issue relates to the inability of the team to identify markers of VilCre neg cells other than the fact that these cells are Olfm4+ and quiescent. Nonetheless, for the reasons stated above the manuscript should reach the target audience within the research community, if the authors can address the specific points below related to issues with methodology as well as defining more precisely the characteristics and growth requirements of adult spheroid cultures.

Thank you for this positive analysis of our study.

Major comments

The main conclusion of the study is that Vil-Cre neg cells are rare quiescent Olfm4+ crypt cells. If this is the case, then standard EDTA treatment should release these cells as well. Consequently, spheroids should also emerge from isolated crypts grown in the absence of ENR. If this is not the case how do the authors explain this?

We have tried hard to generate spheroids by culturing EDTA organoids in medium lacking ENR and by treating EDTA organoids with collagenase/dispase, without success. Therefore, we are left with the conclusion that spheroid-generating cells must be more tightly attached to the matrix than those released by EDTA, and that it is their release from this attachment by collagenase that triggers a regeneration-like phenotype. This hypothesis is supported by several models of regeneration in other tissues as indicated in our references (Gilbert et al., 2010; Machado et al., 2021; Montarras et al., 2005).

From the text the authors appear to suggest that growth of adult spheroids is dependent initially on "material" released by collagenase/dispase treatment. An obvious candidate would be mesenchymal cells, which are known to secrete factors such as Wnts and PGE2 that drive spheroid morphology. To test this, the authors should treat spheroid cultures with Porcupine and/or PGE2 inhibitors.

We followed similar reasoning, considering that spheroids express strongly Ptgs1 ,2 (Figure 3A). We thought their phenotype might be maintained by autocrine prostaglandin action. We tested aspirin, a Ptgs inhibitor, which was without effect on the spheroid phenotype. Besides, we explored a wide variety of conditions to test whether they would affect the spheroid phenotype [Aspirin-see above, cAMP agonists/antagonists, YapTaz inhibitors (verteporfin and CA3), valproic acid, Notch inhibitors (DAPT, DBZ, LY511455), all-trans retinoic acid, NFkB inhibitors (TCPA, BMS), TGFbeta inhibitor (SB431542)]. As these results were negative, we did not include them in the manuscript.

• If these inhibitors block growth then this would suggest that either stromal cells or autocrine signalling involving these pathways is important. Overall, more in-depth analysis of the growth requirements of adult spheroids is required.*

Figure 1d indicates that adult spheroids can be propagated for at least 10 passages. The abstract mentions they are "immortal". The text itself does not address this issue. More precise information as to how long spheroids can be propagated is required. If these cultures can be propagated for 10 passages or more it becomes important to determine what nutrients/mitogens in the basal media are driving growth? Alternatively, what is the evidence that spheroid cultures are completely devoid of mesenchymal cells. The text only mentions that "Upon replating, these spheroids could be stably cultured free of mesenchymal cells (Fig.1B)". No validation is shown to support this.

We agree that "immortal" is not a good way to characterize our spheroids, as also pointed out by referee nr 1. We have changed that in the text, indicating the maximal number of replating we tested was 26 and replacing immortal by stably replatable. Of note, the spheroids could frozen/thawed and recultured many times.

Related to the question whether mesenchymal cells could still contaminate the spheroid cultures, we can provide the following answers:

• No fibroblasts could be seen in replated cultures and multiple spheroids could be repeatedly propagated from a single starting spheroid.
• The bulk RNAseq experiment comparing organoids to ENR or BCM cultured spheroids show, despite expression of several mesenchymal markers (see matrisome in Fig3), absence of significant expression of Pdgfra (see in revision plan folder for CP20Millions results from the raw data of new suppl table 2, with Clu, Tacstd2 and Alpi shown as controls).
• Regarding the nutrients/mitogens in the medium driving spheroid growth, we did not explore the point further than showing that they grow in basal medium (i.e. advanced DMEM), given that the presence of Matrigel makes it difficult to pinpoint what is really needed. In Figure 2, the authors describe the growth requirements for adult spheroids and indicate that spheroids grown in ENR or EN became dark and shrink. The representative images showing this are clear, but this analysis should be quantified.

Added to the manuscript.

In SF3, the gene expression profile of organoids from the sandwich method only partially overlaps with that of organoids from the old protocol. What are the gene expression differences between the 2 culture systems? Secondly, the sandwich method appears to sustain growth of Tom+ spheroids based on RNAseq and the IF images. This suggest that Vil-Cre negative cells are not necessarily the only source of adult spheroids and thus this experiment seems to indicate that any cell may be converted to grow as a spheroid under the right conditions. These points should be addressed.

Looking back to our data in order to answer the point raised by the referee, we realized that we had inadvertently-compared organoids to ENR-cultured spheroids generated by the first protocol to BCM-cultured spheroids generated by the sandwich method. We have corrected this error in a new version of suppl fig3. This shows increased correspondence between genes up- or downregulated in the spheroids obtained in the two protocols (from 49/48% to 57/57% (Venn diagram on the new figure). We agree that, even after this correction, the spheroids obtained with the two protocols present sizeable differences in their transcriptome. However, considering the very different way these spheroids were obtained and cultured initially, we do not believe this to be unexpected. The important point in our opinion is that the core of the up- and down-regulated genes typical of the de-differentiation phenotype of adult spheroids is very similar, as shown in the heatmap (which was made with the correct samples!). Also, a key observation is that that both kind of spheroids survive and can be replated in basal medium. As already stated, this characteristic is only seen rare cases [spheroids obtained from rare FACS-purified cells (Smith et al 2018) or helminth-infected intestinal tissue (Nusse et al.2018)]. Together with the observation that the majority of them is not traced by VilCre constitutes what we consider the halmark of the spheroids described in our study. As shown in figure 4E (old protocol) and Suppl Fig.3 (sandwich protocol) both red and white spheroids were extremely low in VilCre expression. As stated in the text, the fact that some spheroids are nevertheless red is most probably related to the extreme sensitivity of the Rosa26Tom marker to recombination (Liu et al., 2013), but this does not mean that there are two phenotypically different kind of spheroids. It means that the arbitrary threshold of Rosa26Tom recombination introduces an artificial subdivision of spheroids with no phenotypical significance.

Regarding the point made by the referee that "that any cell may be converted to grow as a spheroid under the right conditions", we agree and have shown with others that organoids acquire indeed a spheroid phenotype when cultured for instance in fibroblasts-conditioned medium (see suppl fig1B and (Lahar et al., 2011; Roulis et al., 2020) quoted in the manuscript). However, these spheroids cannot be propagated in basal medium, and revert to an organoid phenotype when put back in ENR (Suppl fig1B).

*In Figure 4, the authors conclude that spheroids do not originate from Lgr5 cell derived clones even after 30days post Tam induction. Does this suggest that in vivo and under homeostatic conditions VilCre neg cells are derived from a distinct stem cell pool or are themselves a quiescent stem cell. Given the rarity of VilCre neg cells, the latter seems unlikely.

*

Despite their rarity, we believe VilCre-negative cells observed under homeostatic conditions are themselves quiescent stem cells. Actually, if they were derived from a larger stem cell pool, this pool should also be VilCre-negative. And we do not see such larger number of VilCre-neg cells under homeostatic conditions.

The problem with the original assertion is that Lgr5-CreERT mice are mosaic and therefore not all Lgr5+ cells are labelled in this model. "White" spheroids may thus derive from cells that in turn derive from these unlabelled Lgr5 cells.

We had considered the possibility that mosaicism [very low for VilCre (Madison et al., 2002); in the 40-50% range for Lgr5CreERT2 (Barker & Clevers. Curr Protoc Stem Cell Biol. 2010 Chapter 5)] could explain our data. We think, however that we can exclude this possibility on the basis that spheroids do not conform to the expected ratio of unrecombined cells, given the observed level of mosaicism. Indeed, for VilCre, a few percent, at most, of unrecombined cells in the epithelium translates into almost 100% unrecombined spheroids. For Lgr5CreERT2 mice, the mosaicism level is in the range of 40%, which is what we observe for EDTA organoids (Figure 4G), while spheroids were in their vast majority unrecombined.

We have included a discussion about the possible role of mosaicism in the new version.

ATACseq experiments were briefly mentioned in the manuscript but unfortunately little information was extracted from this experiment. What does this experiment reveal about the chromatin landscape of adult spheroids relative to normal organoids?

We only performed this experiment to search for an explanation to the paradoxical absence of expression of the VilCre transgene in spheroids, despite robust expression of endogenous villin (Suppl Fig.4). We chose to show the chromatin landscape of a gene equally expressed in both organoids and spheroids (Krt19), a gene specifically expressed in spheroids (Tacstd2) and the endogenous Villin gene also expressed in both. We believe that the observation of a clear difference in pattern of the chromatin accessibility around the endogenous villin gene in organoids and spheroids provides an explanation to the observed results. The cis regulatory sequences needed for expression of the endogenous villin gene seem to be different in organoids and spheroids, which may explain why the regulatory sequences present in the transgene (only 12.4kb) might not allow expression of the transgene in spheroids. We have added a sentence in the manuscript clarifying this point. Missing is obviously the chromatin landscape around the VilCre transgene, but this is beyond reach in such kind of experiments.

Reviewer #2 (Significance (Required)):

The fact that the authors developed a protocol to reproducibly generate fetal-like spheroids from adult tissue is an important advancement in the field. Previous reports have shown that treatment with various small molecule inhibitors can revert budding organoids into a spheroid morphology, but this manuscript demonstrates that spheroids can also be generated from otherwise untreated cells. This new methodology will provide new tools to dissect the molecular determinants of fetal/regenerative cells in the gut. Based on this, the manuscript should attract a significant amount of attention in the intestinal field.

Reviewer #3 (Evidence, reproducibility and clarity (Required)): CR-2024-02491

An Lgr5-independent developmental lineage is involved in mouse intestinal regeneration

Marefati et al.

Homeostatic maintenance of the intestinal epithelium has long been thought to rely upon Wnt signaling responsive Lgr5-expressing stem cells that reside at the crypt base.

However, myriad reported mechanisms or populations have been reported to underlie epithelial regeneration after injury. Many groups have reported that reacquisition of a fetal- link intestinal phenotype is an import part of the regenerative response, however the originating cell type has not been definitively identified. Herein, the authors demonstrate that cells from adult homeostatic intestine can generate immortal spheroids that resemble fetal spheroids and are derived independent of Lgr5+ intestinal stem cells (ISCs). The authors then draw the conclusion that this indicates that a hierarchical stem cell model applies to regeneration of the intestinal epithelium, in addition to the plasticity model.

• *

Comments:

1. Please indicate what species is used for studies in Fig 1.

All experiments were performed in Mus musculus.

Please clarify if Figure 2 studies utilize Matrigel or not.

Yes

RNA-seq analyses of adult intestinal generated spheroids lack the granularity of single cell analyses and thus it is unclear if this is a homogeneous population or if the population has diversity across it (i.e., enteroids/organoids have a high level of diversity). Many of the conclusions from the RNA-seq study are broad and generalized-for example Fig 3F indicates that markers of the +4 ISC populations (Bmi1, tert, lrig1, hopx) were all expressed similarly in adult spheroids as compared to adult organoids. However, while this may be true in the bulk-RNA-seq analyses, clearly scRNA-seq would provide a better foundation to make this statement, as enteroids/organoids are comprised of heterogeneous subpopulations. . .and it might indicate that these +4 markers have only very low expression in the spheroids. Based upon these concerns, misconclusions are likely to be drawn.

We agree and it would be certainly worthwhile to perform scRNAseq of adult spheroid populations. This would certainly be worth doing in future studies to explore the possible heterogeneity of adult spheroids. We nevertheless believe that our scRNAseq performed on homeostatic intestinal tissue from VilCre/Rosa26Tom mice identify Olfm4-low VilCre-neg cells that are likely at the origin of adult spheroids and display a quite homogenous phenotype.

*The language around Figure 4 results is confusing. Please define "white" and "red". It might be simpler to designate recombined versus not recombined lineage.

*

We have clarified this in the figure.

The hypothesis that collagenase/dispase solution acts as a proxy for injury is not demonstrated and backed by data. Thus, it is difficult to make the conclusion that this approach could represent a "stable avatar" of intestinal regenerating cells. It is clear that subpopulations of crypt-based cells generate spheroids in culture without collagenase/dispase (see the cited reference Smith et al, 2018).

* *Smith et al demonstrate clearly the possibility to obtain spheroids with properties probably similar to ours from EDTA derived intestinal crypt cells. However they need to prepurify them by FACS. Besides, Nusse et al describe spheroids similar to ours after infection of the intestine by helminths (Nusse et al. 2018). In our case, and for most labs preparing enteroids with the EDTA protocol, the result is close to 100% organoids. Even if we treat EDTA organoids with collagenase, we do not obtain spheroids. This brought us to the conclusion that spheroid-generating cells must be more tightly attached to the matrix than CBCs and that it is their release from the matrix that activates the spheroid regeneration-like phenotype. This hypothesis is supported by several models of regeneration in other tissues as indicated in our references (Gilbert et al., 2010; Machado et al., 2021; Montarras et al., 2005)

A study based on the absence of recombination in a VilCre lineage tracing scenario is not well-established to be strong experimental approach, as there are many reasons why recombination may not cells may not be lineage marked. In order to use this system as the authors intend, they first need to demonstrate that villin is not expressed in the discrete cell population that they are targeting. For the presented observational studies, this would be difficult to do. While they do demonstrate differences in chromatin accessibility between cells from organoids versus spheroids (fig s4), some of these differences could merely be due to the bulk analytical nature of the study and the lack of comparing stem cell populations from spheroids to stem cell populations from organoids-since the spheroids are likely homogenous versus the organoids that only have a small fraction of stem cells-and thus represent a mix of stem cell and differentiated cell populations. The authors do not demonstrate that villin protein expression varies in these cells.

If it were found that villin is not expressed in their "novel" population, then one would expect that the downstream use of villin-based recombination would demonstrate the same recombination potential (i.e., Mcl1 would not be recombined). Both recombination studies in Fig 6 are difficult to interpret, and thus it is not clear if these studies support the stated conclusions. Quantification of number of crypts that are negative should be reported as a percentage of recombined crypts.

We are sorry but there seems to be a complete misunderstanding of our data regarding the point raised by the referee. The important point of our initial observation is that despite robust expression of villin in spheroids, the VilCre transgene is not expressed (see figure 4E). This in our opinion makes absence of VilCre expression (or of Rosa marker recombination) a trustful marker of a new developmental lineage. All the data in figure 4 constitute an answer.

*The reasoning about heterogeneity of cell type in organoids versus probable homogeneity of spheroids is well taken. However, as the endogenous villin gene is expressed in all cells of both organoids and spheroids, it is highly significant that only spheroids do not express the transgene. *

We performed the ATACseq experiment to search for an explanation to the paradoxical absence of expression of the VilCre transgene in spheroids, despite robust expression of endogenous villin (Suppl Fig.4). We chose to show the chromatin landscape of a gene equally expressed in both organoids and spheroids (Krt19), a gene specifically expressed in spheroids (Tacstd2) and the endogenous Villin gene also expressed in both. We believe that the observation of a clear difference in pattern of the chromatin accessibility around the endogenous villin gene in organoids and spheroids provides an explanation to the observed results. The cis regulatory sequences needed for expression of the endogenous villin gene seem to be different in organoids and spheroids, which may explain why the regulatory sequences present in the transgene (only 12.4kb) might not allow expression of the transgene in spheroids. We have added a sentence in the manuscript clarifying this point. Missing is obviously the chromatin landscape around the VilCre transgene, but this is beyond reach in such kind of experiments.

*Figure 8 indicates that the cell population identified by scRNA-seq may be quiescent. Companion IF or IHC should be conducted to confirm this finding, as well as other conclusions from the informatics conducted.

*

We agree that additional experiments could be performed to support this point. We are unfortunately not in a position to perform these experiments (see section 4 below).

Clearly the data is intriguing, however, the conclusion is strong and is an over interpretation of the presented data. There are a number of validation or extension data that would enhance the overall interpretation of the study: 1. validation of scRNA-seq or bulk RNA-seq concepts by protein staining of intestinal tissues in the damage model will serve as a secondary observation. 2. identification of the ISC that they are defining is critical and important. There is already the notion that this cell type exists and it has been shown with various different markers. 3. expand the analyses of the fetal-like expression profiling to injured intestines to demonstrate that the lineage negative cells indeed express fetal-like proteins. 4. expand the discussion of the Clu+ cell type. Is this cell the previously described revival cell? If so, how does this body of work provide unique aspects to the field?

We agree that all these suggested experiments could be performed and would be of interest. However, we consider that they would not modify the main message of our study and would only constitute an expansion of the present work. As already stated, we are not in the position to perform them (see section 4).

*There is some level of conflicting data, with the stem population being proliferative in culture stimulated by the stromal cells, but quiescent in vivo and also based upon scRNA- seq data in Fig 9.

*

We do not see any conflict in our observation regarding this point. The observation that cells that are quiescent in vivo become proliferative when subjected to culture (with or without addition of stromal cells) is routinely made in a multitude of cell culture systems. In particular, it has been shown that intestinal tissue dissociation activates the Yap/Taz pathway, resulting in proliferation (Yu et al. Hippo Pathway Regulation of Gastrointestinal Tissues. Annual Review of Physiology, 2015 Volume 77, 201-227).

Many of the findings have been previously reported: Population that grows as spheroids (Figure 2), Population that is Wnt independent (Figure 2), Lgr5 independent regenerative growth of the intestine (figure 3F, Figure 4), Clu+ ISCs drive regeneration (Figure 7).

Whereas these individual findings have indeed been reported, it was in a different context. We strongly disagree with the underlying suggestion that our study would not bring new information. We have identified here a developmental lineage involved in intestinal regeneration that has not been described up to now.

Minor comments:

1. • The statement that spheroids must originate from collagenase/dispase digested material might be an overstatement. As spheroids generation from EDTA treated intestines have been previously reported (Smith et al, 2018). * See answer to point 4 above. *Overall while the study includes an extensive amount of work and different approaches, a foundationally supported novel finding is missing. Many of the statements have already been demonstrated by others in the fields. In addition, one of the most intriguing aspects of the study is that the stromal population impacts this stem cell population, however, interactions and factors stimulating the crosstalk are not addressed.

*

Reviewer #3 (Significance (Required)):

Overal while the study includes an extensive amount of work and different approaches, a foundationally supported novel finding is missing. Many of the statements have already been demonstrated by others in the fields. In addition, one of the most intriguing aspects of the study is that the stromal population impacts this stem cell population, however, interactions and factors stimulating the crosstalk are not addressed.

We can only disagree.

4. Description of analyses that authors prefer not to carry out

• *

We have answered most questions raised by the referees by explaining our view, by clarifying individual points and, in several cases, by providing additional information that was not included in the original manuscript.

In a limited number of cases when additional experiments were suggested, we were unfortunately obliged to write that we are not in a position to perform them. This is because my lab is closing after more than fifty years of uninterrupted activity. There will unfortunately be nobody to perform additional experiments.

Nevertheless, as written by referees 1 and 2, we believe that the revised manuscript, as it stands, contains data that will be of interest to the people in the field and may be the bases for future developments. We hope editors will find interest in publishing it.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

RC-2024-02491

An Lgr5-independent developmental lineage is involved in mouse intestinal regeneration Marefati et al.

Homeostatic maintenance of the intestinal epithelium has long been thought to rely upon Wnt signaling responsive Lgr5-expressing stem cells that reside at the crypt base. However, myriad reported mechanisms or populations have been reported to underlie epithelial regeneration after injury. Many groups have reported that reacquisition of a fetal-link intestinal phenotype is an import part of the regenerative response, however the originating cell type has not been definitively identified. Herein, the authors demonstrate that cells from adult homeostatic intestine can generate immortal spheroids that resemble fetal spheroids and are derived independent of Lgr5+ intestinal stem cells (ISCs). The authors then draw the conclusion that this indicates that a hierarchical stem cell model applies to regeneration of the intestinal epithelium, in addition to the plasticity model.

Comments:

1. Please indicate what species is used for studies in Fig 1.
2. Please clarify if Figure 2 studies utilize Matrigel or not.
3. RNA-seq analyses of adult intestinal generated spheroids lack the granularity of single cell analyses and thus it is unclear if this is a homogeneous population or if the population has diversity across it (i.e., enteroids/organoids have a high level of diversity). Many of the conclusions from the RNA-seq study are broad and generalized-for example Fig 3F indicates that markers of the +4 ISC populations (Bmi1, tert, lrig1, hopx) were all expressed similarly in adult spheroids as compared to adult organoids. However, while this may be true in the bulk-RNA-seq analyses, clearly scRNA-seq would provide a better foundation to make this statement, as enteroids/organoids are comprised of heterogeneous subpopulations. . .and it might indicate that these +4 markers have only very low expression in the spheroids. Based upon these concerns, misconclusions are likely to be drawn.
4. The language around Figure 4 results is confusing. Please define "white" and "red". It might be simpler to designate recombined versus not recombined lineage.
5. The hypothesis that collagenase/dispase solution acts as a proxy for injury is not demonstrated and backed by data. Thus, it is difficult to make the conclusion that this approach could represent a "stable avatar" of intestinal regenerating cells. It is clear that subpopulations of crypt-based cells generate spheroids in culture without collagenase/dispase (see the cited reference Smith et al, 2018).
6. A study based on the absence of recombination in a VilCre lineage tracing scenario is not well-established to be strong experimental approach, as there are many reasons why recombination may not cells may not be lineage marked. In order to use this system as the authors intend, they first need to demonstrate that villin is not expressed in the discrete cell population that they are targeting. For the presented observational studies, this would be difficult to do. While they do demonstrate differences in chromatin accessibility between cells from organoids versus spheroids (fig s4), some of these differences could merely be due to the bulk analytical nature of the study and the lack of comparing stem cell populations from spheroids to stem cell populations from organoids-since the spheroids are likely homogenous versus the organoids that only have a small fraction of stem cells-and thus represent a mix of stem cell and differentiated cell populations. The authors do not demonstrate that villin protein expression varies in these cells. If it were found that villin is not expressed in their "novel" population, then one would expect that the downstream use of villin-based recombination would demonstrate the same recombination potential (i.e., Mcl1 would not be recombined). Both recombination studies in Fig 6 are difficult to interpret, and thus it is not clear if these studies support the stated conclusions. Quantification of number of crypts that are negative should be reported as a percentage of recombined crypts.
7. Figure 8 indicates that the cell population identified by scRNA-seq may be quiescent. Companion IF or IHC should be conducted to confirm this finding, as well as other conclusions from the informatics conducted.
8. Clearly the data is intriguing, however, the conclusion is strong and is an over interpretation of the presented data. There are a number of validation or extension data that would enhance the overall interpretation of the study:
   • a. validation of scRNA-seq or bulk RNA-seq concepts by protein staining of intestinal tissues in the damage model will serve as a secondary observation.
   • b. identification of the ISC that they are defining is critical and important. There is already the notion that this cell type exists and it has been shown with various different markers.
   • c. expand the analyses of the fetal-like expression profiling to injured intestines to demonstrate that the lineage negative cells indeed express fetal-like proteins.
   • d. expand the discussion of the Clu+ cell type. Is this cell the previously described revival cell? If so, how does this body of work provide unique aspects to the field?
9. There is some level of conflicting data, with the stem population being proliferative in culture stimulated by the stromal cells, but quiescent in vivo and also based upon scRNA-seq data in Fig 9.
10. Many of the findings have been previously reported: Population that grows as spheroids (Figure 2), Population that is Wnt independent (Figure 2), Lgr5 independent regenerative growth of the intestine (figure 3F, Figure 4), Clu+ ISCs drive regeneration (Figure 7).

Minor comments:

1. The statement that spheroids must originate from collagenase/dispase digested material might be an overstatement. As spheroids generation from EDTA treated intestines have been previously reported (Smith et al, 2018).

Overall while the study includes an extensive amount of work and different approaches, a foundationally supported novel finding is missing. Many of the statements have already been demonstrated by others in the fields. In addition, one of the most intriguing aspects of the study is that the stromal population impacts this stem cell population, however, interactions and factors stimulating the crosstalk are not addressed.

Significance

Overall while the study includes an extensive amount of work and different approaches, a foundationally supported novel finding is missing. Many of the statements have already been demonstrated by others in the fields. In addition, one of the most intriguing aspects of the study is that the stromal population impacts this stem cell population, however, interactions and factors stimulating the crosstalk are not addressed.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

In this manuscript, the Marefati et al. developed a novel approach to generate spheroids from adult intestinal epithelium using a collagenase/dispase based protocol. Adult spheroids were found to be distinct from classic budding-type organoids normally generated from EDTA based release of the crypt epithelium. Transcriptional profiling indicated that adult spheroids were undifferentiated and similar to regenerating crypts or fetal spheroids. To identify the cell of origin that generates adult spheroids, the authors labelled epithelial cells with VilCreERT-LSL-Tom, VilCre-LSL-GFP and Lgr5CreERT-LSLTom mice. From these experiments the authors conclude that that spheroids are only generated from Vil-Cre negative and Lgr5 negative cells. Next the authors deleted the anti-apoptotic gene Mcl1 using Vil-CreERT mice. This led to a strong apoptotic response throughout the crypt epithelium and tissues processed from knockout mice readily generated spheroids, and in vivo, replenishment of the gut epithelium was mediated by unrecombined cells. In a second model, CBCs were ablated using Lgr5DTR mice and VilCre negative cells were found again to contribute to regeneration of the crypt epithelium. Finally based on the absence of Vil-Cre reporter activity, the authors were able to sort out and perform scRNAseq to profile VilCre negative cells. These cells were found to be quiescent, express the stem cell marker Olfm4 and were also abundant in ribosomal gene expression.

The fact that the authors developed a protocol to reproducibly generate fetal-like spheroids from adult tissue is an important advancement in the field. Previous reports have shown that treatment with various small molecule inhibitors can revert budding organoids into a spheroid morphology, but this manuscript demonstrates that spheroids can also be generated from otherwise untreated cells. This new methodology will provide new tools to dissect the molecular determinants of fetal/regenerative cells in the gut. Based on this, the manuscript should attract a significant amount of attention in the intestinal field.

As pointed out by the authors themselves the study has important limitations that diminish enthusiasm. The primary issue relates to the inability of the team to identify markers of VilCre neg cells other than the fact that these cells are Olfm4+ and quiescent. Nonetheless, for the reasons stated above the manuscript should reach the target audience within the research community, if the authors can address the specific points below related to issues with methodology as well as defining more precisely the characteristics and growth requirements of adult spheroid cultures.

Major comments

The main conclusion of the study is that Vil-Cre neg cells are rare quiescent Olfm4+ crypt cells. If this is the case, then standard EDTA treatment should release these cells as well. Consequently, spheroids should also emerge from isolated crypts grown in the absence of ENR. If this is not the case how do the authors explain this?

From the text the authors appear to suggest that growth of adult spheroids is dependent initially on "material" released by collagenase/dispase treatment. An obvious candidate would be mesenchymal cells, which are known to secrete factors such as Wnts and PGE2 that drive spheroid morphology. To test this, the authors should treat spheroid cultures with Porcupine and/or PGE2 inhibitors. If these inhibitors block growth then this would suggest that either stromal cells or autocrine signalling involving these pathways is important. Overall, more in-depth analysis of the growth requirements of adult spheroids is required.

Figure 1d indicates that adult spheroids can be propagated for at least 10 passages. The abstract mentions they are "immortal". The text itself does not address this issue. More precise information as to how long spheroids can be propagated is required. If these cultures can be propagated for 10 passages or more it becomes important to determine what nutrients/mitogens in the basal media are driving growth? Alternatively, what is the evidence that spheroid cultures are completely devoid of mesenchymal cells. The text only mentions that "Upon replating, these spheroids could be stably cultured free of mesenchymal cells (Fig.1B)". No validation is shown to support this.

In Figure 2, the authors describe the growth requirements for adult spheroids and indicate that spheroids grown in ENR or EN became dark and shrink. The representative images showing this are clear, but this analysis should be quantified.

In SF3, the gene expression profile of organoids from the sandwich method only partially overlaps with that of organoids from the old protocol. What are the gene expression differences between the 2 culture systems? Secondly, the sandwich method appears to sustain growth of Tom+ spheroids based on RNAseq and the IF images. This suggest that Vil-Cre negative cells are not necessarily the only source of adult spheroids and thus this experiment seems to indicate that any cell may be converted to grow as a spheroid under the right conditions. These points should be addressed.

In Figure 4, the authors conclude that spheroids do not originate from Lgr5 cell derived clones even after 30days post Tam induction. Does this suggest that in vivo and under homeostatic conditions VilCre neg cells are derived from a distinct stem cell pool or are themselves a quiescent stem cell. Given the rarity of VilCre neg cells, the latter seems unlikely. The problem with the original assertion is that Lgr5-CreERT mice are mosaic and therefore not all Lgr5+ cells are labelled in this model. "White" spheroids may thus derive from cells that in turn derive from these unlabelled Lgr5 cells.

ATACseq experiments were briefly mentioned in the manuscript but unfortunately little information was extracted from this experiment. What does this experiment reveal about the chromatin landscape of adult spheroids relative to normal organoids?

Significance

The fact that the authors developed a protocol to reproducibly generate fetal-like spheroids from adult tissue is an important advancement in the field. Previous reports have shown that treatment with various small molecule inhibitors can revert budding organoids into a spheroid morphology, but this manuscript demonstrates that spheroids can also be generated from otherwise untreated cells. This new methodology will provide new tools to dissect the molecular determinants of fetal/regenerative cells in the gut. Based on this, the manuscript should attract a significant amount of attention in the intestinal field.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

In this manuscript, Marefati et al report an Lgr5-independent lineage in the regenerating intestine using in vitro organoids and in vivo injury-coupled lineage tracing model. In organoids, collagenase/dispase dissociated resulted in "immortal spheroids" that maintain a cystic and undifferentiated phenotype in the absence of standard growth factors (Rspondin/Noggin/EGF). Bulk RNAseq of spheroids demonstrates downregulation of classical CBC signatures and upregulation of fetal spheroid, mesenchymal, inflammation and regenerative signatures. In mice, Villin-Cre lineage tracing revealed some Villin-negative progenies that lack reporter tracing throughout crypt-villus ribbons after injury. The authors proposed that there is Lgr5-independent population support the regenerative response upon CBC depletion. A major caveat of this study is the identification of this population is based on absence of VilCre expression. It is surprising that there is no characterisation of Lgr5 expression throughout the manuscript whilst claiming of a Lgr5-independent lineage. Although the research question is potentially interesting, the concept of epithelial reprogramming upon injury is well documented in the field. The data generated in this manuscript also seem to be preliminary and lack of detailed characterisation. Below are specific comments.

• Expression of Lgr5 should be properly characterised throughout the manuscript in both organoid models and injury-induced regeneration in vivo.
• An important question is the origin of these "Lgr5-independent" adult spheroids. They look and appear like fetal organoids, which could be induced by injury (e.g. upon collagenase/dispase dissociation). Have the authors tried to culture fetal spheroids in BCM over extensive period of time? Do they behave the same? This would be a great way to directly compare the collagenase/dispase-derived organoids with fetal origin.
• Fig 1C, Why is the replating spheroid culture time different between mesenchymal cells and conditioned medium?
• It is unclear how the bulk RNA-seq data in Fig. 3 were compared. How long were the adult organoids and spheroids cultured for (how many passages)? Were they culture in the same condition of were they in ENR vs BCM? These are important information to consider when interpreting the results. For instance, are Ptgs1 & Ptgs2 expression in adult spheroids the same in ENR vs BCM? Are the gene signatures (regenerative, fetal and YAP) changed in adult spheroids culturing in ENR vs BCM?
• It is stated: "In agreement with their aptitude to grow indefinitely, adult spheroids express a set of upregulated genes overlapping significantly with an "adult tissue stem cell module" [159/721 genes; q value 2.11 e-94) (Fig.S2F)].". What is the definition of "indefinitely"? Are they referring to the Fig 1B where spheroid were passaged to P10? The authors should avoid the term "indefinitely" but use a more specific time scale, e.g. passages, months etc.
• SuppFig 3D: Row Z-Score is missing the "e" in Score.
• Fig 4E: Figure legend says QNRQ instead of CNRQ.
• Fig 4G: The brightfield image of adult spheroids 5 days after 3x TAM injections doesn't look like a spheroid. It seems to be differentiating.
• Fig 4: Most mouse model data are missing the number of mice & their respective age used for organoid isolation.
• Fig 4A-D, H-G: How was fluorescent signal of organoids quantified? How many images? Were there equal numbers of organoids? This all needs to be included in methods/figure legends.
• Figure 4B-D, G-H: Which culturing conditions were used for adult spheroids? Original method or sandwich method?
• Fig 6D-E: Please add the timepoint after DT administration these samples are from. It is not listed in text or figure legend.
• SuppFig 6D: again timepoint is missing.
• SuppFig 6: How were the crypts of these mice (DT WT & DT HE) isolated? Was this via EDTA? Also, what is the timepoint for isolation for these samples? Even if untreated, the timepoint adds context to the data. Please add more context to describing these different experiments, either in the figure legends or methods section.
• SuppFig 6E: The quality of the heatmap resolution is too poor to read gene names.
• Fig.5-7, are the regenerating crypt-villus units fully differentiated or are they maintained in the developmental state? Immunostaining of markers for stem cells (Lgr5), differentiated lineages (Alpi, Muc2, Lyz, ChgA etc.) and fetal state (Sca1, Trop2 etc) should be analysed in those "white" unrecombined crypt-villus units.
• The following text needs clarification:

"The kinetics of appearance of newly formed un-recombined ("white") crypts was studied after a single pulse of DT (Fig.7A). This demonstrated an increase at 48 hours, with further increase at day 10 and stable maintenance at day 30. The presence of newly formed white crypts one month after toxin administration indicates that the VilCre-negative lineage is developmentally stable and does not turn on the transgene during differentiation of the various epithelial lineages occurring after regeneration (Fig.7B). Comment: The "newly formed" is an overstatement, the data doesn't conclude that those are "new" crypts. The end of the sentence states that these "white" crypts form developmentally stable lineages, thus these white crypts at day 30 could originate from the initial injury. There was no characterisation of the various epitheial lineages. Are they fully differentiated? Is Lgr5 expressed one month after toxin administration? Can the VilCre neg lineage give rise to CBCs?

The relationship between white crypt generation and appearance of Clu-positive revival cells (Ayyaz et al., 2019) was then explored. In agreement with others and similar to what happens in the irradiation model, (Ayyaz et al., 2019; Yuan et al., 2023) Clu-positive cells were rare in crypts of untreated mice and their number transiently increased forty-eight hours after a single pulse of DT, and more so after three pulses of DT (Fig.7C,D). Comment: Comparing 1 pulse at day 2 vs 3 pulses at day 5 makes the data hard to interpret. How is the Clu ISH level for 1 pulse at day 5? Are they equivalent?

Clu-positive cells were less frequently observed in white crypts (see "Total" versus "White" in Fig.7C). This fits with the hypothesis that Clu expression marks acutely regenerating crypts and that a proportion of the white crypts are chronically regenerating due to DTR expression in CBCs." Comment: I believe the authors suggested that the discrepancy of less Clu expression in white crypts is due to the ectopic expression of DTR in CBCs causing low grade injury without DT administration. This means that some white crypts could have been formed before the administration of DT, and thus are on a different regenerative timeline compared to the white crypts formed from DT administration. Is there any proof of the chronic regeneration? Immunostaining of chronic regenerative markers such as Sca1, Anxa1 or Yap1 nuclear localization would support the claim. It'd be important to show only the white crypts, but not the RFP+ ones, show regenerative markers. - Fig 7D-E: What are the timepoints of harvest for HE-WT-HE 1 pulse DT mice and HE-HE-HE PBS injected mice? - Fig 8-9: Regarding the CBC-like Olfm4 low population, what is the status of Lgr5? This should be shown in the figure since the argument is that this is an Lgr5-independent lineage. And what about the regenerative, Yap, mesenchymal and inflammatory signatures? Are they enriched in the white crypts similar to the in vitro spheroids?

Significance

Strengths: The study employed a range of in vitro and in vivo models to test the hypothesis.

Limitations: Unfortunately, the models chosen did not provide sufficient evidence to draw the conclusions. Injury induced reprogramming, both in vivo and in vitro, has been well documented in the field. The new message here is to show that such reprogrammed state is continuous rather than transient; instead of regenerating Lgr5+ stem cells, it can continue to differentiate to all cell lineages in Lgr5-independent manner. However, through the manuscript, there was no immunostaining of Lgr5 and other differentiation markers. The conclusion is an overstatement without solid proof.

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

The work would have significant impact in the cilia community, if the conclusion is correct. This reviewer, however, has a concern about the authors concluding the presence/absence of TZ, based on only B9D1 and the H-shaped body among nine doublet microtubules. First, is it really established how the structure of Xenopus embryo TZ is? While Chlamydomonas is well known to have a H-shaped TZ, other species have different form inside the 9+0 doublet, or no feature (Comparison of TZ from various species in Dennis Diener https://doi.org/10.1016/B978-0-12-822508-0.00007-1). Fig.2B of this manuscript shows visible densities in the panel "Pre", but it does not look like an H-shape. The tomogram of TZ before deciliation seems clearer (but judging from wavy MTs and membrane in this tomogram, there could be unevenness of embedding and staining), while the tomogram after deciliation is thin and does not cover the entire width. Therefore it is not sure that absence of TZ can be concluded. If the author claims Xenopus embryo cilia have a H-shaped TZ, they have to provide multiple micrographs (ideally tomogram or serial section TEM to cover the entire TZ structure) and/or past literature on Xenopus embryo TZ. B9D1 is likely a membrane associated protein (according to their deciliation by detergent and mechanical force). This may mean B9D1 is located on or near the membrane, in vicinity to TZ, and thus binds to TZ after the main part of TZ is built. In this case, it is risky to judge presence of TZ based on B9D1. Also in this point, TEM imaging will be helpful to confirm the authors' conclusion.

RESPONSE: We appreciate the reviewer’s thoughtful comments on the loss of TZ upon deciliation and its absence during the initial regeneration period. The reviewer is right in their assessment that the TZ of Xenopus cilia has not been well defined before in any manuscript. We want the reviewer to consider that our goal was not to define the TZ in Xenopus but to study deciliation and how cilia regenerate in a vertebrate model system for the first time. We unexpectedly discovered that cilia are deciliated distal to the basal body at the plasma membrane, and the “H-shaped structure,” similar to TZ, was also removed and did not come back for first hour during regeneration. Given this surprising observation, we felt obliged to study and explain our results. To that end, we explored different resources (antibodies and markers of TZ) and different methods over 6 years trying to define TZ in Xenopus.

Our conclusion about the TZ structure came from multiple lines of evidence from our experiments and published literature, including the similarity in structure compared to other organisms and its physical location in the cilium. Specifically, 1) In a review of the basal bodies, Mitchell indirectly suggested that the electron-dense “H-shaped” structure could be a TZ in Xenopus. 2) The electron-dense “H” shaped structure in Chlamydomonas is similar, if not identical, to that shown in Xenopus cilia. 3) The physical location of TZ is always shown to be distal to the basal body and transition fibers (except in clubmoss Phylloglosum) while proximal to the central pair. The electron-dense “H-shaped” structure in Xenopus fulfills these criteria, suggesting that this structure is the TZ in Xenopus. 4) The TZ bonafide protein B9D1 is localized distal to Chibby, which labels the distal end of the basal body, suggesting that the TZ is localized distal to the basal body. Moreover, the loss of an “H-shaped” structure determined using TEM and tomograms corresponds to the loss of the B9D1 signal, further strengthening the conclusion that the H-shaped structure is the TZ.

We will include serial sectioning and imaging of multiple Xenopus cilia in control and 0hr (after deciliation) to address this reviewer's concerns further. Our preliminary data has suggested that the ciliary membrane is tightened around this electron-dense structure, similar to what has been shown before for other organisms like Chlamydomonas. and thus boosts our confidence that this structure likely corresponds to the TZ in Xenopus.

The reviewer has raised a concern that “the tomogram after deciliation is thin and does not cover the entire width. Therefore, it is not sure that absence of TZ can be concluded”. We note that even if the tomograms do not go through the entire cilium (supplementary videos 2 and 3), it does go through more than the center of cilium as seen by the presence of central pair microtubules and we can observe that the electron-dense “H-shaped” structure is not present in these cilia. Further, in the supplementary videos 5 and 6, even if the tomogram again only covers half of the cilia, we can see the presence of the structure, confirming that our tomograms can demonstrate the presence or absence of the H-shaped structure confidently. We have also provided TEM sections in addition to the tomograms to show the same result.

The Reviewer has commented that “B9D1 is located on or near the membrane, in vicinity to TZ, and thus binds to TZ after the main part of TZ is built”. This reviewer is correct in their assessment. This is why we argue that the presence or absence of B9D1 may be a good marker for understanding the presence or absence of TZ assembly.

TIMELINE: We are performing additional serial TEM in the control and deciliated (0hr.) embryos to address the reviewer’s concern. We will need 1 month to finish these experiments.

Their discussion about length/number of cilia and force generated by cilia is interesting, but in the context of this research, this reviewer is skeptical about its value. The calcium induced deciliation is not a physiological phenomenon, but an artificial event (please correct if I am wrong). The argument how length and number of cilia are regulated upon deciliation makes sense only in case deciliation happens regularly and the species must optimize themselves to survive. The argument about possible passway of protein transport to control ciliary number and length (Line408-) seems, although it is an interesting topic in general, not suitable in this manuscript. For this reviewer's view, it is relatively straightforward to interpret the result of cilia number/length under normal growth, without new protein expression (CHX), with protein degradation blocked. Cilia will extend when components are provided. Growth will slow down when it is exhausted. Existing cilia start degrading, when they lack proteins, which are necessary for turn-over. With the current experimental output, there is no point to describe redistribution of proteins.

RESPONSE: We appreciate the reviewer’s comment; however, we would like to argue that different methods of deciliation have been used in different model systems, such as Chlamydomonas, to study cilia regeneration. Although this reviewer may not find some of the experiments and conclusions appropriate for this manuscript, other research groups have found these results interesting. For example, reviewer 2 states, “To support their observations that cilia length is favored over cilia number under conditions of limiting ciliary precursor availability, the authors use a mathematical model that leads to the conclusion that force generation is optimized by increasing cilia length. This is a convincing conclusion and in agreement with other comparable modeling studies performed in the field.” We have already had great discussions about these results with many cilia researchers at multiple conferences. Therefore, we prefer to keep these experiments and results in the manuscript and let readers come to their own conclusions about their importance.

Minor points:

Line65: do they mean "selected few basal bodies"? – we have removed the word “select”

Line73: extracellular flow is not limited to developmental system. – we have altered the statement to add “growth, development and homeostasis”

Line124: alpha-tubulin signal and SEM image – we have added “and scanning electron microscopy (SEM)”

Line139: Could you define explicitly the two hypotheses? – Now, we have reworded the sentences to clarify the two hypotheses. “Therefore, we considered two hypotheses: First, Xenopus MCCs regenerate cilia or second, Xenopus depend on stem cell-based replacement of damaged MCCs.”

Line164: 10,31-33 are not suitable citation for the location of calcium induced deciliation in Chlamydomonas. cite Sanders and Salisbury JCB 108, 1751 – We have changed the citation.

Line181: Later -> latter – We have changed the text.

Line195: by mechanical shearing, B9D1 remained with cilia. They concluded that TZ stays with the axoneme by deciliation. How can they exclude the possibility that mechanical separation works differently from calcium shock? – We do not intend to claim that both calcium-based and mechanical ripping of cilia from cells adopt the same deciliation mechanism, and we have mentioned in line 193 that ‘we adopted an alternative approach of mechanical deciliation’. Using these two methods as complimentary to each other, our aim was to show that TZ is lost by both ciliation methods. For the calcium method, because the membrane is ripped with detergent, we show the loss of TZ by examining the MCCs devoid of cilia. In the mechanical deciliation protocol, since no detergent is involved, we can examine cilia that are likely to have intact membranes and thus maintain a B9D1 signal.

Line214: 1.33uM -> 1.33um - We have made these changes to the text.

__RESPONSE: __All the minor points in the manuscript are addressed.

Overall, the results are well presented and allow strong conclusions to be drawn. The results are based on both immunofluorescence studies and EM analysis. To support their observations that cilia length is favored over cilia number under conditions of limiting ciliary precursor availability, the authors use a mathematical model that leads to the conclusion that force generation is optimized by increasing cilia length. This is a convincing conclusion, and in agreement with other comparable modeling studies performed in the field. It would be fascinating to be able to measure the flow parameters at the cell surface during cilia regeneration to see whether this regeneration actually leads to an increase in the overall flow or force generated by the cilia. But as the authors explain, this is probably a difficult experiment to carry out and appears to be optional in the context of this study.

__RESPONSE: __We thank the reviewer for recognizing and stating that “the results are well presented and allow strong conclusions to be drawn”. We also want to sincerely thank the reviewer for understanding the technical difficulties in performing these experiments.

The authors are apparently only able to detect a single TZ protein, B9D1, to follow the fate of the TZ during the deciliation and reciliation process. In some ways, this provides an incomplete demonstration that all the TZ is indeed removed during deciliation, although this is supported by EM observations. It also provides a limited understanding of the time course of TZ re-formation during reciliation. Given the limitations of antibody availability, could it be possible to express tagged proteins in the animal cap system to track more TZ proteins? In particular, would it be possible to track for example Cby and NPHP proteins. What is the behavior of Cep290? This would greatly reinforce the conclusions on the molecular reorganisation of the TZ after deciliation and during cilia regeneration.

__RESPONSE: __We appreciate this reviewer’s brilliant questions on understanding the time course of TZ re-formation during reciliation. When we started this project and observed that TZ was lost upon deciliation in our preliminary TEM experiment, our first goal was to confirm this outstanding result. Thus, we did more TEMs and EM tomography, used bonafide TZ protein B9D1 to label the structure, and observed its loss upon deciliation. Taken together, we feel highly confident that TZ is lost upon deciliation. To address this reviewer’s concerns, we will performing additional serial TEMs to confirm the loss of TZ after deciliation.

Our next goal was to understand what the reviewer has mentioned, the TZ assembly time course. We started with TEMs at different time points and again saw a surprising result: TZ assembly was delayed compared to cilia axoneme. We were driven by this question of understanding how cilia “put together” the complex structure of TZ structurally and molecularly using EM and fluorescence data. We first attempted a few antibodies, including B9D1, CEP290, MKS5, and NPHP4, to localize to the TZ in the Xenopus cilia. Despite our efforts with different fixation strategies, only B9D1 appeared to localize to the TZ, whereas others did not give any signal or localized at the basal body. Next, we tried localizing TMEM216, TMEM67, and NPHP4 using fluorescent tags, but we again found the same result: they localized to the basal body but not at the TZ. We are perplexed by this result and are pursuing the reasons behind them. However, these experiments are out of the scope of this paper. We want to note that we have used Chibby in our experiments and that it is not lost upon deciliation (Fig S1). This is because Chibby is a distal transition fiber protein (distal end of basal body) and does not extend up to the transition zone.

TIMELINE: To address the reviewer's concern, we are performing additional serial TEM in the control and deciliated (0hr.) embryos. We will attempt to localize CEP290-GFP, requiring approximately 1 month to finish the experiment. However, we would like to note that we cannot guarantee that this experiment will work, as similar experiments with other TZ markers have failed before.

Minor comments

1. Figure 4: The images are poorly defined, and it is difficult to distinguish individual basal bodies and cilia. Therefore, it is not clear how the authors can confidently quantify the number of basal bodies in each condition to construct the graph at the bottom of the figure. In addition, it would be interesting to label the basal body with a centriolar marker to better define it. - Figure 4 labels the Transition Zone protein B9D1 and cilia marker acetylated tubulin and not basal bodies. The graph represents the number of cells with the presence or absence of elongated B9d1 signal.

2. Figure 5: not clear why the graph on the lower right does not include the control at 3 and 6 hrs? Is it because the number is too high and difficult to quantify? – Yes, the reviewer is right. Cilia become too long and too many to quantify their number reliably.

3. References: I would like to draw the authors' attention to studies of deciliation in Paramecia that could be cited in the introduction or discussion of the conservation of this pathway through evolution. – We have added multiple references to paramecia throughout the manuscript. Specifically, we mention that deciliation and regeneration in unicellular models like paramecia have added to our understanding of ciliogenesis. Line 102 “While it is important to remember that regeneration of cilia may not be identical to de novo assembly, cilia regeneration studies in Chlamydomonas reinhardtii, Paramecium and Tetrahymena etc., have provided significant insights into ciliogenesis, g., cargo transport, the presence of precursor pool, regulation of ciliary gene expression.18,23–26”. Further, we also added the reference to paramecia in results, line164 “Next, we determined the location where the deciliation treatment severed cilia. Unicellular models such as Chlamydomonas, Paramecium and Tetrahymena lose cilia distal to the TZ and below the central pair (CP) microtubules33.”. We also add discussion on the importance of TZ in paramecia, line 203 “Interestingly in Paramecium also a unicellular multiciliated cell, displays constant shedding of cilia when TZ proteins are depleted.25”. These statements have been supported by the following studies that are now cited in the manuscript: Machemer and Ogura 1979 Journal of Cell Physiology (10.1113/jphysiol.1979.sp012990) and Gogenddeau et al., Plos Biology (10.1371/journal.pbio.3000640).

RESPONSE: All the minor points in the manuscript are addressed.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

The manuscript by Rao et al. focuses on determining the mechanism of cilia regeneration using Xenopus mucociliary epithelium. The authors employ a simple yet powerful approach to trigger deciliation of multiciliated cells, enabling them to study the mechanism of cilia regeneration. This research has a significant impact on the field of cilia biology and enhances our understanding of ciliopathies. Through detailed cell biological methodologies, the authors obtained intriguing results, including the finding that deciliation removes the transition zone and that cilia repair precedes the transition zone assembly. Additionally, the authors demonstrate that IFT proteins involved in cilia construction concentrate at selected basal bodies. Although there are open questions that the authors also highlight, this manuscript provides solid, pioneering insights into the process of cilia regeneration in vivo.

Significance

The manuscript characterizes the mechanism of cilia regeneration, providing new insights into processes that could be harnessed to restore ciliary function in patients suffering from chronic respiratory diseases.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary

This manuscript investigates how cilia regenerate in multi-ciliated cells. The authors have exploited an original multi-ciliated cell system derived from the Xenopus embryonic cap and use chemical and mechanical deciliation to understand the different steps of cilia regeneration. In this model, they show that cilia are excised just above the BB and below the ciliary transition zone. Their results indicate that during ciliary regeneration, axoneme reassembly precedes TZ formation and that ciliary reassembly relies on de novo protein synthesis. In the context of limited protein synthesis, cells regenerate fewer cilia, but of almost the same size as control cells, suggesting the existence of a cell control system to maximise force generation. Mathematical modelling of the forces exerted by defined numbers of cilia of different lengths supports this hypothesis.

Major comments

Overall, the results are well presented and allow strong conclusions to be drawn. The results are based on both immunofluorescence studies and EM analysis. To support their observations that cilia length is favored over cilia number under conditions of limiting ciliary precursor availability, the authors use a mathematical model that leads to the conclusion that force generation is optimized by increasing cilia length. This is a convincing conclusion, and in agreement with other comparable modeling studies performed in the field. It would be fascinating to be able to measure the flow parameters at the cell surface during cilia regeneration to see whether this regeneration actually leads to an increase in the overall flow or force generated by the cilia. But as the authors explain, this is probably a difficult experiment to carry out and appears to be optional in the context of this study.

The authors are apparently only able to detect a single TZ protein, B9D1, to follow the fate of the TZ during the deciliation and reciliation process. In some ways, this provides an incomplete demonstration that all the TZ is indeed removed during deciliation, although this is supported by EM observations. It also provides a limited understanding of the time course of TZ re-formation during reciliation. Given the limitations of antibody availability, could it be possible to express tagged proteins in the animal cap system to track more TZ proteins? In particular, would it be possible to track for example Cby and NPHP proteins. What is the behavior of Cep290? This would greatly reinforce the conclusions on the molecular reorganisation of the TZ after deciliation and during cilia regeneration.

Minor comments

Figure 4: The images are poorly defined and it is difficult to distinguish individual basal bodies and cilia. It is therefore not clear how the authors can confidently quantify the number of basal bodies in each condition to construct the graph at the bottom of the figure. In addition, it would be interesting to label the basal body with a centriolar marker to better define the basal body.

Figure 5: not clear why the graph on the lower right does not include the control at 3 and 6 hrs? Is it because the number is too high and difficult to quantify?

References: I would like to draw the authors' attention to studies of deciliation in Paramecia that could be cited in the introduction or discussion of the conservation of this pathway through evolution.

Significance

The mechanisms of deciliation and re-ciliation have mostly been studied in protozoa (Chlamydomonas, Paramecia) or in primary ciliated cell cultures. Only a few studies have described deciliation in multiciliated cells, such as sea urchins, or physiological deciliation in the oviduct. The Xenopus deciliation system described here has already been used to determine the dynamics of IFT proteins during ciliogenesis or to define the ciliary proteome. In this study, the authors go one step further by describing more precisely which part of the cilium is shed upon induction of deciliation and the dynamics of the recruitment of the Tip and of the TZ proteins.

This study provides a completely new perspective on the deciliation process:

1. the authors show that, contrary to what is generally accepted from protozoan studies, the deciliation process, in Xenopus multiciliated cells, expels the TZ, leaving only the basal body in the cell;
2. While ciliogenesis is described in various models to begin with the formation of the TZ, in this Xenopus system the TZ maturates after the onset of axonemal elongation, calling into question the precise function of the TZ in axonemal elongation. The observations could be further strengthened by analyzing more TZ proteins to better understand the time course of events involved in the deciliation-reciliation program.

The protocol used to deciliate Xenopus multiciliated cells has been described in previous manuscripts. Its use here reveals striking differences in the deciliation-reconciliation pathways from what is known in the field. It provides new conceptual perspectives for researchers working on the basic mechanisms of ciliogenesis. Note that, as a geneticist and specialist in ciliogenesis using various model organisms, I am not fully competent to critically evaluate the mathematical models developed in this study.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

In this manuscript entitled "Machanisms of cilia regeneration in Xenopus multiciliated epithelium in vivo", the authors mostly focus on the question, whether TZ (transition zone of cilia) plays an essential role for ciliogenesis during cilia regeneration in multiciliated cells. They used Xenopus embryo as a system to examine this question. While cilia regeneration has been actively studies in unicellular green algae, Chlamydomonas reinhardtii, the mechanism of cilia regeneration is not known yet. Their approach is to investigate cells after deciliation by calcium shock, based on a TZ protein B9D1, as well as ultrastructure observation using conventional electron microscopy.

The authors observed loss of signal from B9D1 and H-shaped objects, which is typical for TZ, upon deciliation induced by calcium and also during the following re-growth of cilia. Based on these experiments they concluded that TZ formation is not necessary for cilia regeneration in multiciliated cells, differently from Chlamydomonas. They further conducted experiments to pursue source of component proteins for re-generation. They compared CHX-treated cells (lacking new protein production) and CHX/MG132 (reduced protein degradation) treated cells to find how the massive amount of protein components upon re-ciliation for multiple cilia will be supplied and regulated. This reviewer found the results of the experiments clearly presented and conducted properly.

The work would have significant impact in the cilia community, if the conclusion is correct. This reviewer, however, has a concern about the authors concluding the presence/absence of TZ, based on only B9D1 and the H-shaped body among nine doublet microtubules. First, is it really established how the structure of Xenopus embryo TZ is? While Chlamydomonas is well known to have a H-shaped TZ, other species have different form inside the 9+0 doublet, or no feature (Comparison of TZ from various species in Dennis Diener https://doi.org/10.1016/B978-0-12-822508-0.00007-1). Fig.2B of this manuscript shows visible densities in the panel "Pre", but it does not look like an H-shape. The tomogram of TZ before deciliation seems clearer (but judging from wavy MTs and membrane in this tomogram, there could be unevenness of embedding and staining), while the tomogram after deciliation is thin and does not cover the entire width. Therefore it is not sure that absence of TZ can be concluded. If the author claims Xenopus embryo cilia have a H-shaped TZ, they have to provide multiple micrographs (ideally tomogram or serial section TEM to cover the entire TZ structure) and/or past literature on Xenopus embryo TZ. B9D1 is likely a membrane associated protein (according to their deciliation by detergent and mechanical force). This may mean B9D1 is located on or near the membrane, in vicinity to TZ, and thus binds to TZ after the main part of TZ is built. In this case, it is risky to judge presence of TZ based on B9D1. Also in this point, TEM imaging will be helpful to confirm the authors' conclusion.

Their discussion about length/number of cilia and force generated by cilia is interesting, but in the context of this research, this reviewer is skeptical about its value. The calcium induced deciliation is not a physiological phenomena, but an artificial event (please correct if I am wrong). The argument how length and number of cilia are regulated upon deciliation makes sense only in case deciliation happens regularly and the species must optimize themselves to survive. The argument about possible passway of protein transport to control ciliary number and length (Line408-) seems, although it is an interesting topic in general, not suitable in this manuscript. For this reviewer's view, it is relatively straightforward to interpret the result of cilia number/length under normal growth, without new protein expression (CHX), with protein degradation blocked. Cilia will extend when components are provided. Growth will slow down when it is exhausted. Existing cilia start degrading, when they lack proteins, which are necessary for turn-over. With the current experimental output, there is no point to describe redistribution of proteins.

Minor points:

Line65: do they mean "selected few basal bodies"?

Line73: extracellular flow is not limited to developmental system.

Line124: alpha-tubulin signal and SEM image

Line139: Could you define explicitly the two hypotheses?

Line164: 10,31-33 are not suitable citation for the location of calcium induced deciliation in Chlamydomonas. cite Sanders and Salisbury JCB 108, 1751

Line181: Later -> latter

Line195: by mechanical shearing, B9D1 remained with cilia. They concluded that TZ stays with the axoneme by deciliation. How can they exclude the possibility that mechanical separation works differently from calcium shock?

Line214: 1.33uM -> 1.33um

Significance

The work would have significant impact in the cilia community, if the conclusion is correct. Their discussion about length/number of cilia and force generated by cilia is interesting, but in the context of this research, this reviewer is skeptical about its value.

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Manuscript number: RC-2024-02470

Corresponding author(s): Milán, Somogyvári; Csaba, Sőti

1. General Statements

We thank both Reviewers for their constructive comments, and we hope that our reply clarifies the concerns and the revised manuscript will be recommended for publication in Review Commons affiliated journals.

2. Description of the planned revisions

As the first major concern, Reviewer #1 raised the question of resolving the link between SIR-2.1 and HSF-1.

In order to address this issue, we plan to utilize a two-way approach:

• We plan to check the acetylation status of C. elegans HSF-1 using Mass Spectrometry.
• We aim to evaluate changes in its promoter binding by utilizing a form of Chromatin Immuno-Precipitation combined with RT-qPCR.
• As an alternative approach, we plan to utilize the cbp-1GoF mutant strain MH2430, that was described to acetylate HSF-1 and therefore change its transactivational function (Barrett & Westerheide, 2022). We'd like to see whether this can achieve a phenocopy of the sir-2.1(null) genetic background - on the level of the lipolysis phenotype and HSF-1-acetylation. These experiments are to be performed in both wildtype and sir-2.1-silenced animals under fed and starved conditions.

The second issue raised by Reviewer #1 was to confirm that the miR-53 activity on atgl-1 3' UTR is crucial for the described phenotype.

Our planned solution for this issue is to create novel C. elegans strains that expresses green fluorescent protein under the regulation of the atgl-1-promoter and with either the wildtype 3'UTR of atgl-1 attached or a mutated one, to which mir-53-binding is not possible. Through fluorescence microscopy experiments involving fed and starved animals, we hope to be able to sufficiently assess the necessity of mir-53 activity to changes in atgl-1 expression and function.* *

The following minor concern of Reviewer #1 regarding the Results section are addressed here:

(b) Supporting our ORO results with using the transgene idrIs1[dhs-3p::dhs-3::GFP] to label lipid droplets.

After acquiring the LIU1 strain harboring the aforementioned transgene, we plan to validate some of the results gained from ORO experiments.

• *

Reviewer #2 brought into our attention several concerns that need to be addressed:

1- Firstly, it was mentioned that the overabundance of histograms and the lack of indications on the representative images makes it difficult for the reader to assess the information on the panels.

We plan to include more images of stained worms while also indicating the changes that the histograms are meant to show. We hope these efforts will make our results more convincing.

2- Reviewer #2 mentioned that some of the results shown in our manuscript was already published in Zaarur N et al, 2019.

We thank the Reviewer for bringing this issue to our attention. We'd like to however point out that the mentioned paper by Zaarur et al. is indeed referenced in two places in the Introduction of our manuscript: first, when highlighting that longevity pathways, fat metabolism and lifespan determination are interconnected (line 56), and second, indicating that ATGL-1 mediates longevity in response to dietary restriction and reduced insulin-like signaling (line 65). The regulation of ATGL-1 by starvation and of lipid mobilization by ATGL-1 are not among the novel results of this study. The novelty of our data lies mainly in HSF-1 being involved through specific microRNAs - which to the best of our knowledge has not yet been published. We agree with Reviewer #2 that the visual representation of the data in Zaarur et al. is more pleasing, therefore we plan to incorporate representative images here as well.

5- Selection of mircoRNA genes

It was rightly pointed out by Reviewer #2 that mir-53 was reported to be upregulated by HSF-1 upon heat-shock, while there's no mention of it behaving so upon starvation. However, we considered examining it a potentially worthwhile direction given that in C. elegans it is a common observation that various stresses lead to the activation of similar/overlapping stress-response pathways. As seen at Figure 5E, our data supports the idea that mir-53 expression responds to starvation, as its pre-miRNA levels are elevated by starvation in sir-2.1-silenced animals. As for mir-60 and mir-75, we indeed do not have evidence for them being regulated by HSF-1, however their mutants have been associated with reduced body fat content (Brosnan, Palmer & Zuryn, 2021), thus we considered them also to be potential candidates for the role of mediator in the observed lipolysis phenomenon. We aim to make our reasons for choosing these specific miRs clearer in the manuscript.

3. Description of the revisions that have already been incorporated in the transferred manuscript

In accordance with the minor concerns of Reviewer #1, the following changes have already been made to the manuscript:

In the Abstract, two sentences were changed: (a) "In starving worms, a SIR-2.1-dependent suppression of specific HSF-1 transcriptional activity leads to the inhibition of lipolysis through mitigating miR-53-expression" & (b) "potential crosstalk" was introduced to the last sentence in order to better reflect the nature of our results.

In Introduction, the start of the first sentence was changed to "Lipids are a diverse group of cellular constituents".

In Results, the suggested changes in the C. elegans nomenclature were performed (a), where we substituted "sir-2.1 knockout" with "sir-2.1(null)" and "hsf-1 knockout" with "hsf-1(null)".

The following concerns of Reviewer #2 are addressed in the text of the transferred manuscript:

4- Stage of synchronized populations in the starvation protocol.

We thank the Reviewer for highlighting this issue. The life-stage of the animals should indeed be specified at this particular method. It may have been omitted due to RNAi treatments being described just above, where it is mentioned that L4 animals are washed onto RNAi plates where they spend 2 days. Any starvation protocol starts only after these RNAi treatments, since almost all our experiments include some form of RNAi. Therefore, in any trials that do not have RNAi in them, we still only applied starvation from 2 days after L4 stage - for comparability's sake. This issue is clarified in the revised manuscript.

4. Description of analyses that authors prefer not to carry out

The third major concern mentioned by Reviewer #1 is the epistatic relationships between our novel regulatory pathway and the KIN-1/PKA.

We'd like to thank Reviewer #1 for turning our attention towards the issue. We feel that the phosphorylation of ATGL-1 by KIN-1 - most probably on Serine 303 - was well established in Lee JH et al, 2014, as seen at Figure 5B - which, naturally, must occur downstream from any transcriptional regulation done by the SIR-2.1-HSF-1-MIR-53 axis. Nevertheless, it would be interesting to see if a non-phosphorylatable ATGL-1 will not support lipolysis upon starvation even if its mRNA expression is activated - which is something that was not tested for in Lee JH et al, 2014. However, since this was not the main subject of our manuscript - more of an addition to ATGL-1-regulation - while our work focused on the regulatory axis going through HSF-1, we do not consider it crucial to perform further experiments aimed at the KIN-1/PKA-mediated regulation of ATGL-1.

The following minor concerns of Reviewer #1 regarding the Results section are addressed here:

(c) Fatty acid profiling in the different experimental conditions.

Even though we feel the potential significance of such data, it does not fit under the scope of our current study. We feel this to be better fitting for a later, follow-up research project.

(d) Concern about double RNAi treatment in Figure 2D.

In such particular experiments, where the nematode strain is already a mutant one (where RNAi can only affect intestinal cells), it would require time-consuming crossings with the sir-2.1 and hsf-1 mutant animals. For this reason, we opted to use double RNAi, since according to literature - as well as to our previous experience - double RNAi can be a reliable method to silence the expression of two genes simultaneously. Since the effectivity of RNAi can be influenced by dosage, we compensated for this by mixing the single RNAi treatments with Empty Vector containing bacteria. The results themselves show the silencing-treatments to be effective - to a similar extent as the single RNAi treatments seen in Figure 2A.

(e) Improving statistical power.

We agree with Reviewer #1 that in some cases the addition of biological replicates may have the potential to strengthen our conclusions. We argue however, that even though in each case we applied statistical post-hoc tests in order to avoid a type I error, going above the customary 3 biological replicates at each and every experiment would increase the probability of such an error occurring.

The following concerns of Reviewer #2 are addressed here:

3- Reviewer #2 inquired about RNAi efficiency and the usage of knock-out mutants.

Here, we'd like to highlight that throughout the manuscript we utilized sir-2.1(null) mutants in Fig. 1A-B, Fig. 2A-B, Fig. 3D & Fig. S3A; while using hsf-1(null) mutants in Fig. 2A, Fig. 3B & Fig. 8D-F - among other mutants and transgenics. The list of strains can be found in Supplementary Table 1. Regarding the hsf-1(RNAi), it is a strain used for silencing hsf-1-expression reliably in the past by the lab of origin at ELTE University (Barna et al. 2012 BMC Developmental Biology). The silencing of hsf-1 does not lead to any noticeable phenotypes, but it does fully eliminate hsp-70-induction by heat-shock or starvation as shown on Fig. 5A-B.

6- KIN-1 & KIN-2's role and place in lipolysis regulation

As Reviewer #2 pointed out, both a mutation in kin-1 and kin-2 seemingly lead to the inhibition of lipolysis. However, since kin-2 codes for the regulatory subunit of KIN-1/PKA, a Loss of Function mutation in it leads to a constitutively active PKA - which in turn is expected (among other outcomes) to continuously phosphorylate and thus stabilize ATGL-1. In accordance with this, loss of KIN-1 resulted in an inability to utilize lipid-reserves - therefore the ORO staining levels of these mutants remained similar to wildtype and fed state even upon starvation - while loss of KIN-2 lead to a significantly decreased basal lipid staining, that could not be further decreased by starvation. We argue that the lack of any effect of sir-2.1(RNAi) (or hsf-1(RNAi)) on these phenotypes, while atgl-1(RNAi) was able to revert the kin-2(null)-related basal lipid loss, strongly supports the epistatic relationships proposed.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Review

An intestinal Sir2-HSF1-ATGL1 pathway regulates lipolysis in C. elegans

In the manuscript by Somogyvári et al., the authors focus on the differences between the fed and the fasted state using C. elegans. In particular the authors find that in the fasted state, the C. elegans SIRT1 ortholog, SIR-2.1, activates lipolysis by upregulation of ATGL-1. Further studies show that in fed worms regulation occurs in the intestine by HSF-1, ATGL-1, and the microRNA system. In contrast, in fasted worms, SIR-2.1 functions with the miR-53 microRNA to affect lipolysis and hsf-1. Further experiments attempt to implicate protein kinase A and proteostasis. Ultimately, the authors attempt to invoke a model for stress resilience and aging. Overall, the data as presented is not very convincing. All of the data is presented as histograms which show only mild effects. Images that are shown are not convincing. Some of the data has been previously published (mentioned below). Therefore, the manuscript needs extensive revisions prior to resubmission and should address the comments below.

1. why is most of the data presented as a histogram?

Why are there not representative images that help readers examine the results? /

For example figure 1A does not really show anything but could guide the reader. The worm images throughout the manuscript do not give any indication of what the authors want the data to show the reader. 2. some of the data has already been published.

Mol Metab. 2019 Sep:27:75-82. doi: 10.1016/j.molmet.2019.07.001. Epub 2019 Jul 5. Nava Zaarur et al. fig 1

' ATGL-1 is up-regulated by fasting of C. elegans. (A) Wild type (N2) and atgl-1::gfp worms w

control (Fed) and Fasted groups and stained with Oil Red O.<br /> (B) Triglyceride content was measured in Fed and Fasted groups of N2 and atgl-1::gfp worms. (C) RNA was extracted from Fed and Fasted groups, and atgl-1 mRNA levels were measured by qRT-PCR; actin-1/3 was used for normalization. (D) Fed and Fasted L4 stage atgl-1::gfp worms were visualized by fluorescence microscopy (200X, equal exposure times). Bar e 50 mM. (E) Quantification of the results shown in panel D by ImageJ (10 randomly selected worms per group). '

this is not referenced or discussed and more convincing than simply a histogram 3. - why is there no analysis with mutants and simply Rnai? for example why is sir2 mutant not used.? - does the rnai show any phenotypes? ex hsf-1 rnai = hsf mutant? - Do you know the knockdown efficiency for the rnai clones? 4. Starvation protocol

560 Synchronized populations were washed 3 times with M9 buffer and placed either on 561 plates containing bacterial food source, or empty plates for 18 hours.

- what stage were the Synchronized populations?

1. 1. Brunquell, J., Snyder, A., Cheng, F. & Westerheide, S. D. HSF-1 is a regulator of 699 miRNA expression in Caenorhabditis elegans. PLoS One 12, 1-24 (2017) This is the reference used to define the connection to micrornas. However, this manuscript describes miRNAs induced by heat shock. How does heat shock connect to starvation? The fed or the fasted state? Overall, the rationale for the specific microRNAs shown in the manuscript example mir-53 is unclear.

2. Figure 6. The protein kinase A KIN-1 affects lipolysis and ATGL-1 function 330 downstream from SIR-2.1 and HSF-1.

   • there is no difference between kin-1 knockout and Kin-2 knock out-so how does one say that it is only Kin-1?
   • where are the differences between fig 6b and 6c?
   • 'The complete 306 inhibition of lipolysis in the absence of sir-2.1 or kin-1 suggests that Sir2 and PKA pathways 307 are equally indispensable and cooperate in lipolysis regulation in the wildtype'
   • does data really show this? ?- not much difference between kin-1 and kin-2- can you really separate the requirements?

Significance

Overall, the data as presented is not very convincing. All of the data is presented as histograms which show only mild effects. Images that are shown are not convincing. Some of the data has been previously published (mentioned below). Therefore, the manuscript needs extensive revisions prior to resubmission and should address the comments below.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary.

This study elucidates the contribution of sirtuin 1 ortholog SIR-2.1 in lipid mobilization upon starvation in the nematode Caenorhabditis elegans. The authors claim that HSF-1 controls the expression of adipose triglyceride lipase ATGL-1 in C. elegans gut. Furthermore, they show that SIR-2.1 modulates ATGL-1 activity by regulating the expression of microRNA miR-53 in an HSF-1-dependent manner. The manuscript also describes the interplay between SIR-2.1/HSF-1 and protein kinase A (KIN-1/PKA) in modulating ATGL-1 activity and proteostasis. Finally, the authors claim that lipid mobilization correlates with HSF-1-dependent proteostasis according to the feeding state of the organism.

Major concerns.

This manuscript consists of at least three parts that are relatively connected: - The impact of SIR-2.1 deficiency on ATGL-1 expression. Here, the main novelty is that HSF-1-dependent regulation of miR-53 defines atgl-1 expression during starvation. - The contribution of KIN-1/PKA in lipid mobilization and ATGL-1 activity downstream SIR-2.1 and HSF-1. This link was partially described in a few previous studies (e.g., Lee JH et al, 2014).<br /> - The contribution of HSF-1 in intestinal proteotoxic stress and fat metabolism. The role of HSF-1 in proteostasis has been well documented, whereas its participation to lipid metabolism has been described in a few studies (e.g., Oleson BJ, 2024).

The authors provide novel findings that give a better picture of the signaling cascade regulating these biological processes.

1. However, this study does not conclusively resolve the link between SIR-2.1 and HSF-1. Does Sirtuin 1 influence HSF-1 through histone deacetylation and, therefore, HSF-1 deposition to target genes? Or does Sirtuin regulate HSF-1 acetylation state and therefore its activity? The authors attempted to address these questions with some experiments (Figure 5), however the data are indirect evidence.
2. Furthermore, microRNAs have multiple targets with various biological functions. Although the authors provide first line of evidence demonstrating the impact of miR-53 on starvation-induced lipolysis, it may be important to confirm that the miR-53 activity on atgl-1 3' UTR is crucial for the described phenotype. Thus, the authors may consider to generate C. elegans strains carrying an atgl-1 3' UTR that is not recognized by the endogenous miR-53 (OPTIONAL).
3. The role of KIN-1/PKA and ATGL-1 was previously reported (Lee JH et al, 2014) as mentioned in the manuscript. In the submitted manuscript, the authors tried to link KIN-1/PKA, ATGL-1, SIR-2.1 and HSF-1. The authors suggest that "KIN-1 acts downstream from the SIR-2.1 pathway" and "KIN-1 acts downstream of ATGL-1 post-transcriptional regulation". Most of the authors' conclusions are based on RNAi experiments. Could the authors support their claims by providing evidence that the downstream substrates are differentially posttranslationally modified according to the experimental conditions (starvation vs feeding)? Apart from RNAi methods, could the authors support their claims by using non-phosphorylatable ATGL-1 mutants?

Minor concerns.

Abstract.

(a) "SIR-2.1 suspends a miR-53-mediated suppression...". Please adjust the text to make it more understandable.

(b) "Our findings reveal a crosstalk between proteostasis and lipid/energy metabolism, which may modulate stress resilience and aging.". Which evidence do the authors have that this newly identified crosstalk influences aging? Figure 8 is not sufficient to make such a strong claim.

Introduction.

I would encourage the authors to re-word some of their sentences. For example, "lipids are diverse constitutes" sounds strange.

Results.

(a) Please, keep in mind the internationally accepted C. elegans nomenclature. For example, substitute "sir-2.1 knockout" with "sir-2.1(null)".

(b) The authors used Oil Red O (ORO staining) to assess lipid content in nematodes. However, the method has a few limitations and the accurate assessments of fat stores may be variable across experiments. One option is that the authors corroborate their findings with another approach. For example, they may consider to use the transgene idrIs1[dhs-3p::dhs-3::GFP] to label lipid droplets in intestinal cells.

(c) The authors assessed free-fatty acid content in fed and starved animals. It may be informative to report the individual fatty acid molecules that are mobilized in the different experimental conditions.

(d) It is always difficult to obtain reproducible results by using two RNAi clones (Figure 2D). The authors should corroborate their results with sir-2.1 and hsf-1 mutant worms.

(e) For some of the experiments, the statistics may be improved. Since some panels show tendency towards statistical significance (e.g., 8F), it may be important that the authors strengthen their analyses with additional biological replicates. This would help to consolidate their findings and conclusions.

Significance

This study reports how Sirtuin 1 can modulate ATGL-1 expression by regulating a microRNA (miR-53). It remains unclear if it is through a direct interaction or via epigenetic remodelling of histone acetylation of target genes. By building up on previous studies, the authors provide additional molecular players that take part in lipid mobilisation during starvation.. The audience can be defined as "specialised" and "basic research".

My fields of expertise are: metabolism, aging and epigenetics. I work with mice and C. elegans.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

1. General Statements [optional]

This section is optional. Insert here any general statements you wish to make about the goal of the study or about the reviews.

We thank the reviewers for their insights and helpful suggestions on the manuscript. Based on these, we have prepared a revision plan for this manuscript, which is outlined below. We believe these revisions will improve the overall quality of the manuscript.

2. Description of the planned revisions

Insert here a point-by-point reply that explains what revisions, additional experimentations and analyses are planned to address the points raised by the referees.

• *

Reviewer #1

(Evidence, reproducibility and clarity (Required)):

Summary:

This study builds on previous work from the same group, where they use Drosophila photoreceptors as a model system to investigate the role or ER-plasma membrane contact sites in an in vivo setting. The authors recently described a role of the ER-PM contact site protein dEsyt in regulating photoreceptor function in Drosophila. In this follow-up study, they explore whether this function of dEsyt is connected Ca2+ signaling downstream of photoreceptor activation. Using a dEsyt mutant that should be unable to bind Ca2+, they find that Ca2+ to some extent is required for dEsyt localization, membrane contact site formation and photoreceptor function.

Major comments:

The use of photoreceptor cells in Drosophila is an elegant model system that enable studies of membrane contact sites and associated proteins in a native condition. The data presented by the authors clearly shows that these structures are important for photoreceptor function, and that dEsyt plays a role at these sites. However, this was already known from previous studies by the same group. When it comes to whether these contacts are sensing Ca2+ changes and if these changes are acting through dEsyt, which is the focus of the current manuscript, the results are unclear to me and would need to be clarified by the authors both in text and with new experiments.

1) What is the role of cellular Ca2+ signaling in the regulation of dEsyt function? There are several aspects here that needs to be clarified. 1) How is WT dEsyt localization regulated by Ca2+? This could for example be evaluated in the mutant flies used in Fig. 1 (trpl302; trp343), where lack of light-induced Ca2+ influx would be predicted to result in a localization of dEsyt that resembles that observed for dEsytCaBM. 2) Is Ca2+ important for dEsyt localization, lipid exchange or both? The authors express a version of dEsyt with mutation made in all three C2 domains. In mammalian E-Syts, Ca2+ binding to the C2A domain is important for lipid exchange while binding to C2C (in E-Syt1) is important for interactions with lipids in the plasma membrane. Using more carefully designed mutants will allow the authors to determine how Ca2+ regulates dEsyt function in vivo. In addition, the authors must show experimentally that the mutant dEsytCaBM is unable to bind Ca2+ (could e.g. be done by acute Ca2+ changes in the cell-based model used in Fig. 3). Writing that "This transgene carrying a total of nine mutations should render the protein unable to bind calcium" (p. 6, line 173) is not sufficient.

1) How is WT dEsyt localization regulated by Ca2+?

We agree that further experimental evidence would be helpful in establishing the significance of cellular Ca2+ signaling in the control of dEsyt function. As suggested by the reviewer, the localization of wild type dEsyt will be examined in the mutants: norpAP24 (PLC null mutant) and trpl302; trp343 (protein null mutants of TRPL and TRP channels respectively) in which light induced calcium influx is eliminated. These data will be included in the revision.

2) Is Ca2+ important for dEsyt localization, lipid exchange or both?

We have already performed experiments to address the question of how important calcium binding to dEsyt is for lipid transport at the ER-PM interface in Drosophila photoreceptors. This results indicate a previously unexpected role for lipid exchange and will be included in the revision.

3). Writing that "This transgene carrying a total of nine mutations should render the protein unable to bind calcium" (p. 6, line 173) is not sufficient.

We concur with the reviewers that at present we do not have experimental data to demonstrate that dEsytCaBM can't bind Ca2+. However, as Reviewer 4 pointed out, it will be challenging to demonstrate this experimentally. A direct proof would only come from measurements of the calcium binding affinity of dEsyt (which involves protein purification that is beyond the scope of the current work). An indirect demonstration would be any cellular or in vivo experiment. In addition to the in silico analysis already included in Fig 2 C-F, we propose the following to provide additional evidence to strengthen our in silico analysis: Use AlphaFold model to demonstrate that the arrangement of the calcium binding residues in the C2 domain of dEsyt is compatible with Ca2+ binding.

2) The localization of dEsyt shown in Fig. 3B is a bit confusing. First of all, I would recommend including markers of the ER and the plasma membrane, because without these it is difficult to make statements about the localization of dEsyt to these structures.

As suggested, to better appreciate the localization of dEsyt in photoreceptors, we will perform colocalization of dEsyt with markers of the PM (Rhabdomere) and ER (Sub Microvillar Cisternae).

Second, it appears that WT dEsyt localize to the reticular ER, and that the CaBM version localize to the plasma membrane. This is somewhat opposite to mammalian ESyts, where mutations that prevent Ca2+ binding either had no effect (for ESyt2) or prevented (for ESyt1) the interaction with the plasma membrane. It also appears different from the localization in vivo (Fig. 3C). Clarifying this will be important. It will also be important to connect this localization to changes in Ca2+ and not just to the localization of a mutant that may or may not be deficient in Ca2+ binding (see comment above).

In considering this comment, we need to bear in mind the following:

• Mammalian cells have three genes that encode for Esyt: Esyt 1, 2 and 3 whereas the Drosophila genome encodes only a single gene for Esyt.
• In terms of sequence similarity and structure, dEsyt and hEsyt2 are very similar. However, in contrast to hEsyt2 and hEsyt3, which localize to the plasma membrane (PMID: 17360437), dEsyt acts like hEsyt1 and localizes to the ER-PM junctions.
• A single study (PMID: 27065097) has shown that the SMP domain of Esyt1 can transfer lipids in an in vitro assay. In our studies, we have noted an unexpected function for the SMP domain of dEsyt for in vivo function as measured through phenotypes in the eye (data will be presented in the revised ms).
• While knocking out the single dEsyt in Drosophila photoreceptor neurons results in phenotypes (Nath et.al PMID: 32716137) to date, knocking out all three Esyts in mammalian cell culture models or mice has not revealed an in vivo Bearing these points in mind it may not be reasonable to expect every observation on mammalian Esyt to be recapitulated in the fly system or vice versa. 3) I don't fully understand the time course of events. The authors show that dEsytCaBM is mislocalized already at day 1 in dark-reared flies (Fig. 3C) but this mislocalization is not accompanied by a change in MCS density or gap distance, and consistently does not influence the localization of RDGB. The authors next expose the flies to constant light illumination to trigger Ca2+ dependent signaling, and this leads to mislocalization of RDGB, perhaps indicating changes in MCS (this is not shown). From these results it is difficult to know what the role of dEsyt is. It would be necessary to also show a control where Ca2+ signaling is not induced, e.g. a parallel dark-control (same number of days but no illumination).

It is important to remember that even complete loss of Esyt does not result in altered MCS or mislocalization of RDGB on day 1 post eclosion. This has been published by us previously (Nath et.al PMID: 32716137). Since we show in this manuscript that dEsytCaBM exerts a dominant negative effect when expressed in wild type and phenocopies dEsytKO, one might expect expression of dEsytCaBM to also lead to altered MCS density and mislocalization of RDGB by 6D constant light.

Bearing this in mind, we will incorporate the following data in the manuscript: Addition of MCS density in dEsytKO photoreceptors at Day1 in Figure 3C.

• Electron Microscopy to check MCS density in Rh1>dEsytCaBM at Day 6CL with appropriate control genotypes.
• Confocal Imaging: RDGB staining in Rh1>dEsytCaBM- Day 6CD reared flies with appropriate control genotypes- dark control where only reduced Ca2+ signaling is induced due to dark noise or spontaneous PLC activation. This is particularly important given that the authors show in Fig. 1 that preventing Ca2+ influx had a dramatic impact on MCS density even at day 1 (which is in sharp contrast to dEsytCaBM-expressing flies, that show normal morphology at day 1, which rather implies that dEsyt is not a major Ca2+ effector).

In thinking about this comment, it is important to bear in mind the details of the experimental paradigm in use in each of the experiments while drawing comparisons between the observed results. It is to be noted that throughout the manuscript dEsytCaBM is expressed selectively in photoreceptors using the Rhodopsin enhancer which drives expression of the transgene during late eye development. By contrast, in germ line mutant strains such as trpl302;trp343 the channels are blocked throughout development. Thus the phenotypes of trpl302;trp343 might be broader than that of expressing dEsytCaBM. Therefore, mutating the calcium binding residues of dEsyt and expressing it using Rh1 enhancer at a specific developmental time window might not have the same impact on the contact site density as completely blocking the major calcium permeable channels, TRP and TRPL that is important to sustain the ongoing phototransduction cascade all through the development.

4) The experiments done in dEsyt KO flies are important, and here the authors show that dEsyt1 could to some extent rescue all phenotypes. Some results are a bit puzzling. For example, dEsyt1CaBM localization in dEsyt1 KO flies is identical to that of WT dEsyt (Fig. 5C), which is in sharp contrast to the data shown in Fig. 3C. What is the reason for this? I would have anticipated the opposite (i.e. that in WT flies, dEsytCaBM can form dimers with endogenous dEsyt through SMP-domain interactions which may have an impact on its localization and the function of endogenous dEsyt, but that in the dEsyt KO cells, dEsytCaBM would show a different localization due to the lack of endogenous dEyt to interact with). It is important to clarify as one of the major observations here is that dEsytCaBM no longer localize to MCS. Since the CaBM version of dEsyt could rescue, to some extent, MCS density and delay photoreceptor degeneration, this implies that Ca2+ may not be required for regulation of dEsyt function or that the mutant is still able to partially bind to Ca2+.

The localization shown in Fig 5C is not of dEsytCaBM in dEsytKO photoreceptors but the localization of RDGB in Rh1>dEsytCaBM; dEsytKO at Day 1 (Figure 5C i) and as a function of age and illumination- Day 6CL (Figure 5C ii).

One experiment that would help the authors determining the function of dEsyt in vivo would be to use a mutant that lacks functional SMP domain (ideally also with and without mutations in the C2-domains).

There is information available to address the question of how the lipid binding module, SMP is important to render dEsyt functional at the ER-PM interface in Drosophila photoreceptors. The same will be included in the revision.

5) PLC activation typically couples to rapid signaling and involved hydrolysis of PIP2 and release of Ca2+ from the ER. Mammalian Esyts also require PIP2 for plasma membrane binding (through interactions with C2-domains), so constitutive PLC activity would be expected to impair ESyt localization to MCS. Here, the authors expose flies for days of constant illumination. How does this influence plasma membrane PIP2 levels and could this be of relevance for how data is interpreted?

This is an interesting question from the reviewer. However, we would like to clarify the fact that constitutive activation of PLC is different from constant activation of PLC during illumination. Flies have robust mechanisms for controlling PLC turnover and PIP2 levels during continuous illumination and Ca2+ is a key regulator of this process; the underlying mechanisms have been described by Raghu and Hardie in multiple past papers (PMID: 11343651, PMID: 15355960). This is why, apart from adaptation, flies grown in constant light for many days do not show electrophysiological defects and neither do they undergo retinal degeneration. We will however measure the kinetics of PIP2 resynthesis in (i) wild type (Day 1 vs Day 6CD vs Day 6CL) and (ii) Control, Rh1>dEsyt and Rh1>dEsytCaBM (Day 1 vs Day6CL). This might reveal some interesting insight into the mutants.

Do the authors know whether the CaBM mutant has reduced affinity for PIP2?

The ability of wild type dEsyt to bind PIP2 has not been determined. We will test this and if it does so, the impact of CaBM on PIP2 binding can be tested.

Minor comments:

• The overexpression of WT dEsyt had a dramatic impact on MCS density and gap distance, while expression of dEsytCaBM did not. If these contacts are important for photoreceptor function, is it not surprising that such a dramatic change in photoreceptor structure was without effect on function? This should be further discussed. The establishment of more contact sites and reduction in contact site distance in Rh1>dEsyt::GFP photoreceptors is likely indicative of the proposed tethering role of the protein at the ER-PM MCS. Increase in contact site density or reduction in distance need not directly parallel to the increase in the levels of MCS proteins that are expressed at these contact sites to enhance the ongoing signal transduction. We will test this idea proposed by the reviewer and include the following data in a revision to strengthen our statement:

• RDGB levels in control vs Rh1>dEsyt::GFP - Western blot

• Electroretinograms from the genotypes indicated above as a functional readout of the ongoing signaling cascade.
• PIP2 kinetics in control vs Rh1>dEsyt::GFP to understand if establishing more contact sites can enhance the replenishment of the lipid at the PM. 2) How is quantification of MCS density and gap distance influenced by retinal degeneration (e.g. induced by dEsyt KO)?

Wherever we have analyzed MCS density or gap distance, these experiments have been done in flies at ages prior to the onset of retinal degeneration defined as collapse of the microvilli of the rhabdomere. Therefore, our measurements of MCS density and gap in this paper are not affected by retinal degeneration.

3) The graphical abstract is a bit confusing. It seems to suggest that changes in dEsyt is a consequence of ageing and does not show any role of this protein in photoreceptor function. I think that the abstract could be improved to more clearly highlight the findings in the manuscript. For example, it doesn't at all show the difference in localization between WT and CaBM.

We will modify the graphical abstract.

4) P. 5, line 135 the authors state that "The tethering and lipid transfer activity of mammalian Esyts are reported to be influenced by Ca2+". This is a massive understatement. Ca2+ is a critical regulator of Esyt function in mammalian cells.

The statement will be modified.

5) In figure legend 1B and C: correct µM to µm.

Changes will be incorporated as per the suggestion.

6) In figure legend 2A: should be red rectangles and not black rectangles.

Changes will be incorporated as per the suggestion.

7) In Fig. 2B: specify which isoform of human ESyt that is shown.

Changes will be incorporated as per the suggestion.

8) In Fig. 2C: do the authors mean D374 or D384 (as indicated in Fig. 2A)?

Changes will be incorporated as per the suggestion; the residue is D374.

Significance

Light-induced signal transduction in photoreceptor cells involves Ca2+ influx and signaling and also depends on correct formation of ER-plasma membrane contact sites. In mammalian cells, the Esyts (esp. Esyt1 and Esyt2) localize to ER-PM contacts in a Ca2+-dependent manner, and the ion has dual effects in both enriching the protein at the membrane contact sites and in promoting lipid transport. Mammalian Esyts form homo- and heterodimers, and the properties of the dimers depends on their composition (PMID: 26202220). Drosophila only have one Esyt (dEsyt) which is structurally most similar to mammalian Esyt2, and the authors have previously shown how this protein is required for photoreceptor function (PMID: 32716137), although the role of Ca2+ was not investigated in that study. However, an earlier study has shown that mutations of all Ca2+-coordinating residues in dEsyt impairs protein function in Drosophila neurons (PMID: 28882990), so a similar Ca2+-dependence in the retina would be expected. The results from the present study confirm the requirement of Ca2+ signaling for dEsyt function, and extends this Ca2+-dependent regulation to also involve photoreceptor-induced Ca2+ signaling, which corroborates many other studies showing the requirement of Ca2+ signaling for the regulation of Esyt function in mammalian cells (e.g. PMID: 23791178; PMID: 27065097; PMID: 29222176; PMID: 26202220; PMID: 24183667; PMID: 30589572). As such, the results from this study represent an incremental step towards understanding Esyt function in vivo. These results would be of greatest interest to researchers working of photoreceptor function, and of some interest to a broader audience working on membrane contact sites and signal transduction. My own background is in mammalian cell biology, with a focus on lipid and Ca2+ signaling and inter-organelle communication. I have limited understanding of the model system used here (Drosophila photoreceptor cells).

---

We would like to provide an alternative perspective on the reviewer’s view that “As such, the results from this study represent an incremental step towards understanding Esyt function in vivo.”

We are well aware of the content in several studies of Esyt in mammalian cells including the ones cited by the reviewer (e.g. PMID: 23791178; PMID: 27065097; PMID: 29222176; PMID: 26202220; PMID: 24183667; PMID: 30589572). These have been cited in our manuscript. However, it is important to recognize that each of these studies is an analysis of the properties of mammalian Esyt as a molecule in the context of Ca2+. However, none of these studies addresses the key question of whether the regulation of Esyt by Ca2+ is important for cellular function or to support cell physiology. The reason for this is quite straightforward and well known in the field. To date, there is no cellular or physiological phenotype that is reported to depend on endogenous Esyt function in mammalian cellular or animal models. As an illustrative example, deletion of all three mammalian Esyt does not affect cell signalling (PMID 23791178) including Ca2+ signalling and a triple knockout of all three Esyt in mice (PMID: 27348751) has no discernable phenotype.

By contrast, deletion of the single Esyt gene in Drosophila results in robust phenotypes in adult photoreceptors (PMID: 32716137). Using these phenotypes, in this manuscript we study the importance of Ca2+ dependent regulation of cellular functions mediated by dEsyt. Therefore, this study fills an important unfilled gap in establishing the mechanism by which dEsyt proteins regulate cellular functions in vivo, in a Ca2+ dependent manner. We respectfully ask that this not be caricatured as an incremental step.

---

Reviewer #2

Evidence, reproducibility and clarity

Esyt is a C domain (a Ca2+ binding domain) containing protein that localizes to the ER-MCS, playing a role in ER-mitochondria tethering and lipid transfer. At the same time, proteins at the ER-MCS are well-positioned to sense changing levels of Ca2+. Previous studies reported that loss of Esyt in Drosophila causes a loss of ER-PM integrity and retinal degeneration. Here, the authors report the consequence of disrupting the Esyt C domain in Drosophila photoreceptor cells. They used in-silico strategies to identify the Ca2+ contacting residues within the C domain and generated transgenic flies containing either the wild type or the Esyt-CaBM mutants. They show that the wild type transgene rescues several Esyt KO phenotypes in the Drosophila photoreceptors. In some cases, they report dominant negative effects of Esyt-CaBM overexpression.

This is a straightforward structure-function analysis of the Esyt C domain. Overall, the experiments are well executed. At the same time, a few aspects of the manuscript could be further improved. For example, the authors analyze multiple aspects of photoreceptor integrity. In some cases, they show that the mutant Esyt transgene shows dominant negative effects. In others, there is no evidence or even a partial function. Clarifying these points could be helpful. Below are a few specific points for the authors' consideration:

Major Comments

1. RDGB is a protein that localizes to the ER-MCS. Esyt-CABM-GFP expression causes RDGB mis-localization even in the presence of wild type Esyt expression, suggestive of a dominant negative effect (Fig. 4C). But Esyt CaBM-GFP expression doesn't seem to have a dominant negative effect on contact site distance (Fig. 4D). Are the authors not seeing a dominant negative effect because they didn't examine older flies? Or, is there a distinct effect of Esyt CaBM on RDGB localization and contact site distance? If there is a distinct effect, what is the reason? As the reviewer correctly mentions, we are not seeing a dominant negative effect of dEsytCaBM::GFP expression on contact site distance because we didn't examine older flies.

Dominant negative effect of dEsytCaBM on the wild type protein is observed in all phenotypes analyzed. The contact site distance analysis shown in the paper is done on day 1 old constant dark reared flies. Contact site distance exhibited by dEsytCaBM is like that of dEsytKO photoreceptors at day 1 post eclosion. dEsyt deprived photoreceptors are comparable to its wild type counterpart at Day 1 in all aspects of phototransduction (PMID: 32716137). But as a function of age and illumination, the dEsytKO photoreceptors exhibit progressive loss in contact site integrity, followed by induction of retinal degeneration and RDGB mis-localisation (PMID: 32716137). These observations are consistent in dEsytCaBM.

During the revision, the following experiments will be included to strengthen this statement:

• Add the MCS density and gap distance in dEsytKO photoreceptors at Day1 in Figure 3C.
• Electron Microscopy to check MCS density and distance in Rh1>dEsytCaBM at Day 6CL with appropriate control genotypes.

Esyt-CABM-GFP partially rescues the Esyt KO phenotype in retinal degeneration (Fig 6). This is surprising since cellular assays in Fig 4 show a failure of Esyt-CaBM to localize to ER-MCS. The results here contrast with earlier data showing that Esyt-CABM has dominant negative effects. How will the authors interpret the results? Is it possible that Esyt-CAMB still has some residual Ca2+ binding activity? Alternatively, does this result imply that Esyt can still function (albeit at lower capacity) without binding Ca2+? Is there Esyt function unrelated to ER-MCS site maintenance when it comes to its role in retinal degeneration? A reasonable explanation is warranted.

Partial rescue of dEsytKO phenotypes by Rh1>dEsytCaBM; dEsytKO photoreceptors indicate that apart from calcium sensing, there might be another function for dEsyt at the ER-PM interface which is yet to be discovered.

---

Minor Comments:

Figure legends refer to "SMC" (I am guessing they are referring to Sub microvillar cisternae) without defining it in the text.

Changes will be incorporated as per the suggestion.

---

Significance

This study will be of interest to those generally interested in the ER mitochondria contact sites. The main significance here is in dissecting the role of the C-domain within the Esyt protein. The authors demonstrate a physiological role using Drosophila photoreceptors as a model.

We thank the reviewer for appreciating the significance of our study which seeks to show the in vivo significance of the Ca2+ regulation of dEsyt for in vivo function.

__Reviewer #3 __

(Evidence, reproducibility and clarity (Required)):

Summary

In the present work, the authors explore the role of Ca2+ binding to Esyt in the regulation of ER-PM contact sites using drosophila photoreceptors as a model system. By expressing in wild type or in EsytKO flies a mutated version of dEsyt which is predicted to lose Ca2+ binding, they highlight a potential role of Ca2+ binding to Esyt in the regulation of ER-PM contact sites density and the development of rhabdomeres. The data clearly show the effect of Esyt mutant during development of photoreceptors in Drosophila. However, as discussed below, one essential missing point is the experimental proof that the mutant has indeed lost its ability to bind Ca2+, and that PIP2 binding is not perturbed.

Major comments

1. One major comment is the lack of experimental proof that the EsytCABM mutant is indeed unable to bind Ca2+. The MIB tool only gives a prediction and it is not sufficient to prove their statements throughout the manuscript on the requirement of Ca2+ binding for the regulation of MCS. We understand the reviewer’s comment that this manuscript does not contain experimental data demonstrating that dEsytCaBM does not bind Ca2+. However, as Reviewer 4 pointed out, it will be challenging to demonstrate this experimentally. A direct proof would likely come from measurements of the calcium binding affinity of dEsyt (which involves protein purification that is beyond the scope of this work). An indirect demonstration would be any cellular or in vivo experiment oar any additional in silico analysis. To provide additional indirect evidence to address this question, we will:

2. Use the AlphaFold model to demonstrate that the arrangement of the calcium binding residues in dEsyt is compatible with Ca2+

3. Evaluate if the wild type dEsyt is mislocalized in the photoreceptors upon eliminating the calcium entry to these specialized sensory neurons. The localization of wild type dEsyt will be examined in the mutants: norpAP24 (PLC null mutant) and trpl302; trp343 (protein null mutant of TRPL and TRP channels respectively) in which light induced calcium influx is eliminated. Moreover, they should check experimentally the potential differences in the capacity of EsytCABM mutant to bind PI(4,5)P2, which can potentially perturb its subcellular localization.

As recommended by the reviewer, it is important to determine the PIP2 binding capacity of dEsytCaBM. The ability of wild type dEsyt to bind PIP2 has not been determined. We will test this and if it does so, the impact of CaBM on PIP2 binding can be tested.

Figure 1A: the legend on the right side of the scheme is missing. On the left, RDGB and dEsyt don't associate with the PM.

Changes will be incorporated as per the suggestion.

line 125: the authors should describe more precisely the Trp mutant that they used.

The text will be modified.

Concerning the quantification of MCS density done throughout the paper, can the authors mention what they considered as an MCS, in other words, what distance they defined as the maximal distance between the ER and the PM.

We used fixation methods that allow enhanced membrane preservation and better visualization of membranes and MCS (PMID: 2496206). Such images allowed us to quantify the fraction of SMC that are present at the base of the microvilli in each ultrathin section of a photoreceptor. The MCS is the dark stretch that can be seen at the base of the rhabdomere in each TEM image (PMID: 32716137). Contact site distance measured is the absolute distance between the visible demarcation of the PM and SMC as indicated by the yellow arrows in Figure 4D iii, vi, and ix.

Figure 3: the localization of Esyt and EsytCABM in S2R cells and in vivo is not precisely analyzed: a co-staining with PM and ER markers should be added in order to state the localization at ER-PM MCS or at apical PM.

As suggested, to better understand the compartmental localization of dEsyt in photoreceptors, we will use markers of PM (Rhabdomere) and ER (Sub Microvillar Cisternae) and conduct co-localization assays.

line 181: the authors should precise in which membrane compartments Esyt is localized.

The text will be modified.

line 185-187: the conclusion here doesn't seem to fit the data, as the EsytCABM mutant looks enriched at ER-PM contact sites.

As previously answered, we will remark on whether there is an enrichment of dEsytCaBM at the ER-PM contact sites following the co-localization experiment that is recommended in Q5.

a paragraph on the production of Drosophila transgene mutants should be added to the Mat et Med section.

The text will be added as suggested.

considering the phenotypes observed for the EsytCABM mutant in vivo, the authors should provide an analysis of the level of expression of the exogenous proteins Esyt and EsytCABM by western blot in the different backgrounds. EsytCABM seems to be expressed at lower levels in Figure 3C.

As per the suggestion, western blot analysis will be conducted and better representative confocal images depicting the protein levels will be added in the manuscript.

Fig 4D: considering the perturbation of RDGB localization observed at Day 6, the authors should analyze the organization of MCS by TEM at Day 6, in addition to Day 1.

We agree that to support the observation of RDGB mis-localization, the decrease in contact site integrity as a function of age and illumination (Day6CL) should be evaluated in Rh1>dEsytCaBM photoreceptors. The manuscript revision will include data from this experiment.

the EsytCABM mutant exhibits strong dominant negative effects, but rescues completely or partially some of the phenotypes of Esyt KO: could the authors discuss and provide some hypothesis on this apparent discrepancy?

We are unsure what the reviewer means by “apparent discrepancy”. When dEsytCaBM is expressed in wild type photoreceptors, it exhibits a strong dominant negative effect presumably by inhibiting the function of wild type dEsyt protein.

dEsytKO is a protein null allele. Therefore, when dEsytCaBM is expressed in the dEsytKO background it does not exert a dominant negative effect as there is no wild type protein to interact with. The partial rescue of dEsytKO phenotypes by Rh1>dEsytCaBM; dEsytKO photoreceptors likely indicates that calcium binding is not the sole factor affecting dEsyt function at the ER-PM interface.

lines 230-233: the sentence is not clear. I don't see any consistency between data in Figure 5B, showing only very partial rescue by EsytCABM, and the data in Figure 5C (ii) showing complete rescue of RDGB localization by EsytCABM.

The time point (six days of continuous light exposure following eclosion) at which RDGB localization was analyzed becomes extremely important in thinking about this reviewer comment. If we look at the degeneration kinetics depicted in figure 5B, we can see that neurodegeneration begins in both dEsytKO and Rh1>dEsytCaBM on Day 8 post-eclosion; prior to which, on Day 6, RDGB is mislocalized from the base. However, in Rh1>dEsytCaBM; dEsytKO, the onset of degeneration is delayed, and the photoreceptors show intact structure until Day 8 or Day 10, and measurable retinal degeneration begins on Day 12. This may be the reason why, RDGB continues to be correctly localized in Rh1>dEsytCaBM; dEsytKO at Day 6CL.

Figure 6D: could the authors comment the increase of MCS density observed in Esyt-GFP expressing flies.

Esyt is proposed to function as a tether that connects the ER and PM (PMID: 23791178; PMID: 27065097; PMID: 29222176), bringing them closer together. Based on this idea, perhaps by expressing dEsyt::GFP we are drawing the membranes together thus establishing more MCS.

on several TEM images, some pictures illustrating different conditions look very similar, as if they were serial cuts: Fig 1B (Day 1 and Day 14), Fig 4D (Rh1 and Rh1>dEsytCABM::GFP), Fig 6B Day 1 and Day 14 and Fig 6C Day 1. Could the authors check if there was a mistake with these pictures?

The images are not taken from serial sections of the same TEM block as is evident from the arrangement of nucleus of each photoreceptor cell. As mentioned in the figure legends, all experiments are carried out using 3 independent blocks (N=3 fly heads) prepared from each genotype and 10 photoreceptors from each block/ fly retinae are used for quantification of contact site density/ contact site distance. Aside from the arrangement of the accessory cells and cellular nuclei, the TEM images will appear very similar since Drosophila photoreceptor neurons are symmetrically arranged, with around 700–800 ommatidia per eye each comprising 8 photoreceptors.

Minor comments:

• lines 84-88 : the sentence is not clear. Besides, the authors should precise what they mean by "extra-cellular Ca2+ influx enhance ER-PM contact sites". Which parameter exactly has been shown to be regulated by Ca2+?

The paper by Idevall-Hagren et al. proposes that following store operated Ca2+ influx, Esyt1 translocates to ER-PM junctions and the number of ER-PM contact sites increases. Please refer to this section of the publication from Idevall-Hagren et al. (2015) (PMID: 26202220):

“As detected by TIRF microscopy, the depletion of Ca2+ from the lumen of the ER occurring under these conditions led to a progressive accumulation of ER‐anchored STIM1 at the PM, where it activates Orai Ca2+ channels (Fig 4C). Subsequent addition of 1–10 mM Ca2+ to the extracellular medium, either in the absence or in the presence of SERCA inhibitors, caused a massive increase in cytosolic Ca2+ (SOCE) through the activated Ca2+ channels (Figs 4A and EV4D–G). Such increase induced a very robust translocation of E‐Syt1 to the PM (Figs 4B and EV4D–G), which, in the absence of SERCA inhibition (i.e., when a reversible inhibitor of the SERCA pump had been washed out), preceded the dissociation of STIM1 and the inactivation of SOCE (Fig 4D). Inspection of TIRF microscopy images during the manipulation showed that E‐Syt1 does not form new contacts but populates and expands contacts previously occupied by STIM1.”

• lines 108-110: can you give the reference?

Reference for the localization of dEsyt to ER-PM MCS is Nath et.al PMID PMID: 32716137

Reference for the localization of TRP and TRPL at the microvillar plasma membrane: Numerous primary research papers have shown this- for example see review PMID: 11557987, PMID: 22487656

• line 189: the authors should summarize the findings in one sentence. "Functional activity" would refer to lipid transfer.

The text will be modified as per the suggestion.

Reviewer #3 (Significance (Required)):

General assessment

The work relies on a model system that enables the exploration of the role of Esyt in vivo, in a fundamental process highly regulated during development. The data clearly show the effect of Esyt mutant during development of photoreceptors in Drosophila but as discussed before, some experimental evidences are missing to completely prove the statements.

Advance

This work brings new insights in the functional role of lipid transfer during development and explores how the dialog between lipid transfer and Ca2+ flux can influence MCS organization. The interesting points that could be explored in the paper are the effects of a Ca2+ influx on Esyt and EsytCABM localization, and on their lipid transfer activity.

Audience

This work would be of interest for the membrane contact sites community and for the Developmental biology community.

We thank the reviewer for highlighting the significance of our work and the clarity of the data. Additional data to address the points they have raised will be provided.

__Reviewer #4 __

(Evidence, reproducibility and clarity (Required)):

In this study, Nath et al., aim at understanding the role of dESyt Ca2+ binding activity on ER-PM MCS in D. melanogaster photoreceptors. Using a combination of transmission electron microscopy and fluorescence microscopy, the authors explore the ability of a dESyt mutant, supposedly unable to bind Ca2+ (based on homology with the human ortholog hESyt2), to recapitulate the function of the wild type version of the protein in establishing ER-PM MCS and modulating their density.

Findings:

1) MCS density depends on the activity of TRP and TRPL channels in aging photoreceptors.

2) Mutation of dESyt Ca2+ binding residues (dEsytCaBM::GFP) leads to a gross mis-localization of the protein, even in the presence of the endogenous protein.

3) Overexpression of the mutant affects the structure of photoreceptors upon constant illumination.

4) After 6 days of continuous illumination, RDGB is mis-localized in cells overexpressing dEsytCaBM::GFP.

5) Overexpressed dEsytCaBM::GFP fails to reduce the distance between ER and PM, meaning it fails to establish ER-PM contract sites, while overexpressed dEsyt::GFP show reduced MCS distance. Overexpressed dEsyt::GFP also leads to a 10% increase in MCS density compared to WT or cells expressing dEsytCaBM::GFP.

6) dEsytCaBM::GFP is not able to rescue the light dependent retinal degeneration of dESytKO, although it slightly delays the onset, but is able to rescue RDGB localization at day 6 of constant illumination.

7) Examining MCS density in dESytKO cells, rescues with dEsyt::GFP and dEsytCaBM::GFP show a slightly higher MCS density than dESytKO at day 1. At day 14, ER-PM MCS were non-existent in dESytKO, unchanged in dEsyt::GFP and reduced by 20% in dEsytCaBM::GFP compared to day1.

Specific comments:

My field of expertise is biochemistry and structural biology (including cellular cryo-electron tomography), but I have no experience with drosophila biology, so I am not able to judge the drosophila work per se.

While I find the confocal microscopy experiments compelling, I have some reservations regarding the quantification of the TEM images (MCS distances and density) as it was done manually, and therefore, to some extent subjective, especially, when differences between conditions are in the order of 10%. I would have found the quantification more convincing if done systematically, i.e. segmenting the MCS and computationally measuring distances and densities. Otherwise, the authors could expand a little bit on how their methodology is accurate.

As the reviewer correctly mentions, the quantification will be more convincing if done systematically, i.e. segmenting the MCS and computationally measuring distances and densities. For MCS measurements, we have experimented with the segmentation method using ImageJ and Imaris. As mentioned in the answer to Q4 of reviewer 3, we used fixation methods that allow enhanced membrane preservation and better visualization of membranes and MCS (Matsumoto‐Suzuki et al, 1989). However, this staining method does not selectively stain the ER which is part of the MCS but all the ER. Due to this, automated segmentation poses significant challenges.

The primary drawback of the segmentation method is that, in the process of training the software to predict/detect distinct cellular compartments, it recognizes all ER membranes, including SMC as well as the ER that is not part of the MCS. As a result, the software's minimum distance calculation may be between PM and SMC or PM and generic ER, which does not help the analysis we wish to perform. Similarly, to determine the contact site distance in images with obscure ER and PM boundaries, the software uses the border it can identify—which is typically inside the rhabdomere rather than at its edge. For the contact site density measurements, software is not able to distinguish between ER and pigment granules close to the rhabdomere as the gray scale value for both these compartments are comparable.

Advantages of manual approach:

To account for potential effects of photoreceptor depth on contact site density and distance, we have analyzed TEM sections obtained directly from the nuclear plane of the photoreceptors to calculate both contact site density and distance. Additionally, by utilizing the freehand line tool, manual analysis enables us to define the length of each little section of the MCS and the base of the rhabdomere. The entire length of the MCS at the base is then calculated by adding each segment together. An illustration of how the manual analysis is done will be included as part of methods in the revision.

Another point is whether the levels of expression of dESyt proteins (dESyt-GFP and dESytCABM-GFP) are comparable. In the overexpression experiments, what are the expression levels of the constructs compared to the endogenous protein? The authors should provide e.g. a Western blot.

As per the suggestion, western blot analysis will be conducted to compare the expression levels of the constructs utilized to the endogenous protein.

Concerning the modelling, while I do think that the identification of dESyt Ca2+ binding residues is correct (the sequence alignment is convincing and the sequence identity is very high), and that most likely the structural arrangement will be conserved, homology modelling (using MODELLER with a single reference) leads to models highly similar to the input reference (in particular when the sequence identity is very high). Therefore, rmsd will necessarily be low and the side chain arrangement of conserved residues will be identical. This is unlikely to happen, as protein structures will not be identical despite high sequence conservation. In addition, a crystal structure is a snapshot of a protein conformation that is favorable for crystal formation. It would have been more interesting to use an AlphaFold model and show that the arrangement on the residues is compatible with Ca2+ binding (i.e., the C positions are similar).

We agree with the reviewer that the data presented to demonstrate the inability of dEsytCaBM to bind Ca2+ is inadequate as is also pointed out by other reviewers. It would be crucial to prove this using multiple approaches. As suggested AlphaFold model will be used to answer the same.

Minor comments:

Line 102: indicate what PI and PA stand for (I don't think that there is a need for acronyms when they are not reused in the text later on).

Changes will be incorporated as per the suggestion.

Line 217-219: "When the same experimental set was examined for MCS density, we discovered that the density enhanced by 10% in Rh1>dEsyt::GFP while being comparable between wild type and dEsytCaBM::GFP flies." The authors don't comment on this finding. Does that imply that increase in the protein levels leads to increase in MCS density?

Yes. Increase in wild type dEsyt protein levels can establish more contact sites as well as reduce the contact site distance which further elucidates the protein's role in functional tethering as mentioned in line 215 as proposed by previous studies in other models (PMID: 23791178; PMID: 27065097; PMID: 29222176).

Lines 298-302: "...implying that dEsytCaBM exerts a dominant negative effect on wild type dEsyt. One possible mechanism for the phenotypes exhibited by dEsytCaBM expression in wild type cells is suggested by the findings of a structural and mass spectrometry investigation of hEsyt2 that reveals that the SMP domain dimerizes to create a 90Å long cylinder to facilitate the transfer of lipids (Schauder et al., 2014)." It is not clear to me what the authors suggest here: because of the dimerisation between wild type and mutant, the mutant has a negative effect or that the SMP dimerization is somehow impaired in dEsytCaBM?

SMP domain of Esyt proteins have previously been shown to dimerize (PMID: 23791178, PMID: 24847877). They are known to form either homodimers or heterodimers in mammalian system where there are three genes that code for the protein (Esyt1, 2 and 3). In Drosophila, since it is just one gene that codes for the protein, our hypothesis is that one copy of the functional wild type gene dimerizes with the CaBM mutant and thereby render the wild type gene product nonfunctional.

Line 304-305: "...protein expression was restricted to the cell body rather than the presynaptic terminals...". I am not sure that this is correct. The fact that a protein is localizing to a compartment does not mean that its expression is restricted to that compartment (one should measure mRNA levels to conclude this).

The statement is based on the findings made by Kikuma et al, 2017 (PMID: 28882990) when they tried to understand the role of dEsyt at the NMJs.

In figure 1B legend, indicate what SMC stands for (the acronym should be indicated in figure 1A legend).

The text will be added as suggested.

In figure 2A legend Ca binding in black box but in red boxes in figure.

Changes will be incorporated as per the suggestion.

**Referees cross-commenting**

I agree with the other reviewers that one of the premise of this study relies on the loss of calcium binding by the dESyt mutant and this is not experimentally proven by the authors. However, I find that this will be difficult to prove in vivo. Only measurements of dESyt calcium binding affinity would constitute a direct proof (which requires protein purification. Any in vivo or cellular experiment would be an indirect proof. I believe that based on the high sequence conservation with ESyt proteins, the calcium binding residues have been correctly identified.

Reviewer #4 (Significance (Required)):

ESyt proteins are known ER-PM tethers involved in lipid transfer at MCS in a Ca2+ dependent manner. Contrary to yeast and mammals, that have several ESyt orthologs, D. melanogaster has only one ESyt, making it an ideal model to study ESyt function in vivo. It has been previously shown that proper localization of ESyt at MCS depends on Ca2+ concentration: ESyts are anchors to the ER but translocate to the PM in response to elevation of Ca2+ levels in the cytosol (Fernández-Busnadiego et al., 2015). The finding that an ESyt mutant unable to bind calcium is not localized properly is therefore not surprising. The link between RDGB, a protein known to localize at MCS, and ESyt has been shown before but to my knowledge Nath et al., show for the first time that RDBG localization at MCS is directly dependent on the Ca2+ binding activity of ESyt. In addition, the authors convincingly demonstrate that the Ca2+ binding activity of dESyt is necessary to maintain the structure of aging photoreceptors.

The main finding of this study is that the Ca2+ binding activity of dESyt regulates the density of ER-PM MCS in photoreceptors. If true (see my comment below), that would be a novel finding, although the authors don't propose any mechanistic explanation for this.

3. Description of the revisions that have already been incorporated in the transferred manuscript

Please insert a point-by-point reply describing the revisions that were already carried out and included in the transferred manuscript. If no revisions have been carried out yet, please leave this section empty.

We haven't made any changes to the manuscript yet. However, we will be able to implement the changes mentioned in the pointwise response to reviewers above.

4. Description of analyses that authors prefer not to carry out

Please include a point-by-point response explaining why some of the requested data or additional analyses might not be necessary or cannot be provided within the scope of a revision. This can be due to time or resource limitations or in case of disagreement about the necessity of such additional data given the scope of the study. Please leave empty if not applicable.

• *

We feel that experiments to directly determine the calcium binding of dEsyt and the loss of this in dEsytCaBM are beyond the scope of this study. This is because of the huge work to heterologously express and purify the protein. We have proposed alternate ways to strengthen this conclusion.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #4

Evidence, reproducibility and clarity

In this study, Nath et al., aim at understanding the role of dESyt Ca2+ binding activity on ER-PM MCS in D. melanogaster photoreceptors. Using a combination of transmission electron microscopy and fluorescence microscopy, the authors explore the ability of a dESyt mutant, supposedly unable to bind Ca2+ (based on homology with the human ortholog hESyt2), to recapitulate the function of the wild type version of the protein in establishing ER-PM MCS and modulating their density.

Findings:

1. MCS density depends on the activity of TRP and TRPL channels in aging photoreceptors.
2. Mutation of dESyt Ca2+ binding residues (dEsytCaBM::GFP) leads to a gross mis-localization of the protein, even in the presence of the endogenous protein.
3. Overexpression of the mutant affects the structure of photoreceptors upon constant illumination.
4. After 6 days of continuous illumination, RDGB is mis-localized in cells overexpressing dEsytCaBM::GFP.
5. Overexpressed dEsytCaBM::GFP fails to reduce the distance between ER and PM, meaning it fails to establish ER-PM contract sites, while overexpressed dEsyt::GFP show reduced MCS distance. Overexpressed dEsyt::GFP also leads to a 10% increase in MCS density compared to WT or cells expressing dEsytCaBM::GFP.
6. dEsytCaBM::GFP is not able to rescue the light dependent retinal degeneration of dESytKO, although it slightly delays the onset, but is able to rescue RDGB localization at day 6 of constant illumination.
7. Examining MCS density in dESytKO cells, rescues with dEsyt::GFP and dEsytCaBM::GFP show a slightly higher MCS density than dESytKO at day 1. At day 14, ER-PM MCS were non-existent in dESytKO, unchanged in dEsyt::GFP and reduced by 20% in dEsytCaBM::GFP compared to day1.

Specific comments:

My field of expertise is biochemistry and structural biology (including cellular cryo-electron tomography), but I have no experience with drosophila biology, so I am not able to judge the drosophila work per se. While I find the confocal microscopy experiments compelling, I have some reservations regarding the quantification of the TEM images (MCS distances and density) as it was done manually, and therefore, to some extent subjective, especially, when differences between conditions are in the order of 10%. I would have found the quantification more convincing if done systematically, i.e. segmenting the MCS and computationally measuring distances and densities. Otherwise, the authors could expand a little bit on how their methodology is accurate.

Another point is whether the levels of expression of dESyt proteins (dESyt-GFP and dESytCABM-GFP) are comparable. In the overexpression experiments, what are the expression levels of the constructs compared to the endogenous protein? The authors should provide e.g. a Western blot.

Concerning the modelling, while I do think that the identification of dESyt Ca2+ binding residues is correct (the sequence alignment is convincing and the sequence identity is very high), and that most likely the structural arrangement will be conserved, homology modelling (using MODELLER with a single reference) leads to models highly similar to the input reference (in particular when the sequence identity is very high). Therefore, rmsd will necessarily be low and the side chain arrangement of conserved residues will be identical. This is unlikely to happen, as protein structures will not be identical despite high sequence conservation. In addition, a crystal structure is a snapshot of a protein conformation that is favorable for crystal formation. It would have been more interesting to use an AlphaFold model and show that the arrangement on the residues is compatible with Ca2+ binding (i.e., the C positions are similar).

Minor comments:

Line 102: indicate what PI and PA stand for (I don't think that there is a need for acronyms when they are not reused in the text later on).

Line 217-219: "When the same experimental set was examined for MCS density, we discovered that the density enhanced by 10% in Rh1>dEsyt::GFP while being comparable between wild type and dEsytCaBM::GFP flies." The authors don't comment on this finding. Does that imply that increase in the protein levels leads to increase in MCS density?

Lines 298-302: "...implying that dEsytCaBM exerts a dominant negative effect on wild type dEsyt. One possible mechanism for the phenotypes exhibited by dEsytCaBM expression in wild type cells is suggested by the findings of a structural and mass spectrometry investigation of hEsyt2 that reveals that the SMP domain dimerizes to create a 90Å long cylinder to facilitate the transfer of lipids (Schauder et al., 2014)." It is not clear to me what the authors suggest here: because of the dimerisation between wild type and mutant, the mutant has a negative effect or that the SMP dimerization is somehow impaired in dEsytCaBM?

Line 304-305: "...protein expression was restricted to the cell body rather than the presynaptic terminals...". I am not sure that this is correct. The fact that a protein is localizing to a compartment does not mean that its expression is restricted to that compartment (one should measure mRNA levels to conclude this).

In figure 1B legend, indicate what SMC stands for (the acronym should be indicated in figure 1A legend).

In figure 2A legend Ca binding in black box but in red boxes in figure.

Referees cross-commenting

I agree with the other reviewers that one of the premise of this study relies on the loss of calcium binding by the dESyt mutant and this is not experimentally proven by the authors. However, I find that this will be difficult to prove in vivo. Only measurements of dESyt calcium binding affinity would constitute a direct proof (which requires protein purification. Any in vivo or cellular experiment would be an indirect proof. I believe that based on the high sequence conservation with ESyt proteins, the calcium binding residues have been correctly identified.

Significance

ESyt proteins are known ER-PM tethers involved in lipid transfer at MCS in a Ca2+ dependent manner. Contrary to yeast and mammals, that have several ESyt orthologs, D. melanogaster has only one ESyt, making it an ideal model to study ESyt function in vivo. It has been previously shown that proper localization of ESyt at MCS depends on Ca2+ concentration: ESyts are anchors to the ER but translocate to the PM in response to elevation of Ca2+ levels in the cytosol (Fernández-Busnadiego et al., 2015). The finding that an ESyt mutant unable to bind calcium is not localized properly is therefore not surprising. The link between RDGB, a protein known to localize at MCS, and ESyt has been shown before but to my knowledge Nath et al., show for the first time that RDBG localization at MCS is directly dependent on the Ca2+ binding activity of ESyt. In addition, the authors convincingly demonstrate that the Ca2+ binding activity of dESyt is necessary to maintain the structure of aging photoreceptors.

The main finding of this study is that the Ca2+ binding activity of dESyt regulates the density of ER-PM MCS in photoreceptors. If true (see my comment below), that would be a novel finding, although the authors don't propose any mechanistic explanation for this.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Summary

In the present work, the authors explore the role of Ca2+ binding to Esyt in the regulation of ER-PM contact sites using drosophila photoreceptors as a model system. By expressing in wild type or in EsytKO flies a mutated version of dEsyt which is predicted to lose Ca2+ binding, they highlight a potential role of Ca2+ binding to Esyt in the regulation of ER-PM contact sites density and the development of rhabdomeres. The data clearly show the effect of Esyt mutant during development of photoreceptors in Drosophila. However, as discussed below, one essential missing point is the experimental proof that the mutant has indeed lost its ability to bind Ca2+, and that PIP2 binding is not perturbed.

Major comments

• One major comment is the lack of experimental proof that the EsytCABM mutant is indeed unable to bind Ca2+. The MIB tool only gives a prediction and it is not sufficient to prove their statements throughout the manuscript on the requirement of Ca2+ binding for the regulation of MCS. Moreover, they should check experimentally the potential differences in the capacity of EsytCABM mutant to bind PI(4,5)P2, which can potentially perturb its subcellular localization.
• Figure 1A: the legend on the right side of the scheme is missing. On the left, RDGB and dEsyt don't associate with the PM.
• line 125: the authors should describe more precisely the Trp mutant that they used.
• concerning the quantification of MCS density done throughout the paper, can the authors mention what they considered as an MCS, in other words, what distance they defined as the maximal distance between the ER and the PM.
• Figure 3: the localization of Esyt and EsytCABM in S2R cells and in vivo is not precisely analyzed: a co-staining with PM and ER markers should be added in order to state the localization at ER-PM MCS or at apical PM.
• line 181: the authors should precise in which membrane compartments Esyt is localized.
• line 185-187: the conclusion here doesn't seem to fit the data, as the EsytCABM mutant looks enriched at ER-PM contact sites.
• a paragraph on the production of Drosophila transgene mutants should be added to the Mat et Med section.
• considering the phenotypes observed for the EsytCABM mutant in vivo, the authors should provide an analysis of the level of expression of the exogenous proteins Esyt and EsytCABM by western blot in the different backgrounds. EsytCABM seems to be expressed at lower levels in Figure 3C.
• Fig 4D: considering the perturbation of RDGB localization observed at Day 6, the authors should analyze the organization of MCS by TEM at Day 6, in addition to Day 1.
• the EsytCABM mutant exhibits strong dominant negative effects, but rescues completely or partially some of the phenotypes of Esyt KO: could the authors discuss and provide some hypothesis on this apparent discrepancy?
• lines 230-233: the sentence is not clear. I don't see any consistency between data in Figure 5B, showing only very partial rescue by EsytCABM, and the data in Figure 5C (ii) showing complete rescue of RDGB localization by EsytCABM.
• Figure 6D: could the authors comment the increase of MCS density observed in Esyt-GFP expressing flies.
• on several TEM images, some pictures illustrating different conditions look very similar, as if they were serial cuts: Fig 1B (Day 1 and Day 14), Fig 4D (Rh1 and Rh1>dEsytCABM::GFP), Fig 6B Day 1 and Day 14 and Fig 6C Day 1. Could the authors check if there was a mistake with these pictures?

Minor comments:

• lines 84-88 : the sentence is not clear. Besides, the authors should precise what they mean by "extra-cellular Ca2+ influx enhance ER-PM contact sites". Which parameter exactly has been shown to be regulated by Ca2+?
• lines 108-110: can you give the reference?
• line 189: the authors should summarize the findings in one sentence. "Functional activity" would refer to lipid transfer.

Significance

General assessment

The work relies on a model system that enables the exploration of the role of Esyt in vivo, in a fundamental process highly regulated during development. The data clearly show the effect of Esyt mutant during development of photoreceptors in Drosophila but as discussed before, some experimental evidences are missing to completely prove the statements.

Advance

This work brings new insights in the functional role of lipid transfer during development and explores how the dialog between lipid transfer and Ca2+ flux can influence MCS organization. The interesting points that could be explored in the paper are the effects of a Ca2+ influx on Esyt and EsytCABM localization, and on their lipid transfer activity.

Audience

This work would be of interest for the membrane contact sites community and for the Developmental biology community.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Esyt is a C domain (a Ca2+ binding domain) containing protein that localizes to the ER-MCS, playing a role in ER-mitochondria tethering and lipid transfer. At the same time, proteins at the ER-MCS are well-positioned to sense changing levels of Ca2+. Previous studies reported that loss of Esyt in Drosophila causes a loss of ER-PM integrity and retinal degeneration. Here, the authors report the consequence of disrupting the Esyt C domain in Drosophila photoreceptor cells. They used in-silico strategies to identify the Ca2+ contacting residues within the C domain and generated transgenic flies containing either the wild type or the Esyt-CaBM mutants. They show that the wild type transgene rescues several Esyt KO phenotypes in the Drosophila photoreceptors. In some cases, they report dominant negative effects of Esyt-CaBM overexpression.

This is a straightforward structure-function analysis of the Esyt C domain. Overall, the experiments are well executed. At the same time, a few aspects of the manuscript could be further improved. For example, the authors analyze multiple aspects of photoreceptor integrity. In some cases, they show that the mutant Esyt transgene shows dominant negative effects. In others, there is no evidence or even a partial function. Clarifying these points could be helpful. Below are a few specific points for the authors' consideration:

Major

1. RDGB is a protein that localizes to the ER-MCS. Esyt-CABM-GFP expression causes RDGB mis-localization even in the presence of wild type Esyt expression, suggestive of a dominant negative effect (Fig. 4C). But Esyt CaBM-GFP expression doesn't seem to have a dominant negative effect on contact site distance (Fig. 4D). Are the authors not seeing a dominant negative effect because they didn't examine older flies? Or, is there a distinct effect of Esyt CaBM on RDGB localization and contact site distance? If there is a distinct effect, what is the reason?
2. Esyt-CABM-GFP partially rescues the Esyt KO phenotype in retinal degeneration (Fig 6). This is surprising since cellular assays in Fig 4 show a failure of Esyt-CaBM to localize to ER-MCS. The results here contrast with earlier data showing that Esyt-CABM has dominant negative effects. How will the authors interpret the results? Is it possible that Esyt-CAMB still has some residual Ca2+ binding activity? Alternatively, does this result imply that Esyt can still function (albeit at lower capacity) without binding Ca2+? Is there Esyt function unrelated to ER-MCS site maintenance when it comes to its role in retinal degeneration? A reasonable explanation is warranted.

Minor:

Figure legends refer to "SMC" (I am guessing they are referring to Sub microvillar cisternae) without defining it in the text.

Significance

This study will be of interest to those generally interested in the ER mitochondria contact sites. The main significance here is in dissecting the role of the C-domain within the Esyt protein. The authors demonstrate a physiological role using Drosophila photoreceptors as a model.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary:

This study builds on previous work from the same group, where they use Drosophila photoreceptors as a model system to investigate the role or ER-plasma membrane contact sites in an in vivo setting. The authors recently described a role of the ER-PM contact site protein dEsyt in regulating photoreceptor function in Drosophila. In this follow-up study, they explore whether this function of dEsyt is connected Ca2+ signaling downstream of photoreceptor activation. Using a dEsyt mutant that should be unable to bind Ca2+, they find that Ca2+ to some extent is required for dEsyt localization, membrane contact site formation and photoreceptor function.

Major comments:

The use of photoreceptor cells in Drosophila is an elegant model system that enable studies of membrane contact sites and associated proteins in a native condition. The data presented by the authors clearly shows that these structures are important for photoreceptor function, and that dEsyt plays a role at these sites. However, this was already known from previous studies by the same group. When it comes to whether these contacts are sensing Ca2+ changes and if these changes are acting through dEsyt, which is the focus of the current manuscript, the results are unclear to me and would need to be clarified by the authors both in text and with new experiments.

1. What is the role of cellular Ca2+ signaling in the regulation of dEsyt function? There are several aspects here that needs to be clarified. 1) How is WT dEsyt localization regulated by Ca2+? This could for example be evaluated in the mutant flies used in Fig. 1 (trpl302; trp343), where lack of light-induced Ca2+ influx would be predicted to result in a localization of dEsyt that resembles that observed for dEsytCaBM. 2) Is Ca2+ important for dEsyt localization, lipid exchange or both? The authors express a version of dEsyt with mutation made in all three C2 domains. In mammalian E-Syts, Ca2+ binding to the C2A domain is important for lipid exchange while binding to C2C (in E-Syt1) is important for interactions with lipids in the plasma membrane. Using more carefully designed mutants will allow the authors to determine how Ca2+ regulates dEsyt function in vivo. In addition, the authors must show experimentally that the mutant dEsytCaBM is unable to bind Ca2+ (could e.g. be done by acute Ca2+ changes in the cell-based model used in Fig. 3). Writing that "This transgene carrying a total of nine mutations should render the protein unable to bind calcium" (p. 6, line 173) is not sufficient.
2. The localization of dEsyt shown in Fig. 3B is a bit confusing. First of all, I would recommend including markers of the ER and the plasma membrane, because without these it is difficult to make statements about the localization of dEsyt to these structures. Second, it appears that WT dEsyt localize to the reticular ER, and that the CaBM version localize to the plasma membrane. This is somewhat opposite to mammalian ESyts, where mutations that prevent Ca2+ binding either had no effect (for ESyt2) or prevented (for ESyt1) the interaction with the plasma membrane. It also appears different from the localization in vivo (Fig. 3C). Clarifying this will be important. It will also be important to connect this localization to changes in Ca2+ and not just to the localization of a mutant that may or may not be deficient in Ca2+ binding (see comment above).
3. I don't fully understand the time course of events. The authors show that dEsytCaBM is mislocalized already at day 1 in dark-reared flies (Fig. 3C) but this mislocalization is not accompanied by a change in MCS density or gap distance, and consistently does not influence the localization of RDGB. The authors next expose the flies to constant light illumination to trigger Ca2+ dependent signaling, and this leads to mislocalization of RDGB, perhaps indicating changes in MCS (this is not shown). From these results it is difficult to know what the role of dEsyt is. It would be necessary to also show a control where Ca2+ signaling is not induced, e.g. a parallel dark-control (same number of days but no illumination). This is particularly important given that the authors show in Fig. 1 that preventing Ca2+ influx had a dramatic impact on MCS density even at day 1 (which is in sharp contrast to dEsytCaBM-expressing flies, that show normal morphology at day 1, which rather implies that dEsyt is not a major Ca2+ effector).
4. The experiments done in dEsyt KO flies are important, and here the authors show that dEsyt1 could to some extent rescue all phenotypes. Some results are a bit puzzling. For example, dEsyt1CaBM localization in dEsyt1 KO flies is identical to that of WT dEsyt (Fig. 5C), which is in sharp contrast to the data shown in Fig. 3C. What is the reason for this? I would have anticipated the opposite (i.e. that in WT flies, dEsytCaBM can form dimers with endogenous dEsyt through SMP-domain interactions which may have an impact on its localization and the function of endogenous dEsyt, but that in the dEsyt KO cells, dEsytCaBM would show a different localization due to the lack of endogenous dEyt to interact with). It is important to clarify as one of the major observations here is that dEsytCaBM no longer localize to MCS. Since the CaBM version of dEsyt could rescue, to some extent, MCS density and delay photoreceptor degeneration, this implies that Ca2+ may not be required for regulation of dEsyt function or that the mutant is still able to partially bind to Ca2+. One experiment that would help the authors determining the function of dEsyt in vivo would be to use a mutant that lacks functional SMP domain (ideally also with and without mutations in the C2-domains).
5. PLC activation typically couples to rapid signaling and involved hydrolysis of PIP2 and release of Ca2+ from the ER. Mammalian Esyts also require PIP2 for plasma membrane binding (through interactions with C2-domains), so constitutive PLC activity would be expected to impair ESyt localization to MCS. Here, the authors expose flies for days of constant illumination. How does this influence plasma membrane PIP2 levels and could this be of relevance for how data is interpreted? Do the authors know whether the CaBM mutant has reduced affinity for PIP2?

Minor comments:

1. The overexpression of WT dEsyt had a dramatic impact on MCS density and gap distance, while expression of dEsytCaBM did not. If these contacts are important for photoreceptor function, is it not surprising that such a dramatic change in photoreceptor structure was without effect on function? This should be further discussed.
2. How is quantification of MCS density and gap distance influenced by retinal degeneration (e.g. induced by dEsyt KO)?
3. The graphical abstract is a bit confusing. It seems to suggest that changes in dEsyt is a consequence of ageing and does not show any role of this protein in photoreceptor function. I think that the abstract could be improved to more clearly highlight the findings in the manuscript. For example, it doesn't at all show the difference in localization between WT and CaBM.
4. P. 5, line 135 the authors state that "The tethering and lipid transfer activity of mammalian Esyts are reported to be influenced by Ca2+". This is a massive understatement. Ca2+ is a critical regulator of Esyt function in mammalian cells.
5. In figure legend 1B and C: correct µM to µm.
6. In figure legend 2A: should be red rectangles and not black rectangles.
7. In Fig. 2B: specify which isoform of human ESyt that is shown.
8. In Fig. 2C: do the authors mean D374 or D384 (as indicated in Fig. 2A)?

Significance

Light-induced signal transduction in photoreceptor cells involves Ca2+ influx and signaling and also depends on correct formation of ER-plasma membrane contact sites. In mammalian cells, the Esyts (esp. Esyt1 and Esyt2) localize to ER-PM contacts in a Ca2+-dependent manner, and the ion has dual effects in both enriching the protein at the membrane contact sites and in promoting lipid transport. Mammalian Esyts form homo- and heterodimers, and the properties of the dimers depends on their composition (PMID: 26202220). Drosophila only have one Esyt (dEsyt) which is structurally most similar to mammalian Esyt2, and the authors have previously shown how this protein is required for photoreceptor function (PMID: 32716137), although the role of Ca2+ was not investigated in that study. However, an earlier study has shown that mutations of all Ca2+-coordinating residues in dEsyt impairs protein function in Drosophila neurons (PMID: 28882990), so a similar Ca2+-dependence in the retina would be expected. The results from the present study confirm the requirement of Ca2+ signaling for dEsyt function, and extends this Ca2+-dependent regulation to also involve photoreceptor-induced Ca2+ signaling, which corroborates many other studies showing the requirement of Ca2+ signaling for the regulation of Esyt function in mammalian cells (e.g. PMID: 23791178; PMID: 27065097; PMID: 29222176; PMID: 26202220; PMID: 24183667; PMID: 30589572). As such, the results from this study represent an incremental step towards understanding Esyt function in vivo. These results would be of greatest interest to researchers working of photoreceptor function, and of some interest to a broader audience working on membrane contact sites and signal transduction. My own background is in mammalian cell biology, with a focus on lipid and Ca2+ signaling and inter-organelle communication. I have limited understanding of the model system used here (Drosophila photoreceptor cells).

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Reviewer #1 (Evidence, reproducibility and clarity (Required)):

Summary In this study, Raja et al. found cytoplasmic condensates formed by the treatment of INFγ, investigated components of these condensates and identified p62, NBR1 and PARP14 as their components. INFγ treatment induced PARP14 expression, and PAPR14 inhibitor treatment inhibited condensation formation, suggesting that the amount of PARP14 and its enzymatic activity are important for the condensate formation. The ADPr-positive p62 condensates were independent of autophagic degradation, and proteasomal activity was required for their formation.

Major comment 1. The finding that the ubiquitin-proteasome, but not autophagy activity, is indispensable for the formation of p62 condensates is of interest. However, the molecular mechanism by which the ubiquitin-proteasome system (UPS) is involved in the regulation of the PARP14-p62 condensate is still unclear. Which step(s) is the UPS involved?

We appreciate the reviewer's acknowledgment of the novelty of our studies on the requirement of the ubiquitin-proteasome system (UPS) in the formation of ADPr-containing PARP14/p62 condensates. We have demonstrated that condensation formation requires the first and last steps of the UPS using two distinct classes of inhibitors: (1) TAK243 inhibits the E1 enzyme by forming a covalent adduct with ubiquitin that mimics the ubiquitin-adenylate complex, thereby blocking the initial step of ubiquitin conjugation (Fig. 6F-G). (2) Three different proteasome inhibitors with varying degrees of selectivity—MG132, epoxomicin, and Bortezomib—block the final step of the UPS by inhibiting the 26S proteasome (Fig. 6B, S6D). Given that blocking the early steps of the ubiquitin conjugation pathway or the late stages of the UPS inhibits the formation of ADPr condensates, we deduce that an active UPS is required. We will explore the involvement of additional steps in the UPS.

The p62 condensate serves as a scaffold for autophagosome formation through the assembling autophagy receptors including NBR1 and TAX1BP1, followed by recruiting ATG proteins such as FIP200. While ADPr-positive p62 condensates also contain NBR1 and polyubiquitinated proteins, they are unrelated to autophagic degradation. It is unclear what factors govern autophagy-independent function.

As we have identified the requirement of active UPS in regulating these condensates, determining the factors that govern autophagy-independent functions, though interesting, is beyond the scope of this manuscript. Our data indicate that the formation of ADPr-containing condensates, which include p62, other autophagy receptors such as NBR1, and polyubiquitinated proteins, but lack the autophagasome membrane protein LC3B (Fig. 3B, 5E-I, and S5D). Notably. this condensate formation is not inhibited by treatment with Bafilomycin A1 and chloroquine, which target the final step of autophagy involving lysosome interaction (Fig. 6A). In response to the reviewer's comments, we will further investigate whether these condensates also include other autophagy receptors, such as TAX1BP1, as well as the downstream autophagosome protein FIP200. Additionally, we will genetically deplete the critical autophagy factor ATG5 to confirm orthogonally that the formation of these condensates is indeed independent of autophagy.

The authors claim that the amount of PARP14 and its MAR activity are essential for the condensate formation. However, all experiments were performed only with PARP14 inhibitors, and further validation is needed. If the importance of PARP14 activity is to be directly demonstrated, experiments in which an enzyme activity mutant is introduced into PARP14 KO cells are needed.

We would like to clarify that we have not only used PARP14 chemical inhibitors to reduce MAR activity but also employed PROTAC to reduce the amount of PARP14 (Fig. 1H). Both approaches demonstrated that the inhibition of either the amount or MAR activity of PARP14 is critical for condensate formation. Additionally, we demonstrated that condensate formation is reduced upon PARP14 knockdown using siRNA and shRNA, as well as CRISPR-mediated knockout (Fig. 1G and S1C-H).

Furthermore, we showed that transient transfection of a PARP14 mutant deficient in ADP-ribosylhydrolase activity into U2OS cells leads to the formation of ADPr condensates that colocalize with PARP14, independent of IFNγ treatment. Notably, a subset of condensates—particularly the larger ones—that contain both PARP14 and ADPr showed strong colocalization with p62 (Fig. 4G). Treatment with PARP14 MAR activity inhibitor under these conditions resulted in the disappearance of ADPr/PARP14 condensates while p62 bodies remained (Fig. 4H), further indicating that ADPr enrichment in p62 bodies depends on the MAR activity of PARP14. To further confirm the dependence on MAR activity, we have now repeated the experiment using a PARP14 mutant deficient in MAR activity. PARP14 and ADPr condensates were not observed upon expression of this mutant, indicating that condensate formation depends on PARP14 MAR activity.

In Figure 2a, the heatmap alone is insufficient. Neither errors nor statistical comparisons are indicated.

We will incorporate our statistical data presented in Fig. S2A-D into Fig. 2A.

The statistical analysis of Figure S2 is inappropriate; instead of t-tests, multiple comparisons should be used to compare three or more groups.

We will perform multiple comparison analyses, as suggested.

Minor comment 1. What percentage of p62 condensates upon INFγ treatment are ADPr positive? Are all p62 bodies seen with INFγ stimulation unrelated to autophagy?

We will perform the quantification, as suggested.

Is ADPr condensation a PARA14-specific phenomenon? PARP9 and PARP12 were also upregulated by INFγ treatment. Are these factors also involved in condensate formation?

Amongst all catalytically active PARPs, ADPr condensation requires only PARP14. Russo et al., J Biol Chem 2021, have shown that genetic knockout of PARP9 affects the formation of ADPr condensates; however, PARP9 is catalytically inactive as an ADP-ribosyltransferase. Ribeiro et al., EMBO 2024, have further confirmed the requirement of PARP9 by siRNA knockdown and have also shown that condensate formation does not require PARP12. Based on the reviewers' comments, we will independently confirm this observation by performing knockdown experiments.

Figure 4D appears to be immunoprecipitation (IP) under non-denaturing conditions. If so, it is not possible to distinguish whether the MAR signal is derived from p62 or from the p62 interacting proteins (the associated ubiquitinated substrates). IP experiments should be performed under denaturing conditions.

We will perform denaturing IP or other experiments to confirm the p62 modification upon IFNγ treatment. Additionally, we would like to note that following our submission, Kubon et al. reported in a bioRxiv preprint that p62 is ADP-ribosylated in a PARP14-dependent manner upon treatment with type I interferon, IFNβ. This finding is consistent with our study involving type II interferon, IFNγ.

In Figure 5B, which band is HO1, the upper or lower?

Both bands are HO1, as shown by Biswas et al., J Biol Chem 2014. One band appears at 28 kDa and the other at 32 kDa. The 32-kDa isoform is predominantly constitutive in the cytoplasm, whereas the 28-kDa HO-1 is predominant and primarily localized to the nucleus.

There is no image for ubiquitin in S5D.

Our original statement, “However, when inhibited with the mTOR inhibitor Torin-1, autophagy is induced, leading to increased autophagosome formation marked by LC3B on the membranes, which facilitates the recruitment of p62 and ubiquitinated proteins (Fig. S5D),” contained a misplaced figure citation. The correct statement should be: “However, when inhibited with the mTOR inhibitor Torin-1, autophagy is induced, leading to increased autophagosome formation marked by LC3B on the membranes (Fig. S5D), which facilitates the recruitment of p62 and ubiquitinated proteins.” Our intention was to show that there are conditions, such as Torin-1 treatment, where p62 and LC3B colocalize.

Right panel in Figure 4F shows only IFγ + RBN, which should show all data sets in the same panel.

Given the complexity of the three conditions with extensive data points and error bars on the FRAP experiments, we aim to present the data clearly. Instead of merging the panel into one figure, we initially provided a summary table in Figure 4F. However, in response to the reviewer's comments, we will provide the composite image that includes all data sets in the same panel.

Reviewer #1 (Significance (Required)):

Liquid droplets, which have continuously being identified in cells, are a hot topic in cell biology. Droplet formation, structure, molecular dynamics, and degradation, as well as their abnormalities and disease development due to genetic mutation and stress, are of wide-ranging interest from basic to pathological aspects. Therefore, this research has the potential to attract interest from a wide range of fields.

General assessment Overall, the data are clear and the phenomenon is of interest. However, the molecular mechanism and biological significance of the condensate formation is unknown; It is unclear why proteasome activity is required for the formation of PARP14-mediated ADP ribosylation. It is also unclear what the consequences are for the cell if the ADPr-positive condensates are not formed. Thea authors should address these general and important issues and provide the data If not all.

We thank the reviewer for acknowledging that our condensate investigation is timely and important in cell biology and for recognizing that “data are clear and the phenomenon is of interest”. As mentioned in the Discussion, these condensates can be reversed by the SARS-CoV-2 macrodomain in lung A549 cells, whose activity to remove ADP-ribosylation is critical for viral replication and pathogenesis, indicating the biological significance of these condensates. In addition, similar IFNγ conditions can induce PARP14 expression in melanoma, where PARP14 inhibition resensitizes these cancers to immunotherapy. Given that these ADPr condensates are also observed in A375 melanoma cells beyond lung cells (Fig. S3B), this provides additional context to investigate their biological significance in the future.

We would like to note that we have already made significant advances by (1) revealing the identity of these condensates as related to p62 bodies (Fig. 3-5), (2) defining the responsible ADP-ribosyltransferase as PARP14 (Fig. 1-2), and (3) determining the requirements for condensation through ubiquitin-proteasome system (Fig. 6). The proposed exploration of the functional consequences and significance is beyond the scope of this manuscript. However, we will further define the mechanistic involvement of which step of ubiquitin-proteasome system.

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

This manuscript investigates the formation of a novel cellular structure or condensate, similar to p62 bodies, that includes PARP14 and p62. The interferon-induced PARP14-mediated ADP-ribosylation of p62 in these condensates depends on an active ubiquitin-proteasome system. These condensates are characterized by the presence of PARP14 and ADPr and include some, but not all, components of the p62 bodies. Furthermore, their formation depends on both ubiquitin activation and proteasome activity, but it is unaffected by autophagy inhibition, unlike conventional p62 bodies.

The Introduction provides a well-delineated context of condensates, highlighting the importance of post-translational modifications in responding to environmental changes.

Although the manuscript is well-organized with apparent logical development, there are weaknesses that diminish the impact of the reported data. A more accurate review of the structures, both from a morphological and quantitative perspective, would strengthen the conclusions and the overall impact of this work. Additionally, while the authors have analyzed the contribution of PARP14 to condensate formation, the biological significance of these structures remains unclear. For instance, performing MS (mass spectrometry) analysis on the described structures could help identify their composition and functions.

The methodologies used in this study are standard for molecular and cellular biology research, including immunofluorescence assays, transient transfections, immunoprecipitations, and fluorescence recovery after photobleaching (FRAP) assays. These methods are described in detail and can be reproduced.

Below, please find a list of comments and suggestions to enhance the robustness of the data:

We thank the reviewer for acknowledging the logical progression of the manuscript and the detailed, reproducible methods. As detailed below, we will perform super-resolution microscopy experiments to examine morphology, improve data quantification, and conduct proteomics experiments to identify how the p62 interactome changes upon IFNγ treatment.

Major Points

1. The IF analyses are central to the conclusions reported and are employed for each of the inhibitors or other tools used to investigate the formation of these condensates. The quality of the IF images needs to be improved; the shape and contacts of the condensates should be analyzed using either super-resolution or EM microscopy, or preferably both. The lack of morphometry and quantification from cell populations needs to be addressed for all experiments. These analyses are needed to support the claim that the condensates presented in this study are indeed novel structures, rather than being transient aggregates of a different nature.

We believe the reviewer may be referring to the size of the images rather than their quality, as they are of high resolution when zoomed in. However, we agree that larger images would enhance clarity. We will be discreet in choosing the figures to present, given that we have over 550 panels in the main figures and over 250 in the supplemental figures. We will resize our figures to ensure the images are clearly visible.

We will perform Airyscan imaging to provide super-resolution images for a better understanding of the morphology of these condensates. We will perform the morphometry quantification (circularity and ellipticity). For quantification, we would like to point out that nearly all of our experiments were analyzed from at least 4 fields, each containing 20-50 cells (depending on the magnification of 20x or 40x). In response to the reviewer's comment, we will also provide quantification on a per-cell basis.

Our data indicate that these are novel ADPr-containing condensates that colocalize with PARP14, p62, NBR1, and ubiquitin, but not LC3B (Fig. 1F, 3B, 5E-F, 5I). These structures are inducible by IFN treatment and can be inhibited by even 1 hour of PARP14 inhibition (Fig. 1A-B, 2D) They are dependent on PARP14 induction and its ADP-ribosyltransferase activity (Fig. 1G, 2B, 2D). These structures are not protein aggregates, as evidenced by their lack of staining with ProteoStat Dye (Fig. S6A), which stains for unfolded proteins.

The claim that PARP14 is essential for the formation of condensates requires support by the analyses indicated above. Minor points regarding Fig. 1 are indicated below. I suggest performing KD of PARP9 and/or PARP12 (whose expression is increased upon IFN treatment) and checking ADPr condensates to validate the central role of PARP14.

ADPr condensation requires PARP14, as demonstrated by multiple genetic depletion techniques (siRNA/shRNA/CRISPR; Fig. 1G, S1C-H) and chemical inhibitors (catalytic and PROTAC; Fig. 1H, 2B, 2D). We will provide additional image analyses to support the claim. In addition, Russo et al., J Biol Chem 2021, have shown that genetic knockout of PARP9 affects the formation of ADPr condensates; however, PARP9 is catalytically inactive as an ADP-ribosyltransferase. Ribeiro et al., EMBO 2024, have further confirmed the requirement of PARP9 by siRNA knockdown and have also shown that condensate formation does not require PARP12. Based on the reviewers' comments, we will independently confirm this observation by performing knockdown experiments.

According to the text, "PARP14 was pulled down by ADP-ribose binding Af1521 macrodomain following IFNγ treatment (Fig. 2H)", but the legend to the figure says otherwise. A Pan-ADPr binding reagent (MABE1016) is reported in the figure. Although the conclusion is similar for the results obtained with these two tools (but they must be described and reported properly), it is still insufficient to claim that PARP14 is ADP-ribosylated. This point should be at least discussed.

We apologize for the confusion. The Pan-ADPr binding reagent (MABE1016) is a His-tagged recombinant Af1521 macrodomain that binds to ADP-ribosylated protein (Gibson et al., Biochemistry 2017). Therefore, we used Ni-NTA resin to pull down His-tagged AF1521 for the subsequent analysis of PARP14. We will revise the text and figure legend for Fig. 2H to clarify this. We will further probe the eluted PARP14 is indeed ADP-ribosylated by western blot. Consistent with our studies, Kar et al. and Ribeiro et al. (EMBO J. 2024) also recently reported that PARP14 is ADP-ribosylated upon IFNγ treatment in A549 cells. Additionally, Higashi et al. (J. Proteome Res. 2019) reported PARP14 ADP-ribosylation in IFNγ-treated macrophage cells. We will include these references in our Discussion.

I have difficulties analyzing the colocalization with the different organelles, even enlarging the images as much as possible. In most cases, only one condensate per image is shown. Continuities with the nuclear envelope appear in some cases: has this been investigated?

We have provided images of at least two cells containing multiple cytoplasmic condensates in Figure S3A. We believe part of the confusion arises from the staining pattern of ADPr in the nucleus, which colocalizes with splicing speckles. However, this nuclear staining was not altered by IFNγ treatment. Therefore, we have focused on the cytoplasmic condensates in this study and will clarify this focus in the main text. However, in response to the reviewer’s question, we will also visit the possibilities of enrichment of signals at the nuclear envelope. For each colocalization study, we have analyzed at least four fields, each containing 20-30 cells. To further strengthen our claim for colocalization analyses, we will now provide quantification on a per-cell basis.

Minor Points

1. Fig. 1A: The DAPI images at 3 and 6 hours are reversed. Additionally, for Fig. S1a and Fig. 1A, please include quantifications.

We apologize for the oversight and will provide quantification.

1. Fig. 1B: Check PARP14 levels (and other IFN-PARPs) under the same experimental conditions.

As suggested, we will assess PARP9, PARP12, and PARP14 levels.

Fig. 1H: Explain why PARP14 IF staining is still visible upon RBN012811 treatment, while it is completely lost in WB analysis or upon PARP14 siRNA treatment (Fig. 1G). In addition, please include IF quantifications.

To clarify, for Figure 1H, the images provided correspond to those from IFNγ-treated cells while the western blot data include both with and without IFNγ treatment. We would like to point out that RBN012811 treatment indeed shows a similar dose-dependent signal with increasing hours of treatment, comparable to the western blot results. Specifically, we observed a small amount of PARP14 remaining at the 1-hour timepoint on western blots and IF images, with the highest intensity observed compared to other timepoints. In addition, we believe part of the discrepancy is due to background staining by PARP14 antibodies. Therefore, we will examine the level of background staining in PARP14 KO cells and provide corresponding IF quantification.

Fig. 2C: Please include quantifications.

We will provide quantification.

1. Fig. 2E: The RBN treatment time is not indicated. Please include this information in the figure legend.

We apologize for the oversight and will add the treatment time (24 h) to the figure legend.

Fig. 2G: I am not convinced about the PARP14 staining. IF images do not show an increase in PARP14 levels, while WB analysis shows a strong increase in PARP14 protein levels (see Fig. 2E). Moreover, the RBN treatment time was not indicated; please include it in the figure legend. Does RBN alone affect PARP14 localization? The reported picture shows only 2 cells, each with a different subcellular localization of PARP14. As previously suggested, quantifications are required.

When presenting the data, our aim was to show the pattern rather than the relative intensity difference. Therefore, we used the autocontrast image function across different conditions, which resulted in an apparent change in pattern even with weak signals in control or RBN-treated samples. To address this, we will ensure the images presented across different conditions have the same exposure and are shown with consistent image contrast parameters. We will also include quantification of the condensates. Additionally, we apologize for the oversight and will add the RBN treatment time to the figure legend.

Fig. 3B: Pearson's correlation coefficient (PCC) is reported for n=3. The images show one condensate per cell. Under these conditions, the number of cells analyzed should be at least 100 for each experiment. Additionally, the PCC between PARP14 and p62 at steady state is shown to be 60% (which is quite high). However, the IF pictures do not support this quantification. Can the authors provide higher-resolution pictures? Does PARP14 always co-localize with p62? Lines 207-208 state: "these findings suggest that PARP14 is localized to p62 bodies upon IFNγ treatment when ADP-ribosylation occurs." According to the PCC value, the two proteins co-localize even in the absence of IFN. Can the authors clarify this aspect?

We apologize for the inaccurate description. The data should be n=4, representing different fields, each containing 20-30 cells (as indicated by the number of dots in the original graph panels in Fig. 3B). The Pearson's correlation coefficient was calculated across the cells, instead of focusing on the condensates—we will provide additional analyses on condensate colocalization analyses. We will also provide larger images and quantification to indicate the level of PARP14 colocalization with p62. For PARP14, we did not see a significant number of PARP14 condensates in control cells; PARP14 condensates were seen only after IFNγ treatment. A fraction of PARP14 condensates did not colocalize with p62. We will provide detailed quantification analyses.

Fig. S3B: Please include quantifications.

Quantification will be provided

Fig. S3C: How was the condensate size quantified? It would be useful to show a quantification mask.

We apologize for the omission in the Method Section. The condensate size quantification was performed with ImageJ. The nuclei were first identified with DAPI staining and masked out from the ADPr channel. The image was then thresholded with the “Maximum Entropy” method from ImageJ and the “Analyze Particles” function was used to identify condensates with size larger than 1 pixel. Quantification mask will be provided.

Fig. 3D: Does p62 KD affect PARP14 localization? The reported picture shows only 2 cells, each with different staining of PARP14.

p62 KD reduced the number of PARP14 condensates but did not change their localization. We will provide a representative image with more cells to illustrate this effect more clearly and provide quantification on the change in number of condensates.

Fig. 4A: Please quantify the PARP14 co-IP signal with p62, normalized to PARP14 total levels. In the reported WB, it is difficult to see the interaction between PARP14 and p62 in untreated conditions. Please provide clearer WB.

As suggested, we will quantify the PARP14 co-IP signal by normalizing it with PARP14 input levels. Additionally, we will provide a clearer WB in the revised manuscript.

Additionally, I would expect an increased interaction between PARP14 and p62 upon IFN treatment due to PARP14 recruitment to p62 condensates, not just because of increased PARP14 levels. Since the authors show that PARP14 is not recruited to ADPr condensates upon RBN treatment (Fig. 2G), why is the interaction between p62 and PARP14 so high under RBN treatment?

RBN treatment inhibits PARP14 catalytic activity but simultaneously increases PARP14 levels, as first described by Schenkel et al., Cell Chem Biol 2021. Western blot data indicate that the interaction between PARP14 and p62 is independent of this activity and instead depends on PARP14 protein levels. However, the formation of ADPr/PARP14-containing condensates requires catalytically active PARP14. Based on these data, we conclude that the colocalization of p62 and PARP14 depends on the catalytic activity of PARP14, which is reflected by its ADP-ribosylation.

Fig. 4C: Please quantify WB signals of ADP-ribosylated p62 for the different conditions analyzed. ADP-ribosylation of p62 is still present in cells lacking PARP14. Are there other enzymes that can modify p62? Moreover, the authors state: "We observed an increased MARylation of p62 upon IFNγ treatment" (line 230); is this dependent on the increase in PARP14 levels or the translocation of PARP14 to ADPr condensates? Quantifications should help clarify this aspect.

WB signals will be quantified. We agree with the reviewer's observation regarding the presence of ADP-ribosylated p62 in PARP14 KO cells. The basal levels of ADP-ribosylated p62 may be due to other PARP enzymes. However, PARP14 is critical for the increase in ADP-ribosylation under IFNγ treatment, as the increase was not observed in PARP14 KO cells. Given that PARP14 inhibitors increase PARP14 levels, we interpret that the increase in p62 MARylation requires an increase in active PARP14 levels, not just its total level. Since PARP14 activity is crucial for the localization of PARP14 to p62 condensates and its enrichment of ADPr signals, it is possible, as suggested by the reviewer, that the increase in MARylation of p62 is dependent on the translocation of PARP14 to the structure. However, the field currently lacks the tools to disrupt p62 bodies without knocking down p62 to definitively test whether colocalization is required for the MARylation increase.

Fig. 4G: Quantificationsare required.

Quantification will be provided

Reviewer #2 (Significance (Required)): The role of the PARP family in cellular processes is a very active and rapidly growing field. New information about the organization of PARPs in the nucleus, cytosol, or different types of bodies/structures is certainly relevant to the field. However, the present study is too preliminary at the moment to be considered highly relevant. Both the data analysis and conclusions need to be carefully reviewed. After major revisions, the manuscript might be of general interest if well contextualized within the fields of post-translational modification and protein degradation processes. It would remain in any case interesting for the field of ADP ribosylation. We thank the reviewer for recognizing the significance of our work in the rapidly evolving field of PARP biology. We apologize for the lack of clarity that we indeed quantified over 100 cells across at least 4 fields of images for the data reported. To further address the concerns raised, we will provide additional cell-based quantification to strengthen our claims. Furthermore, we will enhance the contextualization of our findings within the broader frameworks of post-translational modification and protein degradation processes in the Introduction and Discussion sections.

Reviewer #3 (Evidence, reproducibility and clarity (Required)):

In this manuscript, titled "Interferon-Induced PARP14-Mediated ADP-Ribosylation in p62 Bodies Requires an Active Ubiquitin-Proteasome System", Raja et al. perform fluorescence microscopy assays and molecular analyses on cultured cells ex vivo to further our understanding of ADP-ribose accumulations that form in the cytoplasm in response to IFγ stimulation. Guided by the operon-like linkage and co-expression patterns of PARP14, PARP9, and DTX3L and recent reports describing a SARS-CoV-2-dissolved cytoplasmic body induced by interferon-induced that is rich in ADP-ribose and the PARP9/DTX3L heterodimer form following IFγ stimulation, the authors provide clarity in this manuscript regarding two knowledge gaps - (i) what catalyzes mono(ADP-ribosyl)ation within these structures (as PARP9 lacks ADP-ribosyltransferase activity) and (ii) how these foci/condensates relate to similarly composed autophagy-associate "p62 bodies" that have been previously described. Using a combination of genetic depletion and inhibitor-based approaches, the authors show that these ADPr-condensates rely on the catalytic activity of PARP14. The authors also show that while these ADPr-condensates share componentry with "p62 bodies" like ubiquitin and p62 itself, these foci are distinct accumulations, as they lack both LC3B and sensitivity to autophagy inhibitors and require an active ubiquitn-mediated proteosomal degradation system.

While this report represents a more incremental advance in our understanding of these cell signaling structures, especially considering a pair of very recently published and similar reports (Kar et al., EMBO J [2024] and Chaves Ribiero et al EMBO J [2024]), the work here is well-written and reasoned and complements these works with some novelty and distinction to that reported literature. The experiments are definitive and of high quality and the authors interpretations/conclusions are largely well-supported by the results. Thus, it is my opinion that this work is appropriate for publication with predominantly minor revisions (outlined below) and a few more substantive experimental additions.

We thank the reviewer for recognizing the high quality of our work and for acknowledging that our interpretations are well-supported by the results. We appreciate that the reviewer deemed our work ready for publication with minor revisions. We believe the reviewer’s perception of our work as incremental arises because two related studies were published in June, after we submitted our work to Review Commons in May. According to the Scooping Protection Policy, our work should still be considered novel. The publication of these studies in EMBO J, which reported the discovery of PARP14 in ADPr-containing condensate formation, highlights the significance of our research. However, we further contribute to this discovery by demonstrating the critical role of PARP14 using multiple genetic manipulation techniques (siRNA, shRNA, and CRISPR-Cas9) and chemical inhibitors (catalytic and PROTAC), indicating the rigor of our studies. Moreover, we not only investigate PARP14 but also define the identity of these condensates related to p62 bodies and establish the requirement of an active ubiquitin-proteasome system for their formation.

Major Comment:

A major claim and novelty reported here is that the ADPr condensates are distinct from "p62 bodies". The evidence to support this rely largely on differences in their sensitivity to pharmacological treatments as well as somewhat subtle differences in FRAP recovery in p62 condensates after IFNgamma treatment. But, this claim would be better supported with more comprehensive mapping of differences in the componentry or functional outcomes of these condensates. The authors might consider:

-Mass spectrometry against p62 (a common component) in standard "p62 bodies" and ADPr Condensates, followed by IF to confirm significantly different composition in, what is argued here, these distinct structures. -Fine mapping of concentration dependence of components that give rise to these distinct condensates as has been demonstrated in papers like Riback et al Nature 2020 and others. -Methodology of the author's choosing to decipher functional outcomes from these condensates followed by demonstration that components unique to ADPr condensates are dispensable for functioning "p62 bodies" and, vice versa, components unique to "p62 bodies" are dispensable for ADPr Condensate function.

As rightly pointed out by the reviewer, our studies indicate the alteration in the composition and dynamics of p62 bodies upon IFNγ treatment. This was assessed using immunofluorescence against various known components of p62 bodies (Fig. 3B, 5E-I), quantification of the condensate size (Fig. S3C), and p62 mobility assessment by photokinetic experiments (Fig. 3C, 4F). In considering reviewer’s suggestions, we will perform p62 interactome studies with and without IFNγ treatment to identify potential changes. Additionally, we will analyze the concentration dependence of ADPr for condensate formation. However, we believe that investigating the functional outcomes is beyond the scope of this manuscript.

Minor Comments:

Overall, the representative microscopy images are far too small. For the benefit of future readers, please consider enlarging these images.

More of the quantitation of microscopy images, with accompanying statistics, that are found in abundance in the supplemental material should find their way into the main figures of the manuscript. This will give room for larger and more reader-friendly representative microscopy images in the main figures/text as discussed briefly above.

We appreciate the reviewer’s suggestions and rightly pointed out that our quantification and statistics were in supplementary materials. Given that we have provided over 800 image panels, we will restructure the manuscript so that the cell biology information is more readily available. We will move our quantification data and statistics currently in the supplementary materials to the main figures. Additionally, we will provide larger and more reader-friendly representative microscopy images in the main text as suggested.

Can the authors test whether or not the condensates are purely driven by mono(ADP-ribosyl)ation? Or does poly(ADP-ribose) co-occupy these condensates and play a substantive role?

We have tested for the presence of poly(ADP-ribose) in the condensates and found it is not present. We will provide the supporting data.

The manuscript would benefit from discussing very recent and related reports (Kar et al., EMBO J [2024] and Chaves Ribiero et al EMBO J [2024]), that I suspect were not available at the time of submission.

Yes, the reviewer is correct that our submission preceded the publication of these related reports. In light of the co-discovery, we will add a section to discuss their findings.

IFNalpha and IFNbeta, which are used in Figure S1, do not appear as reagents in Table 1.

We apologize for the oversight and will add the information on IFNa and IFNb to Table 1.

On lines 113-114, it would seem more appropriate to describe the increase of PARP-14 as statistically significant and largest in magnitude. "most significant" would just mean lowest p-value, which I expect is different that the authors intend here.

We thank the reviewer's suggestion will modify the text as follows:

"PARP9, PARP12, and PARP14 were statistically significantly upregulated at 6 and 24 hours post-treatment, with PARP14 showing the largest increase in mRNA expression levels (Fig. 1D and S1B)."

In Figure S1, better care should be taken to crop and align the western blots.

We thank the reviewer for pointing this out, and we will properly align and crop the western blots in Figure S1.

On line 154, it may be more appropriate to describe ITK as a "weaker" inhibitor of PARP14 relative to PARP11. It certainly is effective as an inhibitor (Figs. S2A and S2G) and its unclear how the authors (or anyone would) define what qualities make it "weak".

We thank the reviewer's suggestion and will modify the text as follows:

“…—specifically RBN012579 (hereafter RBN) and ITK7 (a potent PARP11 inhibitor with inhibitory effects on PARP14 that are weaker than RBN)—…"

The multiple bands for PARP14 in Figure 3E should be addressed. Why does this differ from other blots from the same cells?

We believe that the multiple bands seen can be due to insufficient blocking and using a different lot of PARP14 antibodies. We will address the issue by performing a new experiment with proper blocking conditions and using the same lot of PARP14 antibody as other blots. It should be noted that that variation is also observed in the reagent website: https://www.scbt.com/p/parp-14-antibody-c-1

Reviewer #3 (Significance (Required)):

I expect the advances in this work will appeal more to specialists who are interested in ADP-ribosylation as a signaling molecule and to those engaged in biotechnological efforts to drug immunological responses.

The advances reported here are incremental. The ADPr condensates that form in response to IFNgamma, the involvement of PARP9/DTX3L, and very recently the involvement of PARP14 and its MARylation activity are all known. Less known is the notion that this condensate is distinct from other kinds of "bodies", which is a clear point of novelty, especially if buttressed by the authors as suggested in this review.

We agree with the reviewer that our work is significant for multiple fields, including ADP-ribosylation and immunology. The perception of our work as incremental likely stems from the publication of two related recent studies in June, after our submission in May. According to the Scooping Protection Policy in Review Commons, our work should remain novel in editorial consideration. More importantly, the back-to-back EMBO J studies highlight the importance of reporting the critical role of PARP14 in ADPr-containing condensate formation. We further contribute to this discovery in three aspects: (1) we rigorously demonstrate the critical role of PARP14 through multiple genetic techniques (siRNA, shRNA, and CRISPR-Cas9) and chemical inhibitors (catalytic and PROTAC), (2) we reveal the identity of these condensates as related to p62 bodies, and (3) we define their requirement for an active ubiquitin-proteasome system.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

In this manuscript, titled "Interferon-Induced PARP14-Mediated ADP-Ribosylation in p62 Bodies Requires an Active Ubiquitin-Proteasome System", Raja et al. perform fluorescence microscopy assays and molecular analyses on cultured cells ex vivo to further our understanding of ADP-ribose accumulations that form in the cytoplasm in response to IFγ stimulation. Guided by the operon-like linkage and co-expression patterns of PARP14, PARP9, and DTX3L and recent reports describing a SARS-CoV-2-dissolved cytoplasmic body induced by interferon-induced that is rich in ADP-ribose and the PARP9/DTX3L heterodimer form following IFγ stimulation, the authors provide clarity in this manuscript regarding two knowledge gaps - (i) what catalyzes mono(ADP-ribosyl)ation within these structures (as PARP9 lacks ADP-ribosyltransferase activity) and (ii) how these foci/condensates relate to similarly composed autophagy-associate "p62 bodies" that have been previously described. Using a combination of genetic depletion and inhibitor-based approaches, the authors show that these ADPr-condensates rely on the catalytic activity of PARP14. The authors also show that while these ADPr-condensates share componentry with "p62 bodies" like ubiquitin and p62 itself, these foci are distinct accumulations, as they lack both LC3B and sensitivity to autophagy inhibitors and require an active ubiquitn-mediated proteosomal degradation system.

While this report represents a more incremental advance in our understanding of these cell signaling structures, especially considering a pair of very recently published and similar reports (Kar et al., EMBO J [2024] and Chaves Ribiero et al EMBO J [2024]), the work here is well-written and reasoned and complements these works with some novelty and distinction to that reported literature. The experiments are definitive and of high quality and the authors interpretations/conclusions are largely well-supported by the results. Thus, it is my opinion that this work is appropriate for publication with predominantly minor revisions (outlined below) and a few more substantive experimental additions.

Major Comment:

A major claim and novelty reported here is that the ADPr condensates are distinct from "p62 bodies". The evidence to support this rely largely on differences in their sensitivity to pharmacological treatments as well as somewhat subtle differences in FRAP recovery in p62 condensates after IFNgamma treatment. But, this claim would be better supported with more comprehensive mapping of differences in the componentry or functional outcomes of these condensates. The authors might consider:

• Mass spectrometry against p62 (a common component) in standard "p62 bodies" and ADPr Condensates, followed by IF to confirm significantly different composition in, what is argued here, these distinct structures.
• Fine mapping of concentration dependence of components that give rise to these distinct condensates as has been demonstrated in papers like Riback et al Nature 2020 and others.
• Methodology of the author's choosing to decipher functional outcomes from these condensates followed by demonstration that components unique to ADPr condensates are dispensable for functioning "p62 bodies" and, vice versa, components unique to "p62 bodies" are dispensable for ADPr Condensate function.

Minor Comments:

Overall, the representative microscopy images are far too small. For the benefit of future readers, please consider enlarging these images.

More of the quantitation of microscopy images, with accompanying statistics, that are found in abundance in the supplemental material should find their way into the main figures of the manuscript. This will give room for larger and more reader-friendly representative microscopy images in the main figures/text as discussed briefly above.

Can the authors test whether or not the condensates are purely driven by mono(ADP-ribosyl)ation? Or does poly(ADP-ribose) co-occupy these condensates and play a substantive role?

The manuscript would benefit from discussing very recent and related reports (Kar et al., EMBO J [2024] and Chaves Ribiero et al EMBO J [2024]), that I suspect were not available at the time of submission.

IFNalpha and IFNbeta, which are used in Figure S1, do not appear as reagents in Table 1.

On lines 113-114, it would seem more appropriate to describe the increase of PARP-14 as statistically significant and largest in magnitude. "most significant" would just mean lowest p-value, which I expect is different that the authors intend here.

In Figure S1, better care should be taken to crop and align the western blots.

On line 154, it may be more appropriate to describe ITK as a "weaker" inhibitor of PARP14 relative to PARP11. It certainly is effective as an inhibitor (Figs. S2A and S2G) and its unclear how the authors (or anyone would) define what qualities make it "weak".

The multiple bands for PARP14 in Figure 3E should be addressed. Why does this differ from other blots from the same cells?

Significance

I expect the advances in this work will appeal more to specialists who are interested in ADP-ribosylation as a signaling molecule and to those engaged in biotechnological efforts to drug immunological responses.

The advances reported here are incremental. The ADPr condensates that form in response to IFNgamma, the involvement of PARP9/DTX3L, and very recently the involvement of PARP14 and its MARylation activity are all known. Less known is the notion that this condensate is distinct from other kinds of "bodies", which is a clear point of novelty, especially if buttressed by the authors as suggested in this review.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

This manuscript investigates the formation of a novel cellular structure or condensate, similar to p62 bodies, that includes PARP14 and p62. The interferon-induced PARP14-mediated ADP-ribosylation of p62 in these condensates depends on an active ubiquitin-proteasome system. These condensates are characterized by the presence of PARP14 and ADPr and include some, but not all, components of the p62 bodies. Furthermore, their formation depends on both ubiquitin activation and proteasome activity, but it is unaffected by autophagy inhibition, unlike conventional p62 bodies.

The Introduction provides a well-delineated context of condensates, highlighting the importance of post-translational modifications in responding to environmental changes.

Although the manuscript is well-organized with apparent logical development, there are weaknesses that diminish the impact of the reported data. A more accurate review of the structures, both from a morphological and quantitative perspective, would strengthen the conclusions and the overall impact of this work. Additionally, while the authors have analyzed the contribution of PARP14 to condensate formation, the biological significance of these structures remains unclear. For instance, performing MS (mass spectrometry) analysis on the described structures could help identify their composition and functions.

The methodologies used in this study are standard for molecular and cellular biology research, including immunofluorescence assays, transient transfections, immunoprecipitations, and fluorescence recovery after photobleaching (FRAP) assays. These methods are described in detail and can be reproduced.

Below, please find a list of comments and suggestions to enhance the robustness of the data:

Major Points

1. The IF analyses are central to the conclusions reported and are employed for each of the inhibitors or other tools used to investigate the formation of these condensates. The quality of the IF images needs to be improved; the shape and contacts of the condensates should be analyzed using either super-resolution or EM microscopy, or preferably both. The lack of morphometry and quantification from cell populations needs to be addressed for all experiments. These analyses are needed to support the claim that the condensates presented in this study are indeed novel structures, rather than being transient aggregates of a different nature.
2. The claim that PARP14 is essential for the formation of condensates requires support by the analyses indicated above. Minor points regarding Fig. 1 are indicated below. I suggest performing KD of PARP9 and/or PARP12 (whose expression is increased upon IFN treatment) and checking ADPr condensates to validate the central role of PARP14.
3. According to the text, "PARP14 was pulled down by ADP-ribose binding Af1521 macrodomain following IFNγ treatment (Fig. 2H)", but the legend to the figure says otherwise. A Pan-ADPr binding reagent (MABE1016) is reported in the figure. Although the conclusion is similar for the results obtained with these two tools (but they must be described and reported properly), it is still insufficient to claim that PARP14 is ADP-ribosylated. This point should be at least discussed.
4. I have difficulties analyzing the colocalization with the different organelles, even enlarging the images as much as possible. In most cases, only one condensate per image is shown. Continuities with the nuclear envelope appear in some cases: has this been investigated?

Minor Points

1. Fig. 1A: The DAPI images at 3 and 6 hours are reversed. Additionally, for Fig. S1a and Fig. 1A, please include quantifications.
2. Fig. 1B: Check PARP14 levels (and other IFN-PARPs) under the same experimental conditions.
3. Fig. 1H: Explain why PARP14 IF staining is still visible upon RBN012811 treatment, while it is completely lost in WB analysis or upon PARP14 siRNA treatment (Fig. 1G). In addition, please include IF quantifications.
4. Fig. 2C: Please include quantifications.
5. Fig. 2E: The RBN treatment time is not indicated. Please include this information in the figure legend.
6. Fig. 2G: I am not convinced about the PARP14 staining. IF images do not show an increase in PARP14 levels, while WB analysis shows a strong increase in PARP14 protein levels (see Fig. 2E). Moreover, the RBN treatment time was not indicated; please include it in the figure legend. Does RBN alone affect PARP14 localization? The reported picture shows only 2 cells, each with a different subcellular localization of PARP14. As previously suggested, quantifications are required.
7. Fig. 3B: Pearson's correlation coefficient (PCC) is reported for n=3. The images show one condensate per cell. Under these conditions, the number of cells analyzed should be at least 100 for each experiment. Additionally, the PCC between PARP14 and p62 at steady state is shown to be 60% (which is quite high). However, the IF pictures do not support this quantification. Can the authors provide higher-resolution pictures? Does PARP14 always co-localize with p62? Lines 207-208 state: "these findings suggest that PARP14 is localized to p62 bodies upon IFNγ treatment when ADP-ribosylation occurs." According to the PCC value, the two proteins co-localize even in the absence of IFN. Can the authors clarify this aspect?
8. Fig. S3B: Please include quantifications.
9. Fig. S3C: How was the condensate size quantified? It would be useful to show a quantification mask.
10. Fig. 3D: Does p62 KD affect PARP14 localization? The reported picture shows only 2 cells, each with different staining of PARP14.
11. Fig. 4A: Please quantify the PARP14 co-IP signal with p62, normalized to PARP14 total levels. In the reported WB, it is difficult to see the interaction between PARP14 and p62 in untreated conditions. Please provide clearer WB. Additionally, I would expect an increased interaction between PARP14 and p62 upon IFN treatment due to PARP14 recruitment to p62 condensates, not just because of increased PARP14 levels. Since the authors show that PARP14 is not recruited to ADPr condensates upon RBN treatment (Fig. 2G), why is the interaction between p62 and PARP14 so high under RBN treatment?
12. Fig. 4C: Please quantify WB signals of ADP-ribosylated p62 for the different conditions analyzed. ADP-ribosylation of p62 is still present in cells lacking PARP14. Are there other enzymes that can modify p62? Moreover, the authors state: "We observed an increased MARylation of p62 upon IFNγ treatment" (line 230); is this dependent on the increase in PARP14 levels or the translocation of PARP14 to ADPr condensates? Quantifications should help clarify this aspect.
13. Fig. 4G: Quantifications are required.

Significance

The role of the PARP family in cellular processes is a very active and rapidly growing field. New information about the organization of PARPs in the nucleus, cytosol, or different types of bodies/structures is certainly relevant to the field.

However, the present study is too preliminary at the moment to be considered highly relevant. Both the data analysis and conclusions need to be carefully reviewed. After major revisions, the manuscript might be of general interest if well contextualized within the fields of post-translational modification and protein degradation processes. It would remain in any case interesting for the field of ADP ribosylation.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary

In this study, Raja et al. found cytoplasmic condensates formed by the treatment of INFγ, investigated components of these condensates and identified p62, NBR1 and PARP14 as their components. INFγ treatment induced PARP14 expression, and PAPR14 inhibitor treatment inhibited condensation formation, suggesting that the amount of PARP14 and its enzymatic activity are important for the condensate formation. The ADPr-positive p62 condensates were independent of autophagic degradation, and proteasomal activity was required for their formation.

Major comment

1. The finding that the ubiquitin-proteasome, but not autophagy activity, is indispensable for the formation of p62 condensates is of interest. However, the molecular mechanism by which the ubiquitin-proteasome system (UPS) is involved in the regulation of the PARP14-p62 condensate is still unclear. Which step(s) is the UPS involved?
2. The p62 condensate serves as a scaffold for autophagosome formation through the assembling autophagy receptors including NBR1 and TAX1BP1, followed by recruiting ATG proteins such as FIP200. While ADPr-positive p62 condensates also contain NBR1 and polyubiquitinated proteins, they are unrelated to autophagic degradation. It is unclear what factors govern autophagy-independent function.
3. The authors claim that the amount of PARP14 and its MAR activity are essential for the condensate formation. However, all experiments were performed only with PARP14 inhibitors, and further validation is needed. If the importance of PARP14 activity is to be directly demonstrated, experiments in which an enzyme activity mutant is introduced into PARP14 KO cells are needed.
4. In Figure 2a, the heatmap alone is insufficient. Neither errors nor statistical comparisons are indicated.
5. The statistical analysis of Figure S2 is inappropriate; instead of t-tests, multiple comparisons should be used to compare three or more groups.

Minor comment

1. What percentage of p62 condensates upon INFγ treatment are ADPr positive? Are all p62 bodies seen with INFγ stimulation unrelated to autophagy?
2. Is ADPr condensation a PARA14-specific phenomenon? PARP9 and PARP12 were also upregulated by INFγ treatment. Are these factors also involved in condensate formation?
3. Figure 4D appears to be immunoprecipitation (IP) under non-denaturing conditions. If so, it is not possible to distinguish whether the MAR signal is derived from p62 or from the p62 interacting proteins (the associated ubiquitinated substrates). IP experiments should be performed under denaturing conditions.
4. In Figure 5B, which band is HO1, the upper or lower?
5. There is no image for ubiquitin in S5D. Right panel in Figure 4F shows only IFγ + RBN, which should show all data sets in the same panel.

Significance

Liquid droplets, which have continuously being identified in cells, are a hot topic in cell biology. Droplet formation, structure, molecular dynamics, and degradation, as well as their abnormalities and disease development due to genetic mutation and stress, are of wide-ranging interest from basic to pathological aspects. Therefore, this research has the potential to attract interest from a wide range of fields.

General assessment

Overall, the data are clear and the phenomenon is of interest. However, the molecular mechanism and biological significance of the condensate formation is unknown; It is unclear why proteasome activity is required for the formation of PARP14-mediated ADP ribosylation. It is also unclear what the consequences are for the cell if the ADPr-positive condensates are not formed. Thea authors should address these general and important issues and provide the data If not all.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

The authors do not wish to provide a response at this time.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

The manuscript by Balachandra and Amodeo presents Bellymount-Pulsed Tracking as a technique for continuous long-term imaging of Drosophila oogenesis. This approach modifies the existing Bellymount technique by exposing restrained female flies to pulses of CO2 anesthesia in combination with image acquisition. Flies that survived the restraint were kept alive for many hours by addition of a liquid diet in the restraint apparatus. This allowed for imaging and tracking of egg chamber development over longer time periods than capable with ex vivo culturing methods. However, the authors did report a 40% mortality rate and decreased fecundity compared to unrestrained flies. Using this method the authors were able to image and measure the growth rate of developing egg chambers in living flies, and capture events like vitellogenesis which relies on the interactions of multiple organ systems.

This technique is a notable contribution to the fly community, as it could be useful for studying processes that require interactions between multiple tissues and organs, as well as for long-term imaging of other internal structures in the adult fly. The significance is somewhat reduced due to the relatively high mortality rate and the decreased fecundity and egg chamber growth rate reported. However, the authors should be commended for their diligence in documenting the limitations of the procedure, as this now provides a strong jumping off point to improve the technique if it becomes widely adopted by the fly community. Overall, the experiments appear to have been carefully performed and the manuscript is clearly written. However, there are several issues that should be addressed prior to publication.

Major concerns

1. The movies of egg chamber development are challenging to interpret. They could be improved by the addition of timestamps and other annotations. Having multiple example movies of the same process would also be valuable. It could be helpful to potential users of this technique to show the process the authors used for identifying the same egg chamber between such long time points.
2. Figure 4 - Given that the Bellymount PT technique slows oogenesis and reduces egg chamber growth in vitellogenic stages (Figure 3E), it is possible that Bellymount PT slows yolk protein uptake. It would be important to establish a baseline for how much to expect yolk protein levels to change across stages to compare to measurements obtained with Bellymount PT. It would be a relatively simple experiment to show the change in yolk protein uptake across stages in fixed samples. This could also be performed for His2Av dynamics during nurse cell dumping.
3. Movie 11 - The authors propose that Bellymount-PT can be used to visualize the process of border cell migration. However, there is no obvious movement of the cluster relative to the nurse cell nuclei over the course of the 3 hour long movie. The authors should either show a better movie of border cell migration, or remove this claim from the manuscript.
4. Movie 13 - The authors claim that they see egg chamber rotation continue in stage 9 and 10 egg chambers. This movie is not convincing. There is also very strong evidence in the literature that egg chamber rotation ends at stage 8. Chen et al., Cell Reports, 2017 showed using a method that tracks follicle cell migration in vivo that rotational migration ends during stage 8. The only movement of follicle cells after stage 8 is due to the epithelial reorganization that occurs during the posterior movement of the follicle cells as the stretch cells flatten. Additionally, after stage 8 follicle cells lose their circumferentially oriented actin protrusions that drive rotation. This claim should be removed from the manuscript.

Minor comments

1. Line 104 - The authors mention that CO2 affects fertility in flies. They should also reference Sustar et al., Genetics, 2023 and Zimmerman and Berg, PLoS One, 2024 for wider ranging effects of CO2 on oogenesis.
2. Line 244 - Although it is true that the original paper describing egg chamber rotation reported that it starts at 5, subsequent studies from multiple labs have confirmed that it begins much earlier. First shown by Cetera et al., Nature Communications, 2014 but later confirmed by Bilder, Dahmann, and Mirouse labs. Chen et al., Cell Reports, 2016 has even published a movie of an egg chamber initiating rotation as it buds from the germarium.
3. Figures of egg chambers are generally oriented anterior on the left and posterior on the right. Reorienting all the figures would be challenging, so the recommendation is to be clear in the figure legends the orientation of the images. This is important given they are shown in different orientations in Figure 1 than throughout the rest of the paper, and also will be helpful for readers who may not be familiar with the structure of the ovary/egg chambers.
4. Figure 1B and Methods line 334 - Should "Rely" be "Relay"?
5. Figure 1E - Oocyte nuclei are missing from the diagrams of stage 7, 13 and 14 egg chambers. Also, "G" looks like a figure panel label, could just say Germarium
6. Figure 3F-H - "Stagee" should be "Stage"
7. Figure 4B - Why is the fluorescence for egg chamber #6 so much higher than the others? It makes the slopes of the other samples hard to see.
8. Figure 4D,E,G - For clarity, the labeled boxes should be the same color as the lines on the associated graphs. In line 790 "Note the steady increase of H2Av in all three regions as it exits the nurse cell nuclei" - this is not actually shown without the nurse cell nuclei average intensity being on the graph as well.
9. Line 787 - "Note the flow of H2Av" - "flow" is not actually shown in these static images. Consider a more precise description.

Referee Cross-commenting

The other reviewers make several excellent points. We personally feel that it is beyond the scope of this initial report to ask the authors to show that they can see all aspects of oogenesis with this technique. If the method becomes widely adopted by the oogenesis community, individual researchers can optimize it to suit the exact process they want to study. If the authors want to claim they can see a particular process, it needs to be well documented and convincing. For example, we agree that the movies that claim to show egg chamber rotation (both during established stages and later) and border cell migration need to be improved or the claims need to be removed. However, we feel that the authors have documented enough other interesting processes to make the study worthy of publication. Likewise, asking the authors to determine the minimal time window that can be used for imaging could take months of open-ended work and is something that could be better tackled by subsequent users depending on the requirements of the biological process they want to study. It seems better to get the work out into the public sooner rather than later so that improvements can be crowd sourced.

Finally, although Flp-out clones were used for cell tracking in the original Belly mount paper, this technique will be less effective during the first half of oogenesis when the egg chamber is rotating, as the clone is likely to rotate into and out of sight between imaging time points.

Significance

This technique is a notable contribution to the fly community, as it could be useful for studying processes that require interactions between multiple tissues and organs, as well as for long-term imaging of other internal structures in the adult fly. The significance is somewhat reduced due to the relatively high mortality rate and the decreased fecundity and egg chamber growth rate reported. However, the authors should be commended for their diligence in documenting the limitations of the procedure, as this now provides a strong jumping off point to improve the technique if it becomes widely adopted by the fly community. Overall, the experiments appear to have been carefully performed and the manuscript is clearly written. However, there are several issues that should be addressed prior to publication.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary:

The authors describe an improvement of the Bellymount imaging method for internal tissues of the fly's abdomen. They are able to increase the total duration of the imaging by introducing pulsed anesthesia. This allows the immobilized flies to take up food in between the imaging; this increases survival rate and allows for longer total imaging times. The authors illustrate the technique by tracking the development of egg chambers.

Major Points

• The Bellymount PT method results in decreased fecundity, which might affect the processes (oogenesis) the authors looked at. Indeed, the authors conclude that "oogenesis is not completely stalled under the Bellymount-PT protocol" (line 140). The authors do provide some data indicating that egg chambers develop (Fig. 2G,H; Fig. 3F,H), in particular a stage 10 egg chamber proceeding to a stage where dorsal appendages seem to form. However, for early stage egg chambers this is less convincing. The egg chambers show an increase in (cross-sectional) area, however, what is the evidence that they also mature? For example, during egg chamber maturation, the ratio of oocyte/nurse cell volume changes, follicle cells re-arrange, etc. The authors should test whether any of these characteristics can be observed in egg chambers imaged using Bellymount PT. This may include the imaging of egg chambers in which both nuclei and plasma membranes are visualized.
• A potential advantage of the Bellymount PT method is the ability to follow the dynamics of processes. A current drawback, however, is the rather low temporal resolution as the fly needs to wake up between single images. The authors should provide an estimate for the minimal possible cycle time and should test whether flies imaged at 10 minutes interval show lower survival/fecundity than flies imaged at 2 hours interval.
• The authors claim that they can track on a cellular level (based on nuclei), but it is unclear how accurate the tracking is. Especially cell tracking over very long times might be challenging here, as the time delay between two time points is big. The authors should test the accuracy of their tracking, potentially by creating Flip-out clones and using them as a control.
• The authors show that they can visualize cell membranes (Moesin-GFP, Fig. 2C). Tracking cells over time based on their membranes would greatly widen the applicability of the method as it would enable to analyze the complex cellular dynamics during egg chamber maturation. The authors should test whether cells can be tracked over time (e.g. using Moesin-GFP) using their technique.
• Movie 11. The authors claim that they can capture border cell migration. However, it is unclear whether the border cells actually migrate towards posterior. The authors should track and quantitatively analyze the migration path of the border cells in their movies.
• Movie 12. The authors claim that they can observe egg chamber rotation. However, it is unclear whether the egg chambers actually rotate. The authors should track cells and quantify the angular velocity of movement.

Minor Points

• Please move the labels of the scale bars to the legends.
• The figures (especially 2 and 3) would benefit from a clearer structuring. Moving part of them to supplementary figures would also help.
• "stage" typo in figure 3

Significance

The authors describe here an improvement of an existing technique. The advantage of the improved technique is the longer imaging time, which potentially allows users to track cells/organelles/proteins over time. However, tracking requires the user to connect single time points with each other, which is somewhat unclear at this time. Moreover, the potential applicability (and significance) of the technique would be widened if visualization and tracking of cell membranes/organelles/vesicles would be possible. With these further optimizations, the technique would add a useful tool to the Drosophila community.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary

The Drosophila ovary is an established model system for many aspects of development and cell biology. In vitro culture of live ovaries has provided valuable insight, yet these methods do not accurately mimic oogenesis in vivo for some stages. Here the authors develop a new method that allows for sustained imaging of ovaries in intact flies, maintaining normal physiology.

The method provides a valuable addition to the field. Processes such as growth, cell migration, egg chamber rotation, yolk uptake and nurse cell dumping can be observed in the intact fly. Time lapse and 3D reconstruction provide valuable tools. While the detail/resolution of the images is not as good as ex vivo or fixed samples, the ability to maintain normal development and homeostasis provides a novel advantage. The figures and movies are well-presented and sufficient detail is provided in the methods.

Major comments

1. Why do the authors think that growth is slowed? The imaging process or the trapping/anesthesia of the fly? For example, if the frequency of imaging was varied, it could reveal whether it was the actual imaging that affected development. Did the length of time the fly had been in the trap make a difference? The sentence on lines 190-191 is not clear.
2. In Movie 6, the nurse cell nuclear shape does not look normal - more ovoid than round. Perhaps some settings are off in the 3D reconstruction.
3. Movie 11 - why do the border cells seem stalled?
4. There is no discussion of the earliest stages of oogenesis. Is it possible to see egg chambers forming from the germarium?

Minor comments

1. It would be helpful to mention if the egg chambers stay in similar locations or move around - is it challenging to locate the same egg chamber after 2 hours?
2. Are any egg chambers degenerating? This could indicate stress in the fly.
3. In Figure 4D, release of HisAV into the cytoplasm is described. Similar release of nuclear proteins was described by Cooley et al. 1992 so this paper could be cited.
4. At 321 minutes in Figure 4D, a large nucleus is apparent in the oocyte. Is this an oocyte nucleus or evidence for nurse cell translocation to the oocyte as described in Ali-Murthy et al. 2021?

Significance

The technique provides a significant advance to the field, extending the time period currently possible to image ovaries through the Belly Mount method. It will immediately benefit researchers working on the ovary but could be extended to many other tissues in the fly abdomen such as the gut and tumor models.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

The authors do not wish to provide a response at this time.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Summary:

Parkin, a E3 ubiquitin ligase, is involved in the clearance of damaged mitochondrial via mitophagy. Upon mitochondrial damage, the activated Parkin ubiquitinates many mitochondrial substrates, leading to the recruitment of mitophagy effectors. However, the mechanism of substrate recognition by Parkin is still not known.

In this manuscript, Koszela et al. utilized diverse biochemical assays and biophysical approaches, combined with AlphaFold prediction, to identify a conserved region in the flexible linker between the Ubl and RING0 domains of Parkin that recognizes mitochondrial GTPase Miro1 via a stretch of hydrophobic residues and is critical for its ubiquitination activity on Miro1. This manuscript reveals the mechanisms by which Parkin recognizes and ubiquitinates substrate Miro1, providing a biochemical explanation for the presence of Parkin at the mitochondrial membrane prior to activation by mitochondrial damage. This study also provides insights into mitochondrial homeostasis and may facilitate new therapeutic approaches for Parkinson's disease.

Major Comments:

• The authors should expand the background introduction to include the biological function of Miro1, the domain architecture of Miro1 and more context of Miro1 K572 ubiquitination in mitophagy.
• Figure 1B is confusing. Due to the presence of various bands, it is hard to assign specific bands in each lane. In addition, there are various unlabeled bands that makes things unclear. The authors should include loading controls to clearly discern pParkin, Ube1, Ube2L3, and all substrates.
• In Figure 1B, it was not possible to identify the ubiquitination bands of E2 enzyme UBE2L3 and the E1 enzyme UBE1. Please indicate these bands on the gel.
• Since ubiquitinated Miro1 and Mfn1 are similar in molecular weight (Fig. 1b), the authors should show a western blot against the Miro1 and Mfn1 tag as done in the supplementary information, At least for the competition assays involving both Miro1 and Mfn1.
• The conclusion that Miro1 is pParkin's preferred substrate is not convincing. In the competition assay used to show substrate preference, Miro1 is at a five-fold higher concentration than the other substrates and 25-fold higher than FANCI/D2. This would ultimately drive pParkin's interaction with Miro1. This is further highlighted by the fact that it adding Mfn1 in excess has a similar effect. The competition assay should be done at equimolar concentrations of Miro1 and substrate. More convincing would be a competition assay where substrate ubiquitination is quantified at several different concentrations of Miro1.
• In Figure 1F, it is unclear what is defined as "high" or "low" ubiquitination levels statistically. Some of the changes in ubiquitination levels are extremely subtle (ex. mitoNEET and FancI/D2 in the presence and absence of Miro1 and Mfn1). In some cases, I find it extremely difficult to tell if there is any change in the ubiquitination levels when comparing lanes containing excess of different substrates. I would like to see band quantifications of this experiment in triplicate to support the conclusions drawn from the competition assay.
• The authors used both unmodified and phosphorylated Parkin for the crosslinking experiments and observe no difference in the intensity of the bands. However, this is not sufficient to draw any conclusion about the affinity between phosphorylated Parkin and Miro1 (which was done in lines 341-343). The authors should comment on why they did not test pParkin binding with Miro1, especially given the statement:

"In our assays in the absence of pUb, pParkin must interact with its substrates without the action of pUb, likely through 158 transient, low affinity interactions" - The reference to Parkin115-124 as a "Substrate Targeting Region (STR)" is misleading. This would imply that this motif in Parkin is responsible for general substrate recognition when there is no direct evidence of this. In Figure 5F, the authors create a synthetic peptide based off the STR sequence. Although this sequence was effective in inhibiting the ubiquitination of Miro1, it was ineffective against Mfn1. This would indicate that Mfn1 relies on a completely different set of interactions for ubiquitination by Parkin. I suggest that the authors tone down the language in describing this region and rename this region (perhaps "Miro1 Targeting Region (MTR)"?). - The authors appear to confuse plDDT and PAE scores in Figure 5B. The PAE describes the expected positional error of each residue in the model. The plot should be colored in terms of Expected Position Error (Ångstrom), not plDDT scores.

Minor Comments:

• Figure 1A would benefit from a schematic showing the domain architecture. If the goal is to appreciate the length of the linker, then showing the actual amino acid length would be beneficial.
• In Supplementary Figure 2D, the authors performed the MST experiment with His6-Smt3-tagged Parkin. The group had previously shown that the presence of the tag artificially interferes with autoubiquitination, potentially by forming intramolecular interactions. The SEC, Native Page, and ITC data of untagged Parkin with Miro1 provide sufficient evidence that the interaction between the two are weak. The authors should consider removing the MST data, since they are not congruent with the other experiments.
• The ITC data in Supplementary Figure 2C look promising. It would be nice if the authors could try to quantify the Kd of their STR peptides to Miro1
• Are STR peptides 1 and/or 2 unable to inhibit ubiquitination of other Parkin substrates besides Mfn1? Do these other substrates utilize the STR for recognition? AlphaFold modeling may provide some insight on Parkin recognition of other substrates.
• The authors shold consider using AlphaFold3 to model the interaction of pParkin with Miro1 compares to unmodified Parkin.
• Please label the protein names in Figure 4A for a better presentation.
• Page 2, line 37. "...by a 65-residue flexible region (linker) to a unique to Parkin RING0 domain..." should be "...by a 65-residue flexible region (linker) to a unique Parkin RING0 domain...". The second "to" should be omitted.
• Page 3, Line 48: "fulfill", not "fulfil"
• Page 5, line 110. In sentence, "...phosphorylation at Ser65 of Parkin...", it is better to explicitly state that this phosphorylation happens on the Parkin Ubl domain.
• Page 7, line151. Figure 1F should be Figure 1G.
• Page 11, line 241. In sentence "...Miro1 residues R263, R265 and D228...", do the authors mean R261 and not R265?

Significance

Parkin is an E3 ubiquitin ligase that is activated to ubiquitinate diverse substrates on the mitochondrial membrane in response to mitochondrial damage, thereby recruiting mitophagy effectors. This study reveals the mechanisms by which Parkin recognizes and ubiquitinates Miro1, providing insights into mitochondrial homeostasis and facilitating new therapeutic approaches for Parkinson's disease.

Readers with a background in protein ubiquitination and mitochondrial homeostasis might be interested in this study. My expertise includes protein ubiquitination and structural biology. However, I do not have sufficient expertise to evaluate the NMR experiments in this manuscript.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Koszela et al. have submitted this manuscript demonstrating the molecular mechanism of interaction between Parkin and one of its known substrates, Miro1. While the interaction and ubiquitination of Miro1 by Parkin (and it's role in mitochondrial quality control) has been known since 2011, as demonstrated by the Schwarz group and others, the mechanism of action has remained unknown. The ability of Parkin to ubiquitinate multiple proteins upon mitochondrial damage has indeed led many groups to speculate that Parkin is a promiscuous E3 ligase upon activation; this manuscript tries to provide a rationale for the interaction with one of its known substrates through a combination of biochemical and biophysical studies.

The authors demonstrate that Miro1 is efficiently ubiquitinated in in vitro biochemical assays in comparison to a few mitochondrial and non-mitochondrial proteins in an attempt to show that Miro1 is a preferred substrate for Parkin. Cross-linking coupled with mass spectrometry, SAXS and NMR experiments were used to provide compelling evidence for a direct and specific interaction between Parkin and Miro1. Molecular modelling using Colabfold and biochemical assays with mutants of the proposed interaction site were then used to provide further proof for the specificity of the interaction. This interaction is shown to occur between the conserved a.a.115-122 (referred to in this study as STR; located in the linker connecting the Ubl to RING0) and the EF domain of Miro1. Interestingly, the authors show that peptides corresponding to 115-122 competitively inhibit ubiquitination of Miro1 by Parkin. Overall, this article constitutes an important addition to our understanding of Parkin's mechanism of action. However, some of the key claims remain unsubstantiated, as described below.

Major issues:

1. In line 151 the authors claim, 'these data strongly support the hypothesis that Miro1 is the preferred substrate of pParkin...'. Arguably, the biggest issue with this study is the lack of substantial proof that Miro1 is the preferred parkin substrate in a cellular or physiological context. This claim cannot be made based on a biochemical assay with three other proteins. The Harper group has performed in-depth proteomics studies on the kinetics of Parkin-mediated ubiquitination and proposed that VDACs and Mfn2 (among a few others) are most efficiently ubiquitinated upon mitochondrial damage in induced neurons (Ordureau et al, 2018,2020). Interestingly, neither of these papers have been mentioned by the authors in this manuscript. The Trempe group has shown that Mfn2 is efficiently targeted by Parkin through mitochondrial reconstitution assays and proximity ligation assays (Vranas et al, 2022). The authors need to substantiate their claim through cellular or mitochondrial assays to prove that Miro1 is the preferred physiological substrate of Parkin. Cellular experiments also account for cellular abundance and proximity of Parkin to the substrate, which is not possible in biochemical assays of the kind presented here. In the absence of strong experimental proof for this claim, these claims should be tampered down to Miro1 being "the preferred substrate compared to the other proteins in this assay", and the manuscript should focus more on the molecular mechanism of interaction between Miro1 and Parkin.
2. In addition to the point above, the authors do not describe the rationale for specifically choosing Mfn1 and MitoNEET for their comparison with Miro1 as substrates. Interestingly, Miro1, MitoNEET and Mfn1 are not among the most efficiently ubiquitinated substrates of Parkin (Ordureau et al, 2018). Additionally, the authors have used a construct of Mfn1 that lacks the full HR1 domain for their assays. Previously, it has been shown that the HR1 of mitofusins is targeted by Parkin (McLelland et al. 2018). Can the authors prove that their Mfn1 construct is as efficiently ubiquitinated as full-length Mfn1 by Parkin? If it is not possible to obtain soluble full-length Mfn1 or other membrane proteins for these assays, then I strongly recommend the authors should perform mitochondrial reconstitution assays as others have performed previously (Vranas et al, 2022) and use this opportunity to also report the ubiquitination kinetics of multiple mitochondrial substrates compared to Miro1 to make a more compelling case for substrate preference.
3. The authors show that both pParkin-Miro1 and Parkin-Miro1 complexes can be captured by chemical cross-linking. It is well-established in the field that pUbl binds to RING0 (Gladkova et al, 2018) (Sauve et al, 2018) while non-phosphorylated Ubl binds RING1 (Trempe et al, 2013). The Komander group has also shown that the ACT (adjacent to the STR) element binds RING2 in the activated Parkin structure (Gladkova et al, 2018). This suggests that STR could occupy different positions in the Parkin and pParkin. The authors have only reported the cross-link/MS data and model of the Parkin-Miro1 complex. Arguably, the pParkin-Miro1 data is just as, if not more, relevant given that pParkin represents the activated form the ligase. The authors need to robustly establish that Miro1 binds to the STR element in both cases by demonstrating the following:

A. Mass spectrometry data from cross-linked pParkin-Miro1 complex suggesting the same interaction site.

B. Colabfold modelling with the pParkin structure to show that Miro1 would bind to the same element. 4. Does Parkin only bind to Miro1, or can it bind to Miro2 as well? Are there differences between the binding site and Ub target sites between the two proteins? The author should also show experimentally if both proteins get ubiquitinated as efficiently by Parkin and if the STR element is involved in recognizing both proteins. Interestingly, the Harper group reports that Miro2 gets more efficiently ubiquitinated than Miro1 (Ordureau et al, 2018). 5. In Figure 5D, the level of unmodified Miro1 seems to be similar in assays with WT or I122Y Parkin, though the former seems to form longer chains while the latter forms shorter chains. Is there an explanation for this? Perhaps, the authors need to perform this assay at shorter time points to show that there is more unmodified Miro1 remaining when treated with I122Y Parkin (and similarly for the L221R mutant of Miro1)? Also, why is the effect of Miro1 L221R and Parkin I122Y not additive?

Minor comments:

1. The authors should report the full cross-linking/MS data report from Merox including the full peptide table and decoy analysis report.
2. The authors should report statistics for the fit of the Colabfold model to the experimental SAXS curve.
3. Why is the Parkin-Miro1 interaction only captured by NMR and not by ITC? The authors should at least attempt to show the interaction of the STR peptide with Miro1 by an orthogonal technique like ITC.
4. The authors should report the NMR line broadening data quantitatively i.e. reporting the reduction in signal intensity for the peaks upon peptide Miro1 binding to quantitatively demonstrate that the 115-122 peak intensity reduction is more significant than other regions.
5. Figure 4 (structure figure) and B (PAE plot) should be annotated with the names of domains and elements in Parkin and Miro1 to make these figures clearer and more informative.

Referees cross-commenting

I am in agreement with reviewers 1 and 2. Both of them raise valid and interesting points in their reviews.

Specifically, I would like to highlight the following:

1. Reviewer 1 makes a very good point (5/6) highlighting that L119A does not impair Parkin recruitment in the previously reported study. I second this concern and believe that the authors need to re-frame their discussion and make it much more nuanced with regards to the role of Miro1-Parkin interaction in mitophagy (if any at all). Additionally, the authors should also note that previous studies in the field from the Youle group (Narendra et al, 2008) and multiple other groups have shown a complete absence of Parkin recruitment to healthy mitochondria. Parkin recruitment to healthy mitochondria hence remains a controversial idea at best, with no evidence for it outside of Parkin overexpression systems (Safiulina et al, 2018) which can also lead to artifacts. The discussion should take all major studies/observations into account to propose a more nuanced picture of the role of Parkin-Miro1 interaction. Perhaps, this interaction plays more of a role in mitochondrial quarantine (Wang et al. 2011) as suggested by the Schwarz group than in Parkin recruitment?
2. Reviewer 3 raises a valid concern about the lack of quantification in ubiquitination assays and alludes to the difficulty in visualizing ubiquitination of multiple proteins. That was a concern I also had but did not include in my review. Perhaps, the authors should also show western blots for each of the protein (in a time course experiment) demonstrating the difference in ubiquitination kinetics of each of proteins instead of busy SDS-PAGE gels for the assay.

Significance

The key strength of this study is the strong biophysical evidence of a direct interaction between Parkin and Miro1 and the discovery of the Miro1 binding site on Parkin. The biophysical and biochemical experiments in this study have been well-designed and executed. The evidence for a specific interaction between Parkin and Miro1 has been provided through multiple approaches. The authors should be commended for this effort. The biggest limitation of this study is the lack of proof that Miro1 is the preferred Parkin substrate in a cellular/physiological context since in biochemical assays Parkin can ubiquitinate multiple proteins non-specifically. Substrate preference claims need to be established in more physiologically relevant experimental settings.

Overall, the study represents a mechanistic advance in terms of our understanding of the interaction between Parkin and one of its substrates i.e. Miro1, showing that Parkin can indeed specifically bind its substrates before targeting them for ubiquitination. This might also inspire others to investigate the molecular mechanism of action of Parkin with other substrates. This paper would likely appeal specialized audiences i.e. biochemists and structural biologists studying Parkin in mitochondrial quality control.

Reviewer expertise: Expert biochemist and biophysicist with a number highly cited works in the field of mitochondrial quality control and Parkin.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

The manuscript by Koszela et al. explores the substrate preference of the Parkinson's disease associated ubiquitin E3-ligase, Parkin. They conclude that Miro1 is a preferred substrate of Parkin and go on to further characterize a binding site of Parkin to Miro1 using a range of biochemical approaches. This site is identical to previously reported (see point 2). The experimental work is strong with many high-quality assays supporting their ideas; however, there are several major points that should be considered:

1. The majority (perhaps all) of their biochemical work on Miro1 uses a truncated form of Miro1 lacking the first GTPase domain. It isn't at all clear why this is the case, as no justification is given. Moreover, functional full-length Miro1 has been purified in several papers (e.g. PMID: 33132189). If ubiquitination kinetics are different between the full-length and truncated form of Miro1, this would call into question the significance of the findings in vivo, where the truncation does not exist.
2. As the manuscript is currently written, there are areas which do not do justice to previous work. Firstly, the authors state throughout the manuscript that no previous work has identified a binding interface between Parkin and one of its substrates, e.g., in the abstract "no substrate interaction site in Parkin has been reported". This is not true as a recent paper already described the binding interface (DOI: 10.1038/s44318-024-00028-1). "we identify a conserved region in the flexible linker", again this interface is identical to that identified previously. Therefore, this study does not "identify" this interface. Given the timing, it is likely that this discovery has been "scooped" by the previous study, but since the present study goes much further in the biochemical characterization of the interface, it would not diminish the paper's importance to rewrite it, giving proper credit where due. Secondly, the authors spend a large part of their discussion speculating on the significance of non-activated Parkin being able to bind Miro e.g., "Importantly, our results suggest that Parkin can interact with Miro1 independently of its activation state, as Parkin phosphorylation does not detectably increase its interaction with Miro1...". Again, this was already known as Parkin has been shown to be recruited to mitochondria upon Miro1 overexpression in the absence of PINK1 (DOI: 10.15252/embj.201899384 and DOI: 10.15252 /embj. 2018100715). The further biochemical characterisation of the Parkin-Miro1 interaction is important and therefore, in both cases the work contained within the manuscript is still a significant contribution, which should, however, be properly discussed in the light of published work.
3. The Miro L221R mutation is used to disrupt Miro-Parkin interaction. Yet, this non-conservative mutation in the midst of a folded domain might have other effects, like affecting calcium binding or preventing the folding of the domain. This is not tested. The complementary Parkin-I122Y used for the same purpose decreases but does not abolish Parkin-Miro1 binding. Parkin-L119A is proposed to abolish the Parkin-Miro1 interaction. The inclusion of this mutant might be important to fully ascertain the role of Parkin-Miro1 binding in Miro1 ubiquitination.
4. The effect of Miro competition on other substrates' ubiquitylation is marginal and its reproducibility is questionable (whether mitoNEET ubiquitylation is affected at all in figure 1G is unclear. This blot is anyway over processed with an unnaturally uniform grey background). If the authors wish to make a point about it, these experiments should be repeated and quantified. Moreover, since the model is that the specific Miro-Parkin interaction is involved, the mutants above should be used in the same competition experiments and shown to be unable to compete.
5. Related to the previous point, one important factor about the kinetics that the authors do not discuss is how any of it relates to mitophagy in vivo. There very well might be a slight intrinsic preference at a given concentration of substrate and Parkin; however, how this plays out in the cell is not clear, e.g., Miro1 may be many times more, or less, abundant than Mfn1, and so a preference might not have much of an effect. So ubiquitination kinetics would need to be considered in a broader cellular context.
6. Related to the above point, the authors state "Parkin translocation was diminished upon L119A mutation, supporting the importance of the Parkin Miro1-interacting site in mitophagy.". However, the study (not cited but which this reviewer assumes to be DOI: 10.1038/s44318-024-00028-1 since the L119A mutation has only ever been used here) finds no change in Parkin recruitment upon damage. So, it cannot be used to support "the importance of the Parkin Miro1-interacting site in mitophagy".

Referees cross-commenting

The reviews align well together with many overlaping point and similar assessment of the significance. Reviewer 2 brings in interesting points pertaining to literature that we were not aware of, explaining why we didn't make these points.

One comment on reviewer's 3 last major points

It does not appear that there is a confusion between pIDDT and PAE scores. The plot is coloured according to PAE (which is a residue x residue 2D matrix, figure 4B), while the protein ribbon is coloured according to pIDDT, which is a 1D per-residue confidence score.

Significance

This study provides an in-depth in vitro assessment of a specific binding interface between the E3-ligase Parkin and one of its substrate Miro1. Although this interface has been recently described, this study goes well beyond previous knowledge by showing that the interface is important for complete Miro ubiquitylation by Parkin, therefore showing that interactions involving unstructured linkers participate in substrate recognition by the E3-ligase. The importance of this interaction remains to be assessed in vivo. This study is of interest to basic mitochondrial dynamics, quality control and mitophagy researcher as well as translational Parkinson's Disease researchers.

The reviewer's expertise is in mitochondrial membrane dynamics.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Manuscript number: RC- 2024-02497

Corresponding author(s): Tourriere, Hélene and Maraver, Antonio

1. General Statements [optional]

We sincerely thank the Editors and Reviewers for the time devoted to our manuscript. We found their critiques interesting and very helpful. After careful examination and thanks to a large collaborative effort, we will be able to answer to all the reviewers’ comments by adding significantly new experimental data.

We are also encouraged by the positive comments of the Reviewers:

“This manuscript will likely engage oncologists who investigate the chemotherapy-resistant mechanisms of platinum compounds in NSCLC treatment” (Reviewer 1);

“Overall, the authors have conducted experiments that sufficiently elucidate their claims, and the description of the experiments is detailed.”; and “Overall, this work unveils a novel mechanism for Notch activation in response to platinum chemotherapy, providing a renewed outlook on overcoming chemotherapy resistance in NSCLC” (Reviewer 2).

We are also aware that both reviewers agreed that there is room for improvement, and we are sure that upon accomplishment of all proposed experiments both reviewers will be fully satisfied.

Please bear in mind that although it was known that platinum-based chemotherapy induced the Notch pathway in lung cancer cells, the underlying molecular mechanism was largely unknown. Thanks to the molecular dissection we performed in our study, we propose an innovative treatment for patients with lung cancer, the main cause of death by cancer in the world. Hence, we agree with both reviewers that our study will be appealing for a large number of cancer researchers, and we feel it will be also the case for those interested in DNA damage, Notch and MDM2 pathways.

2. Description of the planned revisions

Reviewer #1 (Evidence, reproducibility and clarity (Required)):

Summary: This manuscript from Maraver and co-authors investigates the putative resistance mechanisms that hinder the efficacy of platinum-based therapies (e.g., carboplatin) against non-small cell lung carcinoma (NSCLC). Using in vitro lung cancer cell lines, shRNA-based knockdown, and exogenous overexpression systems, the research describes a DNA damage-induced resistance mechanism involving the NOTCH signaling pathway and the E3 ligase MDM2. The authors show that carboplatin treatment induces DNA damage and promotes ATM activation, which in turn activates the NOTCH signaling pathway via ubiquitination and stabilization of the Notch Intracellular Domain (NICD). New findings include the MDM2-mediated ubiquitination and stabilization of NICD. Using in vivo NSCLC-PDX models, they demonstrate that combining carboplatin with Notch and MDM2 inhibitors can enhance tumor killing, suggesting that targeting the MDM2/NICD axis in conjunction with carboplatin may be a viable therapeutic alternative. Furthermore, they show that NICD and MDM2 levels are elevated among tumor samples from chemo-resistant patients. Consistent with these findings, high MDM2 levels correlate with poor progression-free survival (PFS) in NSCLC patients.

[Authors] We thank this reviewer for her/his fair summary of our work that highlights our new findings.

Major comments:

Some of the key conclusions may not be convincing.

[Authors] We understand the concerns that reviewer might have and we are sure that upon accomplishment of all experiments detailed below, she/he will be convinced that the manuscript will be ready for publication.

1. One significant weakness of the manuscript is the lack of exploration into the underlying mechanism of how MDM2 mediates the stabilization of NICD. While the observation of MDM2-mediated NICD stabilization is intriguing, it is important to provide a more convincing explanation for the reviewers. This could be achieved by offering a detailed molecular mechanism, especially considering that MDM2 typically targets proteins for degradation.

[Authors] After reading this reviewer’s comment, we realize we did a poor job discussing better the previous study demonstrating that MDM2 induced ubiquitination on NICD but not for degradative purposes (Pettersson et al., 2013). In particular, they performed it using a mutated form of ubiquitin in lysine 48, i.e., the K48R mutant. Like this, the authors of this seminal study demonstrated that MDM2 was still able to induce ubiquitination in NICD, and hence it was not degradative.

Still, and to confirm that this is the case also upon DNA damage, we will perform experiments using same K48R mutant to formally prove that MDM2 upon DNA damage does not ubiquitinate NICD via lysine 48-linked polymers, and hence it is not degradative. Even more, upon discussion with Laetitia Linares, author of our study and long-lasting expert in ubiquitination (for instance see (Riscal et al., 2016) and (Arena et al., 2018)), we will use another ubiquitin mutant in lysine 63. This different type of ubiquitination does not mark proteins for degradation but promote an association of the targeted protein with DNA helping for DNA repair (Liu et al., 2018). Using a ubiquitin mutated in this lysine, i.e., K63R, this type of ubiquitination cannot occur. Taking into account that we observe NICD increase ubiquitination upon DNA damage, the use of K63R will be very informative.

Hence, we will repeat experiments of current Figure 3A with the same WT ubiquitin as before, and now also with K48R and K63R mutants. Even more, we will also include mutant forms of ubiquitin which can only form ubiquitin chains on lysine 48 (K48 only) or lysine 63 (K63 only) and we anticipate that in the presence of K48 only mutant, NICD will not be ubiquitinated upon DNA damage, while the use of K63 only mutant will be very useful. All these data will be part of the new Figure 3A.

Of note, Dr Linares has all tools required to perform these experiments and hence we will start them soon.

Another weakness lies in the unclear role and the underlying mechanism of ATM in the MDM2-mediated NICD stabilization. While the data presented (Fig. 3B, 3C) suggest that carboplatin could elevate MDM2 levels for NICD stabilization, a more precise method to induce MDM2 overexpression specifically for targeting NICD is required. It appears that ATM plays a crucial role in this regulatory process. The following questions must be addressed: Does ATM induce the phosphorylation of MDM2 for its protein stabilization and/or E3 ligase activity?

[Authors] There are several points here.

For the first one, the use of a more precise method to induce MDM2 overexpression, it is exactly what we did in Figure 4A, i.e., ectopic expression of MDM2 to demonstrate that MDM2 is sufficient to increase NICD levels.

For the second one, i.e., the phosphorylation status of MDM2 by ATM in our system, we will perform different experiments. There are up to six proposed residues in MDM2 to be phosphorylated by ATM upon DNA damage: S386, S395, S407, T419, S425, and S429 (Cheng et al., 2011). Among all of them, S395 is the most well-known and again Dr Linares has interesting tools we will use to answer to this specific reviewer’s point. We will use an MDM2 mutant harboring an aspartate instead of the serine in this position, i.e., S395D, that mimics the serine 395 phosphorylation induced by ATM upon DNA damage. We will use this mutant together with the WT and 464A MDM2 proteins already used, and if this residue is important in our phenotype, total levels of NICD will be even higher and/or localize more in the nuclei when compared with WT MDM2. All these new data will appear as the new Figure 4A __and new Figure 4B__.

Furthermore, we will also use an antibody that recognizes this phosphorylation site by WB after carboplatin treatment and it will be part of the new Figure 3B.

Finally, we will also express WT MDM2 and purify it by immunoprecipitation in different experimental conditions: steady state, upon carboplatin treatment and also in combination of carboplatin and ATM inhibitor, to perform phospho-proteomics analysis upon all these conditions. Of note, and to show the feasibility of this approach, the proteomic platform at Biocampus in Montpellier has experience using this technique (Kassouf et al., 2019).

The combination therapy of carboplatin with MDM2 and NICD inhibitors may lack compelling rationale (see below).

[Authors] This is a very important point but we discuss it below, where more information is provided by the reviewer. Still, we anticipate we will perform a new in vivo experiment to answer to this point.

In lines 275-276, the authors stated that their preclinical data establish the enhancement of carboplatin's therapeutic effect in NSCLC in vivo through MDM2-NICD axis inhibition. However, it's important to note that this finding remains preliminary at this stage.

[Authors] We consider that our statement is not exaggerated, but we will tone down the message as proposed by the reviewer in the next submission.

Minor comments:

1. The observed loss of NICD during ATMi + carboplatin treatment in Figures 2A and 2B raises the question of whether ATM regulates the gene transcription of NOTCH. In addition to the CHX assay conducted in Figures 2C and 2D, quantifying NOTCH mRNA upon ATM inhibition could provide further insights. Alternatively, referencing relevant studies on this topic may strengthen the discussion.

[Authors] This is an interesting experiment and we will perform it.

In Figures 4A and 4B, the noticeable discrepancy between the exogenous expression of wild-type (WT) MDM2 and catalytically inactive MDM2-464A raises concerns. It is essential to consider if the reduced ubiquitination and stability of NICD might be attributed to varying levels of MDM2-464A in the cells rather than its catalytic inactivity. While p53 ubiquitination was utilized as a control, ensuring comparable levels of MDM2 and MDM2-464A expression could enhance the experimental rigor. Compared to the smear poly-ubiquitination bands observed for MDM2 in Figure 4B, the ubiquitination of NICD appears simpler. What distinguishes the feature of MDM2-mediated NICD ubiquitination? Could it potentially involve mono-ubiquitination?

[Authors] The point of the reviewer is well taken, and importantly, as mentioned above in main point 2, we will repeat these experiments and will appear as new Figure 4A and new Figure 4B.

Regarding the type of ubiquitination, as explained in detail in major point 1 to same reviewer, we will fully characterize the type of ubiquitination on NICD induced by DNA damage, and we will confirm that MDM2 is required for this specific ubiquitination in future new Figure 4C where we will overexpress the required ubiquitin forms and WT MDM2.

In Figure 5A, the authors need to consider conducting additional NOTCH-associated factors to definitively demonstrate the activation of NOTCH signaling beyond HES1. Alternatively, in Figure 5B, the NICD Western blot could be complemented by detecting HES1 or other NOTCH-associated factors.

[Authors] To answer to this particular point, we will test for other downstream targets of Notch as NRARP and it will appear as part of new Figure 5C.

In Figures 5C and 5D, crucial control groups are missing, specifically mice treated solely with SP141+DBZ, carboplatin+SP141, and SP141+DBZ. It is essential to include these groups to demonstrate that the enhanced tumor killing results from the combination of carboplatin with SP141 and/or DBZ, rather than from SP141 and DBZ alone. Furthermore, in addition to the currently used NSCLC-PDX model harboring the p53 (P151R) mutation, it would be informative to include a NSCLC-PDX model expressing WT p53.

[Authors] This is a crucial point in this rebuttal as mentioned before in major point 3 and we detail it in here.

We did only 3 groups because preliminary data indicated that SP141 in combination with carboplatin was not showing any benefit compared to carboplatin alone while upon combination of carboplatin with Notch inhibition there was only a slight increase in therapeutic carboplatin benefit but otherwise not very potent, and for simplicity we preferred to don’t show these data. But, after reading this point from Reviewer 1, even if we will propose later only the triple combination for patients, we clearly need to demonstrate that the other combinations are not potent enough or not at all.

The reviewer asked to include: “SP141+DBZ, carboplatin+SP141, and SP141+DBZ”. We imagine that she/he meant: SP141+DBZ, carboplatin+SP141, and carboplatin +DBZ, that together with the vehicle, carboplatin and carboplatin+SP141+DBZ makes 6 groups of treatments. Putting together the 8 mice devoted for tumor growth and survival, plus 4 mice for the acute treatment for IHC and WB purposes (for current Figures 5A and 5B) makes a total of 72, that is a substantial number of mice. Of note, since we performed the in vivo experiment presented in the current manuscript, a new Notch inhibitor called nirogacestat, appear in the market being the first in class Notch inhibitor to treat solid cancer patients (desmoid tumors) after demonstrating a significant therapeutic effect in clinical trials (Gounder et al., 2023).

Hence, we will take advantage of the repetition of this experiment to substitute this new molecule instead of DBZ, that is an interesting molecule for preclinical research, but without any clinical relevance. Therefore, the use of nirogacestat will further increase the medical impact of our data. Importantly, nirogacestat is better tolerated than DBZ, meaning that mice can be treated for longer periods of time and we propose in here to treat up to 12 weeks. Finally, after discussion with Quentin Thomas, author of the manuscript and clinical researcher in the lab, we will provide 4 carboplatin cycles as it is proposed today to NSCLC patients in an attempt of getting closer to the clinical setting. In particular we will provide carboplatin to mice on weeks 1, 4, 7 and 10, while treating with MDM2 inhibitor (SP141) and Notch inhibitor (nirogacestat) from Monday to Friday for the 12 weeks.

This experiment will be long and will require an important use of resources both human and financial, but we are sure that the effect in tumor growth and survival will be more dramatic than the one presented now.

On the contrary and as explained in the 4th subheading part of this “revision plan”, including another 72 mice to treat a p53 proficient NSCLC PDX, when we already demonstrated in vitro that p53 is not required for the phenotype described in this study, for us it is totally unfeasible by ethical reasons, i.e., the use of animals in research (please see below for further details).

All the new data will appear as new Figure 5 (B to E). For new Figure 5A please see below the major comment 2 of Reviewer 2.

Though beyond the current study's scope, in the discussion section, the authors may want to propose or hypothesize on how MDM2-mediated NICD stabilization contributes to carboplatin resistance. This could provide valuable insights for future research directions.

[Authors] We will discuss this part as proposed by the reviewer.

In the Western blot results, the total ATM and ATR controls were absent.

[Authors] The reviewer is totally right and we will repeat experiments to include all the totals as requested.

Authors may choose to include a graphical abstract at the end of their study to visually illustrate the mechanisms they have described.

[Authors] Very good idea thanks, we will do it.

Reviewer #1 (Significance (Required)):

Advance: The authors aim to present a novel perspective on the resistance mechanisms to platinum compounds in NSCLC therapy. They explore platinum compounds-induced DNA damage, ATM activation, and MDM2-mediated stabilization of the active form of NOTCH (NICD). However, to strengthen their claims, they must provide more conclusive results.

Audience: This manuscript will likely engage oncologists who investigate the chemotherapy-resistant mechanisms of platinum compounds in NSCLC treatment, as well as scientists specializing in NOTCH and MDM2 pathways. However, the manuscript's central claims lack robust support from the available data, and the current approaches employed are not sufficiently thoughtful and rigorous; there is room for improvement.

My expertise is molecular medicine, cancer biology, and epigenetics.

[Authors] We want to thank again this reviewer for her/his helpful comments that will increase the impact and the relevance of our study while keeping the original message.

We are also very satisfied when she/he said: “This manuscript will likely engage oncologists who investigate the chemotherapy-resistant mechanisms of platinum compounds in NSCLC treatment”.

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

In this manuscript, Sara Bernardo et al. investigated the molecular mechanisms underlying the activation of the Notch signaling in response to DNA damage induced by platinum-based chemotherapeutic agents in non-small cell lung cancer (NSCLC). They demonstrated that carboplatin treatment induces DNA double-strand breaks (DSBs) and stabilizes NICD, a process dependent on ATM and mediated by MDM2. In vivo experiments in patient-derived xenografts (PDX) showed that inhibition of NICD and MDM2 enhanced platinum effectiveness. Furthermore, clinical analysis revealed a correlation between MDM2 expression and poor prognosis in NSCLC patients treated with platinum compounds, emphasizing the clinical relevance of the MDM2-NICD axis in platinum resistance.

[Authors] We thank this reviewer for her/his nice synopsis of our study.

Major comments:

Overall, the authors have conducted experiments that sufficiently elucidate their claims, and the description of the experiments is detailed. However, there is still room for the improvement.

[Authors] We are very pleased that reviewer finds our experimental work “…sufficiently elucidate their claims, and the description of the experiments is detailed.” And we are sure that after all the new experiments we are proposing in here, she/he will be fully satisfied.

1.The finding that MDM2 promoted NICD stability through non degradative ubiquitination is interesting and in line with a previous study. As it is also known that NICD is regulated by various post-translational modifications, including ubiquitination that promotes NICD degradation. It is unclear what's the potential difference between these two types of ubiquitination. For example, do these two differ in specific ubiquitination sites? Can the authors provide some discussion?

[Authors] We agree with the reviewer and hence we will perform a new set of experiments to determine the role of 2 key lysine residues in the ubiquitin protein promoting either degradation or DNA binding. As explained in detail in major point 1 from reviewer 1, we will determine if DNA damage promotes ubiquitination in position 48, i.e., to degrade, or in position 63, i.e., to facilitate the binding to DNA for repairing upon DNA damage, or in any of these 2 positions. And as mentioned above, we will then confirm that MDM2 is responsible of the specific ubiquitination type we will uncover. We are sure that the reviewer will be satisfied by these new data once is generated.

As for the specific ubiquitination sites in NICD, there are up to 17 lysine residues susceptible of being ubiquitinated. Hence unveiling what residues are targeted by MDM2 and if they differ from others inducing degradation as those promoted by the E3 ligase FBXW7, we feel is out of the scope of the current manuscript. Still, we will discuss all this part as kindly proposed by the reviewer.

Could the overexpression of MDM2 or NICD lead to carboplatin resistance in A549 or H358 cells?

[Authors] This is a very interesting experiment and prompted by the reviewer’s comment we started the subcloning of inducible NICD into lentiviral vectors to generate stable cells and test the carboplatin sensitivity in presence of different levels of NICD. These new data will be the new Figure 5A.

The trends observed in the western blot data within the manuscript appear inconsistent. While the authors propose that NICD levels increased upon incubation with carboplatin, the discrepancy arises when considering the NICD levels without cycloheximide (CHX) treatment in Figure 1E, where no significant elevation is observed (Lane 6 vs. Lane 1).

[Authors] The point of the reviewer is well taken. Please bear in mind that in here we are handling several signaling pathways that interact among them while having each one different kinetics. Our finding of increased NICD upon carboplatin treatment is highly consistent in vitro and in vivo, but it is true that in the experiment mentioned by the reviewer is not obvious, probably due to some kinetic issue. We are repeating this experiment to have the increased in NICD upon carboplatin as it is in the rest of the manuscript (up to 9 times only in main figures).

The quality of western blots needs to be improved, especially Fig. 1C and S1C, also Figure 3B. Moreover, the NICD western blot sometimes appears as one band and sometimes as two bands. Please provide an explanation. If possible, please quantify the bands in western blots.

[Authors] We agree with the reviewers that not all WB have the same quality and we will repeat some of them to homogenize the quality all over the manuscript, and particularly, we will repeat the ones kindly pointed out by the reviewer.

The two bands it is something we also noticed and we will pay attention while reproducing the WB, since it might be related to discrepancies in the percentage of acrylamide. If this is not the case, i.e., upon repetition we still observe in some conditions and not in others, we will provide explanations for this in the new submission as kindly proposed by the reviewer.

Finally, and also as proposed by the reviewer we will quantify the WB bands.

Please provide a necessary discussion on whether the targeted treatment approach towards the MDM2-NICD axis is applicable to all patients or only to those with high expression of MDM2/NICD.

[Authors] In the discussion of the current manuscript, we focused into the MDM2 high expression subset of patients for this issue, but in the next submission we will enlarge to patients with high levels of NICD also.

How to interpret the significance of the simultaneous increase in NICD ubiquitination and stability mediated by MDM2? Please provide a relevant discussion.

[Authors] We will provide strong experimental data to go beyond discussion (please see above the experiments with ubiquitin mutants), but we will also provide discussion of this particular point.

In Figure 5B, please also check the level of MDM2. In Figure 5C, carboplatin appears to have little impact on tumor growth. How to explain the increase of Ki-67 in the carboplatin treatment group in Figure 5A?

[Authors] We will measure also levels of MDM2 in the future new Figure 5C as requested by the reviewer.

As for the interesting observation of the Ki67, since we will repeat the whole experiment, we will pay special attention to this point if ever it is repeated. Should be this the case, we will elaborate an explanation.

Minor comments:

1.Please include scale bars in Figure 1B and Supplemental Figure 1B.

[Authors] We thank the reviewer for this comment. We will include the scale bars where required.

2.Figure 5D, the P values of the survival curve should be indicated in the figures.

[Authors] We will include the P values in the future new Figure 5E.

3.The presentation of survival curve data in Figures 5D and 6A should be consistent.

[Authors] The point of the reviewer is well taken and we will use Prism to draw the PFS for patients in Figure 6A as we did for the mice in current Figure 5D.

4.It seems that supplemental figure 2 is missing.

[Authors] We actually jumped from supplemental figure 1 to 3 because we do not have any associated supplemental figure to main Figure 2. We will clarify this point in the next submission.

5.Please carefully check the spelling of the entire text, for example, on page 20, line 426 it should be 'western'. Also, please spell out the abbreviations DDR and ATM.

[Authors] We will double check all spelling and provide the abbreviations kindly suggested by the reviewer.

6.The abbreviation for Cleaved caspase 3 should be CC3.

[Authors] We thank the reviewer for this information, we will use CC3 in the next submission.

Reviewer #2 (Significance (Required)):

Notch signaling is associated with the occurrence and development of non-small cell lung cancer (NSCLC). Previous study indicates that the expression of Notch protein is significantly higher in NSCLC tissues compared to normal tissues (PMID: 31170211). Additionally, the upregulation of Notch1 is correlated with higher tumor grades, lymph node metastasis, tumor-node-metastasis (TNM) staging, and poor prognosis (PMID: 25996086). Abnormal activation of Notch signaling pathway is frequently observed in chemotherapy-resistant NSCLC, and some studies have aimed to address NSCLC drug resistance via modulating Notch signaling (PMID: 30087852, 38301911). This manuscript firstly proposes that MDM2-mediated stabilization of NICD upon DNA damage plays a major role in NSCLC response to platinum chemotherapy. It further suggests that targeting the MDM2-NICD axis could prove to be an effective therapeutic strategy. Overall, this work unveils a novel mechanism for Notch activation in response to platinum chemotherapy, providing a renewed outlook on overcoming chemotherapy resistance in NSCLC. This manuscript will attract those interested in the mechanisms of chemotherapy resistance and novel treatment approaches.

[Authors] We sincerely thank the reviewer for finding that our “…work unveils a novel mechanism for Notch activation in response to platinum chemotherapy, providing a renewed outlook on overcoming chemotherapy resistance in NSCLC”. We are also very satisfied when she/he says: “This manuscript will attract those interested in the mechanisms of chemotherapy resistance and novel treatment approaches.”

Finally, we are convinced that the reviewer will appreciate all the new proposed experimental data, and also that upon finishing all experiments, she/he will think that the manuscript will be suitable for publication.

3. Description of the revisions that have already been incorporated in the transferred manuscript

For simplicity, we decided to introduce all changes in next submission upon conclusion of all experimental approaches proposed above.

4. Description of analyses that authors prefer not to carry out

While we will perform almost all experiments proposed by reviewers, there is one we feel is not possible to do due to ethical reasons. Reviewer 1 wanted us to perform a new in vivo experiment with the same PDX using up to 6 treatment groups. We use 8 mice per condition (for tumor growth and survival) plus 4 for the “acute” treatment for WB and IHC purposes, hence 12 mice x 6 groups = 72 mice, and we will perform this experiment as indicated above and proposed by the reviewer.

On the contrary, the reviewer asked us also to repeat the same experiment with a PDX p53 proficient. While we understand the possible interest, since we demonstrated in vitro that p53 is not required for the protective phenotype of MDM2 and Notch upon DNA damage, we honestly believe that using another 72 mice to confirm this aspect in vivo, is against the rational use of animals in research going against the 3Rs rule. Hence, we will not perform this experiment unless Editors believe is strictly required.

REFERENCES

Arena, G., Cisse, M. Y., Pyrdziak, S., Chatre, L., Riscal, R., Fuentes, M., Arnold, J. J., Kastner, M., Gayte, L., Bertrand-Gaday, C., et al. (2018). Mitochondrial MDM2 Regulates Respiratory Complex I Activity Independently of p53. Mol Cell 69, 594-609 e598.

Cheng, Q., Cross, B., Li, B., Chen, L., Li, Z., and Chen, J. (2011). Regulation of MDM2 E3 ligase activity by phosphorylation after DNA damage. Mol Cell Biol 31, 4951-4963.

Gounder, M., Ratan, R., Alcindor, T., Schoffski, P., van der Graaf, W. T., Wilky, B. A., Riedel, R. F., Lim, A., Smith, L. M., Moody, S., et al. (2023). Nirogacestat, a gamma-Secretase Inhibitor for Desmoid Tumors. N Engl J Med 388, 898-912.

Kassouf, T., Larive, R. M., Morel, A., Urbach, S., Bettache, N., Marcial Medina, M. C., Merezegue, F., Freiss, G., Peter, M., Boissiere-Michot, F., et al. (2019). The Syk Kinase Promotes Mammary Epithelial Integrity and Inhibits Breast Cancer Invasion by Stabilizing the E-Cadherin/Catenin Complex. Cancers (Basel) 11.

Liu, P., Gan, W., Su, S., Hauenstein, A. V., Fu, T. M., Brasher, B., Schwerdtfeger, C., Liang, A. C., Xu, M., and Wei, W. (2018). K63-linked polyubiquitin chains bind to DNA to facilitate DNA damage repair. Sci Signal 11.

Pettersson, S., Sczaniecka, M., McLaren, L., Russell, F., Gladstone, K., Hupp, T., and Wallace, M. (2013). Non-degradative ubiquitination of the Notch1 receptor by the E3 ligase MDM2 activates the Notch signalling pathway. Biochem J 450, 523-536.

Riscal, R., Schrepfer, E., Arena, G., Cisse, M. Y., Bellvert, F., Heuillet, M., Rambow, F., Bonneil, E., Sabourdy, F., Vincent, C., et al. (2016). Chromatin-Bound MDM2 Regulates Serine Metabolism and Redox Homeostasis Independently of p53. Mol Cell 62, 890-902.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

In this manuscript, Sara Bernardo et al. investigated the molecular mechanisms underlying the activation of the Notch signaling in response to DNA damage induced by platinum-based chemotherapeutic agents in non-small cell lung cancer (NSCLC). They demonstrated that carboplatin treatment induces DNA double-strand breaks (DSBs) and stabilizes NICD, a process dependent on ATM and mediated by MDM2. In vivo experiments in patient-derived xenografts (PDX) showed that inhibition of NICD and MDM2 enhanced platinum effectiveness. Furthermore, clinical analysis revealed a correlation between MDM2 expression and poor prognosis in NSCLC patients treated with platinum compounds, emphasizing the clinical relevance of the MDM2-NICD axis in platinum resistance.

Major comments:

Overall, the authors have conducted experiments that sufficiently elucidate their claims, and the description of the experiments is detailed. However, there is still room for the improvement.

1.The finding that MDM2 promoted NICD stability through non degradative ubiquitination is interesting and in line with a previous study. As it is also known that NICD is regulated by various post-translational modifications, including ubiquitination that promotes NICD degradation. It is unclear what's the potential difference between these two types of ubiquitination. For example, do these two differ in specific ubiquitination sites? Can the authors provide some discussion? 2. Could the overexpression of MDM2 or NICD lead to carboplatin resistance in A549 or H358 cells? 3. The trends observed in the western blot data within the manuscript appear inconsistent. While the authors propose that NICD levels increased upon incubation with carboplatin, the discrepancy arises when considering the NICD levels without cycloheximide (CHX) treatment in Figure 1E, where no significant elevation is observed (Lane 6 vs. Lane 1). 4. The quality of western blots needs to be improved, especially Fig. 1C and S1C, also Figure 3B. Moreover, the NICD western blot sometimes appears as one band and sometimes as two bands. Please provide an explanation. If possible, please quantify the bands in western blots. 5. Please provide a necessary discussion on whether the targeted treatment approach towards the MDM2-NICD axis is applicable to all patients or only to those with high expression of MDM2/NICD. 6. How to interpret the significance of the simultaneous increase in NICD ubiquitination and stability mediated by MDM2? Please provide a relevant discussion. 7. In Figure 5B, please also check the level of MDM2. In Figure 5C, carboplatin appears to have little impact on tumor growth. How to explain the increase of Ki-67 in the carboplatin treatment group in Figure 5A?

Minor comments:

1.Please include scale bars in Figure 1B and Supplemental Figure 1B. 2.Figure 5D, the P values of the survival curve should be indicated in the figures. 3.The presentation of survival curve data in Figures 5D and 6A should be consistent. 4.It seems that supplemental figure 2 is missing. 5.Please carefully check the spelling of the entire text, for example, on page 20, line 426 it should be 'western'. Also, please spell out the abbreviations DDR and ATM. 6.The abbreviation for Cleaved caspase 3 should be CC3.

Significance

Notch signaling is associated with the occurrence and development of non-small cell lung cancer (NSCLC). Previous study indicates that the expression of Notch protein is significantly higher in NSCLC tissues compared to normal tissues (PMID: 31170211). Additionally, the upregulation of Notch1 is correlated with higher tumor grades, lymph node metastasis, tumor-node-metastasis (TNM) staging, and poor prognosis (PMID: 25996086). Abnormal activation of Notch signaling pathway is frequently observed in chemotherapy-resistant NSCLC, and some studies have aimed to address NSCLC drug resistance via modulating Notch signaling (PMID: 30087852, 38301911). This manuscript firstly proposes that MDM2-mediated stabilization of NICD upon DNA damage plays a major role in NSCLC response to platinum chemotherapy. It further suggests that targeting the MDM2-NICD axis could prove to be an effective therapeutic strategy. Overall, this work unveils a novel mechanism for Notch activation in response to platinum chemotherapy, providing a renewed outlook on overcoming chemotherapy resistance in NSCLC. This manuscript will attract those interested in the mechanisms of chemotherapy resistance and novel treatment approaches.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary:

This manuscript from Maraver and co-authors investigates the putative resistance mechanisms that hinder the efficacy of platinum-based therapies (e.g., carboplatin) against non-small cell lung carcinoma (NSCLC). Using in vitro lung cancer cell lines, shRNA-based knockdown, and exogenous overexpression systems, the research describes a DNA damage-induced resistance mechanism involving the NOTCH signaling pathway and the E3 ligase MDM2. The authors show that carboplatin treatment induces DNA damage and promotes ATM activation, which in turn activates the NOTCH signaling pathway via ubiquitination and stabilization of the Notch Intracellular Domain (NICD). New findings include the MDM2-mediated ubiquitination and stabilization of NICD. Using in vivo NSCLC-PDX models, they demonstrate that combining carboplatin with Notch and MDM2 inhibitors can enhance tumor killing, suggesting that targeting the MDM2/NICD axis in conjunction with carboplatin may be a viable therapeutic alternative. Furthermore, they show that NICD and MDM2 levels are elevated among tumor samples from chemo-resistant patients. Consistent with these findings, high MDM2 levels correlate with poor progression-free survival (PFS) in NSCLC patients.

Major comments:

Some of the key conclusions may not be convincing.

1. One significant weakness of the manuscript is the lack of exploration into the underlying mechanism of how MDM2 mediates the stabilization of NICD. While the observation of MDM2-mediated NICD stabilization is intriguing, it is important to provide a more convincing explanation for the reviewers. This could be achieved by offering a detailed molecular mechanism, especially considering that MDM2 typically targets proteins for degradation.
2. Another weakness lies in the unclear role and the underlying mechanism of ATM in the MDM2-mediated NICD stabilization. While the data presented (Fig. 3B, 3C) suggest that carboplatin could elevate MDM2 levels for NICD stabilization, a more precise method to induce MDM2 overexpression specifically for targeting NICD is required. It appears that ATM plays a crucial role in this regulatory process. The following questions must be addressed: Does ATM induce the phosphorylation of MDM2 for its protein stabilization and/or E3 ligase activity?
3. The combination therapy of carboplatin with MDM2 and NICD inhibitors may lack compelling rationale (see below).
4. In lines 275-276, the authors stated that their preclinical data establish the enhancement of carboplatin's therapeutic effect in NSCLC in vivo through MDM2-NICD axis inhibition. However, it's important to note that this finding remains preliminary at this stage.

Minor comments:

1. The observed loss of NICD during ATMi + carboplatin treatment in Figures 2A and 2B raises the question of whether ATM regulates the gene transcription of NOTCH. In addition to the CHX assay conducted in Figures 2C and 2D, quantifying NOTCH mRNA upon ATM inhibition could provide further insights. Alternatively, referencing relevant studies on this topic may strengthen the discussion.
2. In Figures 4A and 4B, the noticeable discrepancy between the exogenous expression of wild-type (WT) MDM2 and catalytically inactive MDM2-464A raises concerns. It is essential to consider if the reduced ubiquitination and stability of NICD might be attributed to varying levels of MDM2-464A in the cells rather than its catalytic inactivity. While p53 ubiquitination was utilized as a control, ensuring comparable levels of MDM2 and MDM2-464A expression could enhance the experimental rigor. Compared to the smear poly-ubiquitination bands observed for MDM2 in Figure 4B, the ubiquitination of NICD appears simpler. What distinguishes the feature of MDM2-mediated NICD ubiquitination? Could it potentially involve mono-ubiquitination?
3. In Figure 5A, the authors need to consider conducting additional NOTCH-associated factors to definitively demonstrate the activation of NOTCH signaling beyond HES1. Alternatively, in Figure 5B, the NICD Western blot could be complemented by detecting HES1 or other NOTCH-associated factors.
4. In Figures 5C and 5D, crucial control groups are missing, specifically mice treated solely with SP141+DBZ, carboplatin+SP141, and SP141+DBZ. It is essential to include these groups to demonstrate that the enhanced tumor killing results from the combination of carboplatin with SP141 and/or DBZ, rather than from SP141 and DBZ alone. Furthermore, in addition to the currently used NSCLC-PDX model harboring the p53 (P151R) mutation, it would be informative to include a NSCLC-PDX model expressing WT p53.
5. Though beyond the current study's scope, in the discussion section, the authors may want to propose or hypothesize on how MDM2-mediated NICD stabilization contributes to carboplatin resistance. This could provide valuable insights for future research directions.
6. In the Western blot results, the total ATM and ATR controls were absent.
7. Authors may choose to include a graphical abstract at the end of their study to visually illustrate the mechanisms they have described.

Significance

Advance: The authors aim to present a novel perspective on the resistance mechanisms to platinum compounds in NSCLC therapy. They explore platinum compounds-induced DNA damage, ATM activation, and MDM2-mediated stabilization of the active form of NOTCH (NICD). However, to strengthen their claims, they must provide more conclusive results.

Audience: This manuscript will likely engage oncologists who investigate the chemotherapy-resistant mechanisms of platinum compounds in NSCLC treatment, as well as scientists specializing in NOTCH and MDM2 pathways. However, the manuscript's central claims lack robust support from the available data, and the current approaches employed are not sufficiently thoughtful and rigorous; there is room for improvement."

My expertise is molecular medicine, cancer biology, and epigenetics.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Manuscript number: RC-2024-02394

Corresponding author(s): Altman, Brian J

1. General Statements [optional]

We thank all three Reviewers for their insightful and helpful feedback and suggestions. We strongly believe that addressing these comments has now resulted in a much-improved manuscript. We appreciate that the Reviewers found the manuscript "interesting" with "valuable insights and... obvious novelty", "an important study that is well-done", and "an important understanding of the crosstalk between cancer cells and immune cells as well as the understanding of how the TME disrupts circadian rhythms". All three Reviewers requested a significant revision, which we provide here. We carefully and completely responded to each Reviewer question or suggestion, in most cases with new experiments and text, and in a very few cases with changes or additions to the Discussion section. This includes new data in seven of the original Figures and Supplementary Figures, and one new main Figure and three new Supplementary Figures. Highlights of these new data include testing the role of low pH in cancer cell supernatant on macrophage rhythms, and analysis of single-cell RNA-sequencing data for heterogeneity in macrophage circadian gene expression. Additional experiments were also performed that were not included in the manuscript, and these data are presented in this Response. A detailed point-by-point response to each comment is included below with excerpts of the data and updated text for the reviewers. Please note that the PDF version of this Response includes images of the new Figures inserted in to the manuscript.

2. Point-by-point description of the revisions

__Reviewer #1 __

Evidence, reproducibility and clarity

The manuscript by Knudsen-Clark et al. investigates the novel topic of circadian rhythms in macrophages and their role in tumorigenesis. The authors explore how circadian rhythms of macrophages may be influenced by the tumor microenvironment (TME). They utilize a system of bone marrow-derived macrophages obtained from transgenic mice carrying PER2-Luciferase (PER2-Luc), a trackable marker of rhythmic activity. The study evaluates how conditions associated with the TME, such as polarizing stimuli (to M1 or M2 subtype), acidic pH, and elevated lactate, can each alter circadian rhythms in macrophages. The authors employ several approaches to explore macrophage functions in cancer-related settings. While the manuscript presents interesting findings and may be the first to demonstrate that tumor stimuli alter circadian rhythms in macrophages and impact tumor growth, it lacks a clear conclusion regarding the role of altered circadian rhythms in suppressing tumor growth. Several discrepancies need to be addressed before publication, therefore, the manuscript requires revision before publication, addressing the following comments:

We thank Reviewer #1 for the comments regarding the quality of our work and are pleased that the Reviewer finds that this manuscript "presents interesting findings and may be the first to demonstrate that tumor stimuli alter circadian rhythms in macrophages and impact tumor growth". We have addressed all comments and critiques from Reviewer #1 below. To summarize, we added new data on how different macrophage polarization states affect media pH (Supplementary Figure 4), further characterized gene expression in our distinct macrophage populations (Supplementary Figure 1), provided clarity in the data and text on the universal nature of Clock Correlation Distance (CCD) across macrophage populations (Figure 6), included human tumor-associated macrophage (TAM) data for CCD (Figure 7) analyzed single-cell RNA-sequencing data of TAMs to demonstrate heterogeneity in circadian gene expression (Figure 9), and used tumor-conditioned media to show that low pH still affects macrophage rhythms in this context *Supplementary Figure 5". Thanks to the helpful suggestions of the Reviewer, we also made numerous clarifications and fixed a critical referencing error that the Reviewer identified.

Major comments: 1. It is well known that pro-inflammatory macrophages primarily rely on glycolysis during inflammation, exhibiting dysregulated tricarboxylic acid (TCA) cycle activity. These pro-inflammatory macrophages are commonly referred to as 'M1' or pro-inflammatory, as noted in the manuscript. In contrast, M2 macrophages, or pro-resolution macrophages, are highly dependent on active mitochondrial respiration and oxidative phosphorylation (OXPHOS). Given that M1 macrophages favor glycolysis, they create an acidic environment due to elevated lactate levels and other acidifying metabolites. However, the study does not address this effect. The authors' hypothesis revolves around the acidic environment created by glycolytic tumors, yet they overlook the self-induced acidification of media when culturing M1 macrophages. This raises the question of how the authors explain the reduced circadian rhythms observed in pro-inflammatory macrophages in their study, while low pH and higher lactate levels enhance the amplitude of circadian rhythms. I would encourage the authors to incorporate the glycolytic activity of pro-inflammatory macrophages into their experimental setup. Otherwise the data look contradictory and misleading in some extent.

We appreciate the important point Reviewer #1 made that macrophages polarized toward a pro-inflammatory phenotype such as those stimulated with IFNγ and LPS (M1 macrophages) prioritize metabolic pathways that enhance glycolytic flux, resulting in increased release of protons and lactate as waste products from the glycolysis pathway. In this way, polarization of macrophages toward the pro-inflammatory phenotype can lead to acidification of the media, which may influence our observations given that we are studying the effect of extracellular pH on rhythms in macrophages. To address this point, we have performed additional experiments in which we measured pH of the media to capture changes in media pH that occur during the time in which we observe changes in rhythms of pro-inflammatory macrophages.

In line with the documented enhanced glycolytic activity of pro-inflammatory macrophages, the media of pro-inflammatory macrophages is acidified over time, in contrast to media of unstimulated or pro-resolution macrophages. Notably, while pH decreased over time in the pro-inflammatory group, the pH differential between the pH7.4, pH6.8, and pH6.5 sample groups was maintained over the period in which we observe and measure changes in circadian rhythms of pro-inflammatory macrophages. Additionally, media that began at pH 7.4 was acidified only to pH 7 by day 2, above the acidic pH of 6.8 or 6.5. As a result, there remained a difference in pH between the two groups (pH 7.4 and pH 6.5) out to 2 days consistent with the changes in rhythms that we observe between these two groups. This indicates that the difference in circadian rhythms observed in pro-inflammatory macrophages cultured at pH 7.4 compared to pH 6.5 were indeed due to the difference in extracellular pH between the two conditions. We have incorporated these data, shown below, into Supplementary Figure 4 and added the following discussion of these data to the Results section:

"In line with their documented enhanced glycolytic capacity, pro-inflammatory macrophages acidified the media over time (Supplementary Figure 4C). Notably, while pH of the media the pro-inflammatory macrophages were cultured in decreased over time pH, the pH differential between the pH 7.4, pH 6.8, and pH 6.5 samples groups of pro-inflammatory macrophages was maintained out to 2 days, consistent with the changes in rhythms that we observe and measure between these groups."

The article examines the role of circadian rhythms in tumor-associated macrophages, yet it lacks sufficient compelling data to support this assertion. Two figures, Figure 7 and Figure 9, are presented in relation to cancer. In Figure 7, gene expression analysis of Arg1 (an M2 marker) and Crem (a potential circadian clock gene) is conducted in wild-type macrophages, BMAL1-knockout macrophages with dysregulated circadian rhythms, and using publicly available data on tumor-associated macrophages from a study referenced as 83. However, it is noted that this referenced study is actually a review article by Geeraerts et al. (2017) titled "Macrophage Metabolism as Therapeutic Target for Cancer, Atherosclerosis, and Obesity" published in Frontiers in Immunology. This raises concerns about the reliability of the results. Furthermore, comparing peritoneal macrophages from healthy mice with macrophages isolated from lung tumors is deemed inaccurate. It is suggested that lung macrophages from healthy mice and those from mice with lung tumors should be isolated separately for a more appropriate comparison. Consequently, Figure 7B is further questioned regarding how the authors could compare genes from the circadian rhythm pathway between these non-identical groups. As a result, the conclusion drawn from these data, suggesting that tumor-associated macrophages exhibit a gene expression pattern similar to BMAL1-KO macrophages, is deemed incorrect, affecting the interpretation of the data presented in Figure 8.

We thank Reviewer #1 for pointing out our error in the reference provided as the source of the TAM data used for CCD in Figure 7. While we took care to provide the GEO ID for the data set (GSE188549) in the Methods section, we mistakenly cited Geeraerts (2017) Front Immunol when we should have cited Geeraerts (2021) Cell Rep. We have corrected this citation error in the main text.

We also appreciate Reviewer #1's concern that we are comparing circadian gene expression of peritoneal macrophages to tumor-associated macrophages derived from LLC tumors, which are grown ectopically in the flank for the experiment from which the data set was produced. To ensure an accurate comparison of gene expression, we downloaded the raw FASTQ files from each dataset and processed them in identical pipelines. Our main comparison between these cell types is Clock Correlation Distance (CCD), which compares the pattern of co-expression of circadian genes (Shilts et al PeerJ 2018). CCD was built from multiple mouse and human tissues to be a "universal" tool to compare circadian rhythms, and designed to compare between different tissues and cell types. Each sample is compared to a reference control built from these multiple tissues. To better convey this concept to readers to give confidence the suitability of CCD for comparing data sets across different tissues, we have added the reference control to Figure 7 (now Figure 6B), We have also expanded our analysis to include bone marrow-derived macrophages, to further demonstrate that the organization of clock gene co-expression is not specific to peritoneal macrophages; we have added this data to Figure 7 (now Figure 6C,D). Finally, we have included an abbreviated explanation of the points made above in the results section.

Due to the universal nature of the CCD tool, we disagree with Reviewer #1's assertion that "the conclusion drawn from these data, suggesting that tumor-associated macrophages exhibit a gene expression pattern similar to BMAL1-KO macrophages, is deemed incorrect". Indeed, this finding mirrors findings in the original CCD paper, which showed that tumor tissues universally exhibit a disordered molecular clock as compared to normal tissue. Notably, the original CCD paper also compared across cell and tumor types.

As an additional note to the review, we would like to clarify that nowhere in the manuscript do we propose that Crem is a potential circadian clock gene. We are clear throughout the manuscript that we are using Crem as a previously established biomarker for acidic pH-sensing in macrophages. Please see below for the modified Figure and text.

"To understand the status of the circadian clock in TAMs, we performed clock correlation distance (CCD) analysis. This analysis has previously been used to assess functionality of the circadian clock in whole tumor and in normal tissue[102]. As the circadian clock is comprised of a series of transcription/translation feedback loops, gene expression is highly organized in a functional, intact clock, with core clock genes existing in levels relative to each other irrespective of the time of day. In a synchronized population of cells, this ordered relationship is maintained at the population level, which can be visualized in a heatmap. CCD is designed to compare circadian clock gene co-expression patterns between different tissues and cell types. To accomplish this, CCD was built using datasets from multiple different healthy tissues from mouse and human to be a universal tool to compare circadian rhythms. Each sample is compared to a reference control built from these multiple tissues (Figure 6B)[102]. To validate the use of this analysis for assessing circadian disorder in macrophages, we performed CCD analysis using publicly available RNA-sequencing data from bone marrow-derived macrophages and wild type peritoneal macrophages, as a healthy control for functional rhythms in a synchronized cell population, and BMAL1 KO peritoneal macrophages, as a positive control for circadian disorder[44]."

And in the Discussion:

"Interestingly, analysis of TAMs by clock correlation distance (CCD) presents evidence that rhythms are disordered in bulk TAMs compared to other macrophage populations (Figure 6). CCD is one of the most practical tools currently available to assess circadian rhythms due to its ability to assess rhythms independent of time of day and without the need for a circadian time series, which is often not available in publicly available data from mice and humans[102]."

If the authors aim to draw a clear conclusion regarding the circadian rhythms of tumor-associated macrophages (TAMs), they may need to analyze single-sorted macrophages from tumors and corresponding healthy tissues. Such data are publicly available (of course not in #83)

We agree with Reviewer #1 that while our interpretation of the data is that there may be heterogeneity in circadian rhythms of tumor-associated macrophages, we cannot prove this without assessing circadian rhythms at the single cell level. While single-cell RNA-sequencing data of freshly isolated tumor associated macrophages of sufficient read depth for circadian gene expression analysis has historically been unavailable, fortunately a dataset was released recently (May 2024) which we were able to use to address this point. We have analyzed publicly available single-cell RNAseq data of tumor-associated macrophages (GSE260641, Wang 2024 Cell) to determine whether there are differences in expression of circadian clock genes between different TAM populations. We have added these data as a new Figure 9. Please see the figure and updated text below.

"Tumor-associated macrophages exhibit heterogeneity in circadian clock gene expression.

__ Our findings suggested that heterogeneity of the circadian clock may lead to disorder in bulk macrophage populations, but did not reveal if specific gene expression changes exist in tumor-associated macrophages at the single-cell level. To determine whether heterogeneity exists within the expression of circadian clock genes of the tumor-associated macrophage population, we analyzed publicly available single-cell RNA sequencing data of macrophages isolated from B16-F10 tumors[107]. To capture the heterogeneity of macrophage subsets within the TAM population, we performed unbiased clustering (Figure 9A). We then performed differential gene expression to determine if circadian clock genes were differentially expressed within the TAM subpopulations. The circadian clock genes Bhlhe40 (DEC1), Bhlhe41 (DEC2), Nfil3 (E4BP4), Rora (RORα), Dbp (DBP), and Nr1d2 (REV-ERBβ) were significantly (adj.p We next sought to determine whether differences in circadian clock gene expression between TAM subpopulations were associated with exposure to acidic pH in the TME. To this end, we first assessed Crem expression in the TAM subpopulations that were identified by unbiased clustering. Crem expression was significantly higher in TAM clusters 4, 5, and 6 compared to TAM clusters 1-3 and 7-9 (Figure 9C). Clusters were subset based on Crem expression into Crem high (clusters 4-6) and Crem low (clusters 1-3, 7-9) (Figure 9D), and differential gene expression analysis was performed. The circadian clock genes Nfil3, Rora, Bhlhe40, and Cry1 (CRY1) were significantly (adj.p __And in the Discussion:

"Supporting the notion that population-level disorder may exist in TAMs, we used scRNA-sequencing data and found evidence of heterogeneity between the expression of circadian clock genes in different TAM subpopulations (Figure 9A, B). Phenotypic heterogeneity of TAMs in various types of cancer has previously been shown[20, 21, 125, 126], and we have identified distinct TAM subpopulations by unbiased clustering (Figure 9A). Within those TAM subpopulations, we identified differential expression of circadian clock genes encoding transcription factors that bind to different consensus sequences: DEC1 and DEC2 bind to E-boxes, NFIL3 and DBP binds to D-boxes, and RORα and REV-ERBβ binds to retinoic acid-related orphan receptor elements (ROREs)[127, 128]. While little is known about regulation of macrophages by E-box and D-box elements beyond the circadian clock, aspects of macrophage function have been shown to be subject to transcriptional regulation through ROREs[129, 130]. Thus, we speculate that variations in these transcription factors may exert influence on expression of genes to drive diversity between TAM subpopulations. Differential expression of circadian clock genes between TAM subpopulations was also associated with Crem expression (Figure 9C-E), suggesting that exposure of TAMs to acidic pH within the TME can alter the circadian clock. However, there remained significant variation in expression of circadian clock genes within the Crem high and Crem low groups (Figure 9B), suggesting that acidic pH is not the only factor in the TME that can alter the circadian clock. Together, these data implicate the TME in driving heterogeneity in TAM circadian rhythms just as it drives heterogeneity in TAM phenotype.

Interestingly, in contrast to our observations of circadian disorder in TAMs isolated from LLC tumors (Figure 6), rhythmicity in expression of circadian genes was observed in bulk TAMs isolated from B16 tumors[107]. This suggests that circadian rhythms of TAMs are maintained differently in different types of cancer. Notably, both of these observations were at the population level. Upon separation of the B16 TAM population into subsets by unbiased clustering of single-cell RNA sequencing data, we measured differences in expression of circadian clock genes between TAM subpopulations (Figure 9A,B). This suggests that even within a rhythmic TAM population, there is heterogeneity in the circadian clock of TAM subpopulations."

Additionally, it is widely acknowledged that human and mouse macrophages exhibit distinct gene expression profiles, both in vitro and in vivo. While assuming that genes involved in circadian rhythms are conserved across species, the authors could consider extending their funding to include analyses of single-sorted macrophages from cancer patients, such as those with lung cancer or pancreatic ductal adenocarcinoma (PDAC). These experiments would provide relevant insights into TAM biology.

We agree that with Reviewer #1 that ultimately, being able to relate findings in mice to humans is critical. It is important to assess if circadian disorder is observed in TAMs in human cancers as it is for LLC tumor-derived macrophages in mice. To address this point, we have performed CCD using a human data set (GSE116946; Garrido 2020 J Immunother Cancer) suitable for use with CCD (wherein macrophages were isolated from bulk tumor in humans, with a high enough samples size, and not cultured prior to sequencing). We have added these data as a new Figure 7, shown below. Please see the added data and updated text below.

"We next assessed the status of the circadian clock in human TAMs from NSCLC patients. We performed CCD with publicly available RNA-seq data of tumor-adjacent macrophages and tumor-associated macrophages from NSCLC patients, using alveolar macrophages from healthy donors as a control[104, 105]. To assess the contribution of the acidic TME to circadian disorder, we subset TAM NSCLC patient samples into groups (Crem high TAMs and Crem low TAMs) based on median Crem expression. Notably, in macrophages from human NSCLC there was a trend toward disorder in Crem low but not Crem high TAM samples (Figure 7A,B). Additionally, the co-variance among core clock genes observed in alveolar macrophages from healthy donors was absent within Crem low and Crem high TAM samples (Figure 7C). In all, these data indicate that there is population-level disorder in the circadian rhythms of tumor-associated macrophages in humans and mice, suggesting that circadian rhythms are indeed altered in macrophages within the TME."

And in the Discussion:

"Indeed, we observed differences in the circadian clock of Crem low human TAM samples compared to Crem high human TAM samples, suggesting that acidic pH influences circadian disorder in TAMs (Figure 7). Interestingly, Crem low TAM samples exhibited a trend toward disorder while Crem high TAM samples did not. This is of particular interest, as we have observed that acidic pH can enhance circadian rhythms in macrophages, raising the question of whether acidic pH promotes or protects against circadian disorder."

Minor comments: 1. Figure 2C needs clarification. It's unclear why pro-inflammatory macrophages treated with lactic acid would have a shorter amplitude and period, while acidic pH would increase amplitude and period in M2 macrophages.

We thank Reviewer #1 for this important observation. Based on the comment, it is our understanding that the Reviewer is referring to the data in Figure 2 (low pH) compared to Figure 4 (lactate). We also find it very interesting that lactate alters rhythms in a manner distinct from the way in which acidic pH alters rhythms. Reviewer 3 asked for clarification on how lactate affected circadian gene expression in pH 7.4 or 6.5. We have added these data as Figure 4C (data and text below). It is notable that lactate opposing effects on circadian gene expression in pH 6.5, enhancing the effects of low pH in some cases (Nr1d1) while blunting them in other cases (Cry1). This is mentioned in the text.

"Lactate was also observed to alter expression of the circadian clock genes Per2, Cry1, and Nr1d1 over time in BMDMs cultured at pH 6.5, while having more subtle effects at pH 7.4 (Figure 4C). Notably, lactate blunted the effect of pH 6.5 on Cry1 expression, while enhancing the effect of low pH on Nr1d1 expression."

Why these two stimuli alter rhythms differently remains an open question that is discussed in the Discussion section and is prime to be a topic of future investigation. We have added to the Discussion section potential reasons why these conditions may alter rhythms differently, such as the different pathways downstream of sensing these two different conditions. Please see the updated text, below.

"Although lactate polarizes macrophages toward a pro-resolution phenotype similar to acidic pH[30, 93], exposure to lactate had different effects on circadian rhythms - and in some cases, circadian clock gene expression - than exposure to acidic pH (Figure 4). Sensing of lactate occurs through different pathways than acid-sensing, which may contribute to the different ways in which these two stimuli modulate circadian rhythms of macrophages[111]. One previously published finding that may offer mechanistic insight into how phenotype can influence circadian rhythms is the suppression of Bmal1 by LPS-inducible miR-155[54]. It has also been observed that RORα-mediated activation of Bmal1 transcription is enhanced by PPARγ co-activation[112]. In macrophages, PPARγ expression is induced upon stimulation with IL-4 and plays a key role in alternative activation of macrophages, promoting a pro-resolution macrophage phenotype, and supporting resolution of inflammation[113-115]. Such observations prompt the question of whether there are yet-unidentified factors induced downstream of various polarizing stimuli that can modulate expression of circadian genes at the transcriptional and protein levels. Further work is required to understand the interplay between macrophage phenotype and circadian rhythms."

The scale in Figure 2C should be equal for all conditions (e.g., -200).

We appreciate Reviewer #1's preference for the axes to be scaled similarly to enable cross-comparison between graphs. However, due to the different amplitude of pro-inflammatory macrophages compared to the others, we feel that making all axes the same will make it hard to see the rhythms of pro-inflammatory macrophages, hindering the reader's ability to observe the data. Thus, we have put the matched-axis plots, shown below, in Supplementary Figure 4A.

Absolute values of amplitude, damping, and period differ between Figure 1 and Figure 2A, B, C. The authors should explain these discrepancies.

As with many experimental approaches, there is slight variation in absolute values between independent experiments, which Reviewer #1 correctly notes. However, while the absolute values vary slightly, the relationship between the values in each of these conditions remains the same across the panels mentioned by Reviewer #1.

The authors should consider modulating the acidic environment of macrophages in settings more representative of cancer. For example, by adding conditioned media from tumor cells with pronounced glycolysis.

We appreciate Reviewer #1's desire to more closely mimic the tumor microenvironment. To address Reviewer #1's point, we cultured macrophages in RPMI or cancer cell (KCKO) supernatant at pH 6.5 or pH-adjusted to pH 7.4 and assessed rhythms by measuring rhythmic activity of Per2-Luc with LumiCycle analysis. We then compared changes in rhythms between macrophages cultured normal media to cancer cell supernatant in pH-matched conditions to assess how cancer cell-conditioned media may influence circadian rhythms of macrophages, and the contribution of acidic pH. We have added these data, shown below, as a new Supplementary Figure 5, and included a discussion of these data in the manuscript. Please see the new Figure and updated text below.

"Cancer cell supernatant alters circadian rhythms in macrophages in a manner partially reversed by neutralization of pH.

We have observed that polarizing stimuli, acidic pH, and lactate can alter circadian rhythms. However, the tumor microenvironment is complex. Cancer cells secrete a variety of factors and deplete nutrients in the environment. To model this, we cultured BMDMs in RPMI or supernatant collected from KCKO cells, which are a murine model of pancreatic ductal adenocarcinoma (PDAC)[94, 95], at pH 6.5 or neutralized to pH 7.4 (Supplementary Figure 5). Circadian rhythms of BMDMs cultured in cancer cell supernatant at pH 7.4 or pH 6.5 exhibited increased amplitude and lengthened period compared to RPMI control at pH 7.4 or 6.5, respectively, indicating that cancer cell supernatant contains factors that can alter circadian rhythms of BMDMs. Notably, BMDMs cultured in cancer cell supernatant at pH 6.5 had increased amplitude and shortened period compared to BMDMs cultured in cancer cell-conditioned media at pH7.4, indicating that pH-driven changes in rhythms were maintained in BMDMs cultured in cancer cell supernatant. When the pH of cancer cell supernatant was neutralized to pH7.4, the increased amplitude was decreased, and the shortened period was lengthened, indicating that neutralizing acidic pH partially reverses the changes in rhythms observed in macrophages cultured in cancer cell supernatant at pH 6.5. These data further support our observations that acidic pH can alter circadian rhythms of macrophages both alone and in combination with various factors in the TME."

And, in the Discussion:

"We have shown that various stimuli can alter rhythms of macrophages in a complex and contributing manner, including polarizing stimuli, acidic pH, and lactate. TGFβ is produced by a variety of cells within the TME, and was recently identified as a signal that can modulate circadian rhythms[123, 124]. Additionally, when we exposed macrophages to cancer cell-conditioned media, rhythms were modulated in a manner distinct from acidic pH or lactate, with these changes in rhythms partially reversed by neutralization of the cancer cell-conditioned media pH (Supplementary Figure 5). It is conceivable that, in addition to acidic pH, other stimuli in the TME are influencing circadian rhythms to drive population-level disorder that we observed by CCD."

Arg1 alone is not sufficient as an M2 polarization marker. The authors should include additional markers.

We thank Reviewer #1 for bringing up this critical point in experimental rigor. While Arg1 is a commonly-used marker for M2 polarization, Reviewer #1 points out that polarization of macrophages is typically assessed by a full panel of markers characteristic of the M2 state. To address this point, we have expanded our panel to include several other markers of M2 polarization in mice such as Retnla, Ym1, MGL1, and CD206. In response to Reviewer 2's major point 2 and Reviewer 3's major point 4 below, we have also expanded our panel of markers used to assess the M1 polarization state with Tnfa, Il1b. and Il6. We have added these data, shown below, to Supplementary Figure 1 and updated the text appropriately. Please see the new Figure and updated text below.

"Consistent with previous studies, we found that genes associated with anti-inflammatory and pro-resolution programming characteristic of IL-4 and IL-13-stimulated macrophages such as Arg1, Retnla, Chil3 (Ym1), Clec10a (MGL1), and Mrc1 (CD206) were induced in IL-4 and IL-13-stimulated macrophages, but not IFNγ and LPS-stimulated macrophages. In contrast, genes associated with pro-inflammatory activity characteristic of IFNγ and LPS-stimulated macrophages such as Nos2 (iNOS), Tnfa, Il1b, and Il6 were induced in IFNγ and LPS-stimulated macrophages, but not IL-4 and IL-13-stimulated macrophages (Supplementary Figure 1)[28, 30, 65, 71, 74, 75]. This indicates that macrophages stimulated with IL-4 and IL-13 were polarized toward a pro-resolution phenotype, while macrophages stimulated with IFNγ and LPS were polarized toward a pro-inflammatory phenotype."

__ Significance__

While the manuscript provides valuable insights and has obvious novelty, it requires a significant revision

We thank Reviewer #1 for their deep read of our manuscript, and their helpful feedback and suggestions. As shown by the comments above, we are confident we have fully addressed each of the points that were made to result in a much-improved revised manuscript.

__ Reviewer #2 __

Evidence, reproducibility and clarity

Knudsen-Clark et al. showed that the circadian rhythm of bone marrow-derived macrophages (BMDM) can be affected by polarization stimuli, pH of the microenvironment, and by the presence of sodium-lactate. Mechanistically, the acidic pH of cell microenvironment is partly regulated by intracellular cAMP-mediated signaling events in BMDM. The authors also showed that the circadian clock of peritoneal macrophages is also modified by the pH of the cell microenvironment. Using publicly available data, the authors showed that the circadian rhythm of tumor-associated macrophages is similar to that of Bmal1-KO peritoneal macrophages. In a murine model of pancreatic cancer, the authors showed that the tumor growth is accelerated in C57BL/6 mice co-injected with cancer cells and Bmal1-KO BMDM as compared to mice co-injected with cancer cell and wild type BMDM.

We thank Reviewer #2 for their insightful and helpful comments and feedback. Their Review guided key clarifying experiments and additions to the Discussion and Methods. To summarize, we added new data to Supplementary Figure 1 to characterize distinct gene expression in our different polarized macrophage populations, showed in Supplementary Figure 2 that serum shock independently induces cAMP and Icer, discussed the limitations of the artificial polarization models more clearly, and updated our Methods to better explain how macrophages were isolated from the peritoneum. We also quantified multiple immunoblots of pCREB, provided clarity in the Methods and Reviewer-only data on how our protein-extraction protocol isolates nuclear protein, better introduced the BMAL1-KO mouse model, and showed in Supplementary Figure 6 that low pH can induce oscillations in the absence of a serum shock.

Major points of criticism: 1. Nine main figures include different experimental models on a non-systematic manner in the manuscript, and only literature-based correlation is used to link the results each other. The authors used in vitro BMDM and peritoneal cell-based model systems to study the effects of IL4+IL13, IFNg+LPS, low pH, sodium-lactate, adenylate cyclase inhibitors on the circadian clock of macrophages. The link between these microenvironment conditions of the cells is still correlative with the tumor microenvironment: publicly available data were used to correlate the increased expression level of cAMP-activated signaling events with the presence of acidic pH of tumor microenvironment. Notably, the cell signaling messenger molecule cAMP is produced by not only low extracellular pH by activated GPCRs, but also starvation of the cell. The starvation is also relevant to this study, since the BMDM used in the in vitro culture system were starving for 24 hours before the measurement of Per2-Luc expression to monitor circadian rhythm.

        We agree with the important point that Reviewer #2 makes that our synchronization protocol of serum starvation followed by serum shock can impact the cAMP signaling pathway. Indeed, it has previously been shown that serum shock induces phosphorylation of CREM in rat fibroblasts, which is indicative of signaling through the cAMP pathway. To address this point, we have added a schematic of our synchronization protocol to Supplementary Figure 2B for additional clarity. We have also performed additional experiments to test whether cAMP signaling is induced in macrophages by our synchronization protocol. For this, we assessed downstream targets of the cAMP signaling pathway, Icer and pCREB, after serum starvation but before serum shock, and at several time points post-treatment with serum shock (Supplementary Figures 2D,E). We observed that Icer and phosphorylation of Creb are induced rapidly in macrophages upon exposure to serum shock, as early as 10 minutes for pCREB and 1 hour post-exposure for Icer. Notably, this signaling is transient and rapidly returns to baseline, with pCREB levels fully returned to baseline by 2 hours post-treatment, at which time media is replaced and the experiment begins (CT 0). These data, shown below, have been added to Supplementary Figure 2 and a discussion of these data has been added to the manuscript - please see the modified text below.

"The synchronization protocol we use to study circadian rhythms in BMDMs involves a 24-hour period of serum starvation followed by 2 hours of serum shock. It has previously been shown that serum shock can induce signaling through the cAMP pathway in rat fibroblasts[98]. To determine whether the synchronization protocol impacts cAMP signaling in macrophages, we harvested macrophages before and after serum shock. We then assessed Icer expression and phosphorylation of cyclic AMP-response element binding protein (CREB), which occur downstream of cAMP and have been used as readouts to assess induction of cAMP signaling in macrophages[29, 96, 100]. Serum shock of macrophages following serum starvation led to rapid phosphorylation of CREB and Icer expression that quickly returned to baseline (Supplementary Figure 2D,E). This indicates that serum starvation followed by serum shock in the synchronization protocol we use to study circadian rhythms in BMDMs induces transient signaling through the cAMP signaling pathway. "

The definition of pre-resolution macrophages (MF) used across the manuscript could be argued. The authors defined BMDM polarized with IL-4 and IL-13 as pre-resolution MF. Resolution is followed by inflammation, but the IL-4 secretion does not occur in every inflammatory setting. Moreover, IL-4 and IL-13 are secreted during specific tissue environment and immunological settings involving type 2 inflammation or during germinal center reactions of the lymph nodes. • What are the characteristics of pre-resolution macrophages (MF)? The authors indicated that IL-4 and IL-13 cytokines were used to model the pre-resolution macrophages. In which pathological context are these cytokines produced and induce pre-resolution macrophages? IL-4 polarized BMDM can also produce pro-inflammatory protein and lipid mediators as compared to LPS-stimulated BMDM, and IL-4 polarized BMDM still have potent capacity to recruit immune cells and to establish type 2 inflammation.

• The authors showed Arg1 and Vegfa qPCR data from BMDM only. Based on the literature, these MFs are anti-inflammatory cells particularly. Resolution-related MFs followed by acute inflammation are a specific subset of MFs, and the phenotype of pre-resolution MF should be described, referred, and measured specifically.

We thank Reviewer #2 for bringing up this important point that clarity is required in describing our in vitro macrophage models. We chose the most commonly used models of in vitro macrophage polarization in the tumor immunology field, M2 (IL-4+IL-13) and M1 (IFNγ+LPS). These polarization conditions have been used for over two decades in the field, and have been well-characterized to drive a pro-inflammatory (for M1) and pro-resolution or anti-inflammatory (for M2) macrophage phenotype (Murray 2017 Annu Rev Phys). Each of these cell states have similarities in phenotype to pro-inflammatory and pro-resolution (pro-tumorigenic) macrophages found in tumors. In fact, in the literature, pro-inflammatory and pro-resolution TAMs will frequently be categorized as "M1" or "M2", respectively, even though this is a gross oversimplification (Ding 2019 J Immunol, Garrido-Martin 2020 J Immunother Cancer).

As Reviewer #2 points out, IL-4 and IL-13 play a role in inflammatory settings, mediating protective responses to parasites and pathological responses to allergens. Importantly, IL-4 and IL-13 are also key regulators and effectors of resolution and wound repair (Allen 2023 Annu Rev Immunol). In line with this, M2 macrophages show many of the characteristics of pro-resolution programming in their gene expression profile, expressing genes associated with wound healing (ex. Vegf) and immunoregulation (ex. Arg1) (Mantovani 2013 J Pathol). These cells have frequently been used as a model for studying TAMs in vitro, due to the similarity in pro-resolution programming that is dysregulated/hijacked in TAMs (Biswas 2006 Blood). M2 macrophages have also been referred to as anti-inflammatory, and this is in line with their role in the type 2 response driven by IL-4 and IL-13, as this is primarily a response induced by allergy or parasites where tissue damage drives an anti-inflammatory and pro-resolution phenotype in macrophages (Pesce 2009 Plos Pathogens and Allen 2023 Annu Rev Immunol).

We do not assert that these in vitro models recapitulate the macrophage polarization cycle that Reviewer #2 astutely describes, and indeed, stimuli polarizing macrophages in tumor are much more diverse and complex (Laviron 2022 Cell Rep). We also fully agree with Reviewer #2 that, while IL4 and IL13 may exist in the tumor and be secreted by Th2 CD4 T cells (see Shiao 2015 Cancer Immunol Res), there may be multiple reasons why macrophages may be polarized to a pro-resolution, M2-like state in a tumor (in fact, exposure to low pH and lactate each independently do this, as we show in Supplementary Figure 2 and Figure 4, and was previously shown in Jiang 2021 J Immunol and Colegio 2014 Nature). Nonetheless, using the well-described M1 and M2 in vitro models allows our findings to be directly comparable to the vast literature that also uses these models, and to understand how distinct polarization states respond to low pH.

We fully agree with Reviewer #2 that these cells must be defined more clearly in the text. We have taken care to discuss the limitations of using in vitro polarization models to study macrophages in our Limitations of the Study section. To better address Reviewer #2's concern, we have more thoroughly introduced the M2 macrophages as a model, and are clear that that these are type 2-driven macrophages that share characteristics of pro-resolution macrophages. We have also added additional citations to the manuscript, including those highlighted above in our response. Finally, we have expanded our panel to better characterize the IL-4/IL-13 stimulated macrophages using more markers that have been characterized in the literature, in line with both Reviewer #2's comments and that of Reviewer #1 and Reviewer #3. Please see the updated data and text, below.

"As macrophages are a phenotypically heterogeneous population in the TME, we first sought to understand whether diversity in macrophage phenotype could translate to diversity in circadian rhythms of macrophages. To this end, we used two well-established in vitro polarization models to study distinct macrophage phenotypes[5, 60-63]. For a model of pro-inflammatory macrophages, we stimulated macrophages with IFNγ (interferon γ) and LPS (lipopolysaccharide) to elicit a pro-inflammatory phenotype[60, 64]. These macrophages are often referred to as 'M1' and are broadly viewed as anti-tumorigenic, and we will refer to them throughout this paper as pro-inflammatory macrophages[65, 66]. For a model at the opposite end of the phenotypic spectrum, we stimulated macrophages with IL-4 and IL-13[60, 67]. While these type 2 stimuli play a role in the response to parasites and allergy, they are also major drivers of wound healing; in line with this, IL-4 and IL-13-stimulated macrophages have been well-characterized to adopt gene expression profiles associated with wound-healing and anti-inflammatory macrophage phenotypes[68-71]. As such, these macrophages are often used as a model to study pro-tumorigenic macrophages in vitro and are often referred to as 'M2' macrophages; throughout this paper, we will refer to IL-4 and IL-13-stimulated macrophages as pro-resolution macrophages[66, 72, 73]. Consistent with previous studies, we found that genes associated with anti-inflammatory and pro-resolution programming characteristic of IL-4 and IL-13-stimulated macrophages such as Arg1, Retnla, Chil3 (Ym1), Clec10a (MGL1), and Mrc1 (CD206) were induced in IL-4 and IL-13-stimulated macrophages, but not IFNγ and LPS-stimulated macrophages. In contrast, genes associated with pro-inflammatory activity characteristic of IFNγ and LPS-stimulated macrophages such as Nos2 (iNOS), Tnfa, Il1b, and Il6 were induced in IFNγ and LPS-stimulated macrophages, but not IL-4 and IL-13-stimulated macrophages (Supplementary Figure 1)[28, 30, 65, 71, 74, 75]. This indicates that macrophages stimulated with IL-4 and IL-13 were polarized toward a pro-resolution phenotype, while macrophages stimulated with IFNγ and LPS were polarized toward a pro-inflammatory phenotype.

In the Limitations of the Study section, we now write the following:

"Our observations of rhythms in macrophages of different phenotypes are limited by in vitro polarization models. It is important to note that while our data suggest that pro-inflammatory macrophages have suppressed rhythms and increased rate of desynchrony, it remains unclear the extent to which these findings apply to the range of pro-inflammatory macrophages found in vivo. We use IFNγ and LPS co-treatment in vitro to model a pro-inflammatory macrophage phenotype that is commonly referred to as 'M1', but under inflammatory conditions in vivo, macrophages are exposed to a variety of stimuli that result in a spectrum of phenotypes, each highly context-dependent. The same is true for for 'M2'; different tissue microenvironment are different and pro-resolution macrophages exist in a spectrum."

The authors used IFNg and LPS, or IL-4 and IL-13 and co-treatments to polarize BMDM in to type 1 (referred as pro-inflammatory MF) and type 2 (referred as pre-resolution MF) activation state. The comparison between these BMDM populations has limitations, since LPS induces a potent inflammatory response in MF. The single treatment with MF-polarizing cytokines enable a more relevant comparison to study the circadian clock in classically and alternatively activated MF.

We thank Reviewer #2 for bringing up this important point to provide additional clarity on our polarization conditions. The use of IFNγ and LPS to polarize macrophages toward a pro-inflammatory, M1 phenotype, and the use of IL-4 an IL-13 to polarize macrophages toward a pro-resolution, M2 phenotype have been commonly used for over two decades, and thus are well-characterized in the literature (please see Murray 2017 Annu Rev Phys for an extensive review on the history of these polarization models, as well as Hörhold 2020 PLOS Computational Biology, Binger 2015 JCI, McWhorter 2013 PNAS, Ying 2013 J Vis Exp for more recent studies using these models). The use of LPS alone or in combination with IFNγ, and IL-13 along with IL-4, was introduced in 1998 (Munder 1998 J Immunol). This approach was originally designed to mimic what could happen when macrophages were exposed to CD4+ Th1 cells, which produce IFNγ, or Th2 cells, which produce IL-4 and IL-13 (Munder 1998 J Immunol, Murray 2017 Annu Rev Phys). As Reviewer #2 points out, these stimuli induce potent responses, driving macrophages to adopt pro-inflammatory or pro-resolution/anti-inflammatory phenotypes that are two extremes at opposite ends of the spectrum of macrophage phenotypes (Mosser 2008 Nat Rev Immunol). Since our goal was to study rhythms of distinct macrophage phenotypes in vitro, and how TME-associated conditions such as acidic pH and lactate affect their rhythms, these cell states were appropriate for our questions. Thus, the polarization models used in this paper allowed us to achieve this goal. We include a section in the Discussion on the limitations of in vitro polarization models.

"A critical question in understanding the role of circadian rhythms in macrophage biology is determining how different polarization states of macrophages affect their internal circadian rhythms. This is especially important considering that tumor-associated macrophages are a highly heterogeneous population. Our data indicate that compared to unstimulated macrophages, rhythms are enhanced in pro-resolution macrophages, characterized by increased amplitude and improved ability to maintain synchrony; in contrast, rhythms are suppressed in pro-inflammatory macrophages, characterized by decreased amplitude and impaired ability to maintain synchrony (Figure 1). These agree with previously published work showing that polarizing stimuli alone and in combination with each other can alter rhythms differently in macrophages[80, 81]. In a tumor, macrophages exist along a continuum of polarization states and phenotypes[18-21, 24]. Thus, while our characterizations of rhythms in in vitro-polarized macrophages provide a foundation for understanding how phenotype affects circadian rhythms of macrophages, further experiments will be needed to assess macrophages across the full spectrum of phenotypes. Indeed, alteration of rhythms may be just as highly variable and context-dependent as phenotype itself."

There are missing links between the results of showing the circadian rhythm of polarized BMDM, sodium-lactate treated BMDM, and tumor growth. Specifically, do the used pancreatic ductal adenocarcinoma cells produce IL-4 and sodium-lactate? In the LLC-based experimental in silico analysis of tumors, the LLC do not produce IL-4.

Reviewer #2 raises important points about the source of lactate and IL-4 in tumors as relevance for our investigation of how these factors can alter rhythms in macrophages. Tumor-infiltrating Th2 CD4 T cells are potential sources of IL-4 and IL-13 in the tumor (see Shiao 2015 Cancer Immunol Res). Various cells in the tumor can produce lactate. We discuss this in both the Introduction and the Results: poor vascularization of tumors results in hypoxia areas, where cells are pushed toward glycolysis to survive and thus secrete increased glycolytic waste products such as protons and lactate. As lactate is lactic acid, ionized it is sodium l-lactate.

How can the circadian rhythm affect the function of BMDM? The Authors should provide evidence that circadian rhythm affects the function of polarized MF.

We agree with Reviewer #2 that the next step is to determine how altered rhythms influence function of macrophages. This will be the topic of future work, but is outside the scope of this paper. Our contribution with this paper is providing the first evidence that rhythms are altered in the TME and the TME-associated conditions can alter rhythms in macrophages. We have added what is currently known about how circadian rhythms influence macrophages function to the discussion section to facilitate a conversation about this important future direction. Please see the updated text below.

"Considering our observations that conditions associated with the TME can alter circadian rhythms in macrophages, it becomes increasingly important to understand the relevance of macrophage rhythms to their function in tumors. It has been shown that acidic pH and lactate can each drive functional polarization of macrophages toward a phenotype that promotes tumor growth, with acidic pH modulating phagocytosis and suppressing inflammatory cytokine secretion and cytotoxicity[28, 30, 93]. However, how the changes in circadian rhythms of macrophages driven by these conditions contributes to their altered function remains unknown. Current evidence suggests that circadian rhythms confer a time-of-day-dependency on macrophage function by gating the macrophage response to inflammatory stimuli based on time-of-day. As such, responses to inflammatory stimuli such as LPS or bacteria are heightened during the active phase while the inflammatory response is suppressed during the inactive phase. An important future direction will be to determine how changes in circadian rhythms of macrophages, such as those observed under acidic pH or high lactate, influences the circadian gating of their function."

In Figure 3, the authors show data from peritoneal cells. The isolated peritoneal cells are not pure macrophage populations. Based on the referred article in the manuscript, the peritoneal cavity contains more then 50% of lymphocytes, and the myeloid compartment contains 80% macrophages.

Reviewer #2 raises important concerns about the purity of the peritoneal population used in our experiments. We enrich for peritoneal macrophages from the peritoneal exudate cells by removing non-adherent cells in culture. This is described in our Methods section and is a method of isolation that is commonly used in the field, as lymphocytes are non-adherent. In addition to the source cited in the paper within our Methods section (Goncalves 2015 Curr Prot Immunol), please see Layoun 2015 J Vis Exp, de Jesus 2022 STAR Protocols, and Harvard HLA Lab protocol - macrophages enriched in this manner have been shown to be over 90% pure. We have modified our Methods section to make this clear, and added the additional references in this response to this section of our Methods. Please see the modified text below.

"Peritoneal exudate cells were harvested from mice as previously published[137]. To isolate peritoneal macrophages, peritoneal exudate cells were seeded at 1.2*106 cells/mL in RPMI/10% HI FBS supplemented with 100U/mL Penicillin-Streptomycin and left at 37⁰C for 1 hour, after which non-adherent cells were rinsed off[136]. Isolation of peritoneal macrophages using this method has been shown to yield a population that is over 90% in purity[138, 139]. Peritoneal macrophages were then cultured in Atmospheric Media at pH 7.4 or 6.5 with 100μM D-luciferin, and kept at 37⁰C in atmospheric conditions."

The figure legend of Figure 3 describes the effects of pH on the circadian rhythm of bone marrow-derived macrophages ex vivo. Peritoneal macrophages involve tissue resident peritoneal macrophages with yolk sac and fetal liver origin, and also involve small peritoneal MF with bone marrow origin. The altered description of results and figure legends makes confusion.

We are very grateful to Reviewer #2 for pointing out our typo. We have fixed the caption of Figure 3 to properly describe the data as "peritoneal macrophages ex vivo".

In Figure 6C, one single Western blot is shown with any quantification. The authors should provide data of the relative protein level of p-CREB from at least 3 independent experiments. In the Western-blot part of the methods, the authors described that the pellet was discarded after cell lysis. The p-CREB is the activated form of the transcription factor CREB and there is increased binding to the chromatin to regulate gene expression. By discarding the pellet after cell lysis, the chromatin-bond p-CREB could be also removed at the same time.

We thank Reviewer 2 for bringing up this point. We agree that quantification is an important aspect of western blot. We have repeated the experiment again for n=3 and provide quantification of pCREB normalized to total protein. We have added these data, shown below, to Figure 5.

Reviewer #2 also expressed concern that we may not be capturing all of the CREB due to nuclear localization and chromatin binding. We specifically chose the lysis buffer M-Per, which is formulated to lyse the nucleus and solubilize nuclear and chromatin-bound proteins. To demonstrate this, we show in the below Figure to the Reviewer that the nuclear protein p85 is solubilized and readily detectable by western blot using our protein extraction method.

We have also added an additional sentence in the Methods section for clarity - please see the modified text below.

"Cells were lysed using the M-Per lysis reagent (Thermo Scientific, CAT#78501), supplemented with protease and phosphatase inhibitor cocktail (1:100; Sigma, CAT#PPC1010) and phosphatase inhibitor cocktail 2 (1:50; Sigma, CAT#P5726), with 200μM deferoxamine (Sigma, CAT#D9533). M-Per is formulated to lyse the nucleus and solubilize nuclear and chromatin-bound proteins, allowing isolation of nuclear proteins as well as cytosolic proteins. Lysates were incubated on ice for 1 hour, then centrifuged at 17,000 xg to pellet out debris; supernatant was collected."

It is confusing that adenylate-cyclase inhibitor MDL-12 elevated the phospho-CREB levels in BMDM. How can the authors exclude any other inducers of CREB phosphorylation?

We agree with Reviewer #2 that it is surprising pCREB was elevated with MDL-12 treatment alone, and we do indeed think that there are other pathways contributing to this. We have addressed this point in the Discussion - please see the text below.

"The mechanism through which acidic pH can modulate the circadian clock in macrophages remains unclear. Evidence in the literature suggests that acidic pH promotes a pro-resolution phenotype in macrophages by driving signaling through the cAMP pathway[29]. It has previously been shown that cAMP signaling can modulate the circadian clock[99]. However, our data indicated that cAMP signaling was not fully sufficient to confer pH-mediated changes in circadian rhythms of macrophages (Figure 5A,B). Treatment with MDL-12, commonly known as an inhibitor of adenylyl cyclase[29, 117], resulted in suppression of pH-induced changes in amplitude of circadian rhythms but did not inhibit signaling through the cAMP signaling pathway (Figure 5C,D). While MDL-12 is commonly used as an adenylyl cyclase inhibitor, it has also been documented to have inhibitory activity toward phosphodiesterases (PDEs) and the import of calcium into the cytosol through various mechanisms[118, 119]. This is of particular interest, as calcium signaling has also been shown to be capable of modulating the circadian clock[120]. Furthermore, while acid-sensing through GPCRs have been the most well-characterized pathways in macrophages, there remain additional ways in which acidic pH can be sensed by macrophages such as acid-sensing ion channels[121, 122]. Further work is required to understand the signaling pathways through which pH can influence macrophage phenotype and circadian rhythms."

It is described in the methods that BMDM were starving for 24 hours in serum-free culture media followed by serum shock (50% FBS). The cAMP production can be induced during cell starvation which should be considered for the data representation.

We appreciate that Reviewer #2 points out that our synchronization protocol of serum starvation followed by serum shock may impact the cAMP signaling pathway in macrophages, as serum shock has been shown to induce pCREB, a downstream mediator of cAMP signaling, in rat fibroblasts. Indeed, we show in additional experiments performed (in response to Reviewer #2's major comment 1) evidence that cAMP signaling is induced in macrophages following the serum shock phase of our synchronization protocol, as indicated by elevation of Icer and pCREB. As we note above, this induction is transient and returns to baseline by 2 hours post-serum shock, the time at which we replace media and begin our experiments (CT 0).

Despite the transient nature of cAMP induction by our synchronization protocol, we agree wholeheartedly with Reviewer #2 that this must be considered in light of our experimental system in which we are studying the effect of acidic pH on circadian rhythms of macrophages, which in itself induces signaling through the cAMP signaling pathway. To address Reviewer #2's point, we have performed experiments in which we culture unstimulated BMDMs in neutral pH 7.4 or acidic pH 6.5, without prior serum starvation and serum shock (i.e. we do not submit these BMDMs to the synchronization protocol). We then observed circadian rhythms of Per2-Luc by LumiCycle to determine whether acidic pH alters circadian rhythms of BMDMs in the absence of prior serum starvation followed by serum shock. Similar to our observations in Figure 2, circadian rhythms of macrophages at pH 6.5 had increased amplitude and shortened period compared to rhythms of macrophages at pH 7.4. This indicates that pH-driven changes in circadian rhythms observed in our system are not due to the synchronization protocol. The data, shown below, have been placed in a new Supplementary Figure 6, and a discussion of these results has been added to the Results section - please see the updated text below.

"As acidic pH induces signaling through the cAMP pathway, we sought to determine whether acidic pH independently contributed to the pH-driven changes in circadian rhythms we observe in BMDMs. To test this, we omitted the synchronization step and observed BMDM rhythms by LumiCycle when cultured in neutral pH 7.4 or acidic pH 6.8 or pH 6.5 (Supplementary Figure 6). Circadian rhythms of BMDMs cultured at pH 6.5 exhibited similar changes as previously observed, with enhanced amplitude and shortened period relative to BMDMs at pH 7.4. This indicates pH-driven changes observed in circadian rhythms of BMDMs occur in the absence of prior serum starvation and serum shock. "As acidic pH independently induces signaling through the cAMP pathway, we sought to determine whether acid pH could also independently contribute to the pH-driven changes in circadian rhythms we observe in BMDMs. To test this, we omitted the synchronization step and observed BMDM rhythms by LumiCycle when cultured in neutral pH 7.4 or acidic pH 6.8 or pH 6.5 (Supplementary Figure 6). Circadian rhythms of BMDMs cultured at pH 6.5 exhibited similar changes as previously observed, with enhanced amplitude and shortened period relative to BMDMs at pH 7.4. This indicates pH-driven changes observed in circadian rhythms of BMDMs occur in the absence of prior serum starvation and serum shock."

How can the authors explain and prove that the wild type and Bmal1-KO BMDM co-injected with pancreatic cancer cells subcutaneously survive, present, and have effector functions at the same extent in the subcutaneous tissue, before and during tumor growth (Figure 9)? In other words, what kind of MF-derived parameters could be modified by disrupting the circadian rhythm of MF during tumor development? The production of MF-derived regulatory enzymes, cytokines, growth factors are affected by the disrupted circadian clock in MF?

        Review #2 poses the very important question of why we see differences in tumor growth in our co-injection model, and what might be driving it. Of note, co-injection models of tumor growth are commonly used to determine macrophage-specific roles in tumor growth (Colegio 2014 Nature, Mills 2019 Cell Rep, Lee 2018 Nat Comm). We observed that tumor growth is altered when macrophages with disrupted circadian rhythms (BMAL1 KO) are co-injected compared to when macrophages with intact circadian rhythms (WT) are co-injected in a murine model of pancreatic cancer using KCKO cells. Our observation is supported by a previously published paper in which they used a co-injection model of melanoma, which we cite in the manuscript(Alexander 2020 eLife). What drives this difference in tumor growth remains an open question that is the subject of future work and is outside the scope of this paper, which focuses on our discovery that factors associated with the tumor microenvironment can alter circadian rhythms in macrophages. We have included a discussion on what is currently known about how circadian rhythms alter macrophage function, acknowledging that we have yet to answer these important questions and identifying it as interest for future work. Please see the text below.

"Considering our observations that conditions associated with the TME can alter circadian rhythms in macrophages, it becomes increasingly important to understand the relevance of macrophage rhythms to their function in tumors. It has been shown that acidic pH and lactate can each drive functional polarization of macrophages toward a phenotype that promotes tumor growth, with acidic pH modulating phagocytosis and suppressing inflammatory cytokine secretion and cytotoxicity[28, 30, 93]. However, how the changes in circadian rhythms of macrophages driven by these conditions contributes to their altered function remains unknown. Current evidence suggests that circadian rhythms confer a time-of-day-dependency on macrophage function by gating the macrophage response to inflammatory stimuli based on time-of-day. As such, responses to inflammatory stimuli such as LPS or bacteria are heightened during the active phase while the inflammatory response is suppressed during the inactive phase. An important future direction will be to determine how changes in circadian rhythms of macrophages, such as those observed under acidic pH or high lactate, influences the circadian gating of their function. Data from our lab and others suggest that disruption of the macrophage-intrinsic circadian clock accelerates tumor growth, indicating that circadian regulation of macrophages is tumor-suppressive in models of PDAC (our work) and melanoma [109]. This agrees with complementary findings that behavioral disruption of circadian rhythms in mice (through chronic jetlag) disrupts tumor macrophage circadian rhythms and accelerates tumor growth[56]. It remains unclear whether this is through the pro-tumorigenic functions of macrophages such as extracellular matrix remodeling or angiogenesis, through suppression of the anti-tumor immune response, or a combination of both functions. Further work will be needed to tease apart these distinctions."

Minor points of criticism: 1. The figure legends of the graphs and diagrams are missing in Figure 2D,E,F

We thank Reviewer #2 for pointing out that figure legends were missing. We have added legends for Figure 2D,E,F.

The BMAL1-based in vivo murine model of circadian rhythm is not introduced in the manuscript.

We thank Reviewer #2 for bringing to our attention that the BMAL1 KO macrophage model was not well-introduced in the manuscript. To address this point, we have modified the text to better introduce this model. Please see the modified text below.

"As a positive control for circadian clock disruption, we used data from BMAL1 KO peritoneal macrophages [44]. BMAL1 KO macrophages have a genetic disruption of the circadian clock due to the loss of Bmal1, the central clock gene. As a result, circadian rhythms of BMAL1 KO macrophages are disrupted, lacking rhythmicity and downstream circadian regulation of macrophage function (Supplementary Figure 8)[45, 54]. "As a positive control for circadian clock disruption, we used data from BMAL1 KO peritoneal macrophages [44]. BMAL1 KO macrophages have a genetic disruption of the circadian clock due to the loss of Bmal1, the central clock gene. As a result, circadian rhythms of BMAL1 KO macrophages are disrupted, lacking rhythmicity and downstream circadian regulation of macrophage function (Supplementary Figure 8)[45, 54]."__ __

Significance

Knudsen-Clark et al. showed that the circadian rhythm of bone marrow-derived macrophages (BMDM) can be affected by polarization stimuli, pH of the microenvironment, and by the presence of sodium-lactate. Mechanistically, the acidic pH of cell microenvironment is partly regulated by intracellular cAMP-mediated signaling events in BMDM. The authors also showed that the circadian clock of peritoneal macrophages is also modified by the pH of the cell microenvironment. Using publicly available data, the authors showed that the circadian rhythm of tumor-associated macrophages is similar to that of Bmal1-KO peritoneal macrophages. In a murine model of pancreatic cancer, the authors showed that the tumor growth is accelerated in C57BL/6 mice co-injected with cancer cells and Bmal1-KO BMDM as compared to mice co-injected with cancer cell and wild type BMDM.

We are grateful to Reviewer #2 for their very helpful comments and suggestions, which we believe have greatly enhanced the clarity and reproducibility of this manuscript.

__Reviewer #3 (Evidence, reproducibility and clarity (Required)): __

Review for Knudsen-Clark et al.

"Circadian rhythms of macrophages are altered by the acidic pH of the tumor microenvironment"

Knudsen-Clark and colleagues explore the impact of TME alterations on macrophage circadian rhythms. The authors find that both acidic pH and lactate modulate circadian rhythms which alter macrophage phenotype. Importantly, they define circadian disruption of tumor-associated macrophages within the TME and show that circadian disruption in macrophages promotes tumor growth using a PDAC line. This represents an important understanding of the crosstalk between cancer cells and immune cells as well as the understanding of how the TME disrupts circadian rhythms. The study is well-done, however, authors need to address several important points below.

We thank Reviewer #3 for their in-depth and insightful comments and suggestions, which have resulted in a much-improved manuscript. We were pleased that Reviewer #3 found the work to be "an important study that is well-done" and that it "represents an important understanding of the crosstalk between cancer cells and immune cells as well as the understanding of how the TME disrupts circadian rhythms.". In response to Reviewer #3's comments, we have added several new key experiments and changes to the text. To summarize, we added new data to Supplementary Figure 1 to better characterize our macrophage polarization states, showed in Figure 3 that low pH affects peritoneal macrophage circadian gene expression in a similar fashion as bone marrow-derived macrophages, added new data in Figure 4 to show how lactate and low pH affect circadian gene expression over time, and new computational analysis to Figures 6, 7, and Supplementary Figure 9 to probe circadian gene covariance from publicly available data. We also made several key additions to the Discussion to discuss the functional implications of macrophage circadian rhythm disruption by low pH and potential mechanisms of this disruption. Finally, at the request of Reviewer #3, we consolidated several existing Figures and added new data, where appropriate, to existing figures, and we worked to describe new findings succinctly.

Major comments:

• In Figures 3 and 4, the authors can include additional clock genes that can be run by qPCR. This was done in Figure 2 and was a nice addition to the data.

We agree with Reviewer #3's suggestion that an analysis of clock gene expression at the mRNA level would enhance our data in Figures 3 and 4. To address this point, we have performed short time course experiments to assess circadian clock gene expression over time in BMDMs cultured with or without lactate at neutral or acidic pH (for Figure 4). In line with the difference in circadian rhythms of Per2-Luc levels between BMDMs cultured in the presence or absence of lactate which we observed by Lumicycle analysis, we measured changes in expression of the circadian clock genes Per2, Nr1d1, and Cry1 between macrophages cultured with 25 mM sodium-L-lactate compared to those cultured with 0 mM sodium-L-lactate at pH 6.5. We have added these data, shown below, to Figure 4, and updated the manuscript accordingly to discuss these results. Please see below for the new Figure Panel and modified text.

"Lactate was also observed to alter expression of the circadian clock genes Per2, Cry1, and Nr1d1 over time in BMDMs cultured at pH 6.5, while having more subtle effects at pH 7.4 (Figure 4C). Notably, lactate blunted the effect of pH 6.5 on Cry1 expression, while enhancing the effect of low pH on Nr1d1 expression. In all, these data indicate that concentration of lactate similar to that present in the TME can influence circadian rhythms and circadian clock gene expression of macrophages."

As an additional measure to address Reviewer #3's point about Figure 3 (peritoneal macrophages), we have compared expression of circadian clock genes in peritoneal macrophages cultured at neutral pH 7.4 or acidic pH 6.8 for 24 hours using a publicly available RNA-seq data set from Jiang 2021 J Immunol (GSE164697). In line with previous observations in macrophages cultured under acidic compared to neutral pH conditions, including the clock gene expression data from Figure 2 in BMDMs and the Per2-Luc levels observed in peritoneal macrophages in Figure 3, we found that peritoneal macrophages exhibited differences in expression of circadian clock genes when cultured at acidic pH 6.8 compared to neutral pH 7.4. We have added these data, shown below, as Figure 3B, and have updated the manuscript accordingly - please see below for the new Figure panel and modified text.

"To test whether pH-driven changes in circadian rhythms of peritoneal macrophages were reflected at the mRNA level, we compared expression of circadian clock genes in peritoneal macrophages cultured at neutral pH 7.4 or acidic pH 6.8 for 24 hours using publicly available RNA-sequencing data [30]. In line with altered circadian rhythms observed by Lumicycle, peritoneal macrophages cultured at pH 6.8 expressed different levels of circadian clock genes than peritoneal macrophages culture at pH 7.4 (Figure 3B). The trends in changes of gene expression in peritoneal macrophages cultured at pH 6.8 matched what we observed in BMDMs, where low pH generally led to higher levels of circadian clock gene expression (Figure 2D-F). These data support our observations by LumiCycle and indicate that acidic pH drives transcriptional changes in multiple components of the circadian clock. In all, these data are evidence that pH-dependent changes in circadian rhythms are relevant to in vivo-differentiated macrophages."

We have also updated the Methods section appropriately

"FASTQ files from a previously published analysis of peritoneal macrophages cultured under neutral pH 7.4 or acidic pH 6.8 conditions were downloaded from NCBI GEO (accession #GSE164697) [30]."

2) There are far too many figures with minimal data in each. Please consolidate the figures. For example, Figures 1-3 can be fully combined, Figures 4-6 can be combined, and Figures 7-8 can be combined. Additionally, it is unclear if Figure 5 needs to be in the main, it can be moved to the supplement.

We appreciate the preference of Reviewer #3 to see some of the figures consolidated. We have combined Figures 5 and 6 into a single new Figure 5. Additionally, we have added new data from revisions to current figures to increase the amount of data in each figure and minimize the amount of new figures generated. In all, despite the large amount of new data added to the paper in response to Reviewer comments and suggestions (including additional data in Figure 4 and new Figures 6 and 8), our manuscript now contains 10 main Figures, only one more than the initial submission.

3) The observation that conditions like pH and lactate alter macrophage phenotype and rhythmicity are important. However, macrophage phenotype via gene expression does not always correlate to function. It is important for authors to demonstrate the effect of pH or lactate on macrophage function. This can be done using co-culture assays with cancer cells. If these experiments cannot be performed, it is suggested that authors discuss these limitations and consideration in the discussion.

Reviewer #3 correctly points out that changes in phenotype does not always correlate to changes in function. Others have shown that acidic pH and lactate can each alter macrophage phenotype, and also alter macrophage function and the ability to promote tumor growth (please see El-Kenawi 2019 Br J Cancer, Jiang 2021 J Immunol, Colegio 2014 Nature). How changes in rhythms influence macrophage function remains unknown and we agree with Reviewer #3 that this is an important future direction, We have added a section in the Discussion to facilitate the discussion of this important future direction. Please see the text below.

"Considering our observations that conditions associated with the TME can alter circadian rhythms in macrophages, it becomes increasingly important to understand the relevance of macrophage rhythms to their function in tumors. It has been shown that acidic pH and lactate can each drive functional polarization of macrophages toward a phenotype that promotes tumor growth, with acidic pH modulating phagocytosis and suppressing inflammatory cytokine secretion and cytotoxicity[28, 30, 93]. However, how the changes in circadian rhythms of macrophages driven by these conditions contributes to their altered function remains unknown. Current evidence suggests that circadian rhythms confer a time-of-day-dependency on macrophage function by gating the macrophage response to inflammatory stimuli based on time-of-day. As such, responses to inflammatory stimuli such as LPS or bacteria are heightened during the active phase while the inflammatory response is suppressed during the inactive phase. An important future direction will be to determine how changes in circadian rhythms of macrophages, such as those observed under acidic pH or high lactate, influences the circadian gating of their function."

4) On line 119-122, authors describe a method for polarization of macrophages. They then reference one gene to confirm each macrophage polarization state. To more definitively corroborate proper macrophage polarization, authors should perform qPCR for additional target genes that are associated with each phenotype. For example, Socs3, CD68, or CD80 for M1, and CD163 or VEGF for M2. Alternatively, the authors should cite previous literature validating this in vitro polarization model.

We appreciate Reviewer #3's suggestion to better the phenotypic identity of our polarization models with additional canonical markers. To address this point, we have expanded our panel using transcriptional markers commonly used in the murine polarization model for M1 macrophages such as Tnfa, Il6, and Il1b. As discussed in the response to Reviewer #1's minor point 5 and Reviewer #2's major point 2, we have also expanded our panel to include additional markers for M2 such as Vegf, Retnla, Ym1, Mgl1, and CD206. We have added these new data to Supplementary Figure 1. Finally, we have added additional citations for the in vitro polarization models. Please see the modified text and new data, below.

"As macrophages are a phenotypically heterogeneous population in the TME, we first sought to understand whether diversity in macrophage phenotype could translate to diversity in circadian rhythms of macrophages. To this end, we used two well-established in vitro polarization models to study distinct macrophage phenotypes[5, 60-63]. For a model of pro-inflammatory macrophages, we stimulated macrophages with IFNγ (interferon γ) and LPS (lipopolysaccharide) to elicit a pro-inflammatory phenotype[60, 64]. These macrophages are often referred to as 'M1' and are broadly viewed as anti-tumorigenic, and we will refer to them throughout this paper as pro-inflammatory macrophages[65, 66]. For a model at the opposite end of the phenotypic spectrum, we stimulated macrophages with IL-4 and IL-13[60, 67]. While these type 2 stimuli play a role in the response to parasites and allergy, they are also major drivers of wound healing; in line with this, IL-4 and IL-13-stimulated macrophages have been well-characterized to adopt gene expression profiles associated with wound-healing and anti-inflammatory macrophage phenotypes[68-71]. As such, these macrophages are often used as a model to study pro-tumorigenic macrophages in vitro and are often referred to as 'M2' macrophages; throughout this paper, we will refer to IL-4 and IL-13-stimulated macrophages as pro-resolution macrophages[66, 72, 73]. Consistent with previous studies, we found that genes associated with anti-inflammatory and pro-resolution programming characteristic of IL-4 and IL-13-stimulated macrophages such as Arg1, Retnla, Chil3 (Ym1), Clec10a (MGL1), and Mrc1 (CD206) were induced in IL-4 and IL-13-stimulated macrophages, but not IFNγ and LPS-stimulated macrophages. In contrast, genes associated with pro-inflammatory activity characteristic of IFNγ and LPS-stimulated macrophages such as Nos2 (iNOS), Tnfa, Il1b, and Il6 were induced in IFNγ and LPS-stimulated macrophages, but not IL-4 and IL-13-stimulated macrophages (Supplementary Figure 1)[28, 30, 65, 71, 74, 75]. This indicates that macrophages stimulated with IL-4 and IL-13 were polarized toward a pro-resolution phenotype, while macrophages stimulated with IFNγ and LPS were polarized toward a pro-inflammatory phenotype.

5) Several portions of the manuscript are unnecessarily long, including the intro and discussion. Please consolidate the text. The results section is also very lengthy, please consider consolidation.

We appreciate Reviewer #3's preference for a shorter manuscript. The revised manuscript, in response to the many Reviewer comments and requests, contains many new pieces of data, and we have taken care to describe these new data as briefly and simply as possible. In preparation for this Revision, we also removed and shortened several sections of the Results and Discussion where we felt extra explanation was not necessary. We will work with the editor of the journal we submit to ensure the length of the manuscript sections is compliant with the journal's guidelines.

6) The authors find that macrophage phenotype impacts rhythmicity. However, there is no mechanistic understanding of why this occurs. The authors should provide some mechanistic insight on this topic in the discussion.

We agree with Reviewer #3 that while the mechanism by which macrophage phenotype alters rhythms remains unknown, this is an important topic of discussion. While there is some literature on how circadian rhythms modulate inflammatory response (and hints at how it may influence phenotype) in macrophages, there is very little on the converse: how phenotype may influence circadian rhythms. We address this point by expanding on our Discussion - please see the modified text below.

"Elucidating the role of circadian rhythms in regulation of macrophage biology necessitates a better understanding of the crosstalk between phenotype and circadian rhythms. Although lactate polarizes macrophages toward a pro-resolution phenotype similar to acidic pH[30, 93], exposure to lactate had different effects on circadian rhythms - and in some cases, circadian clock gene expression - than exposure to acidic pH (Figure 4). Sensing of lactate occurs through different pathways than acid-sensing, which may contribute to the different ways in which these two stimuli modulate circadian rhythms of macrophages[111]. One previously published finding that may offer mechanistic insight into how phenotype can influence circadian rhythms is the suppression of Bmal1 by LPS-inducible miR-155[54]. It has also been observed that RORα-mediated activation of Bmal1 transcription is enhanced by PPARγ co-activation[112]. In macrophages, PPARγ expression is induced upon stimulation with IL-4 and plays a key role in alternative activation of macrophages, promoting a pro-resolution macrophage phenotype, and supporting resolution of inflammation[113-115]. Such observations prompt the question of whether there are yet-unidentified factors induced downstream of various polarizing stimuli that can modulate expression of circadian genes at the transcriptional and protein levels. Further work is required to understand the interplay between macrophage phenotype and circadian rhythms."

7) The data presented in Figure 9 is very intriguing and arguably the strongest aspect of the paper. To strengthen the point, the authors could repeat this experiment with an additional cell model, another PDAC line or a different cancer line.

We appreciate Reviewer #3's comment about the impact of tumor growth data. Indeed, our finding that deletion of Bmal1 in co-injected macrophages accelerated PDAC growth has been recapitulate by others in different cancer models. This lends strength to our observations. We discuss and cite complementary work on macrophage rhythms and tumor growth in other models of cancer the Discussion, please see below.

"Data from our lab and others suggest that disruption of the macrophage-intrinsic circadian clock accelerates tumor growth, indicating that circadian regulation of macrophages is tumor-suppressive in models of PDAC (our work) and melanoma [109]. This agrees with complementary findings that behavioral disruption of circadian rhythms in mice (through chronic jetlag) disrupts tumor macrophage circadian rhythms and accelerates tumor growth[56]."

Minor Comments:

1) Data is Figure 2 is interesting and the impact on circadian rhythms is clear based on changes in amplitude and period. However, though the impact on period and amplitude is clear from Figures 2A-C, the changes in circadian gene expression are less clear. For instance, though amplitude is up in 2B, amplitude is suppressed in 2C. However, that does not appear to be reflected in the gene expression data in Figures 2E and F. The authors should comment on this.

Reviewer #3 correctly points out that there appear to be discrepancies between the LumiCycle data in Figure 2 and the circadian gene expression data in Figure 2. This discrepancy is perhaps unsurprising given that the gene expression data is only a short time course over 12 hours, while the LumiCycle data are collected over a course of 3 days. The gene expression data do not allow us to determine changes in period or rhythm. Another point of interest is that it's been shown that circadian regulation occurs on many different levels (transcriptional, post-transcriptional, translational, post-translational). As result of this, circadian patterns observed in gene transcripts don't always match those of their encoded proteins; just the same, circadian patterns of proteins aren't always reflected in their encoding gene transcripts (Collins 2021 Genome Res). Due to this multi-level regulation, we propose that the results of the LumiCycle analysis, which measures PER2-Luc levels, are a more robust readout of rhythms because they are further downstream of the molecular clock than transcriptional readouts. That said, observing changes at both the protein (by Lumicycle) and transcriptional level confirm that all components of the clock are altered by acidic pH, even if the way in which they are altered appears to differ. We have incorporated the points we raised above into the Results section.

Please see the modified text below.

"Low pH was also observed to alter the expression of the circadian clock genes Per2, Cry1, and Nr1d1 (REV-ERBα) over time across different macrophage phenotypes, confirming that multiple components of the circadian clock are altered by acidic pH (Figure 2D-F). Notably, the patterns in expression of circadian genes did not always match the patterns of PER2-Luc levels observed by LumiCycle. This is perhaps unsurprising, as circadian rhythms are regulated at multiple levels (transcriptional, post-transcriptional, translational, post-translational); as a result, circadian patterns observed in circadian proteins such as PER2-Luc do not always match those of their gene transcripts[77]."

2) On line 156-158, authors describe damping rate. I believe the authors are trying to say that damping rate increases as the time it takes cells to desynchronize decreases and vice versa. However, this point needs to be better explained.

We thank Reviewer #3 for bringing to our attention that this was not communicated clearly in the text. We have adjusted our explanation to be clearer. Please see the modified text below.

"Damping of rhythms in most free-running cell populations (defined as populations cultured in the absence of external synchronizing stimuli) occurs naturally as the circadian clocks of individual cells in the population become desynchronized from each other; thus, damping can be indicative of desynchrony within a population[84]. The damping rate increases as the time it takes for rhythms to dissipate decreases; conversely, as damping rate decreases as the time it takes for rhythms to dissipate increases."

3) Data presented in Figures 3 and 4 are different in terms of the impact of changing the pH. The source of the macrophages is different, but the authors could clarify this further.

We thank Reviewer #3 for this comment. Our conclusion is that the impact of low pH is largely similar in Figure 3 (peritoneal macrophages) and Figure 4 (BMDMs exposed to low pH and lactate). In both Figures 3 and 4, exposure to acidic pH by culturing macrophages at pH 6.5 increased amplitude, decreased period, and increased damping rate compared to macrophages cultured at neutral pH 7.4.

4) For heatmaps shown in Figures 7 and 8, please calculate covariance and display asterisks where P We thank Reviewer #3 for the excellent suggestion to use an additional approach to asses circadian clock status in samples by measuring co-variance in the circadian clock gene network. To address this point, we have performed weighted gene co-expression network analysis (WGCNA) to calculate covariance, as was originally performed in Chun and Fortin et al Science Advances 2022. For the samples analyzed in Figure 7 (now Figure 6), we have added these data to the figure. We have applied this analysis to a new set of human data that we analyzed and added it to the new Figure 7. Finally, for the samples analyzed in Figure 8, we have added these data as a new Supplementary Figure 9. Please see the data and modified text below.

Figure 6

"Weighted gene co-expression network analysis (WGCNA) has been used as an alternate approach to measure the co-variance between clock genes and thus assess bi-directional correlations among the core clock gene network in healthy tissue and tumor samples [103]. In line with the circadian disorder observed by CCD, while many bi-directional correlations among the core clock gene network were significant and apparent in wild type peritoneal macrophages, these relationships were altered or abolished within BMAL1 KO peritoneal macrophages and TAM samples, and in some cases replaced by new relationships (Figure 6E). This indicates that there is population-level disorder in the circadian rhythms of tumor-associated macrophages in murine lung cancer."

Figure 7

"We next assessed the status of the circadian clock in human TAMs from NSCLC patients. We performed CCD with publicly available RNA-seq data of tumor-adjacent macrophages and tumor-associated macrophages from NSCLC patients, using alveolar macrophages from healthy donors as a control[104, 105]. To assess the contribution of the acidic TME to circadian disorder, we subset TAM NSCLC patient samples into groups (Crem high TAMs and Crem low TAMs) based on median Crem expression. Notably, in macrophages from human NSCLC there was a trend toward disorder in Crem low but not Crem high TAM samples (Figure 7A,B). Additionally, the co-variance among core clock genes observed in alveolar macrophages from healthy donors was absent within Crem low and Crem high TAM samples (Figure 7C). In all, these data indicate that there is population-level disorder in the circadian rhythms of tumor-associated macrophages in humans and mice, suggesting that circadian rhythms are indeed altered in macrophages within the TME."

Supplementary Figure 9

"CCD score worsened as populations became increasingly desynchronized, with the 12hr desynchronized population having a significantly worse CCD score than synchronized, homogenous macrophage population (Figure 8C). This indicates that as circadian rhythms of individual macrophages within a population become more different from each other, circadian disorder increases at the population-level. This is further supported by WGCNA, which revealed that the significant co-variance of circadian clock genes in the synchronized population was progressively altered and lost as the population is increasing desynchronized to 12 hours (Supplementary Figure 9)."

Reviewer #3 (Significance (Required)):

This is an important study that is well-done. It is the feeling of the reviewer that the study warrants a revision, at the discretion of the editor. The study represents an important understanding of the crosstalk between cancer cells and immune cells as well as the understanding of how the TME disrupts circadian rhythms.

We thank Reviewer #3 for their comments regarding the impact and significance of our work. As shown by the comments above, we are confident we have fully addressed each of the points that were made to result in a much-improved revised manuscript.

---

---

---

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Review for Knudsen-Clark et al. "Circadian rhythms of macrophages are altered by the acidic pH of the tumor microenvironment"

Knudsen-Clark and colleagues explore the impact of TME alterations on macrophage circadian rhythms. The authors find that both acidic pH and lactate modulate circadian rhythms which alter macrophage phenotype. Importantly, they define circadian disruption of tumor-associated macrophages within the TME and show that circadian disruption in macrophages promotes tumor growth using a PDAC line. This represents an important understanding of the crosstalk between cancer cells and immune cells as well as the understanding of how the TME disrupts circadian rhythms. The study is well-done, however, authors need to address several important points below.

Major comments:

1. In Figures 3 and 4, the authors can include additional clock genes that can be run by qPCR. This was done in Figure 2 and was a nice addition to the data.
2. There are far too many figures with minimal data in each. Please consolidate the figures. For example, Figures 1-3 can be fully combined, Figures 4-6 can be combined, and Figures 7-8 can be combined. Additionally, it is unclear if Figure 5 needs to be in the main, it can be moved to the supplement.
3. The observation that conditions like pH and lactate alter macrophage phenotype and rhythmicity are important. However, macrophage phenotype via gene expression does not always correlate to function. It is important for authors to demonstrate the effect of pH or lactate on macrophage function. This can be done using co-culture assays with cancer cells. If these experiments cannot be performed, it is suggested that authors discuss these limitations and consideration in the discussion.
4. On line 119-122, authors describe a method for polarization of macrophages. They then reference one gene to confirm each macrophage polarization state. To more definitively corroborate proper macrophage polarization, authors should perform qPCR for additional target genes that are associated with each phenotype. For example, Socs3, CD68, or CD80 for M1, and CD163 or VEGF for M2. Alternatively, the authors should cite previous literature validating this in vitro polarization model.
5. Several portions of the manuscript are unnecessarily long, including the intro and discussion. Please consolidate the text. The results section is also very lengthy, please consider consolidation.
6. The authors find that macrophage phenotype impacts rhythmicity. However, there is no mechanistic understanding of why this occurs. The authors should provide some mechanistic insight on this topic in the discussion.
7. The data presented in Figure 9 is very intriguing and arguably the strongest aspect of the paper. To strengthen the point, the authors could repeat this experiment with an additional cell model, another PDAC line or a different cancer line.

Minor Comments:

1. Data is Figure 2 is interesting and the impact on circadian rhythms is clear based on changes in amplitude and period. However, though the impact on period and amplitude is clear from Figures 2A-C, the changes in circadian gene expression are less clear. For instance, though amplitude is up in 2B, amplitude is suppressed in 2C. However, that does not appear to be reflected in the gene expression data in Figures 2E and F. The authors should comment on this.
2. On line 156-158, authors describe damping rate. I believe the authors are trying to say that damping rate increases as the time it takes cells to desynchronize decreases and vice versa. However, this point needs to be better explained.
3. Data presented in Figures 3 and 4 are different in terms of the impact of changing the pH. The source of the macrophages is different, but the authors could clarify this further.
4. For heatmaps shown in Figures 7 and 8, please calculate covariance and display asterisks where P < 0.001. This will more clearly demonstrate the loss of co-variance between the clock network as a result of clock disruption, the TME, and cell desynchrony.

Significance

This is an important study that is well-done. It is the feeling of the reviewer that the study warrants a revision, at the discretion of the editor. The study represents an important understanding of the crosstalk between cancer cells and immune cells as well as the understanding of how the TME disrupts circadian rhythms.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Knudsen-Clark et al. showed that the circadian rhythm of bone marrow-derived macrophages (BMDM) can be affected by polarization stimuli, pH of the microenvironment, and by the presence of sodium-lactate. Mechanistically, the acidic pH of cell microenvironment is partly regulated by intracellular cAMP-mediated signaling events in BMDM. The authors also showed that the circadian clock of peritoneal macrophages is also modified by the pH of the cell microenvironment. Using publicly available data, the authors showed that the circadian rhythm of tumor-associated macrophages is similar to that of Bmal1-KO peritoneal macrophages. In a murine model of pancreatic cancer, the authors showed that the tumor growth is accelerated in C57BL/6 mice co-injected with cancer cells and Bmal1-KO BMDM as compared to mice co-injected with cancer cell and wild type BMDM.

Major points of criticism: 1. Nine main figures include different experimental models on a non-systematic manner in the manuscript, and only literature-based correlation is used to link the results each other. The authors used in vitro BMDM and peritoneal cell-based model systems to study the effects of IL4+IL13, IFN+LPS, low pH, sodium-lactate, adenylate cyclase inhibitors on the circadian clock of macrophages. The link between these microenvironment conditions of the cells is still correlative with the tumor microenvironment: publicly available data were used to correlate the increased expression level of cAMP-activated signaling events with the presence of acidic pH of tumor microenvironment. Notably, the cell signaling messenger molecule cAMP is produced by not only low extracellular pH by activated GPCRs, but also starvation of the cell. The starvation is also relevant to this study, since the BMDM used in the in vitro culture system were starving for 24 hours before the measurement of Per2-Luc expression to monitor circadian rhythm. 2. The definition of pre-resolution macrophages (MF) used across the manuscript could be argued. The authors defined BMDM polarized with IL-4 and IL-13 as pre-resolution MF. Resolution is followed by inflammation, but the IL-4 secretion does not occur in every inflammatory setting. Moreover, IL-4 and IL-13 are secreted during specific tissue environment and immunological settings involving type 2 inflammation or during germinal center reactions of the lymph nodes. • What are the characteristics of pre-resolution macrophages (MF)? The authors indicated that IL-4 and IL-13 cytokines were used to model the pre-resolution macrophages. In which pathological context are these cytokines produced and induce pre-resolution macrophages? IL-4 polarized BMDM can also produce pro-inflammatory protein and lipid mediators as compared to LPS-stimulated BMDM, and IL-4 polarized BMDM still have potent capacity to recruit immune cells and to establish type 2 inflammation. • The authors showed Arg1 and Vegfa qPCR data from BMDM only. Based on the literature, these MFs are anti-inflammatory cells particularly. Resolution-related MFs followed by acute inflammation are a specific subset of MFs, and the phenotype of pre-resolution MF should be described, referred, and measured specifically. 3. The authors used IFN and LPS, or IL-4 and IL-13 and co-treatments to polarize BMDM in to type 1 (referred as pro-inflammatory MF) and type 2 (referred as pre-resolution MF) activation state. The comparison between these BMDM populations has limitations, since LPS induces a potent inflammatory response in MF. The single treatment with MF-polarizing cytokines enable a more relevant comparison to study the circadian clock in classically and alternatively activated MF. 4. There are missing links between the results of showing the circadian rhythm of polarized BMDM, sodium-lactate treated BMDM, and tumor growth. Specifically, do the used pancreatic ductal adenocarcinoma cells produce IL-4 and sodium-lactate? In the LLC-based experimental in silico analysis of tumors, the LLC do not produce IL-4. 5. How can the circadian rhythm affect the function of BMDM? The Authors should provide evidence that circadian rhythm affects the function of polarized MF. 6. In Figure 3, the authors show data from peritoneal cells. The isolated peritoneal cells are not pure macrophage populations. Based on the referred article in the manuscript, the peritoneal cavity contains more then 50% of lymphocytes, and the myeloid compartment contains 80% macrophages. 7. The figure legend of Figure 3 describes the effects of pH on the circadian rhythm of bone marrow-derived macrophages ex vivo. Peritoneal macrophages involve tissue resident peritoneal macrophages with yolk sac and fetal liver origin, and also involve small peritoneal MF with bone marrow origin. The altered description of results and figure legends makes confusion. 8. In Figure 6C, one single Western blot is shown with any quantification. The authors should provide data of the relative protein level of p-CREB from at least 3 independent experiments. In the Western-blot part of the methods, the authors described that the pellet was discarded after cell lysis. The p-CREB is the activated form of the transcription factor CREB and there is increased binding to the chromatin to regulate gene expression. By discarding the pellet after cell lysis, the chromatin-bond p-CREB could be also removed at the same time. 9. It is confusing that adenylate-cyclase inhibitor MDL-12 elevated the phospho-CREB levels in BMDM. How can the authors exclude any other inducers of CREB phosphorylation? 10. It is described in the methods that BMDM were starving for 24 hours in serum-free culture media followed by serum shock (50% FBS). The cAMP production can be induced during cell starvation which should be considered for the data representation. 11. How can the authors explain and prove that the wild type and Bmal1-KO BMDM co-injected with pancreatic cancer cells subcutaneously survive, present, and have effector functions at the same extent in the subcutaneous tissue, before and during tumor growth (Figure 9)? In other words, what kind of MF-derived parameters could be modified by disrupting the circadian rhythm of MF during tumor development? The production of MF-derived regulatory enzymes, cytokines, growth factors are affected by the disrupted circadian clock in MF?

Minor points of criticism: 1. The figure legends of the graphs and diagrams are missing in Figure 2D,E,F 2. The BMAL1-based in vivo murine model of circadian rhythm is not introduced in the manuscript.

Significance

Knudsen-Clark et al. showed that the circadian rhythm of bone marrow-derived macrophages (BMDM) can be affected by polarization stimuli, pH of the microenvironment, and by the presence of sodium-lactate. Mechanistically, the acidic pH of cell microenvironment is partly regulated by intracellular cAMP-mediated signaling events in BMDM. The authors also showed that the circadian clock of peritoneal macrophages is also modified by the pH of the cell microenvironment. Using publicly available data, the authors showed that the circadian rhythm of tumor-associated macrophages is similar to that of Bmal1-KO peritoneal macrophages. In a murine model of pancreatic cancer, the authors showed that the tumor growth is accelerated in C57BL/6 mice co-injected with cancer cells and Bmal1-KO BMDM as compared to mice co-injected with cancer cell and wild type BMDM.

Major points of criticism:

1. Nine main figures include different experimental models on a non-systematic manner in the manuscript, and only literature-based correlation is used to link the results each other. The authors used in vitro BMDM and peritoneal cell-based model systems to study the effects of IL4+IL13, IFN+LPS, low pH, sodium-lactate, adenylate cyclase inhibitors on the circadian clock of macrophages. The link between these microenvironment conditions of the cells is still correlative with the tumor microenvironment: publicly available data were used to correlate the increased expression level of cAMP-activated signaling events with the presence of acidic pH of tumor microenvironment. Notably, the cell signaling messenger molecule cAMP is produced by not only low extracellular pH by activated GPCRs, but also starvation of the cell. The starvation is also relevant to this study, since the BMDM used in the in vitro culture system were starving for 24 hours before the measurement of Per2-Luc expression to monitor circadian rhythm.
2. The definition of pre-resolution macrophages (MF) used across the manuscript could be argued. The authors defined BMDM polarized with IL-4 and IL-13 as pre-resolution MF. Resolution is followed by inflammation, but the IL-4 secretion does not occur in every inflammatory setting. Moreover, IL-4 and IL-13 are secreted during specific tissue environment and immunological settings involving type 2 inflammation or during germinal center reactions of the lymph nodes.
   • What are the characteristics of pre-resolution macrophages (MF)? The authors indicated that IL-4 and IL-13 cytokines were used to model the pre-resolution macrophages. In which pathological context are these cytokines produced and induce pre-resolution macrophages? IL-4 polarized BMDM can also produce pro-inflammatory protein and lipid mediators as compared to LPS-stimulated BMDM, and IL-4 polarized BMDM still have potent capacity to recruit immune cells and to establish type 2 inflammation.
   • The authors showed Arg1 and Vegfa qPCR data from BMDM only. Based on the literature, these MFs are anti-inflammatory cells particularly. Resolution-related MFs followed by acute inflammation are a specific subset of MFs, and the phenotype of pre-resolution MF should be described, referred, and measured specifically.
3. The authors used IFN and LPS, or IL-4 and IL-13 and co-treatments to polarize BMDM in to type 1 (referred as pro-inflammatory MF) and type 2 (referred as pre-resolution MF) activation state. The comparison between these BMDM populations has limitations, since LPS induces a potent inflammatory response in MF. The single treatment with MF-polarizing cytokines enable a more relevant comparison to study the circadian clock in classically and alternatively activated MF.
4. There are missing links between the results of showing the circadian rhythm of polarized BMDM, sodium-lactate treated BMDM, and tumor growth. Specifically, do the used pancreatic ductal adenocarcinoma cells produce IL-4 and sodium-lactate? In the LLC-based experimental in silico analysis of tumors, the LLC do not produce IL-4.
5. How can the circadian rhythm affect the function of BMDM? The Authors should provide evidence that circadian rhythm affects the function of polarized MF.
6. In Figure 3, the authors show data from peritoneal cells. The isolated peritoneal cells are not pure macrophage populations. Based on the referred article in the manuscript, the peritoneal cavity contains more then 50% of lymphocytes, and the myeloid compartment contains 80% macrophages.
7. The figure legend of Figure 3 describes the effects of pH on the circadian rhythm of bone marrow-derived macrophages ex vivo. Peritoneal macrophages involve tissue resident peritoneal macrophages with yolk sac and fetal liver origin, and also involve small peritoneal MF with bone marrow origin. The altered description of results and figure legends makes confusion.
8. In Figure 6C, one single Western blot is shown with any quantification. The authors should provide data of the relative protein level of p-CREB from at least 3 independent experiments. In the Western-blot part of the methods, the authors described that the pellet was discarded after cell lysis. The p-CREB is the activated form of the transcription factor CREB and there is increased binding to the chromatin to regulate gene expression. By discarding the pellet after cell lysis, the chromatin-bond p-CREB could be also removed at the same time.
9. It is confusing that adenylate-cyclase inhibitor MDL-12 elevated the phospho-CREB levels in BMDM. How can the authors exclude any other inducers of CREB phosphorylation?
10. It is described in the methods that BMDM were starving for 24 hours in serum-free culture media followed by serum shock (50% FBS). The cAMP production can be induced during cell starvation which should be considered for the data representation.
11. How can the authors explain and prove that the wild type and Bmal1-KO BMDM co-injected with pancreatic cancer cells subcutaneously survive, present, and have effector functions at the same extent in the subcutaneous tissue, before and during tumor growth (Figure 9)? In other words, what kind of MF-derived parameters could be modified by disrupting the circadian rhythm of MF during tumor development? The production of MF-derived regulatory enzymes, cytokines, growth factors are affected by the disrupted circadian clock in MF?

Minor points of criticism:

1. The figure legends of the graphs and diagrams are missing in Figure 2D,E,F
2. The BMAL1-based in vivo murine model of circadian rhythm is not introduced in the manuscript.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

The manuscript by Knudsen-Clark et al. investigates the novel topic of circadian rhythms in macrophages and their role in tumorigenesis. The authors explore how circadian rhythms of macrophages may be influenced by the tumor microenvironment (TME). They utilize a system of bone marrow-derived macrophages obtained from transgenic mice carrying PER2-Luciferase (PER2-Luc), a trackable marker of rhythmic activity. The study evaluates how conditions associated with the TME, such as polarizing stimuli (to M1 or M2 subtype), acidic pH, and elevated lactate, can each alter circadian rhythms in macrophages. The authors employ several approaches to explore macrophage functions in cancer-related settings. While the manuscript presents interesting findings and may be the first to demonstrate that tumor stimuli alter circadian rhythms in macrophages and impact tumor growth, it lacks a clear conclusion regarding the role of altered circadian rhythms in suppressing tumor growth. . Several discrepancies need to be addressed before publication, therefore, the manuscript requires revision before publication, addressing the following comments:

Major comments:

1. It is well known that pro-inflammatory macrophages primarily rely on glycolysis during inflammation, exhibiting dysregulated tricarboxylic acid (TCA) cycle activity. These pro-inflammatory macrophages are commonly referred to as 'M1' or pro-inflammatory, as noted in the manuscript. In contrast, M2 macrophages, or pro-resolution macrophages, are highly dependent on active mitochondrial respiration and oxidative phosphorylation (OXPHOS). Given that M1 macrophages favor glycolysis, they create an acidic environment due to elevated lactate levels and other acidifying metabolites. However, the study does not address this effect. The authors' hypothesis revolves around the acidic environment created by glycolytic tumors, yet they overlook the self-induced acidification of media when culturing M1 macrophages. This raises the question of how the authors explain the reduced circadian rhythms observed in pro-inflammatory macrophages in their study, while low pH and higher lactate levels enhance the amplitude of circadian rhythms. I would encourage the authors to incorporate the glycolytic activity of pro-inflammatory macrophages into their experimental setup. Otherwise the data look contradictory and misleading in some extent.
2. The article examines the role of circadian rhythms in tumor-associated macrophages, yet it lacks sufficient compelling data to support this assertion. Two figures, Figure 7 and Figure 9, are presented in relation to cancer. In Figure 7, gene expression analysis of Arg1 (an M2 marker) and Crem (a potential circadian clock gene) is conducted in wild-type macrophages, BMAL1-knockout macrophages with dysregulated circadian rhythms, and using publicly available data on tumor-associated macrophages from a study referenced as 83. However, it is noted that this referenced study is actually a review article by Geeraerts et al. (2017) titled "Macrophage Metabolism as Therapeutic Target for Cancer, Atherosclerosis, and Obesity" published in Frontiers in Immunology. This raises concerns about the reliability of the results. Furthermore, comparing peritoneal macrophages from healthy mice with macrophages isolated from lung tumors is deemed inaccurate. It is suggested that lung macrophages from healthy mice and those from mice with lung tumors should be isolated separately for a more appropriate comparison. Consequently, Figure 7B is further questioned regarding how the authors could compare genes from the circadian rhythm pathway between these non-identical groups. As a result, the conclusion drawn from these data, suggesting that tumor-associated macrophages exhibit a gene expression pattern similar to BMAL1-KO macrophages, is deemed incorrect, affecting the interpretation of the data presented in Figure 8.
3. If the authors aim to draw a clear conclusion regarding the circadian rhythms of tumor-associated macrophages (TAMs), they may need to analyze single-sorted macrophages from tumors and corresponding healthy tissues. Such data are publicly available (of course not in #83)
4. Additionally, it is widely acknowledged that human and mouse macrophages exhibit distinct gene expression profiles, both in vitro and in vivo. While assuming that genes involved in circadian rhythms are conserved across species, the authors could consider extending their funding to include analyses of single-sorted macrophages from cancer patients, such as those with lung cancer or pancreatic ductal adenocarcinoma (PDAC). These experiments would provide relevant insights into TAM biology.

Minor comments:

1. Figure 2C needs clarification. It's unclear why pro-inflammatory macrophages treated with lactic acid would have a shorter amplitude and period, while acidic pH would increase amplitude and period in M2 macrophages.
2. The scale in Figure 2C should be equal for all conditions (e.g., -200).
3. Absolute values of amplitude, damping, and period differ between Figure 1 and Figure 2A, B, C. The authors should explain these discrepancies.
4. The authors should consider modulating the acidic environment of macrophages in settings more representative of cancer. For example, by adding conditioned media from tumor cells with pronounced glycolysis.
5. Arg1 alone is not sufficient as an M2 polarization marker. The authors should include additional markers.

Significance

While the manuscript provides valuable insights and has obvious novelty, it requires a significant revision

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Reviewer #1

Major comments

1. In Fig. 3B, the dramatic (~20%) change in EGFR mobility upon EGF stimulation (i.e. from fast-mobile to confined) implies EGFR-EGF binding in excess of what is typically reported in the literature. How do the authors reconcile this and are there features of their cell model/analysis pipeline that are artificially contributing to this observation?

We tagged the cytoplasmic tail of EGFR with GFP (EGFR-GFP) (p. 11, l.198-201) instead of using antibody-mediated labeling, which has been used in previous studies. Compared with antibody-mediated labeling in the extracellular region of EGFR, a greater change in EGFR mobility upon EGF stimulation was observed, which is consistent with previous studies (Hiroshima et al., 2018; Yasui et al., 2018) (p. 11, l.209-211). The differences in labeling may cause differences in the characteristics of EGFR or binding efficiency to EGF. We have mentioned this possibility in the Discussion section (p. 25, l.501-505).

The detection of PIP2 nanodomains in the plasma membrane is somewhat controversial, especially using the PH domain of PLCd to detect PIP2 or using similar strategies. The recent study by Pacheco et al (JCB 2023, PMID: 36416724) uses a variety of measurements of fluorescent labeling of PIP2 by protein-based biosensors (similar to this study) and concludes that PIP2 is free diffusing in the plasma membrane, which would be inconsistent with PIP2 nanodomains. Pacheco et al propose that while engagement of PIP2 to effectors via the inositol headgroup may serve to immobilize this lipid, allowing clustering, the use of relatively large protein domains as fluorescent ligands that bind to the PIP2 headgroup to track PIP2, as performed here, displaces any intrinsic clustering mechanism, leading to free diffusion of PIP2. How can the clustering observed here for PIP2 be reconciled? Is it possible that additional, non lipid-based interactions function alongside PH domain-PIP2 interactions as a form of coincidence detection? It would be quite helpful to support the data shown in this manuscript with a different PIP2 binding domain, such as the Tubby domain used by Pacheco et al. It would not be necessary to repeat all experiments with such a complementary probe, but some key experiments that assess the apparent clustering of PIP2 would be important to consider repeating with this complementary PIP2 probe.

As suggested by the reviewer, we performed SMLM and SMT analyses using TubbyC. No significant differences were observed between PLCd-PH and TubbyC probes. These results are shown in Figs. 2C (p. 9, l.167-172), and S3E and S3F (p. 12, l.226-229).

It is unclear if and how stimulation with EGF or overexpression of synaptojanin modulates PIP2 in the plasma membrane. Some studies found that EGF stimulation does not change PIP2 levels in the PM, including Delos Santos et al. (Mol Biol Cell. 2017, PMID: 28814502). Others found that the regulation of PIP2 levels in the plasma membrane is tightly controlled and the total levels of PIP2 can resist alterations of PIP2 by changes in lipid enzymes (Wills et al. JCS 2023, PMID: 37534432). Hence, it is not clear if the stimulation with EGF or the overexpression of synaptojanin changes plasma membrane PIP2 levels, or may only alter the nanoscale dynamnics of this lipid. If the effects of synaptojanin may be restricted to alterations of the nanoscale organization of PIP2 in the membrane, it would be important to consider that synaptojanin is strongly localized to clathrin-coated pits in the plasma membrane (e.g. Perera et al. PNAS 2007. PMID: 17158794), and that EGFR only exhibits strong recruitment to clathrin-coated pits following EGF stimulation, which would suggest that the non-ligand-bound EGFR is distant to synaptojanin-containing structures. Some consideration of the possibility of broad action of PIP2 depletion vs nanoscale localized effects by these treatments should be considered when interpreting the results of this study.

Here, we show that EGF stimulation decreases the degree of PI(4,5)P2 nanodomain clustering, mainly by reducing the density of small nanodomains. As noted by the reviewer, EGF stimulation does not induce a significant change in PI(4,5)P2 levels in the plasma membrane (Delos Santos et al., 2017). These results suggest that PI(4,5)P2 may be hydrolyzed by PLCγ only in the region around EGFR. In contrast, synaptojanin expression reduced PI(4,5)P2 levels by 45% (Field et al., 2005), leading to defects in cytokinesis (Abe et al., 2012; Field et al., 2005). We previously found that synaptojanin expression diminishes the localization of PI(4,5)P2-binding proteins to the plasma membrane (Abe et al., 2012; Abe et al., 2021). These results suggest that synaptojanin expression decreases the amount of PI(4,5)P2 throughout the plasma membrane rather than in the restricted region around EGFR. We have added this information to the Discussion section (p. 22, l.443-452).

Minor comments

1. Please quantify the extent to which endogenous EGFR was knocked down.

We knocked out rather than knocked down the EGFR gene with CRISPR/Cas9 gene editing. Therefore, there was no intrinsic EGFR in the cells used (Fig. S2A). The detailed methods is described in the Methods section (p. 28, l.546-564).

Fig. 2C - please provide the entire field of view, including the area chosen for the zoomed in images.

As suggested by the reviewer, we have added the entire and enlarged images to the new Fig. 2D (p. 9, l.173-174).

Regarding the time points chosen to measure EGFR area and others. Why were the 1 mins and 2 mins time points chosen to examine EGFR-PIP2 and EGFR-GRB2 interactions, respectively? Is there evidence that these interactions peak at these time points? Alternatively, please provide evidence of their interactions at earlier time points (e.g. 15-30 seconds for EGFR-PIP2 and 1 mins for EGFR-GRB2).

SMT analysis indicated that the colocalization rate of EGFR and PI(4,5)P2 significantly decreased after 0.5 min of EGF stimulation. For EGFR and GRB2, the rate was reduced after 0.5 min of EGF stimulation. These results suggest that the binding of EGF to EGFR and the subsequent reaction occurred within 0.5 min under our experimental conditions. We have added the relevant data to Fig. S4 (p. 12, l.229-234). In addition, we have provided the bivariate H(r) values of EGFR and PI(4,5)P2 at 0.5 min after EGF stimulation, as obtained with SMLM analysis (Fig. 2B, green) (p. 9, l.166-167).

Please demonstrate with immunoblot the extent to which EGFR-EGFP construct can be stimulated by EGF (EGFR Y1068) vs. control cells.

As suggested by the reviewer, we have added the results to the new Fig. S3A. Although we reduced the expression level of EGFR-GFP for SMT, it was phosphorylated to the same level after EGF stimulation as the intrinsic EGFR in the parental cells (p. 11, l.201-203).

Fig. S3B (right) - how do the authors explain the apparent decrease in diffusion coefficient for the fast mobile fraction of EGFR. Presumably, these receptors are not engaged with ligand, so what is causing the decrease in diffusion coefficient?

It is not necessary to assume that the EGFR in the fast mobile fraction was EGF-free. In the Variational Bayesian-Hidden Markov Model analysis, all time points in the single trajectories were assigned to one of the three states, and we frequently observed state transitions within a single trajectory (Hiroshima et al. 2018). These transitions suggest that the difference in the mobility states is not caused solely by ligand binding, as differences in the membrane environment might also influence the mobility. Therefore, ligand-free and ligand-bound molecules have different diffusion coefficients in the same fast-mobile fraction.

Fig. 6A - it is unclear which method was used to probe pAKT (S473) inhibition by wortmannin. Please specify.

The cells were incubated overnight in serum-free medium, treated with 10 µM wortmannin for 1 h, and stimulated with 20 nM EGF. pAKT was detected using anti-AKT (S473) (#4060, Cell Signaling Technology) as the primary antibody. This is described in the Methods section (p. 30, l.591-592; p. 31, l.623-627).

Fig. 6E (middle) - The authors explain that wortmannin treatment causes PIP2 dispersal. This interpretation would be strengthened with a quantification, as another interpretation of the representative image is that wortmannin appears to reduce the abundance of PIP2. This discrepancy requires explanation. There also appears to be an increase in pEGFR Y1068 relative to control.

The extent of PI(4,5)P2 distribution after wortmannin treatment was not a focus of this study; consequently, we deleted the text "wortmannin treatment causes PI(4,5)P2 dispersal." In addition, we replaced the images and changed the pseudocolor to avoid the incorrect impression that the amount of pEGFR was increased in wortmannin-treated cells (Fig. 6E).

Please provide uncropped immunoblot images without contrast adjustment. Some immunoblots appear to have lane-to-lane differences in background.

For all immunoblot images in this study, we did not adjust the contrast or brightness of the original images. The figure source data files show the images before they were trimmed.

Given the conclusion on line 421 re: PI3K localization, can you provide data to support that this is the case (i.e. that PI3K acts at distinct sites on the membrane away from the specific PIP2-EGFR nanodomains)? This should be possible given the methods described in the manuscript.

We want to say that PLCγ hydrolyzes PI(4,5)P2 around EGFR, whereas PI3K phosphorylates PI(4,5)P2 around the heterodimer of ERBB3 and EGFR in the plasma membrane. We have added the results of the SMT analysis, which indicated that the colocalization rate of PLCγ-PI3K was lower than that of EGFR-PLCγ or EGFR-PI3K (Fig. S7E and S7F) (p. 18, l.366-369). We have also added this information to the Discussion section (p. 22, l.428-442).

Likewise, showing whether this specific pool of PIP2-EGFR nanodomains are within or away from the well-characterized EGFR-tetraspanin nanodomains would add value to the interpretation of the results. However, this reviewer notes that this would add significant experiments to the study and this could be considered in future studies.

We were also interested in the local content of PI(4,5)P2 in EGFR-tetraspanin nanodomains. In the revised Discussion section we state that further studies will reveal the relationship between these molecules (p. 24, l.465-468).

Please explicitly state whether statistical considerations were made for multiple comparisons in the methods.

As the reviewer suggested, we have added the description in the Methods section (p.38, l.753-754).

Reviewer #2

Minor comments:

• From the original paper (Rosenbloom et al), it seems that rsKame still requires photoactivation at 405 nm. Was the done here for superresolution imaging? It is not listed in the methods for rsKame.

Similar to the photoactivation of Dronpa (Mizuno et al., 2010), we photoactivated rsKame with a 488 nm laser for excitation and turning off the fluorescence, and we attributed this to the spontaneous recovery of the stochastic turning on of the fluorescence, instead of the illumination at 405 nm. We have added this information to the Methods section (p. 35, l.705-709).

• The stimulation conditions vary throughout, in both EGF concentration and time (1-5 min), possible differences due to various stimulation conditions should be discussed. Furthermore, superresolution samples were fixed after 1 min of EGF stimulation. The lack of EGFR reorganization may be due to the time required for EGF to diffuse to the adherent cell surface. Other superresolution imaging has demonstrated that EGFR forms oligomers on the cell after EGF stimulation (e.g., Mudumbi et al, Cell Reports 2024; Needham et al Nat Comm 2016). Comment if your results are consistent or not with these other works.

For SMT and SMLM analyses throughout this study, the cells were treated with 20 nM EGF, which is sufficiently above the dissociation constant of 2-6 nM (Sugiyama et al., 2023). In the revised manuscript, we examined the time course of colocalization between EGFR and PI(4,5)P2 or GRB2 (Fig. S4) (p. 12, l.229-234). The colocalization rate of EGFR and PI(4,5)P2 decreased significantly by 0.5 min after EGF stimulation and remained low for at least 5 min (Fig. S4A). Under the same experimental conditions, the colocalization rate of EGFR and GRB2 increased by 0.5 min after EGF stimulation and remained high for at least 5 min (Fig. S4B). These results suggest that the binding of EGF to EGFR and the subsequent reaction were almost in a steady state, at least during 0.5-5 min of EGF stimulation.

As noted by the reviewer, Mudumbi et al. showed that the cluster size of EGFR increased after EGF stimulation, which may be different from our results (Fig. 1B). However, their methods and aims were different from those of the SMLM analysis in this study. They used very sparse conditions for single-cluster analysis, whereas in SMLM we used dense conditions to detect the coaggregation of nanodomains, which may contain multiple clusters. We also used sparse conditions for single-molecule imaging (Fig. 4) and observed dimer/oligomer formation with an increase in the fluorescence signal in the single EGF spots. We observed a similar increase in the single-spot fluorescence signal in our previous studies. Our results are at least qualitatively consistent with those reported previously (Mudumbi et al., 2024; Needham et al., 2016). We have added this information to the Discussion section (p. 25, l.506-512).

• Single molecule tracking was performed at room temperature. At what temperature were the superresolution and western blot samples stimulated? Fluidity and organization of the plasma membrane is altered by temperature. Possible caveats should be discussed. If biochemistry was performed at 37 C, then EGFR signaling cannot be correlated between samples dues to faster activation at physiological temperature.

Cells were stimulated with EGF at 25{degree sign}C in all experiments. We have added this information to the Methods section (p. 27, l.525-526).

• Many experiments are performed using transient transfection, with no control for or quantification of expression level. The frequency of EGFR:domain overlap and colocalization during SMT could be dependent on the relative expression levels of proteins/lipid markers. Was this accounted for?

As noted by the reviewer, the expression levels of EGFR and probes differ among cells. To consider the differences in expression levels among cells, we measured the density of particles (particle number/cell area). We then normalized the original colocalization rates, which were calculated using our custom-made software (Yanagawa and Sako, 2021), to the densities of both the EGFR and the probes. We normalized all data for the colocalization rates and presented them as relative colocalization rates (p. 12, l. 220-223; p. 34, l. 669-673)

• Please describe why some experiments performed with HeLa EGFR knockout cells and other with CHO cells?

As noted by the reviewer, we used CHO-K1 cells in the experiment shown in Fig. 4, whereas HeLa cells were used in other experiments. We performed a similar experiment using HeLa cells and we obtained similar results. The results for HeLa cells are presented in Fig. S5A (p. 13, l.256-257).

• The author state that "...dimerized EGFR was mainly found in the immobile fraction..." How did they determine this? This interpretation of the SMT data is important for suggesting that PIP2 EGFR stabilization (Fig. 4), and should therefore be explain/justified.

As shown in Figs. 4A and S5A, SMT analysis revealed that EGFR monomers decreased and dimers increased in the immobile fraction of control CHO-K1 cells after EGF stimulation (Fig. 4A). The changes in oligomer size were smaller in the slow- and fast-mobile fractions than in the immobile fraction. In addition, the fraction size of the immobile state was not reduced after EGF stimulation (Fig. 3B). These results suggest that stable EGFR dimer/oligomers were mostly increased in the immobile fraction after EGF stimulation. In contrast, the colocalization rate in the slow-mobile and fast-mobile fractions decreased after EGF stimulation, whereas the rate in the immobile fraction did not change significantly (Fig. S3D). The PI(4,5)P2 probes employed in this study can detect PI(4,5)P2 near EGFR but they might not bind to PI(4,5)P2 associated with EGFR due to steric hindrance. It is plausible that concentrated PI(4,5)P2 molecules help to stabilize EGFR dimer/oligomers in the immobile fraction. However, we cannot exclude the possibility that the dimer/oligomers in the slow- and fast-mobile fractions were also stabilized by PI(4,5)P2, which was not detected by the PI(4,5)P2 probes. We have added this information to the Discussion section (p.25, l.487-500).

• In Fig 3, it is shown that the immobile fraction of EGFR increases with EGF, while PIP2 diffusion is unchanged. If PIP2 interacts with EGFR and stabilizes dimers, would you expect to also see an increase in the PIP2 immobile fraction?

Our results suggest that the interaction between EGFR and PI(4,5)P2 is transient (Fig. 3D) (p.11, l.218-220). Therefore, we did not observe an increase in the PI(4,5)P2 immobile fraction. We have added a movie to the revised manuscript. Movie1 shows the lateral colization of EGFR and PI(4,5)P2 before and after EGF stimulation. As mentioned above, despite the decrease in PI(4,5)P2 concentration after EGF stimulation, colocalization between EGFR and PI(4,5)P2 was maintained in the immobile fraction.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

The work presented in "Bilateral regulation of EGFR activity and local PI dynamics observed with superresolution microscopy" by Abe et al studies the role of PI(4,5)P2 in EGFR signaling. This is an important question since the interplay of lipids and membrane receptors is known to be important for signaling, but the underlying mechanisms are not fully understood. The authors use multicolor superresolution and single molecule tracking coupled with biochemical approaches to understand how EGFR and PIP2 interplay on the plasma membrane. This work focuses on the biophysical mechanism of EGFR signaling and will be relevant to journals in the areas of biophysics, cell biology, cell signaling and microscopy. Overall, this an important study that identifies PIP2 as playing a functional role in EGFR signaling. However, there are some caveats to the experimental conditions that need to be discussed.

Minor comments:

From the original paper (Rosenbloom et al), it seems that rsKame still requires photoactivation at 405 nm. Was the done here for superresolution imaging? It is not listed in the methods for rsKame.

The stimulation conditions vary throughout, in both EGF concentration and time (1-5 min), possible differences due to various stimulation conditions should be discussed. Furthermore, superresolution samples were fixed after 1 min of EGF stimulation. The lack of EGFR reorganization may be due to the time required for EGF to diffuse to the adherent cell surface. Other superresolution imaging has demonstrated that EGFR forms oligomers on the cell after EGF stimulation (e.g., Mudumbi et al, Cell Reports 2024; Needham et al Nat Comm 2016). Comment if your results are consistent or not with these other works.

Single molecule tracking was performed at room temperature. At what temperature were the superresolution and western blot samples stimulated? Fluidity and organization of the plasma membrane is altered by temperature. Possible caveats should be discussed. If biochemistry was performed at 37 C, then EGFR signaling cannot be correlated between samples dues to faster activation at physiological temperature.

Many experiments are performed using transient transfection, with no control for or quantification of expression level. The frequency of EGFR:domain overlap and colocalization during SMT could be dependent on the relative expression levels of proteins/lipid markers. Was this accounted for?

Please describe why some experiments performed with HeLa EGFR knockout cells and other with CHO cells?

The author state that "...dimerized EGFR was mainly found in the immobile fraction..." How did they determine this? This interpretation of the SMT data is important for suggesting that PIP2 EGFR stabilization (Fig. 4), and should therefore be explain/justified.

In Fig 3, it is shown that the immobile fraction of EGFR increases with EGF, while PIP2 diffusion is unchanged. If PIP2 interacts with EGFR and stabilizes dimers, would you expect to also see an increase in the PIP2 immobile fraction?

Significance

This study address an important question since the interplay of lipids and membrane receptors is known to be important for signaling, but the underlying mechanisms are not fully understood. The authors are able to make a conceptual advance in our understanding of EGFR biology by using advanced imaging techniques that allow for quantification of protein distribution and dynamics on intact cells. The elegant application of superresolution and single molecule tracking are important strengths of this work, while the clever use of receptor mutant and phosphates reveals novel insights. This work focuses on the biophysical mechanism of EGFR signaling and will be relevant to journals in the areas of biophysics, cell biology, cell signaling and microscopy.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

The epidermal growth factor receptor (EGFR) is a central regulator of cell function, with important roles in development as well as tissue homeostasis in adults. The upregulation of EGFR expression or activity drives tumor progression in several types of cancer. This study examines the regulation of EGFR activity by compartmentalization of EGFR into unique plasma membrane nanodomains demarked by enrichment of phosphatidylinositol-4,5-bisphosphate (PIP2) and phosphatidylserine (PS). To do so, EGFR was labeled via C-terminal fusion with rsKame and labeling of PIP2 with PLCdelta-PH fused with PAmCherry or PS with evectin-2-PH-HaloTag. This was examined following fixation, using SMLM and statistical analysis using Ripley's univariate H-function (to determine the size and distribution of each nanocluster) and Ripley's bivariate H-function (to determine the distribution of different nanoclusters relative to one another). This revealed that the overlap of EGFR and PS did not change following stimulation of EGF, but this stimulation did reduce the overlap of EGFR and PIP2.

These experiments in fixed cells were complemented with live cell single molecule imaging studies, tracking EGFR and PIP2. EGF stimulation increased the immobile fraction of EGFR, and decreased the overlap of EGFR and PIP2. Experiments with overexpression of the inositol 5-phosphatase synaptojanin and expression of mutants of EGFR (3RN) with a disrupted putative PIP2 binding site were used to assess the function of PIP2 in EGFR regulation. Expression of synaptojanin and mutation of EGFR (3RN) had similar results in that the confinement, dimerization, and c-terminal pY1068 of EGFR following EGF stimulation was altered. Perturbation of PIP2 availability by expression of a dominant interfering PLCgamma led to a reduction of the EGF-stimulated pT654 on EGFR, known to negatively regulate EGFR activation. From this emerges a model where clustering of EGFR with PIP2 nanodomains at the cell surface is required to support initial EGFR activation, and then the hydrolysis of PIP2 to generate DAG is required for eventual deactivation of EGFR.

The manuscript is by Abe et al. is well-written, presenting a compelling investigation into the spatiotemporal dynamics of PIP2 and its regulatory role in EGFR activity in living cells. The use of super-resolution single-molecule microscopy to visualize the interactions between EGFR and PIP2 nanodomains and single particle tracking methodologies are complimentary, yielding important insights and underscoring the manuscript's contribution to advancing our understanding in this area. The experimental workflow is sophisticated and novel and is strengthened by the careful consideration of critical controls throughout the manuscript. For example, studying the expected Grb2 localization with EGFR, showing the expected gain in spatial correlation of EGFR and Grb2 upon EGF stimulation strengthens the use of this workflow to study interaction of EGFR and other nanoclusters. The results not only enhance our understanding of the intricate mechanisms governing EGFR regulation by lipids but also highlight the importance of PIP2 in the modulation of EGFR dimerization and autophosphorylation. While the experiments conducted in the study are largely well done and of high quality, there are several outstanding issues that must be addressed before considering publication. It is essential that the authors provide further clarification on certain aspects of their methodology for replicability and transparency. Additionally, a more detailed discussion on the implications of their findings within the broader context of cell signaling and potential impacts on related research areas would enhance the manuscript's significance. This could also require some additional experiments to align the observations made in this study with that of previous studies, in particular as it relates to PIP2 dynamics and clustering. Addressing these concerns will not only strengthen the conclusions drawn but also provide the scientific community with a more comprehensive understanding of the complex interplay between putative PIP2 nanodomains and EGFR activity. The resolution of these issues is crucial for the manuscript to fully meet the publication standards of contributing novel and impactful insights to the field.

Major comments

1. In Fig. 3B, the dramatic (~20%) change in EGFR mobility upon EGF stimulation (i.e. from fast-mobile to confined) implies EGFR-EGF binding in excess of what is typically reported in the literature. How do the authors reconcile this and are there features of their cell model/analysis pipeline that are artificially contributing to this observation?

2. The detection of PIP2 nanodomains in the plasma membrane is somewhat controversial, especially using the PH domain of PLCd to detect PIP2 or using similar strategies. The recent study by Pacheco et al (JCB 2023, PMID: 36416724) uses a variety of measurements of fluorescent labeling of PIP2 by protein-based biosensors (similar to this study) and concludes that PIP2 is free diffusing in the plasma membrane, which would be inconsistent with PIP2 nanodomains. Pacheco et al propose that while engagement of PIP2 to effectors via the inositol headgroup may serve to immobilize this lipid, allowing clustering, the use of relatively large protein domains as fluorescent ligands that bind to the PIP2 headgroup to track PIP2, as performed here, displaces any intrinsic clustering mechanism, leading to free diffusion of PIP2. How can the clustering observed here for PIP2 be reconciled? Is it possible that additional, non lipid-based interactions function alongside PH domain-PIP2 interactions as a form of coincidence detection? It would be quite helpful to support the data shown in this manuscript with a different PIP2 binding domain, such as the Tubby domain used by Pacheco et al. It would not be necessary to repeat all experiments with such a complementary probe, but some key experiments that assess the apparent clustering of PIP2 would be important to consider repeating with this complementary PIP2 probe.

3. It is unclear if and how stimulation with EGF or overexpression of synaptojanin modulates PIP2 in the plasma membrane. Some studies found that EGF stimulation does not change PIP2 levels in the PM, including Delos Santos et al. (Mol Biol Cell. 2017, PMID: 28814502). Others found that the regulation of PIP2 levels in the plasma membrane is tightly controlled and the total levels of PIP2 can resist alterations of PIP2 by changes in lipid enzymes (Wills et al. JCS 2023, PMID: 37534432). Hence, it is not clear if the stimulation with EGF or the overexpression of synaptojanin changes plasma membrane PIP2 levels, or may only alter the nanoscale dynamnics of this lipid. If the effects of synaptojanin may be restricted to alterations of the nanoscale organization of PIP2 in the membrane, it would be important to consider that synaptojanin is strongly localized to clathrin-coated pits in the plasma membrane (e.g. Perera et al. PNAS 2007. PMID: 17158794), and that EGFR only exhibits strong recruitment to clathrin-coated pits following EGF stimulation, which would suggest that the non-ligand-bound EGFR is distant to synaptojanin-containing structures. Some consideration of the possibility of broad action of PIP2 depletion vs nanoscale localized effects by these treatments should be considered when interpreting the results of this study.

Minor comments 1. Please quantify the extent to which endogenous EGFR was knocked down. 2. Fig. 2C - please provide the entire field of view, including the area chosen for the zoomed in images. 3. Regarding the time points chosen to measure EGFR area and others. Why were the 1 mins and 2 mins time points chosen to examine EGFR-PIP2 and EGFR-GRB2 interactions, respectively? Is there evidence that these interactions peak at these time points? Alternatively, please provide evidence of their interactions at earlier time points (e.g. 15-30 seconds for EGFR-PIP2 and 1 mins for EGFR-GRB2). 4. Please demonstrate with immunoblot the extent to which EGFR-EGFP construct can be stimulated by EGF (EGFR Y1068) vs. control cells. 5. Fig. S3B (right) - how do the authors explain the apparent decrease in diffusion coefficient for the fast mobile fraction of EGFR. Presumably, these receptors are not engaged with ligand, so what is causing the decrease in diffusion coefficient? 6. Fig. 6A - it is unclear which method was used to probe pAKT (S473) inhibition by wortmannin. Please specify. 7. Fig. 6E (middle) - The authors explain that wortmannin treatment causes PIP2 dispersal. This interpretation would be strengthened with a quantification, as another interpretation of the representative image is that wortmannin appears to reduce the abundance of PIP2. This discrepancy requires explanation. There also appears to be an increase in pEGFR Y1068 relative to control. 8. Please provide uncropped immunoblot images without contrast adjustment. Some immunoblots appear to have lane-to-lane differences in background. 9. Given the conclusion on line 421 re: PI3K localization, can you provide data to support that this is the case (i.e. that PI3K acts at distinct sites on the membrane away from the specific PIP2-EGFR nanodomains)? This should be possible given the methods described in the manuscript. 10. Likewise, showing whether this specific pool of PIP2-EGFR nanodomains are within or away from the well-characterized EGFR-tetraspanin nanodomains would add value to the interpretation of the results. However, this reviewer notes that this would add significant experiments to the study and this could be considered in future studies. 11. Please explicitly state whether statistical considerations were made for multiple comparisons in the methods.

Significance

General strengths:

• This study examines an important question in the cell biology of a key regulator of cell physiology, EGFR. While the mobility and nanodomain clustering of EGFR has emerged as critical for the regulation of this receptor's ligand binding, dimerization, and downstream signaling output, there remains much to be understood about the nanoscale organization of EGFR relative to other signaling lipids and proteins in the plasma membrane. This study examines a novel link between nanodomains demarked by PIP2 and EGFR mobility and signaling.
• This study makes use of sophisticated multiplexed labeling of EGFR and lipids such as PIP2 and PS, along with super-resolution microscopy and single molecule imaging coupled to cutting-edge image quantification.
• Controls are generally well-considered and appropriate to support and validate the experimental workflows that are

General limitations:

• The labeling of PIP2 with fluorescently-labelled protein domains that recognize this lipid has some limitations, as described above in the comments.

Advance: This study fills a knowledge gap in how the central regulator of cell physiology, EGFR, is organized at the cell surface. It is well appreciated that EGFR exhibits confinement at the plasma membrane and that this receptor exhibits nanoscale clustering that regulates receptor function. However, the nature of the nanoscale clusters in which EGFR is detected in the ligand-bound and non-ligand-bound states, and how this defines receptor output is only beginning to be resolved. This study examined how clustering of the lipid PIP2 in the plasma membrane relates to EGFR clusters, and how this may functionally impact EGFR signaling. This fills a knowledge gap of the molecules within the plasma membrane that impact receptor nanoscale clustering and function. This study also advances how mechanisms that impact EGFR nanoscale organization also in turn affect signaling output, which provides compelling evidence for the significance of EGFR-PIP2 interactions (especially with the EGFR mutant that is predicted to have reduced PIP2 interactions).

Audience: This study will be of significant interest to fundamental cell biology researchers in general, and in particular those interested in cell signaling and lipid cell biology.

Reviewer expertise: This reviewer has expertise in cell biology of receptor signaling, phosphoinositides, single-particle tracking, and plasma membrane nanodomains.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

__Reviewer #1 (Evidence, reproducibility, and clarity (Required)):____ __ Summary: Viruses exploit host endoplasmic reticulum (ER)-resident chaperones to support new protein synthesis during viral replication. Here, Najarro et al. study the role of the ER-resident HSP70 family member Binding immunoglobulin protein (BiP) during lytic infection by the Kaposi's sarcoma-associated herpesvirus (KSHV). Using the established doxycycline-inducible lytic reactivation infection model cell line iSLK-BAC16, they showed that KSHV reactivation leads to an upregulation of total BiP protein but not RNA, and is independent of the unfolded protein response. siRNA knockdown or pharmacological inhibition by HA15 of BiP significantly reduced global viral gene expression and infectious virus production. The authors attribute this to at least the reduction of levels of the K1 gene which is required for efficient viral replication. Finally, they showed that HA15 has cytostatic activity in KSHV-transformed B cells and cytotoxic effects in KSHV-infected lymphatic endothelial cells arguing for BiP inhibition as a potential therapeutic strategy to treat KSHV-driven malignancies. The manuscript is well-written and the conclusions were generally supported by the data with a few exceptions below.

Major comments:

• They propose in lines 196-199 that the reduction of K1 from HA15 treatment partially explains the defect in virion production during lytic reactivation. I am not convinced that this statement is fully supported by their data. Reduction of K1 is likely a downstream consequence and not the cause of the inhibition of lytic replication.

We thank the reviewer for this comment. We conducted a more detailed analysis of our RNAseq data in iSLK.219 cells and confirmed the downregulation of the K1 transcript in latently infected cells treated with HA15 (See Fig 3 and Sup Fig 5). It is likely that the drop in transcript levels results from IRE1-mediated degradation in a recently-described process known as RIDDLE (IRE1-mediated RNA decay lacking endomotif), in which IRE1 depletes mRNAs1*. We have included this hypothesis in the discussion. *

Unfortunately, we cannot confirm the downregulation of K1 at the protein level in iSLK.219 cells since the antibodies are highly specific for K1 variants in PEL cells. To overcome this technical limitation, we conducted mass spectrometry analysis of the viral proteome from whole cell lysates of latent and lytic cells undergoing HA15 treatment. While we detect the expected global downregulation of viral proteins in lytic cells treated with HA15, we were not able to detect any viral proteins except for LANA in the latently infected cells, and our detection of several lytic proteins was limited. We speculate that the levels of latent viral proteins expressed in iSLK.219 cells are below the limits of detection of our assay, or that extensive modification of some of these viral proteins may hinder their detection. Due to these limitations, we decided not to include these data in the manuscript.

• Additionally, we note that the lower levels of K1 detected in latent iSLK.219 and TREx-BCBL-1 cells treated with HA15 may affect viral reactivation, which is consistent with findings from the Damania lab showing K1's crucial role in viral replication2.*

• *

• The quantification of the K1 blots in Fig. 3C only has n=2. With subtle differences by eye, large error bars, and no statistical analysis, it is hard to conclude here with confidence. *

We agree with the reviewer. We have moved the K1 blot to the Sup. Fig. 3E and adjusted the text accordingly.* *

• Like K1, ORF45, and K8.1 proteins are similarly decreased at 24 h in Fig. 2E, suggesting that the defect is upstream of K1. Does HA15 affect the amount of endogenous and/or transgene copy of RTA being produced (hence the broader effect in early gene expression at 24h?)?

• **To answer the Reviewer's query, we re-evaluated the impact of HA15 treatment on the activity of dox-inducible RTA. However, we think it is unlikely for HA15 to alter RTA activity since RTA does not enter the secretory pathway. *

To evaluate the activity of RTA in HA15 treated cells, we measured the expression of the viral episome-encoded RFP reporter, driven by the viral PAN promoter4*, at 24h post-doxycycline treatment of iSLK.219 cells. We compared the response of the PAN promoter to RTA in cells treated with or without HA15 at this early timepoint, to avoid any potential confounding effects stemming from elevated endogenous RTA expression at later times post-reactivation. We demonstrate that the levels of RFP in iSLK.219 cells treated with Dox are identical in presence or absence of HA15. This result, included in Sup. Fig. 3, indicates that the activity of RTA, crucial for initiating the lytic cycle in this context, is unaffected by BiP inhibition at early times post reactivation. *

• *

• K1 levels appear to decrease even during latency. Are the other latent proteins also affected? What about latent genome copies?

To address this query, we compared the Log2 fold change of latent transcripts (K1, K2, K12, ORF71, ORF72, ORF73) in the iSLK.219 RNAseq data set (Fig 3). Only the K1 transcript is reduced in HA15-treated cells. We include these data in Sup Fig 5A.

Regarding differences in genome copies, the consistent levels of the viral genome-encoded GFP in HA15 -/+ iSLK-219 cells (Sup Fig 3) indicate no significant changes in the levels of viral genomes at 24h post-treatment (prior to DNA replication). Previous studies by our lab and others show that knockdown of the major latency protein LANA results in episomal loss and lower levels of GFP5*. These results validate the use of GFP fluorescence in iSLK.219 as a proxy for genome copies. *

• *

• Fig. 3C was performed in a PEL cell line which they showed to enter cytostasis upon HA15 treatment (Fig. 5). This cytostasis (rather than K1) may be the root cause of the defect in viral replication as cells could be arrested at a different stage compared to the G2 requirement for lytic replication in PEL cells (Balisteri et al., PLOS Pathogens 2016, PMID: 26891221).

See point 2. below

• The cytostatic effect in PEL cell lines (Fig. 5) should be demonstrated using more direct methods that measure cell cycle (e.g. PI-BrdU).

We thank the reviewer for this comment. While more direct methods to measure the cell cycle stage affected by HA15 treatment will inform on its mechanism of action, these experiments lie outside of the scope of this manuscript and we consider are better suited for future studies on the anticancer properties of HA15. The data presented in Fig. 5 demonstrates that HA15 treatment of PEL cells causes a reduction in cell numbers without cytotoxicity, thus supporting our conclusion of a net negative effect on proliferation rather than cell death. The loss of our LN2 tank and PEL cell lines currently limits our ability to do these more detailed analyses. At the moment, we do not have an accurate estimate of how long it will take to replace these cell lines for our subsequent studies.

• *

• While having an uninfected B cell as a matched negative control for PEL is challenging, primary peripheral B cells (mostly of mature memory B cell stage) may not be the appropriate negative control. PEL cells are of plasma cell lineage which have unusually high protein translation and overloaded ER. The plasma cell lineage may explain the sensitivity of PEL cells to HA15. It is possible that HA15 may be toxic to plasma cells when used as a therapeutic agent.

We agree with the reviewer on the potential impact of HA15 on plasma cell viability. Indeed, HA15 (>2uM) treatment reduces the viability of plasma cell myeloma lines (NCI-H929 and U266 cells), substantiating its use as a potential anti-cancer drug6. Although HA15 has not been tested as a therapeutic agent in humans, studies in mice have demonstrated tolerability without evident toxicity, measured as normal body weight7*. The potential therapeutic application of HA15 for cancer warrants further investigation and is beyond the scope of our manuscript. *

• Does HA15 have cytostatic effects in uninfected or latently infected iSLK cells?

• *

We observed no cytostatic or cytotoxic effects in uninfected or latently infected iSLK cells exposed to up to 30uM of HA15. Although HA15 has been tested on various cancer types8*, it has not been evaluated in Renal Carcinoma Cells (RCC), the cellular background of iSLK.219 cells. The mechanism behind the resistance of these cells to HA15 eludes us, but its link to the cellular background of iSLK.219s merits exploration in future studies. *

Minor comments: 1. Consider changing the title of line 98 to specify cell type since BiP levels do not increase in BCBL-1 (Supp. Fig. 3).

• *

Revised in the manuscript

Fig. 3A may benefit from using z-scores instead of log2TPM so differences are more obvious per gene.

Since the data have already been collected, can the authors include both latent and lytic cells with and without HA15 treatment in Fig. 3A? It may give more information for the reader. *

*We have reanalyzed all the RNAseq data and included a z-score plot for all samples in Fig. 3. We also providing three new supplementary tables with the raw counts, the z-scores for viral genes, and the log2 of the normalized counts.

*

*Reviewer #1 (Significance (Required)):

Significance: Here, the authors convincingly demonstrate the proviral role of the ER chaperone BiP during KSHV reactivation. This manuscript will be relevant to researchers in the gammaherpesvirus field. Although the authors did present some interesting data, the scope is narrow, and mechanistic studies were not pursued that would have added more insight in BiP and/or KSHV biology. For instance, how do BiP protein levels increase during reactivation (is this at the level of RNA sequestration/export, translation, or protein stability?)? How does BiP promote lytic replication?

Field of expertise: KSHV, molecular and cell biology

*

* __Reviewer #2 (Evidence, reproducibility and clarity (Required)): __ Many viruses have complex relationships with cellular ER proteostasis machinery that remain poorly understood. Here Najarro, et al. report on studies of the oncogenic gammaherpesvirus KSHV. They report that the ER chaperone BiP is upregulated in epithelial cells during KSHV lytic replication. Unexpectedly, BiP upregulation is independent of the unfolded protein response, which stimulates transcriptional activation of BiP to meet the protein folding demand in the ER. Using a combination of genetic and pharmacologic approaches (CRISPRi and selective chemical inhibitor) they demonstrate that BiP inhibition interferes with the replication of diverse enveloped viruses including poxviruses and several herpesviruses, and reduces proliferation of KSHV-infected cells.

Figure-by-figure:

Fig. 1: This figure convincingly demonstrates the selective upregulation of BiP at the protein level during the course of KSHV lytic replication, and that KSHV late genes are dispensable for this upregulation. It further demonstrates that BiP is not upregulated at the mRNA level at all during KSHV infection, despite the fact that the UPR-dependent BiP mRNA upregulation pathway (presumably via ATF6 and IRE1) remains functional.

Fig. 2: This figure convincingly demonstrates that BiP ATPase activity is required to support KSHV lytic replication in both epithelial and B cell models on infection, even though it is also clear that BiP is not upregulated in the B cell model.

Fig. 3: This data demonstrates that steady-state levels of KSHV lytic gene products are reduced following HA15-treatment, whereas later gene expression was unaffected. As an interesting side note, v-IL6 bucks the trend of HA15-mediated downregulation of viral mRNA levels, suggesting that it may be regulated in a different manner. One thing that the authors may consider is the report from Drs. Yuan Chang and Patrick Moore (PMID: 12434062) that demonstrated that the v-IL6 gene is transactivated by type I interferon. Considering the poor replication of this virus during HA15 treatment, it may be valuable to investigate IFN production by these cells, and the extent to which it is impacted by inhibition of BiP ATPase activity.*

We thank the reviewer for bringing this report to our attention. We also found intriguing the specific transcriptional upregulation of IL6 in IFN-a treated BCP-1 cells. Although we see a dramatic upregulation of the vIL6 in HA15 treated cells, we still detect the expression of most viral genes, albeit at significantly lower levels than in untreated cells, which indicates that the viral transcriptional program in lytic+HA15 iSLK.219 cells is different from the one seen in IFN-treated BCP-1 cells. Preliminary analyses of the host transcriptome from our RNAseq results show the expression of several ISGs (OAS1, 2 and 3, IFI6, IFIT1, IFIT3, IFITM1) in lytic-untreated iSLK.219 cells, but not in those treated with HA15. Together, these observations substantiate the notion that there is no IFN-driven expression of vIL6 in HA15-treated iSLK.219 cells.

Fig. 4: This figure demonstrates that HA15 has broad, non-cytotoxic, antiviral activity against diverse enveloped viruses.

Figs. 5/6: These figure shows cytotoxic effects of HA15 on latently infected PEL cells, either solely infected with KSHV or co-infected with KSHV and EBV, whereas normal B cells were unaffected. HA15 was also cytotoxic to KSHV infected lymphatic endothelial cells.

**Referees cross-commenting**

I appreciate the insightful comments from Reviewer #1 and Reviewer #3. I think we are largely on the same page. The data is generally supportive of author's conclusions, with a few exceptions that are straightforward to address in revisions. The manuscript is limited in scope, which could also be addressed by additional experimentation if the authors are motivated to explore mechanism in greater depth. Of particular note is the lack of mechanistic insight into how BiP is upregulated at the protein level during lytic replication, if the mRNA is unchanged. The experimental approaches to this are straightforward.

*

*

We appreciate the reviewers' comments on the scope of our study. The mechanism of BiP upregulation remains an outstanding question for the following technical reasons: We hypothesized that the upregulation of BiP may depend on the IRES element present in its 5' UTR9. We tested this hypothesis by transfecting iSLK.219 cells with a bicistronic Renilla-(BiP)IRES-Firefly luciferase reporter from Licursi et. al10*. Unfortunately, for reasons that still elude us, our reactivation rates in transfected cells were consistently low in all of our experiments and therefore, we were not able to measure luciferase changes consistently and reliably. A potential workaround this technical limitation is to use a lentivirus-encoded IRES reporter to a lentiviral vector, as transduction of iSLK.219 cells does not alter viral reactivation, in our experience. At the moment, we do not have access to these reporters due to our lab's move to a different institution, and the first author of our study has started the next stage of their career. Therefore, we will not be able to pursue these experiments in a timely manner. *

• *

*As for the scope of this manuscript, even when the mechanism of BiP upregulation in KSHV infected cells remains unsolved, we consider that the broad-spectrum antiviral effect of BiP inhibition is an exciting finding that advances the field and benefits the virology community-the proteostasis network has been seldomly explored as a potential node for broad-spectrum antiviral intervention. Our results provide important proof-of-concept to continue the investigation of factors involved in protein synthesis, folding and transport as potential targets for the development of versatile broad-spectrum antivirals. *

Reviewer #2 (Significance (Required)):

Strengths: This is a well-written manuscript. The text and figures are clear and accurate and the methods are sufficiently informative that the study can be reproduced. The data generally supports the authors' conclusions. BiP appears to be a druggable target with minimal off-target cytotoxicity in normal, uninfected cells, although this study does not go beyond cell culture studies to validate in vivo.

Weaknesses: The study is somewhat limited in scope. The authors make the case for UPR transcription-independent upregulation of BiP during KSHV infection, and that late gene synthesis is dispensable, but the mechanism is not investigated further.

Point by point discussion:

Could an early KSHV gene product involved in this phenotype be identified by screening an ORF library or viral genome-wide CRISPRi screen?

The question of the viral protein responsible for the upregulation of BiP during lytic infection is indeed a fascinating one. However, we suspect that the mechanism may be not specifically directed to BiP, but rather general modulation of IRES-related translation. Identifying the gene product(s) affected and corroborating IRES involvement is a major undertaking and a long-term goal requiring considerable effort. These analyses are outside the scope of this manuscript, but we will pursue them in the future.

Or, beyond implicating viral factors in the mechanism of BiP upregulation, can some simple biochemical studies be performed to investigate BiP protein? Is the BiP mRNA more efficiently spliced and exported in KSHV infected cells?

Do alternative translation initiation mechanisms like eIF2A play a role in boosting BiP levels during infection?

What is the normal BiP protein turnover mechanism, and is this hindered during KSHV lytic replication? Is BiP AMPylation/de-AMPylation by FICD affected (PMID: 36041787)? These kinds of mechanistic studies are well within reach and would help extend the impact and interest to a broad audience.

We agree on the putative involvement of translation initiation factors like eIF2A on promoting the translation of BiP (see discussion). We tested the effect of siRNA-mediated KD of eIF2A on BiP expression and found that, interestingly, the levels of BiP rose above those of controls in latent iSLK.219 cells (Data included in the manuscript and the discussion has been modified accordingly). This finding aligns with previous reports suggesting that eIF2A may suppress IRES-mediated translation in yeast cells and in mammalian in vitro translation assays. Moreover, Starck et. al11, observed a 50% increase of endogenous BiP levels in HeLa cells transfected with siRNAs against eIF2A, supporting the IRES-suppressor role for eIF2A in mammalian cells. Future work will be required to address the role of eIF2A on BiP translation. These analyses are beyond the scope our manuscript.

Reviewer #3 (Evidence, reproducibility and clarity (Required)):

The manuscript by Najarro et al. investigates the contribution of BiP/GRP78 to double-stranded DNA virus infection, primarily focusing on the oncogenic gammaherpesvirus Kaposi's sarcoma-associated herpesvirus (KSHV). The authors observe that BiP expression is increased in lytic iSLK.219 cells as well as in KSHV-infected LECs. Interestingly, the authors data suggest a post-translational regulation of BiP in the iSLK.219 cells. Using various knockdown approaches and chemical inhibitors the authors demonstrate that inhibition of BiP impacts KSHV reactivation in multiple cells lines. Importantly, the authors also find that BiP inhibition can selectively kill KSHV-infected cells, while sparing primary B cells. Overall, this is a very well controlled and presented manuscript. My comments for the manuscript are minor, and largely cosmetic to aid the presentation of the data.

• Fig 1C, It would be ideal to show that PAA treatment did indeed prevent the virus from entering the late stage of gene expression.

*We have included an immunoblot for K8.1 in Figure 1C to confirm the effect of PFA on arresting the KSHV lytic cycle. *

Sup Fig2, should show KD efficiency of XBP1, same goes for ATF6.

• *

Sup. Fig. 2D shows the expression of XBP1s in NS vs. XBP1KD cells in the presence or absence of Tg. In Sup Fig. 2G we have also included a bar graph showing the efficiency of downregulation of ATF6 mRNA in the presence of the targeting sgRNA.

Sup Fig 3. It is interesting that the authors do not see increased BiP in TREx-BCBL1-RTA cells. A potential caveat is that lytic reactivation in TREx-BCBL1-RTA cells is not as efficient as in iSLK.219 cells. Therefore, it may simply be a result of the reduced population entering the lytic cycle. It may be worth adding a comment regarding this.

• Images of the microscopy for Figure 4 would be useful.

Images have been included in Fig. 4

• Add label of the cell types for Figure 5.

DONE

• Does HSV1, HCMV, or VacV increase BiP expression compared to mock-infected cells?

Yes, we have included a comment on this in the discussion

Reviewer #3 (Significance (Required)):

Overall, this is a very well controlled and presented manuscript.

• *

• *

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

The manuscript by Najarro et al. investigates the contribution of BiP/GRP78 to double-stranded DNA virus infection, primarily focusing on the oncogenic gammaherpesvirus Kaposi's sarcoma-associated herpesvirus (KSHV). The authors observe that BiP expression is increased in lytic iSLK.219 cells as well as in KSHV-infected LECs. Interestingly, the authors data suggest a post-translational regulation of BiP in the iSLK.219 cells. Using various knockdown approaches and chemical inhibitors the authors demonstrate that inhibition of BiP impacts KSHV reactivation in multiple cells lines. Importantly, the authors also find that BiP inhibition can selectively kill KSHV-infected cells, while sparing primary B cells. Overall, this is a very well controlled and presented manuscript. My comments for the manuscript are minor, and largely cosmetic to aid the presentation of the data.

• Fig 1C, It would be ideal to show that PAA treatment did indeed prevent the virus from entering the late stage of gene expression.
• Sup Fig2, should show KD efficiency of XBP1, same goes for ATF6.
• Sup Fig 3. It is interesting that the authors do not see increased BiP in TREx-BCBL1-RTA cells. A potential caveat is that lytic reactivation in TREx-BCBL1-RTA cells is not as efficient as in iSLK.219 cells. Therefore, it may simply be a result of the reduced population entering the lytic cycle. It may be worth adding a comment regarding this.
• Images of the microscopy for Figure 4 would be useful.
• Add label of the cell types for Figure 5.
• Does HSV1, HCMV, or VacV increase BiP expression compared to mock-infected cells?

Significance

Overall, this is a very well controlled and presented manuscript.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Many viruses have complex relationships with cellular ER proteostasis machinery that remain poorly understood. Here Najarro, et al. report on studies of the oncogenic gammaherpesvirus KSHV. They report that the ER chaperone BiP is upregulated in epithelial cells during KSHV lytic replication. Unexpectedly, BiP upregulation is independent of the unfolded protein response, which stimulates transcriptional activation of BiP to meet the protein folding demand in the ER. Using a combination of genetic and pharmacologic approaches (CRISPRi and selective chemical inhibitor) they demonstrate that BiP inhibition interferes with the replication of diverse enveloped viruses including poxviruses and several herpesviruses, and reduces proliferation of KSHV-infected cells.

Figure-by-figure:

Fig. 1: This figure convincingly demonstrates the selective upregulation of BiP at the protein level during the course of KSHV lytic replication, and that KSHV late genes are dispensable for this upregulation. It further demonstrates that BiP is not upregulated at the mRNA level at all during KSHV infection, despite the fact that the UPR-dependent BiP mRNA upregulation pathway (presumably via ATF6 and IRE1) remains functional.

Fig. 2: This figure convincingly demonstrates that BiP ATPase activity is required to support KSHV lytic replication in both epithelial and B cell models on infection, even though it is also clear that BiP is not upregulated in the B cell model.

Fig. 3: This data demonstrates that steady-state levels of KSHV lytic gene products are reduced following HA15-treatment, whereas later gene expression was unaffected. As an interesting side note, v-IL6 bucks the trend of HA15-mediated downregulation of viral mRNA levels, suggesting that it may be regulated in a different manner. One thing that the authors may consider is the report from Drs. Yuan Chang and Patrick Moore (PMID: 12434062) that demonstrated that the v-IL6 gene is transactivated by type I interferon. Considering the poor replication of this virus during HA15 treatment, it may be valuable to investigate IFN production by these cells, and the extent to which it is impacted by inhibition of BiP ATPase activity.

Fig. 4: This figure demonstrates that HA15 has broad, non-cytotoxic, antiviral activity against diverse enveloped viruses.

Figs. 5/6: These figure shows cytotoxic effects of HA15 on latently infected PEL cells, either solely infected with KSHV or co-infected with KSHV and EBV, whereas normal B cells were unaffected. HA15 was also cytotoxic to KSHV infected lymphatic endothelial cells.

Referees cross-commenting

I appreciate the insightful comments from Reviewer #1 and Reviewer #3. I think we are largely on the same page. The data is generally supportive of author's conclusions, with a few exceptions that are straightforward to address in revisions. The manuscript is limited in scope, which could also be addressed by additional experimentation if the authors are motivated to explore mechanism in greater depth. Of particular note is the lack of mechanistic insight into how BiP is upregulated at the protein level during lytic replication, if the mRNA is unchanged. The experimental approaches to this are straightforward.

Significance

Strengths: This is a well-written manuscript. The text and figures are clear and accurate and the methods are sufficiently informative that the study can be reproduced. The data generally supports the authors' conclusions. BiP appears to be a druggable target with minimal off-target cytotoxicity in normal, uninfected cells, although this study does not go beyond cell culture studies to validate in vivo.

Weaknesses: The study is somewhat limited in scope. The authors make the case for UPR transcription-independent upregulation of BiP during KSHV infection, and that late gene synthesis is dispensable, but the mechanism is not investigated further. Could an early KSHV gene product involved in this phenotype be identified by screening an ORF library or viral genome-wide CRISPRi screen? Or beyond implicating viral factors in the mechanism of BiP upregulation, can some simple biochemical studies be performed to investigate BiP protein? Is the BiP mRNA more efficiently spliced and exported in KSHV infected cells? Do alternative translation initiation mechanisms like eIF2A play a role in boosting BiP levels during infection? What is the normal BiP protein turnover mechanism, and is this hindered during KSHV lytic replication? Is BiP AMPylation/de-AMPylation by FICD affected (PMID: 36041787)? These kinds of mechanistic studies are well within reach and would help extend the impact and interest to a broad audience.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary:

Viruses exploit host endoplasmic reticulum (ER)-resident chaperones to support new protein synthesis during viral replication. Here, Najarro et al. study the role of the ER-resident HSP70 family member Binding immunoglobulin protein (BiP) during lytic infection by the Kaposi's sarcoma-associated herpesvirus (KSHV). Using the established doxycycline-inducible lytic reactivation infection model cell line iSLK-BAC16, they showed that KSHV reactivation leads to an upregulation of total BiP protein but not RNA, and is independent of the unfolded protein response. siRNA knockdown or pharmacological inhibition by HA15 of BiP significantly reduced global viral gene expression and infectious virus production. The authors attribute this to at least the reduction of levels of the K1 gene which is required for efficient viral replication. Finally, they showed that HA15 has cytostatic activity in KSHV-transformed B cells and cytotoxic effects in KSHV-infected lymphatic endothelial cells arguing for BiP inhibition as a potential therapeutic strategy to treat KSHV-driven malignancies. The manuscript is well-written and the conclusions were generally supported by the data with a few exceptions below.

Major comments:

1. They propose in lines 196-199 that the reduction of K1 from HA15 treatment partially explains the defect in virion production during lytic reactivation. I am not convinced that this statement is fully supported by their data. Reduction of K1 is likely a downstream consequence and not the cause of the inhibition of lytic replication. Consider revising this statement in light of my comments below:
   • a. The quantification of the K1 blots in Fig. 3C only has n=2. With subtle differences by eye, large error bars, and no statistical analysis, it is hard to draw conclusions from here with confidence.
   • b. Like K1, ORF45 and K8.1 proteins are similarly decreased at 24 h in Fig. 2E suggesting that the defect is upstream of K1. Does HA15 affect the amount of endogenous and/or transgene copy of RTA being produced (hence the broader effect in early gene expression at 24h?)?
   • c. K1 levels appear to decrease even during latency. Are the other latent proteins also affected? What about latent genome copies?
   • d. Fig. 3C was performed in a PEL cell line which they showed to enter cytostasis upon HA15 treatment (Fig. 5). This cytostasis (rather than K1) may be the root cause of the defect in viral replication as cells could be arrested at a different stage compared to the G2 requirement for lytic replication in PEL cells (Balisteri et al., PLOS Pathogens 2016, PMID: 26891221).
2. The cytostatic effect in PEL cell lines (Fig. 5) should be demonstrated using more direct methods that measure cell cycle (e.g. PI-BrdU).
3. While having an uninfected B cell as a matched negative control for PEL is challenging, primary peripheral B cells (mostly of mature memory B cell stage) may not be the appropriate negative control. PEL cells are of plasma cell lineage which have unusually high protein translation and overloaded ER. The plasma cell lineage may explain the sensitivity of PEL cells to HA15. It is possible that HA15 may be toxic to plasma cells when used as a therapeutic agent.
4. Does HA15 have cytostatic effects in uninfected or latently infected iSLK cells?

Minor comments:

1. Consider changing the title of line 98 to specify cell type since BiP levels do not increase in BCBL-1 (Supp. Fig. 3).
2. Fig. 3A may benefit from using z-scores instead of log2TPM so differences are more obvious per gene.
3. Since the data have already been collected, can the authors include both latent and lytic cells with and without HA15 treatment in Fig. 3A? It may give more information for the reader.

Significance

Here, the authors convincingly demonstrate the proviral role of the ER chaperone BiP during KSHV reactivation. This manuscript will be relevant to researchers in the gammaherpesvirus field. Although the authors did present some interesting data, the scope is narrow, and mechanistic studies were not pursued that would have added more insight in BiP and/or KSHV biology. For instance, how do BiP protein levels increase during reactivation (is this at the level of RNA sequestration/export, translation, or protein stability?)? How does BiP promote lytic replication?

Field of expertise: KSHV, molecular and cell biology