l'écriture de l'image du bas est peu intelligible. Je suggère de reéquilibrer les deux figurés en réduisant la taille de la time line et en augmentant celle de l'image avec les bulles

source : Violaine Louvet, Grégory Miura. Introduction sur le code source, les logiciels. Accompagner la préservation et la diffusion des logiciels dans les établissements, Média Normandie; ADBU; Software Heritage, May 2023, Visioconférence, France. ⟨hal-04102897⟩

This code uses multiple independent if statements instead of elif, so all conditions are checked and the grade gets overwritten multiple times.

work on specific parts of your code only. I recognise this wandering even in my small projects.

in software dev you need a sense of how good the code needs to be in comparison to use case / context / impact of failure time available

This line highlights the key advantage of free and open-source operating systems: access to source code, which allows users to study, modify, and redistribute the software.

Suggested Improvement: The hyperlinks are added at the bottom in a smaller font size whereas they could just be embedded in the headings itself considering they are the same words. It would help make the webpage more robust by having a clean code and smoother transition between pages.

Gravierter Metall-Plakette für Stall, Voliere, Käfig oder Terrarium. Der QR-Code führt natürlich ebenfalls direkt zum Patenkontakt.

Gravierter Metall-Anhänger für Hunde- oder Aufschub-Marke für Katzen-Halsband. Der QR-Code führt direkt zu den hinterlegten Informationen und dem Patenkontakt.

Ein QR-Code, die Notfallkarte & das digitale Register sorgen dafür, dass Finder oder Behörden direkt wissen, wen sie kontaktieren müssen.

Der Satz "Dein Tier bleibt keine Minute unversorgt" kann über die 3 Kästen als Überschrift gesetzt werden.

Author response:

The following is the authors’ response to the original reviews.

eLife Assessment

This useful study presents Altair-LSFM, a solid and well-documented implementation of a light-sheet fluorescence microscope (LSFM) designed for accessibility and cost reduction. While the approach offers strengths such as the use of custom-machined baseplates and detailed assembly instructions, its overall impact is limited by the lack of live-cell imaging capabilities and the absence of a clear, quantitative comparison to existing LSFM platforms. As such, although technically competent, the broader utility and uptake of this system by the community may be limited.

We thank the editors and reviewers for their thoughtful evaluation of our work and for recognizing the technical strengths of the Altair-LSFM platform, including the custom-machined baseplates and detailed documentation provided to promote accessibility and reproducibility. Below, we provide point-by-point responses to each referee comment. In the process, we have significantly revised the manuscript to include live-cell imaging data and a quantitative evaluation of imaging speed. We now more explicitly describe the different variants of lattice light-sheet microscopy—highlighting differences in their illumination flexibility and image acquisition modes—and clarify how Altair-LSFM compares to each. We further discuss challenges associated with the 5 mm coverslip and propose practical strategies to overcome them. Additionally, we outline cost-reduction opportunities, explain the rationale behind key equipment selections, and provide guidance for implementing environmental control. Altogether, we believe these additions have strengthened the manuscript and clarified both the capabilities and limitations of AltairLSFM.

Public Reviews:

Reviewer #1 (Public review): 

Summary: 

The article presents the details of the high-resolution light-sheet microscopy system developed by the group. In addition to presenting the technical details of the system, its resolution has been characterized and its functionality demonstrated by visualizing subcellular structures in a biological sample.

Strengths: 

(1) The article includes extensive supplementary material that complements the information in the main article.

(2) However, in some sections, the information provided is somewhat superficial.

We thank the reviewer for their thoughtful assessment and for recognizing the strengths of our manuscript, including the extensive supplementary material. Our goal was to make the supplemental content as comprehensive and useful as possible. In addition to the materials provided with the manuscript, our intention is for the online documentation (available at thedeanlab.github.io/altair) to serve as a living resource that evolves in response to user feedback. We would therefore greatly appreciate the reviewer’s guidance on which sections were perceived as superficial so that we can expand them to better support readers and builders of the system.

Weaknesses:

(1) Although a comparison is made with other light-sheet microscopy systems, the presented system does not represent a significant advance over existing systems. It uses high numerical aperture objectives and Gaussian beams, achieving resolution close to theoretical after deconvolution. The main advantage of the presented system is its ease of construction, thanks to the design of a perforated base plate.

We appreciate the reviewer’s assessment and the opportunity to clarify our intent. Our primary goal was not to introduce new optical functionality beyond that of existing high-performance light-sheet systems, but rather to substantially reduce the barrier to entry for non-specialist laboratories. Many open-source implementations, such as OpenSPIM, OpenSPIN, and Benchtop mesoSPIM, similarly focused on accessibility and reproducibility rather than introducing new optical modalities, yet have had a measureable impact on the field by enabling broader community participation. Altair-LSFM follows this tradition, providing sub-cellular resolution performance comparable to advanced systems like LLSM, while emphasizing reproducibility, ease of construction through a precision-machined baseplate, and comprehensive documentation to facilitate dissemination and adoption.

(2) Using similar objectives (Nikon 25x and Thorlabs 20x), the results obtained are similar to those of the LLSM system (using a Gaussian beam without laser modulation). However, the article does not mention the difficulties of mounting the sample in the implemented configuration.

We appreciate the reviewer’s comment and agree that there are practical challenges associated with handling 5 mm diameter coverslips in this configuration. In the revised manuscript, we now explicitly describe these challenges and provide practical solutions. Specifically, we highlight the use of a custommachined coverslip holder designed to simplify mounting and handling, and we direct readers to an alternative configuration using the Zeiss W Plan-Apochromat 20×/1.0 objective, which eliminates the need for small coverslips altogether.

(3) The authors present a low-cost, open-source system. Although they provide open source code for the software (navigate), the use of proprietary electronics (ASI, NI, etc.) makes the system relatively expensive. Its low cost is not justified.

We appreciate the reviewer’s perspective and understand the concern regarding the use of proprietary control hardware such as the ASI Tiger Controller and NI data acquisition cards. Our decision to use these components was intentional: relying on a unified, professionally supported and maintained platform minimizes complexity associated with sourcing, configuring, and integrating hardware from multiple vendors, thereby reducing non-financial barriers to entry for non-specialist users.

Importantly, these components are not the primary cost driver of Altair-LSFM (they represent roughly 18% of the total system cost). Nonetheless, for individuals where the price is prohibitive, we also outline several viable cost-reduction options in the revised manuscript (e.g., substituting manual stages, omitting the filter wheel, or using industrial CMOS cameras), while discussing the trade-offs these substitutions introduce in performance and usability. These considerations are now summarized in Supplementary Note 1, which provides a transparent rationale for our design and cost decisions.

Finally, we note that even with these professional-grade components, Altair-LSFM remains substantially less expensive than commercial systems offering comparable optical performance, such as LLSM implementations from Zeiss or 3i.

(4) The fibroblast images provided are of exceptional quality. However, these are fixed samples. The system lacks the necessary elements for monitoring cells in vivo, such as temperature or pH control.

We thank the reviewer for their positive comment regarding the quality of our data. As noted, the current manuscript focuses on validating the optical performance and resolution of the system using fixed specimens to ensure reproducibility and stability.

We fully agree on the importance of environmental control for live-cell imaging. In the revised manuscript, we now describe in detail how temperature regulation can be achieved using a custom-designed heated sample chamber, accompanied by detailed assembly instructions on our GitHub repository and summarized in Supplementary Note 2. For pH stabilization in systems lacking a 5% CO₂ atmosphere, we recommend supplementing the imaging medium with 10–25 mM HEPES buffer. Additionally, we include new live-cell imaging data demonstrating that Altair-LSFM supports in vitro time-lapse imaging of dynamic cellular processes under controlled temperature conditions.

Reviewer #2 (Public review): 

Summary: 

The authors present Altair-LSFM (Light Sheet Fluorescence Microscope), a high-resolution, open-source microscope, that is relatively easy to align and construct and achieves sub-cellular resolution. The authors developed this microscope to fill a perceived need that current open-source systems are primarily designed for large specimens and lack sub-cellular resolution or are difficult to construct and align, and are not stable. While commercial alternatives exist that offer sub-cellular resolution, they are expensive. The authors' manuscript centers around comparisons to the highly successful lattice light-sheet microscope, including the choice of detection and excitation objectives. The authors thus claim that there remains a critical need for high-resolution, economical, and easy-to-implement LSFM systems. 

We thank the reviewer for their thoughtful summary. We agree that existing open-source systems primarily emphasize imaging of large specimens, whereas commercial systems that achieve sub-cellular resolution remain costly and complex. Our aim with Altair-LSFM was to bridge this gap—providing LLSM-level performance in a substantially more accessible and reproducible format. By combining high-NA optics with a precision-machined baseplate and open-source documentation, Altair offers a practical, high-resolution solution that can be readily adopted by non-specialist laboratories.

Strengths: 

The authors succeed in their goals of implementing a relatively low-cost (~ USD 150K) open-source microscope that is easy to align. The ease of alignment rests on using custom-designed baseplates with dowel pins for precise positioning of optics based on computer analysis of opto-mechanical tolerances, as well as the optical path design. They simplify the excitation optics over Lattice light-sheet microscopes by using a Gaussian beam for illumination while maintaining lateral and axial resolutions of 235 and 350 nm across a 260-um field of view after deconvolution. In doing so they rest on foundational principles of optical microscopy that what matters for lateral resolution is the numerical aperture of the detection objective and proper sampling of the image field on to the detection, and the axial resolution depends on the thickness of the light-sheet when it is thinner than the depth of field of the detection objective. This concept has unfortunately not been completely clear to users of high-resolution light-sheet microscopes and is thus a valuable demonstration. The microscope is controlled by an open-source software, Navigate, developed by the authors, and it is thus foreseeable that different versions of this system could be implemented depending on experimental needs while maintaining easy alignment and low cost. They demonstrate system performance successfully by characterizing their sheet, point-spread function, and visualization of sub-cellular structures in mammalian cells, including microtubules, actin filaments, nuclei, and the Golgi apparatus.

We thank the reviewer for their thoughtful and generous assessment of our work. We are pleased that the manuscript’s emphasis on fundamental optical principles, design rationale, and practical implementation was clearly conveyed. We agree that Altair’s modular and accessible architecture provides a strong foundation for future variants tailored to specific experimental needs. To facilitate this, we have made all Zemax simulations, CAD files, and build documentation openly available on our GitHub repository, enabling users to adapt and extend the system for diverse imaging applications.

Weaknesses:

There is a fixation on comparison to the first-generation lattice light-sheet microscope, which has evolved significantly since then:

(1) The authors claim that commercial lattice light-sheet microscopes (LLSM) are "complex, expensive, and alignment intensive", I believe this sentence applies to the open-source version of LLSM, which was made available for wide dissemination. Since then, a commercial solution has been provided by 3i, which is now being used in multiple cores and labs but does require routine alignments. However, Zeiss has also released a commercial turn-key system, which, while expensive, is stable, and the complexity does not interfere with the experience of the user. Though in general, statements on ease of use and stability might be considered anecdotal and may not belong in a scientific article, unreferenced or without data.

We thank the reviewer for this thoughtful and constructive comment. We have revised the manuscript to more clearly distinguish between the original open-source implementation of LLSM and subsequent commercial versions by 3i and ZEISS. The revised Introduction and Discussion now explicitly note that while open-source and early implementations of LLSM can require expert alignment and maintenance, commercial systems—particularly the ZEISS Lattice Lightsheet 7—are designed for automated operation and stable, turn-key use, albeit at higher cost and with limited modifiability. We have also moderated earlier language regarding usability and stability to avoid anecdotal phrasing.

We also now provide a more objective proxy for system complexity: the number of optical elements that require precise alignment during assembly and maintenance thereafter. The original open-source LLSM setup includes approximately 29 optical components that must each be carefully positioned laterally, angularly, and coaxially along the optical path. In contrast, the first-generation Altair-LSFM system contains only nine such elements. By this metric, Altair-LSFM is considerably simpler to assemble and align, supporting our overarching goal of making high-resolution light-sheet imaging more accessible to non-specialist laboratories.

(2) One of the major limitations of the first generation LLSM was the use of a 5 mm coverslip, which was a hinderance for many users. However, the Zeiss system elegantly solves this problem, and so does Oblique Plane Microscopy (OPM), while the Altair-LSFM retains this feature, which may dissuade widespread adoption. This limitation and how it may be overcome in future iterations is not discussed.

We thank the reviewer for this helpful comment. We agree that the use of 5 mm diameter coverslips, while enabling high-NA imaging in the current Altair-LSFM configuration, may pose a practical limitation for some users. We now discuss this more explicitly in the revised manuscript. Specifically, we note that replacing the detection objective provides a straightforward solution to this constraint. For example, as demonstrated by Moore et al. (Lab Chip, 2021), pairing the Zeiss W Plan-Apochromat 20×/1.0 detection objective with the Thorlabs TL20X-MPL illumination objective allows imaging beyond the physical surfaces of both objectives, eliminating the need for small-format coverslips. In the revised text, we propose this modification as an accessible path toward greater compatibility with conventional sample mounting formats. We also note in the Discussion that Oblique Plane Microscopy (OPM) inherently avoids such nonstandard mounting requirements and, owing to its single-objective architecture, is fully compatible with standard environmental chambers.

(3) Further, on the point of sample flexibility, all generations of the LLSM, and by the nature of its design, the OPM, can accommodate live-cell imaging with temperature, gas, and humidity control. It is unclear how this would be implemented with the current sample chamber. This limitation would severely limit use cases for cell biologists, for which this microscope is designed. There is no discussion on this limitation or how it may be overcome in future iterations.

We thank the reviewer for this important observation and agree that environmental control is critical for live-cell imaging applications. It is worth noting that the original open-source LLSM design, as well as the commercial version developed by 3i, provided temperature regulation but did not include integrated control of CO2 or humidity. Despite this limitation, these systems have been widely adopted and have generated significant biological insights. We also acknowledge that both OPM and the ZEISS implementation of LLSM offer clear advantages in this respect, providing compatibility with standard commercial environmental chambers that support full regulation of temperature, CO₂, and humidity.

In the revised manuscript, we expand our discussion of environmental control in Supplementary Note 2, where we describe the Altair-LSFM chamber design in more detail and discuss its current implementation of temperature regulation and HEPES-based pH stabilization. Additionally, the Discussion now explicitly notes that OPM avoids the challenges associated with non-standard sample mounting and is inherently compatible with conventional environmental enclosures.

(4) The authors' comparison to LLSM is constrained to the "square" lattice, which, as they point out, is the most used optical lattice (though this also might be considered anecdotal). The LLSM original design, however, goes far beyond the square lattice, including hexagonal lattices, the ability to do structured illumination, and greater flexibility in general in terms of light-sheet tuning for different experimental needs, as well as not being limited to just sample scanning. Thus, the Alstair-LSFM cannot compare to the original LLSM in terms of versatility, even if comparisons to the resolution provided by the square lattice are fair.

We agree that the original LLSM design offers substantially greater flexibility than what is reflected in our initial comparison, including the ability to generate multiple lattice geometries (e.g., square and hexagonal), operate in structured illumination mode, and acquire volumes using both sample- and lightsheet–scanning strategies. To address this, we now include Supplementary Note 3 that provides a detailed overview of the illumination modes and imaging flexibility afforded by the original LLSM implementation, and how these capabilities compare to both the commercial ZEISS Lattice Lightsheet 7 and our AltairLSFM system. In addition, we have revised the discussion to explicitly acknowledge that the original LLSM could operate in alternative scan strategies beyond sample scanning, providing greater context for readers and ensuring a more balanced comparison.

(5) There is no demonstration of the system's live-imaging capabilities or temporal resolution, which is the main advantage of existing light-sheet systems.

In the revised manuscript, we now include a demonstration of live-cell imaging to directly validate AltairLSFM’s suitability for dynamic biological applications. We also explicitly discuss the temporal resolution of the system in the main text (see Optoelectronic Design of Altair-LSFM), where we detail both software- and hardware-related limitations. Specifically, we evaluate the maximum imaging speed achievable with Altair-LSFM in conjunction with our open-source control software, navigate.

For simplicity and reduced optoelectronic complexity, the current implementation powers the piezo through the ASI Tiger Controller, which modestly reduces its bandwidth. Nonetheless, for a 100 µm stroke typical of light-sheet imaging, we achieved sufficient performance to support volumetric imaging at most biologically relevant timescales. These results, along with additional discussion of the design trade-offs and performance considerations, are now included in the revised manuscript and expanded upon in the supplementary material.

While the microscope is well designed and completely open source, it will require experience with optics, electronics, and microscopy to implement and align properly. Experience with custom machining or soliciting a machine shop is also necessary. Thus, in my opinion, it is unlikely to be implemented by a lab that has zero prior experience with custom optics or can hire someone who does. Altair-LSFM may not be as easily adaptable or implementable as the authors describe or perceive in any lab that is interested, even if they can afford it. The authors indicate they will offer "workshops," but this does not necessarily remove the barrier to entry or lower it, perhaps as significantly as the authors describe.

We appreciate the reviewer’s perspective and agree that building any high-performance custom microscope—Altair-LSFM included—requires a basic understanding of (or willingness to learn) optics, electronics, and instrumentation. Such a barrier exists for all open-source microscopes, and our goal is not to eliminate this requirement entirely but to substantially reduce the technical and logistical challenges that typically accompany the construction of custom light-sheet systems.

Importantly, no machining experience or in-house fabrication capabilities are required. Users can simply submit the provided CAD design files and specifications directly to commercial vendors for fabrication. We have made this process as straightforward as possible by supplying detailed build instructions, recommended materials, and vendor-ready files through our GitHub repository. Our dissemination strategy draws inspiration from other successful open-source projects such as mesoSPIM, which has seen widespread adoption—over 30 implementations worldwide—through a similar model of exhaustive documentation, open-source software, and community support via user meetings and workshops.

We also recognize that documentation alone cannot fully replace hands-on experience. To further lower barriers to adoption, we are actively working with commercial vendors to streamline procurement and assembly, and Altair-LSFM is supported by a Biomedical Technology Development and Dissemination (BTDD) grant that provides resources for hosting workshops, offering real-time community support, and developing supplementary training materials.

In the revised manuscript, we now expand the Discussion to explicitly acknowledge these implementation considerations and to outline our ongoing efforts to support a broad and diverse user base, ensuring that laboratories with varying levels of technical expertise can successfully adopt and maintain the Altair-LSFM platform.

There is a claim that this design is easily adaptable. However, the requirement of custom-machined baseplates and in silico optimization of the optical path basically means that each new instrument is a new design, even if the Navigate software can be used. It is unclear how Altair-LSFM demonstrates a modular design that reduces times from conception to optimization compared to previous implementations.

We thank the reviewer for this insightful comment and agree that our original language regarding adaptability may have overstated the degree to which Altair-LSFM can be modified without prior experience. It was not our intention to imply that the system can be easily redesigned by users with limited technical background. Meaningful adaptations of the optical or mechanical design do require expertise in optical layout, optomechanical design, and alignment.

That said, for laboratories with such expertise, we aim to facilitate modifications by providing comprehensive resources—including detailed Zemax simulations, complete CAD models, and alignment documentation. These materials are intended to reduce the development burden for expert users seeking to tailor the system to specific experimental requirements, without necessitating a complete re-optimization of the optical path from first principles.

In the revised manuscript, we clarify this point and temper our language regarding adaptability to better reflect the realistic scope of customization. Specifically, we now state in the Discussion: “For expert users who wish to tailor the instrument, we also provide all Zemax illumination-path simulations and CAD files, along with step-by-step optimization protocols, enabling modification and re-optimization of the optical system as needed.” This revision ensures that readers clearly understand that Altair-LSFM is designed for reproducibility and straightforward assembly in its default configuration, while still offering the flexibility for modification by experienced users.

Reviewer #3 (Public review):

Summary: 

This manuscript introduces a high-resolution, open-source light-sheet fluorescence microscope optimized for sub-cellular imaging. The system is designed for ease of assembly and use, incorporating a custommachined baseplate and in silico optimized optical paths to ensure robust alignment and performance. The authors demonstrate lateral and axial resolutions of ~235 nm and ~350 nm after deconvolution, enabling imaging of sub-diffraction structures in mammalian cells. The important feature of the microscope is the clever and elegant adaptation of simple gaussian beams, smart beam shaping, galvo pivoting and high NA objectives to ensure a uniform thin light-sheet of around 400 nm in thickness, over a 266 micron wide Field of view, pushing the axial resolution of the system beyond the regular diffraction limited-based tradeoffs of light-sheet fluorescence microscopy. Compelling validation using fluorescent beads and multicolor cellular imaging highlights the system's performance and accessibility. Moreover, a very extensive and comprehensive manual of operation is provided in the form of supplementary materials. This provides a DIY blueprint for researchers who want to implement such a system.

We thank the reviewer for their thoughtful and positive assessment of our work. We appreciate their recognition of Altair-LSFM’s design and performance, including its ability to achieve high-resolution, imaging throughout a 266-micron field of view. While Altair-LSFM approaches the practical limits of diffraction-limited performance, it does not exceed the fundamental diffraction limit; rather, it achieves near-theoretical resolution through careful optical optimization, beam shaping, and alignment. We are grateful for the reviewer’s acknowledgment of the accessibility and comprehensive documentation that make this system broadly implementable.

Strengths:

(1) Strong and accessible technical innovation: With an elegant combination of beam shaping and optical modelling, the authors provide a high-resolution light-sheet system that overcomes the classical light-sheet tradeoff limit of a thin light-sheet and a small field of view. In addition, the integration of in silico modelling with a custom-machined baseplate is very practical and allows for ease of alignment procedures. Combining these features with the solid and super-extensive guide provided in the supplementary information, this provides a protocol for replicating the microscope in any other lab.

(2) Impeccable optical performance and ease of mounting of samples: The system takes advantage of the same sample-holding method seen already in other implementations, but reduces the optical complexity.

At the same time, the authors claim to achieve similar lateral and axial resolution to Lattice-light-sheet microscopy (although without a direct comparison (see below in the "weaknesses" section). The optical characterization of the system is comprehensive and well-detailed. Additionally, the authors validate the system imaging sub-cellular structures in mammalian cells.

(3) Transparency and comprehensiveness of documentation and resources: A very detailed protocol provides detailed documentation about the setup, the optical modeling, and the total cost.

We thank the reviewer for their thoughtful and encouraging comments. We are pleased that the technical innovation, optical performance, and accessibility of Altair-LSFM were recognized. Our goal from the outset was to develop a diffraction-limited, high-resolution light-sheet system that balances optical performance with reproducibility and ease of implementation. We are also pleased that the use of precisionmachined baseplates was recognized as a practical and effective strategy for achieving performance while maintaining ease of assembly.

Weaknesses: 

(1) Limited quantitative comparisons: Although some qualitative comparison with previously published systems (diSPIM, lattice light-sheet) is provided throughout the manuscript, some side-by-side comparison would be of great benefit for the manuscript, even in the form of a theoretical simulation. While having a direct imaging comparison would be ideal, it's understandable that this goes beyond the interest of the paper; however, a table referencing image quality parameters (taken from the literature), such as signalto-noise ratio, light-sheet thickness, and resolutions, would really enhance the features of the setup presented. Moreover, based also on the necessity for optical simplification, an additional comment on the importance/difference of dual objective/single objective light-sheet systems could really benefit the discussion.

In the revised manuscript, we have significantly expanded our discussion of different light-sheet systems to provide clearer quantitative and conceptual context for Altair-LSFM. These comparisons are based on values reported in the literature, as we do not have access to many of these instruments (e.g., DaXi, diSPIM, or commercial and open-source variants of LLSM), and a direct experimental comparison is beyond the scope of this work.

We note that while quantitative parameters such as signal-to-noise ratio are important, they are highly sample-dependent and strongly influenced by imaging conditions, including fluorophore brightness, camera characteristics, and filter bandpass selection. For this reason, we limited our comparison to more general image-quality metrics—such as light-sheet thickness, resolution, and field of view—that can be reliably compared across systems.

Finally, per the reviewer’s recommendation, we have added additional discussion clarifying the differences between dual-objective and single-objective light-sheet architectures, outlining their respective strengths, limitations, and suitability for different experimental contexts.

(2) Limitation to a fixed sample: In the manuscript, there is no mention of incubation temperature, CO₂ regulation, Humidity control, or possible integration of commercial environmental control systems. This is a major limitation for an imaging technique that owes its popularity to fast, volumetric, live-cell imaging of biological samples.

We fully agree that environmental control is critical for live-cell imaging applications. In the revised manuscript, we now describe the design and implementation of a temperature-regulated sample chamber in Supplementary Note 2, which maintains stable imaging conditions through the use of integrated heating elements and thermocouples. This approach enables precise temperature control while minimizing thermal gradients and optical drift. For pH stabilization, we recommend the use of 10–25 mM HEPES in place of CO₂ regulation, consistent with established practice for most light-sheet systems, including the initial variant of LLSM. Although full humidity and CO₂ control are not readily implemented in dual-objective configurations, we note that single-objective designs such as OPM are inherently compatible with commercial environmental chambers and avoid these constraints. Together, these additions clarify how environmental control can be achieved within Altair-LSFM and situate its capabilities within the broader LSFM design space.

(3) System cost and data storage cost: While the system presented has the advantage of being opensource, it remains relatively expensive (considering the 150k without laser source and optical table, for example). The manuscript could benefit from a more direct comparison of the performance/cost ratio of existing systems, considering academic settings with budgets that most of the time would not allow for expensive architectures. Moreover, it would also be beneficial to discuss the adaptability of the system, in case a 30k objective could not be feasible. Will this system work with different optics (with the obvious limitations coming with the lower NA objective)? This could be an interesting point of discussion. Adaptability of the system in case of lower budgets or more cost-effective choices, depending on the needs.

We agree that cost considerations are critical for adoption in academic environments. We would also like to clarify that the quoted $150k includes the optical table and laser source. In the revised manuscript, Supplementary Note 1 now includes an expanded discussion of cost–performance trade-offs and potential paths for cost reduction.

Last, not much is said about the need for data storage. Light-sheet microscopy's bottleneck is the creation of increasingly large datasets, and it could be beneficial to discuss more about the storage needs and the quantity of data generated.

In the revised manuscript, we now include Supplementary Note 4, which provides a high-level discussion of data storage needs, approximate costs, and practical strategies for managing large datasets generated by light-sheet microscopy. This section offers general guidance—including file-format recommendations, and cost considerations—but we note that actual costs will vary by institution and contractual agreements.

Conclusion:

Altair-LSFM represents a well-engineered and accessible light-sheet system that addresses a longstanding need for high-resolution, reproducible, and affordable sub-cellular light-sheet imaging. While some aspects-comparative benchmarking and validation, limitation for fixed samples-would benefit from further development, the manuscript makes a compelling case for Altair-LSFM as a valuable contribution to the open microscopy scientific community. 

Recommendations for the authors:

Reviewer #2 (Recommendations for the authors):

(1) A picture, or full CAD design of the complete instrument, should be included as a main figure.

A complete CAD rendering of the microscope is now provided in Supplementary Figure 4.

(2) There is no quantitative comparison of the effects of the tilting resonant galvo; only a cartoon, a figure should be included.

The cartoon was intended purely as an educational illustration to conceptually explain the role of the tilting resonant galvo in shaping and homogenizing the light sheet. To clarify this intent, we have revised both the figure legend and corresponding text in the main manuscript. For readers seeking quantitative comparisons, we now reference the original study that provides a detailed analysis of this optical approach, as well as a review on the subject.

(3) Description of L4 is missing in the Figure 1 caption.

Thank you for catching this omission. We have corrected it.

(4) The beam profiles in Figures 1c and 3a, please crop and make the image bigger so the profile can be appreciated. The PSFs in Figure 3c-e should similarly be enlarged and presented using a dynamic range/LUT such that any aberrations can be appreciated.

In Figure 1c, our goal was to qualitatively illustrate the uniformity of the light-sheet across the full field of view, while Figure 1d provided the corresponding quantitative cross-section. To improve clarity, we have added an additional figure panel offering a higher-magnification, localized view of the light-sheet profile. For Figure 3c–e, we have enlarged the PSF images and adjusted the display range to better convey the underlying signal and allow subtle aberrations to be appreciated.

(5) It is unclear why LLSM is being used as the gold standard, since in its current commercial form, available from Zeiss, it is a turn-key system designed for core facilities. The original LLSM is also a versatile instrument that provides much more than the square lattice for illumination, including structured illumination, hexagonal lattices, live-cell imaging, wide-field illumination, different scan modes, etc. These additional features are not even mentioned when compared to the Altair-LSFM. If a comparison is to be provided, it should be fair and balanced. Furthermore, as outlined in the public review, anecdotal statements on "most used", "difficult to align", or "unstable" should not be provided without data.

In the revised manuscript, we have carefully removed anecdotal statements and, where appropriate, replaced them with quantitative or verifiable information. For instance, we now explicitly report that the square lattice was used in 16 of the 20 figure subpanels in the original LLSM publication, and we include a proxy for optical complexity based on the number of optical elements requiring alignment in each system.

We also now clearly distinguish between the original LLSM design—which supports multiple illumination and scanning modes—and its subsequent commercial variants, including the ZEISS Lattice Lightsheet 7, which prioritizes stability and ease of use over configurational flexibility (see Supplementary Note 3).

(6) The authors should recognize that implementing custom optics, no matter how well designed, is a big barrier to cross for most cell biology labs.

We fully understand and now acknowledge in the main text that implementing custom optics can present a significant barrier, particularly for laboratories without prior experience in optical system assembly. However, similar challenges were encountered during the adoption of other open-source microscopy platforms, such as mesoSPIM and OpenSPIM, both of which have nonetheless achieved widespread implementation. Their success has largely been driven by exhaustive documentation, strong community support, and standardized design principles—approaches we have also prioritized in Altair-LSFM. We have therefore made all CAD files, alignment guides, and detailed build documentation publicly available and continue to develop instructional materials and community resources to further reduce the barrier to adoption.

(7) Statements on "hands on workshops" though laudable, may not be appropriate to include in a scientific publication without some documentation on the influence they have had on implanting the microscope.

We understand the concern. Our intention in mentioning hands-on workshops was to convey that the dissemination effort is supported by an NIH Biomedical Technology Development and Dissemination grant, which includes dedicated channels for outreach and community engagement. Nonetheless, we agree that such statements are not appropriate without formal documentation of their impact, and we have therefore removed this text from the revised manuscript.

(8) It is claimed that the microscope is "reliable" in the discussion, but with no proof, long-term stability should be assessed and included.

Our experience with Altair-LSFM has been that it remains well-aligned over time—especially in comparison to other light-sheet systems we worked on throughout the last 11 years—we acknowledge that this assessment is anecdotal. As such, we have omitted this claim from the revised manuscript.

(9) Due to the reliance on anecdotal statements and comparisons without proof to other systems, this paper at times reads like a brochure rather than a scientific publication. The authors should consider editing their manuscript accordingly to focus on the technical and quantifiable aspects of their work.

We agree with the reviewer’s assessment and have revised the manuscript to remove anecdotal comparisons and subjective language. Where possible, we now provide quantitative metrics or verifiable data to support our statements.

Reviewer #3 (Recommendations for the authors):

Other minor points that could improve the manuscript (although some of these points are explained in the huge supplementary manual): 

(1) The authors explain thoroughly their design, and they chose a sample-scanning method. I think that a brief discussion of the advantages and disadvantages of such a method over, for example, a laserscanning system (with fixed sample) in the main text will be highly beneficial for the users.

In the revised manuscript, we now include a brief discussion in the main text outlining the advantages and limitations of a sample-scanning approach relative to a light-sheet–scanning system. Specifically, we note that for thin, adherent specimens, sample scanning minimizes the optical path length through the sample, allowing the use of more tightly focused illumination beams that improve axial resolution. We also include a new supplementary figure illustrating how this configuration reduces the propagation length of the illumination light sheet, thereby enhancing axial resolution.

(2) The authors justify selecting a 0.6 NA illumination objective over alternatives (e.g., Special Optics), but the manuscript would benefit from a more quantitative trade-off analysis (beam waist, working distance, sample compatibility) with other possibilities. Within the objective context, a comparison of the performances of this system with the new and upcoming single-objective light-sheet methods (and the ones based also on optical refocusing, e.g., DAXI) would be very interesting for the goodness of the manuscript.

In the revised manuscript, we now provide a quantitative trade-off analysis of the illumination objectives in Supplementary Note 1, including comparisons of beam waist, working distance, and sample compatibility. This section also presents calculated point spread functions for both the 0.6 NA and 0.67 NA objectives, outlining the performance trade-offs that informed our design choice. In addition, Supplementary Note 3 now includes a broader comparison of Altair-LSFM with other light-sheet modalities, including diSPIM, ASLM, and OPM, to further contextualize the system’s capabilities within the evolving light-sheet microscopy landscape.

(3) The modularity of the system is implied in the context of the manuscript, but not fully explained. The authors should specify more clearly, for example, if cameras could be easily changed, objectives could be easily swapped, light-sheet thickness could be tuned by changing cylindrical lens, how users might adapt the system for different samples (e.g., embryos, cleared tissue, live imaging), .etc, and discuss eventual constraints or compatibility issues to these implementations.

Altair-LSFM was explicitly designed and optimized for imaging live adherent cells, where sample scanning and short light-sheet propagation lengths provide optimal axial resolution (Supplementary Note 3). While the same platform could be used for superficial imaging in embryos, systems implementing multiview illumination and detection schemes are better suited for such specimens. Similarly, cleared tissue imaging typically requires specialized solvent-compatible objectives and approaches such as ASLM that maximize the field of view. We have now added some text to the Design Principles section that explicitly state this.

Altair-LSFM offers varying levels of modularity depending on the user’s level of expertise. For entry-level users, the illumination numerical aperture—and therefore the light-sheet thickness and propagation length—can be readily adjusted by tuning the rectangular aperture conjugate to the back pupil of the illumination objective, as described in the Design Principles section. For mid-level users, alternative configurations of Altair-LSFM, including different detection objectives, stages, filter wheels, or cameras, can be readily implemented (Supplementary Note 1). Importantly, navigate natively supports a broad range of hardware devices, and new components can be easily integrated through its modular interface. For expert users, all Zemax simulations, CAD models, and step-by-step optimization protocols are openly provided, enabling complete re-optimization of the optical design to meet specific experimental requirements.

(4) Resolution measurements before and after deconvolution are central to the performance claim, but the deconvolution method (PetaKit5D) is only briefly mentioned in the main text, it's not referenced, and has to be clarified in more detail, coherently with the precision of the supplementary information. More specifically, PetaKit5D should be referenced in the main text, the details of the deconvolution parameters discussed in the Methods section, and the computational requirements should also be mentioned. 

In the revised manuscript, we now provide a dedicated description of the deconvolution process in the Methods section, including the specific parameters and algorithms used. We have also explicitly referenced PetaKit5D in the main text to ensure proper attribution and clarity. Additionally, we note the computational requirements associated with this analysis in the same section for completeness.

(5)  Image post-processing is not fully explained in the main text. Since the system is sample-scanning based, no word in the main text is spent on deskewing, which is an integral part of the post-processing to obtain a "straight" 3D stack. Since other systems implement such a post-processing algorithm (for example, single-objective architectures), it would be beneficial to have some discussion about this, and also a brief comparison to other systems in the main text in the methods section. 

In the revised manuscript, we now explicitly describe both deskewing (shearing) and deconvolution procedures in the Alignment and Characterization section of the main text and direct readers to the Methods section. We also briefly explain why the data must be sheared to correct for the angled sample-scanning geometry for LLSM and Altair-LSFM, as well as both sample-scanning and laser-scanning-variants of OPMs.

(6) A brief discussion on comparative costs with other systems (LLSM, dispim, etc.) could be helpful for non-imaging expert researchers who could try to implement such an optical architecture in their lab.

Unfortunately, the exact costs of commercial systems such as LLSM or diSPIM are typically not publicly available, as they depend on institutional agreements and vendor-specific quotations. Nonetheless, we now provide approximate cost estimates in Supplementary Note 1 to help readers and prospective users gauge the expected scale of investment relative to other advanced light-sheet microscopy systems.

(7) The "navigate" control software is provided, but a brief discussion on its advantages compared to an already open-access system, such as Micromanager, could be useful for the users.

In the revised manuscript, we now include Supplementary Note 5 that discusses the advantages and disadvantages of different open-source microscope control platforms, including navigate and MicroManager. In brief, navigate was designed to provide turnkey support for multiple light-sheet architectures, with pre-configured acquisition routines optimized for Altair-LSFM, integrated data management with support for multiple file formats (TIFF, HDF5, N5, and Zarr), and full interoperability with OMEcompliant workflows. By contrast, while Micro-Manager offers a broader library of hardware drivers, it typically requires manual configuration and custom scripting for advanced light-sheet imaging workflows.

(8) The cost and parts are well documented, but the time and expertise required are not crystal clear.Adding a simple time estimate (perhaps in the Supplement Section) of assembly/alignment/installation/validation and first imaging will be very beneficial for users. Also, what level of expertise is assumed (prior optics experience, for example) to be needed to install a system like this? This can help non-optics-expert users to better understand what kind of adventure they are putting themselves through.

We thank the reviewer for this helpful suggestion. To address this, we have added Supplementary Table S5, which provides approximate time estimates for assembly, alignment, validation, and first imaging based on the user’s prior experience with optical systems. The table distinguishes between novice (no prior experience), moderate (some experience using but not assembling optical systems), and expert (experienced in building and aligning optical systems) users. This addition is intended to give prospective builders a realistic sense of the time commitment and level of expertise required to assemble and validate AltairLSFM.

Minor things in the main text:

(1) Line 109: The cost is considered "excluding the laser source". But then in the table of costs, you mention L4cc as a "multicolor laser source", for 25 K. Can you explain this better? Are the costs correct with or without the laser source? 

We acknowledge that the statement in line 109 was incorrect—the quoted ~$150k system cost does include the laser source (L4cc, listed at $25k in the cost table). We have corrected this in the revised manuscript.

(2) Line 113: You say "lateral resolution, but then you state a 3D resolution (230 nm x 230 nm x 370 nm). This needs to be fixed.

Thank you, we have corrected this.

(3) Line 138: Is the light-sheet uniformity proven also with a fluorescent dye? This could be beneficial for the main text, showing the performance of the instrument in a fluorescent environment.

The light-sheet profiles shown in the manuscript were acquired using fluorescein to visualize the beam. We have revised the main text and figure legends to clearly state this.

(4) Line 149: This is one of the most important features of the system, defying the usual tradeoff between light-sheet thickness and field of view, with a regular Gaussian beam. I would clarify more specifically how you achieve this because this really is the most powerful takeaway of the paper.

We thank the reviewer for this key observation. The ability of Altair-LSFM to maintain a thin light sheet across a large field of view arises from diffraction effects inherent to high NA illumination. Specifically, diffraction elongates the PSF along the beam’s propagation direction, effectively extending the region over which the light sheet remains sufficiently thin for high-resolution imaging. This phenomenon, which has been the subject of active discussion within the light-sheet microscopy community, allows Altair-LSFM to partially overcome the conventional trade-off between light-sheet thickness and propagation length. We now clarify this point in the main text and provide a more detailed discussion in Supplementary Note 3, which is explicitly referenced in the discussion of the revised manuscript.

(5) Line 171: You talk about repeatable assembly...have you tried many different baseplates? Otherwise, this is a complicated statement, since this is a proof-of-concept paper. 

We thank the reviewer for this comment. We have not yet validated the design across multiple independently assembled baseplates and therefore agree that our previous statement regarding repeatable assembly was premature. To avoid overstating the current level of validation, we have removed this statement from the revised manuscript.

(6) Line 187: same as above. You mention "long-term stability". For how long did you try this? This should be specified in numbers (days, weeks, months, years?) Otherwise, it is a complicated statement to make, since this is a proof-of-concept paper.

We also agree that referencing long-term stability without quantitative backing is inappropriate, and have removed this statement from the revised manuscript.

(7) Line 198: "rapid z-stack acquisition. How rapid? Also, what is the limitation of the galvo-scanning in terms of the imaging speed of the system? This should be noted in the methods section.

In the revised manuscript, we now clarify these points in the Optoelectronic Design section. Specifically, we explicitly note that the resonant galvo used for shadow reduction operates at 4 kHz, ensuring that it is not rate-limiting for any imaging mode. In the same section, we also evaluate the maximum acquisition speeds achievable using navigate and report the theoretical bandwidth of the sample-scanning piezo, which together define the practical limits of volumetric acquisition speed for Altair-LSFM.

(8) Line 234: Peta5Kit is discussed in the additional documentation, but should be referenced here, as well.

We now reference and cite PetaKit5D.

(9) Line 256: "values are on par with LLSM", but no values are provided. Some details should also be provided in the main text.

In the revised manuscript, we now provide the lateral and axial resolution values originally reported for LLSM in the main text to facilitate direct comparison with Altair-LSFM. Additionally, Supplementary Note 3 now includes an expanded discussion on the nuances of resolution measurement and reporting in lightsheet microscopy.

Figures:

(1) Figure 1 could be implemented with Figure 3. They're both discussing the validation of the system (theoretically and with simulations), and they could be together in different panels of the same figure. The experimental light-sheet seems to be shown in a transmission mode. Showing a pattern in a fluorescent dye could also be beneficial for the paper.

In Figure 1, our goal was to guide readers through the design process—illustrating how the detection objective’s NA sets the system’s resolution, which defines the required pixel size for Nyquist sampling and, in turn, the field of view. We then use Figure 1b–c to show how the illumination beam was designed and simulated to achieve that field of view. In contrast, Figure 3 presents the experimental validation of the illumination system. To avoid confusion, we now clarify in the text that the light sheet shown in Figure 3 was visualized in a fluorescein solution and imaged in transmission mode. While we agree that Figures 1 and 3 both serve to validate the system, we prefer to keep them as separate figures to maintain focus within each panel. We believe this organization better supports the narrative structure and allows readers to digest the theoretical and experimental validations independently.

(2) Figure 3: Panels d and e show the same thing. Why would you expect that xz and yz profiles should be different? Is this due to the orientation of the objectives towards the sample?

In Figure 3, we present the PSF from all three orthogonal views, as this provides the most transparent assessment of PSF quality—certain aberration modes can be obscured when only select perspectives are shown. In principle, the XZ and YZ projections should be equivalent in a well-aligned system. However, as seen in the XZ projection, a small degree of coma is present that is not evident in the YZ view. We now explicitly note this observation in the revised figure caption to clarify the difference between these panels.

(3) Figure 4's single boxes lack a scale bar, and some of the Supplementary Figures (e.g. Figure 5) lack detailed axis labels or scale bars. Also, in the detailed documentation, some figures are referred to as Figure 5. Figure 7 or, for example, figure 6. Figure 8, and this makes the cross-references very complicated to follow

In the revised manuscript, we have corrected these issues. All figures and supplementary figures now include appropriate scale bars, axis labels, and consistent formatting. We have also carefully reviewed and standardized all cross-references throughout the main text and supplementary documentation to ensure that figure numbering is accurate and easy to follow.

eLife Assessment

This important study provides the first putative evidence that alteration of the Hox code in neck lateral plate mesoderm is sufficient to induce ectopic development of forelimb buds at neck level. The authors use both gain-of-function (GOF) and loss-of-function (LOF) approaches in chick embryos to test the roles of Hox paralogy group (PG) 4-7 genes in limb development. The GOF data provide strong evidence that overexpression of Hox PG6/7 genes are sufficient to induce forelimb buds at neck level. However, the experiments using dominant negative constructs are lacking some key controls that are needed to demonstrate the specificity of the LOF effect rendering the work as a whole incomplete.

Reviewer #2 (Public review):

In the original review of this manuscript, I noted that this study provides the first evidence that alteration of the Hox code in neck lateral plate mesoderm is sufficient for ectopic forelimb budding. Their finding that ectopic expression of Hoxa6 or Hoxa7 induces wing budding at neck level, a demonstration of sufficiency, is of major significance. The experiments used to test the necessity of specific Hox genes for limb budding involved overexpression of dominant negative constructs, and there were questions about whether the controls were well designed. The reviewers made several suggestions for additional experiments that would address their concerns. In their responses to those comments, the authors indicated that they would conduct those experiments, and they acknowledged the requests for further discussion of a few points.

In the revised version of the manuscript, the authors have provided additional RNA-seq data in Table 3, which lists 221 genes that are shared between the Hoxa6-induced limb bud and normal wing bud but not the neck. This shows that the ectopic limb bud has a limb-like character. The authors also expanded the discussion of their results in the context of previous work on the mouse. These changes have improved the paper.

The authors elected not to conduct the co-transfection experiments that were suggested to test the ability of Hoxa4/a5 to block the limb-inducing ability of Hoxa6/a7. They also chose not to conduct the additional control experiments that were suggested for the dominant negative studies. The authors' justification for not conducting these experiments is provided in the responses to reviewers.

The paper is improved over the previous version, but the conclusions, particularly regarding the dominant negative experiments, would have been strengthened by the additional experiments that were recommended by the reviewers. Under the current publishing model for eLife, it is the authors' prerogative to decide whether to revise in accordance with the reviewers' suggestions. Therefore, it seems to me that this version of the manuscript is the definitive version that the authors want to publish, and that eLife should publish it together with the reviewers' comments and the authors' responses.

Search feature is clearly displayed with great colour contrast so that the user can easily find it. It is also a big size enhancing accessibility and operation principle. Suggestions are also shown during typing and clicking tab enables you to go through the list which improves navigation and accessibility.

Image of Top Winter Deals - Save up to 40% has Alt text when you hover over the image. It states, "Top Winter Deals. Save up to 40% on fitness, appliances, storage, and more." This further enhances the accessibility of the page following the operable principle.

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

Response to Reviewer 1:

The authors introduce G2PT, a hierarchical graph transformer model that integrates genetic variants (SNPs), gene annotations, and multigenic systems (Gene Ontology) to predict and interpret complex traits.

We thank the reviewer for this accurate summary of our approach and contributions.

Major Comments:

Comment 1-1. Insufficient Specification of Model Architecture: The description of the "hierarchical graph transformer" lacks technical depth. Key implementation details are missing: how node embeddings are initialized for SNPs, genes, and systems; how graph connectivity is defined at each level (e.g., adjacency matrices used in Equations 5-9, the sparsity); justification for the choice of embedding dimension and number of attention heads, including any sensitivity analysis; and the architecture of the feed-forward neural networks (e.g., number of layers, activation functions, and hidden dimensions).

__Reply 1-1. __As requested, we have expanded the technical description of the model architecture, including the hierarchical graph transformer (HiGT), in the Materials and Methods section. Details regarding node initialization and hierarchical connectivity are now included in the new paragraph "Model Initialization and Graph Construction." Specifically, all node embeddings corresponding to SNPs, genes, and ontology-defined systems are initialized using uniform Xavier initialization (Glorot and Bengio, 2010).

We have also clarified our hyperparameter optimization strategy. Learning rate, weight decay, hidden (embedding) dimension, and the number of attention heads were selected via grid search, as summarized in new Supplementary Fig. 8, reproduced below. Based on both performance and computational efficiency, we adopted four attention heads-consistent with the configuration commonly used in academic transformer models (Vaswani et al., 2017) (the original Transformer used eight).

Regarding the feed-forward neural network, we follow the standard Transformer architecture consisting of two position-wise layers with hidden dimension four times larger than the node embedding size and a GeLU nonlinear activation function (Hendrycks and Gimpel, 2016). This configuration is widely established in the literature and functions as an intermediate processing step following attention; therefore, it is not a focus of hyperparameter tuning. All corresponding updates have been incorporated into the revised Methods section for clarity and completeness.

Comment 1-2. No Simulation Studies to Validate Epistasis Detection: The ground truth epistasis interaction should use the ones that have been manually validated by literature. The central claim of discovering epistatic interactions relies heavily on the model's attention mechanism and downstream statistical filtering. However, no simulation studies are presented to validate that G2PT can reliably detect epistasis when ground-truth interactions are known. Demonstrating robust detection of non-additive interactions under varying genetic architectures and noise levels in simulated genotype-phenotype datasets is essential to substantiate the method's core capability.

Reply 1-2. We agree that a simulation of epistasis detection using the G2PT model is a worthy addition to the manuscript. Accordingly, we have now incorporated a new section in the Results titled "Validation of Epistasis through Simulation Studies", which includes two new figures reproduced below (Supplementary Fig. 6 and Fig. 5). We have also added a new Methods section to describe this simulation study under the heading "Epistasis Simulation". These simulation studies show that G2PT recovers epistatic gene pairs with high fidelity when these pairs are coherent with the systems ontology (c.f. 'ontology coherence' in Supplementary Fig. 6, which reflects the probability that both SNPs are assigned to the same leaf system). Furthermore, G2PT outcompetes previous tools, such as PLINK-epistasis, which do not use knowledge of the systems hierarchy in the same way (Supplementary Fig 6b-d). Using simulation parameters consistent with current genome-wide association studies (n = 400,000) and understanding of heritability (h2 = 0.3 to 0.5) (Bloom et al. 2015; Speed and Evans 2023), we find that approximately 10% of all epistatic SNP pairs can be recovered at a precision of 50% (Fig. 5). We have provided the source code for this simulation study in our GitHub repository (https://github.com/idekerlab/G2PT/blob/master/Epistasis_simulation.ipynb)

Comment 1-3. Lack of Justification for Model Complexity and Missing Ablation Insights: While Supplementary Figure 2 presents ablation studies, the manuscript needs to justify the high computational cost (168 GPU hours using 4×A30 GPUs) of the full model. It remains unclear how much performance gain is specifically due to reverse propagation (Equations 8-9), which is claimed to capture biological context. The benefit of using a full Gene Ontology hierarchy versus a flat system list is not quantified. There is also no comparison between bidirectional versus unidirectional propagation. Overall, the added complexity is not empirically shown to be necessary

Reply 1-3. We thank the reviewer for prompting a clearer justification of complexity and ablations. We have now revised the Results to (i) quantify the specific value of the ontology and reverse propagation, and (ii) explain why a flat SNP→system model is computationally and biologically sub-optimal. We have added new ablation results to compare bidirectional (forward+reverse) versus forward-only propagation. Reverse propagation has little effect when epistatic pairs are within one system (ontology coherence ρ=1.0) but substantially improves retrieval when interactions span related systems (e.g., ρ≈0.8) (Figure reproduced below) A flat design scores a dense genes×systems map, ignoring known sparsity (sparse SNP→gene assignments; sparse ontology edges) and losing multi-scale context; our hierarchical formulation restricts computation to observed edges (SNP→gene→system) and aggregates signals across levels, yielding better efficiency and biological fidelity.

Comment 1-4. Non-Equivalent Benchmarking Against PRS Methods: Figure 2 compares G2PT to polygenic risk score (PRS) methods such as LDpred2 and Lassosum, but G2PT is run only on SNPs pre-filtered by marginal association (p-values between 10⁻⁵ and 10⁻⁸), while the PRS methods use genome-wide SNPs. This introduces a strong bias in G2PT's favor by effectively removing noise. A fair comparison would require: (a) running LDpred2 and Lassosum on the same pre-filtered SNP sets as G2PT, or (b) running G2PT on genome-wide or LD-pruned SNP sets. The reported superior performance of G2PT may be driven primarily by this input filtering, not the model architecture.

Reply 1-4. We appreciate the reviewer's concern regarding benchmarking equivalence. In response, we have extended our analyses to include PRS-CS (Ge et al., 2019) and SBayesRC (Zheng et al., 2024), two state-of-the-art Bayesian shrinkage methods comparable to LDpred2 and Lassosum. Although we initially attempted to run LDpred2 and Lassosum under all SNP-filtering conditions, their computational requirements at UK Biobank scale proved prohibitively time consuming. We therefore focused on PRS-CS and SBayesRC, which offer similar modeling principles with greater computational tractability. These methods have now been run at matched SNP-filtering conditions to our original study. The new results demonstrate that G2PT consistently outperforms PRS-CS and SBayesRC (new Fig. 2, reproduced below), indicating that its performance advantage is not solely attributable to SNP pre-filtering but also to its hierarchical attention-based architecture.

Comment 1-5: No Details on Hyperparameter Optimization: Although the manuscript mentions grid search for hyperparameter tuning, it provides no information about which parameters were optimized (e.g., learning rate, dropout rate, weight decay, attention dropout, FFNN dimensions), what search space was explored, or what final values were selected. There is also no assessment of how sensitive the model's performance is to these choices. Better transparency would help facilitate reproducibility

Reply 1-5. We agree with the reviewer and have expanded the manuscript to include full details of hyperparameter optimization. As described in the revised Methods section, we performed a grid search over learning rate {10−3,10−4,10−5} hidden dimension {64,128} and weight decay {0,10−5,10−3}. The results, summarized in Supplementary Fig. 8 (reproduced above), show that model performance is most sensitive to the learning rate, while hidden dimension and weight decay exert more moderate effects. Based on these findings, we selected a learning rate of 10−5, hidden dimension of 64, and weight decay of 10−3 for all subsequent experiments. Although a hidden dimension of 128 slightly improved performance, we adopted 64 to balance predictive accuracy with computational efficiency.

Comment 1-6. Absence of Control for Key Confounders: In interpreting attention scores as reflecting genetic relevance (e.g., the role of the immunoglobulin system), the model includes only age, sex, and genetic principal components as covariates. Important confounders such as BMI, alcohol use, or medication (e.g., statins) have not been controlled for. Since TG/HDL levels are strongly influenced by environment and lifestyle, it is entirely plausible that some high-attention features reflect environmental tagging, not biological causality.

Reply 1-6. In the current framework, we included age, sex, and genetic principal components to account for demographic and population-structure effects, focusing on genetic contributions within a controlled baseline. We acknowledge that non-genetic covariates can influence downstream biological states and may indirectly shape attention at the gene or system level. Accurately modeling such effects requires an extended framework where environmental variables directly modulate gene and system embeddings rather than being implicitly absorbed by the attention mechanism. We have clarified these limitations in the Discussion along with plans to incorporate explicit confounder modeling in future extensions of G2PT.

Comment 1-7. Oversimplified Treatment of SNP-to-Gene Mapping: The SNP-to-gene mapping strategy combines cS2G, eQTL, and nearest-gene annotations, but the limitations of this approach are not adequately addressed. The manuscript does not specify how conflicts between methods are resolved or what fraction of SNPs map ambiguously to multiple genes. Supplementary Figure 2 shows model performance degrades when using only nearest-gene mapping, but there is no systematic analysis of how mapping uncertainties propagate through the hierarchy and affect attention or interpretation.

Reply 1-7. In the revision (Results), we have clarified how conflicts between cS2G, eQTL, and nearest-gene annotations are resolved, and we have reported the proportion of SNPs that map to multiple genes across these three annotation approaches. We note that the hierarchical attention mechanism enables the model to prioritize among alternative gene mappings in a data-driven manner, and this is a major strength of the approach. As shown in Fig. 3 (Results, reproduced below), SNP-to-gene attention weights reveal dominant linkages, reducing the impact of mapping uncertainty on interpretation. We now explicitly describe this mechanism and acknowledge that further work in probabilistic mapping and fine-mapping approaches is a valuable future direction for improving resolution and interpretability.

"For SNPs with several potential SNP-to-gene mappings (Methods), we found that G2PT often prioritized one of these genes in particular due to its membership in a high-attention system. For example, the chr11q23.3 locus contains multiple genes including the APOA1/C3/A4/A5 gene cluster (Fig. 3c) which is well-known to govern lipid transport, an important system for G2PT predictions (Fig. 3a). Due to high linkage disequilibrium in the region, all of its associated SNPs had multiple alternative gene mappings available. For example, SNP rs1145189 mapped not only to APOA5 but to the more proximal BUD13, a gene functioning in spliceosomal assembly (a system receiving substantially lower G2PT attention). Here, the relevant information flow learned by G2PT was from rs1145189 to APOA5 to lipid transport and protein-lipid complex remodeling (Fig. 3c; and conversely, deprioritizing BUD13 as an effector gene for TG/HDL). We found that this particular genetic flow was corroborated by exome sequencing, which implicates APOA5 but not BUD13 in regulation of TG/HDL, using data that were not available to G2PT. Similarly, two other SNPs at this locus - rs518547 and rs11216169 - had potential mappings to their closest gene SIK3, where they reside within an intron, but also to regulatory elements for the more distant lipid transport genes APOC3 and APOA4. Here, G2PT preferentially weighted the mappings to APOC3 and APOA4 rather than to SIK3 (Fig. 3c)."

Comment 1-8. Naive Scoring of System Importance: The method used to quantify the biological relevance of systems (i.e., correlating attention scores with predicted phenotype values) risks circular reasoning. Since the model is trained to optimize prediction, systems that contribute strongly to prediction will naturally show high correlation-even if they are not biologically causal. No comparison is made with established gene set enrichment methods applied to GWAS summary statistics. The approach lacks an independent benchmark to validate that the "important" systems are biologically meaningful.

Reply 1-8. As requested, we compared G2PT's system-level importance scores with results from MAGMA competitive gene-set analysis, an established enrichment approach. This analysis indeed shows significant correlation between the systems identified by the two approaches (ρ = 0.26, p .01; Supplementary Table. 2), reflecting a shared emphasis on canonical lipid processes. We also observed systems detected by G2PT but not strongly detected by MAGMA's linear enrichment model-for example, the lipopolysaccharide-mediated signaling pathway (Kalita et al. 2022)

Comment 1-9. No External Validation to Assess Generalizability. All evaluations are performed using cross-validation within the UK Biobank. There is no assessment of generalizability to independent cohorts or diverse ancestries. Given population structure, genotyping platform, and phenotype measurement variability, external validation is essential before claiming the method is suitable for broader use in polygenic risk assessment.

Reply 1-9. To externally validate the G2PT model requires individual level genotype data with paired TG/HDL measurements, sample size at the scale of the UK Biobank, and GPU access to this data. Thus, we approached the All of Us program, a large and diverse cohort with individual level data and T2D conditions with HbA1C measurements. We first processed the All of Us genotype and phenotype data as we had processed UKBB data (Methods), resulting in 41,849 participants with T2D and 80,491 without T2D across various ethnicities. We then transferred the trained T2D G2PT model to the AoU Workbench and evaluated its performance. The model demonstrated robust discriminative capability with an explained variance of 0.025, as shown in the new Fig. 2d, (reproduced above).

Comment 1-10. Computational Burden and Scalability Are Not Addressed: The paper notes that training the model requires 168 GPU hours on 4×A30 GPUs for just ~5,000 SNPs. However, there is no discussion of whether G2PT can scale to larger SNP sets (e.g., genome-wide imputed data) or more complex biological hierarchies (e.g., Reactome pathways). Without addressing scalability, the model's applicability to real-world, large-scale genomic datasets remains unclear.

Reply 1-10. We have addressed scalability with both engineering optimizations and new scalability experiments. First, we refactored the model to use the xFormer memory-efficient attention for the hierarchical graph transformer (Lefaudeux et al., 2022), which also helps full parallelization of training, reducing bottlenecks. Second, we added a scaling study with progressively increasing SNP count. On 4×A30 GPUs, end-to-end training time for the 5k-SNP setting decreased from 4000 to 400 min. (approximately 7 GPU-hours, ×10). These new results are given in Supplementary Fig. 7, reproduced below.

Minor Comment:

Comment 1-11. Attention Weights as Mechanistic Insight: The paper equates high attention scores with biological importance, for example in highlighting the immunoglobulin system. There is no causal validation showing that altering the highlighted SNPs, genes, or systems has an actual effect on TG/HDL. Attention weights in transformer models are known to sometimes reflect spurious correlations, especially in high-dimensional settings. The correlation between attention scores and predictions (Supplementary Fig. 3a,b) does not constitute biological evidence. The interpretability claims can be restated without supporting functional or causal validation.

Reply 1-11. We thank the reviewer for this thoughtful comment. We agree that attention weights are not causal evidence. In the revision, we (1) reframe attention-based findings as hypothesis-generating rather than mechanistic, and (2) add an explicit limitation noting that correlations between attention scores and predictions do not constitute biological validation.

Response to Reviewer 2:

This manuscript describes the introduction of the Genotype-to-Phenotype Transformer (G2PT), described by the authors as "a framework for modeling hierarchical information flow among variants, genes, multigenic systems, and phenotypes." The authors used the ratio TG/HDL as a trait for proof of concept of this tool.

This is a potentially interesting computational tool of interest to bioinformaticians, computational genomicists, and biologists.

We thank the reviewer for their overall positive assessment of our study.

Comment 2-1. The rationale for choosing the TG/HDL ratio for this proof of concept analysis is not well justified beyond it being a marker for insulin resistance. Overall the use of a ratio may be problematic (see below). Analyses of TG and HDL separately as individual quantitative traits would be of interest. And an analysis of a dichotomous clinical trait (T2DM or CAD) would also be of great interest.

Reply 2-1. We thank the reviewer for this suggestion. In the revised manuscript, we have expanded our analyses beyond the TG/HDL ratio to include TG and HDL as individual quantitative traits (Fig. 2, reproduced below). These additional analyses demonstrate that G2PT captures predictive signals robustly across each lipid component, not solely through their ratio. Furthermore, to address the reviewer's interest in clinical outcomes, we incorporated an analysis of type 2 diabetes (T2D) as a dichotomous trait of direct clinical relevance. Collectively, these results strengthen the rationale for our chosen phenotype and show that the G2PT framework generalizes effectively across quantitative and binary traits, consistently outperforming advanced PRS and machine learning benchmarks.

Comment 2-2. The approach to mapping SNPs to genes does not incorporate the most advanced approaches. This should be described in more detail.

Reply 2-2. We agree that the choice of SNP-to-gene mapping materially affects both performance and interpretability-indeed, our epistasis simulations suggest that more accurate mappings can improve recovery and localization. In this proof-of-concept work we use a straightforward, modular mapping sufficient to demonstrate the modeling framework, and we have clarified this in the Methods. The architecture is designed to plug-and-play alternative SNP-to-gene maps (e.g., eQTL/colocalization-based assignments, promoter-capture Hi-C). A dedicated follow-up study will systematically compare these alternatives and quantify their impact on attribution and downstream discovery.

Comment 2-3. The example of gene prioritization at the A1/C3/A4/A5 gene locus is not particularly illuminating, as the prioritized genes are already well-known to influence TG and HDL-C levels and the TG/HDL ratio. Can the authors provide an example where G2PT prioritized a gene at a locus that is not already a well-known regulator of TG and HDL metabolism?

Reply 2-3. We thank the reviewer for this suggestion. We have revised the manuscript to de-emphasize the well-established APOA1 locus and instead highlight the less expected "Positive regulation of immunoglobulin production" system (Figure 3a,b, Discussion). Here our model prioritizes the gene TNFSF13 based on specific variants that are not previously associated with TG or HDL (e.g., rs5030405, rs1858406, shown in blue). This finding points to an intriguing, non-canonical link between B-cell regulation and lipid metabolism. While full exploration of this finding is beyond the scope of the present methods paper, this example demonstrates G2PT's ability to identify novel, high-priority candidates in atypical systems.

Comment 2-4. The identification of epistatic interactions is a potentially interesting application of G2PT. However, suppl table 1 shows a very limited number of such interactions with even fewer genes, and most of these are well established biological interactions (such as LPL/apoA5). The TGFB1 and FKBP1A interaction is interesting and should be discussed. What is needed for increasing the number of potential interactions, greater power?

Reply 2-4. We are glad the reviewer appreciates the use of the G2PT model to identify epistatic interactions. We have now discussed a potential mechanism of epistasis between TGFB1 and FKBP1A in the protein dephosphorylation system (Discussion). In addition, we have addressed the reviewer's question about statistical power through extensive epistasis simulations (Fig. 5 and Supplementary Fig. 6), which show that G2PT's detection ability scales strongly with sample size-1,000 samples are insufficient, performance improves at 5,000, and power becomes reliable at 100,000. Realistic simulations (Fig. 5b-d) further demonstrate that under biologically plausible architectures, G2PT can robustly recover specific interactions even within complex genetic backgrounds

Comment 2-5. Furthermore, the use of the TG/HDL ratio for the assessment of epistatic interactions may be problematic. For example, if one SNP affected only TG and the other only HDL-C, it would appear to be an epistatic interaction with regard to the ratio, although the biological epistasis may be limited to non-existent.

Reply 2-5. We have greatly expanded the example phenotypes modeled in our study, Please see our reply 2-1 above.

Response to Reviewer 3:

This manuscript by Lee et al provides a sensible and powerful approach to polygenic score prediction. The model aggregates information from SNPs to genes to systems, using a transformer based architecture, which appears to increase predictive performance, produce interpretable outputs of genes and systems that underlie risk, and identify candidates for epistasis tests.

I think the manuscript is clear and well written, and conducted via state-of-the-art approaches. I don't have any concerns regarding the claims that are made.

We thank the reviewer for their very positive assessment of our study.

Major comments:

Comment 3-1. Specifically, lipid based traits are perhaps the most well-powered and the most biologically coherent; they are also very well-studied biologically and thus overrepresented in the gene ontology. It is unclear whether this approach will work as well for a trait like Schizophrenia for which the underlying pathways are not as well captured in existing ontologies. The authors anticipate this in their limitations section, and I am not expecting them to solve every issue with this, but it would be nice to expand the testing a little bit beyond only this one trait.

Reply 3-1. We appreciate the reviewer's suggestion to expand beyond a single lipid trait. In the revised manuscript, we have included analyses of additional phenotypes, including low-density lipoprotein (LDL) and T2D (Fig. 2). These additions demonstrate the broader applicability of our framework beyond a single trait class.

Comment 3-2. It also seems like the authors have not compared their method to the truly latest PRS methods, such as PRS-CSx and SBayesR. I would suggest adding some of the methods shown to be the best from this recent paper: https://www.nature.com/articles/s41598-025-02903-1

Reply 3-2. We agree these are important comparators. Accordingly, we have extended our comparison to include PRS‑CS (Ge et al., 2019) and SBayesRC (Zheng et al., 2024), following its strong performance demonstrated in recent benchmarking studies (see Figure 2 above). We confirmed that G2PT outperforms advanced PRS methods for all TG/HDL ratio, LDL, and T2D phenotypes.

Comment 3-3. Another major comment regards whether this method could be applied to traits with just GWAS summary statistics, rather than individual level data. This would not enable identification of specific methods underlying an individual, but it could still learn SNP based weights that could be mapped to genes and systems that could help explain risk when the model is applied to individuals (kind of like a pretraining step?)

Reply 3-3. We appreciate this suggestion. While SNP weights from GWAS summary statistics could, in principle, serve as informative priors for attention values, incorporating them would require a sophisticated mathematical formulation that is beyond the scope of this study. Our current framework also relies on individual-level genotype and phenotype data to capture multilevel information flow and individual-specific variation.

Minor comments:

Comment 3-4. Why the need to constrain to a small number of SNPs? Is it just computational cost? If so, what would happen as power increases and more SNPs exceed the thresholds used?

Reply 3-4. Yes, it's about computational cost, but we've now modified the code for improved computational efficiency. First, we refactored the model to use the xFormer memory-efficient attention for the hierarchical graph transformer (Lefaudeux et al., 2022), which also helps full parallelization of training, reducing bottleneck effects. Second, we added a scaling study of the impact of varying SNP count. On 4×A30 GPUs, end-to-end training time for the 5k-SNP setting decreased from 65 hours to 7 GPU-hours (×9). We expect performance can potentially increase if more SNPs are provided to the model based on Fig. 2 (reproduced above). With the optimized implementation, users can raise SNP thresholds as power increases; the expected behavior is improved accuracy up to a plateau, while hierarchical sparsity maintains training tractability and ensures well-regularized results.

Comment 3-5. What type of sample size/power does this method require to work well? If others were to use it, how many SNPs/samples would be needed to obtain good performance?

Reply 3-5. To address this comment, we quantified performance as a function of training size by subsampling the cohort and retraining G2PT with identical architecture and SNP set. New Supplementary Fig. 3 (reproduced below) shows monotonic gains with sample size across three representative phenotypes. We found that stable performance is reached by ~100k samples. These trends hold for continuous traits (TG/HDL, LDL) and more modestly for a binary trait (T2D), consistent with lower per-sample information for case-control settings.

C'est un terme qui est déjà très reconnu, donc je ne pense pas que ce soit nécessaire de le remplacer. Toutefois, si on souhaite privilégier une option en français, je proposerais "à code source ouvert".

"However, exactly how many triplets code amino acids and how many have other functions we are unable to say." This makes more sense in relation to a comment I made at the very beginning of the paper. Still, the fact that the idea was even suggested during this period of time is impressive.

"The code is probably degenerate; that is, in general, one particular amino-acid can coded by one of several triplets of bases." I think it's really interesting that Crick and the other researchers apart of this paper were able to glean such an advanced concept for the 1960s.

Author response:

The following is the authors’ response to the latest reviews:

"One remaining question is the interpretation of matching variants with very low stable posterior probabilities (~0), which the authors have analyzed in detail but without fully conclusive findings. I agree with the authors that this event is relatively rare and the current sample size is limited but this might be something to keep in mind for future studies."

Fine-mapping stability – on matching variants with very low stable posterior probability

We thank Reviewer 2 for encouraging us to think more about how low stable posterior probability matching variants can be interpreted. We describe a few plausible interpretations, even though – as Reviewer 2 and we have both acknowledged – our present experiments do not point to a clear and conclusive account.

One explanation is that the locus captured by the variant might not be well-resolved, in the sense that many correlated variants exist around the locus. Thus, the variant itself is unlikely causal, but the set of variants in high LD with it may contain the true causal variant, or it's possible that the causal variant itself was not sequenced but lies in that locus. A comparison of LD patterns across ancestries at the locus would be helpful here.

Another explanation rests on the following observation. For a variant to be matching between top and stable PICS and to also have very small stable PP, it has to have the largest PP after residualization on the ALL slice but also have positive PP with gene expression on many other slices. In other words, failing to control for potential confounders shrinks the PP. If one assumes that the matching variant is truly causal, then our observation points to an example of negative confounding (aka suppressor effect). This can occur when the confounders (PCs) are correlated with allele dosage at the causal variant in a different direction than their correlation with gene expression, so that the crude association between unresidualized gene expression and causal variant allele dosage is biased toward 0.

Although our present study does not allow us to systematically confirm either interpretation – since we found that matching variants were depleted in causal variants in our simulations, violating the second argument, but we also found functional enrichment in analyses of GEUVADIS data though only 17 matching variants with low stable PP were reported – we believe a larger-scale study using larger cohort sizes (at least 1000 individuals per ancestry) and many more simulations (to increase yield of such cases) would be insightful.

———

The following is the authors’ response to the original reviews:

Reviewer #1:

Major comments:

(1) It would be interesting to see how much fine-mapping stability can improve the fine-mapping results in cross-population. One can simulate data using true genotype data and quantify the amount the fine-mapping methods improve utilizing the stability idea.

We agree, and have performed simulation studies where we assume that causal variants are shared across populations. Specifically, by mirroring the simulation approach described in Wang et al. (2020), we generated 2,400 synthetic gene expression phenotypes across 22 autosomes, using GEUVADIS gene expression metadata (i.e., gene transcription start site) to ensure largely cis expression phenotypes were simulated. We additionally generated 1,440 synthetic gene expression phenotypes that incorporate environmental heterogeneity, to motivate our pursuit of fine-mapping stability in the first place (see Response to Reviewer 2, Comment 6). These are described in Results section “Simulation study”:

We evaluated the performance of the PICS algorithm, specifically comparing the approach incorporating stability guidance against the residualization approach that is more commonly used — similar to our application to the real GEUVADIS data. We additionally investigated two ways of “combining” the residualization and stability guidance approaches: (1) running stability-guided PICS on residualized phenotypes; (2) prioritizing matching variants returned by both approaches. See Response to Reviewer 2, Comment 5.

(2) I would be very interested to see how other fine-mapping methods (FINEMAP, SuSiE, and CAVIAR) perform via the stability idea.

Thank you for this valuable comment. We ran SuSiE on the same set of simulated datasets. Specifically, we ran a version that uses residualized phenotypes (supposedly removing the effects of population structure), and also a version that incorporates stability. The second version is similar to how we incorporate stability in PICS. We investigated the performance of Stable SuSiE in a similar manner to our investigation of PICS. First we compared the performance relative to SuSiE that was run on residualized phenotypes. Motivated by our finding in PICS that prioritizing matching variants improves causal variant recovery, we did the same analysis for SuSiE. This analysis is described in Results section “Stability guidance improves causal variant recovery in SuSiE.”

We reported overall matching frequencies and causal variant recovery rates of top and stable variants for SuSiE in Figures 2C&D.

Frequencies with which Stable and Top SuSiE variants match, stratified by the simulation parameters, are summarized in Supplementary File 2C (reproduced for convenience in Response to Reviewer 2, Comment 3). Causal variant recovery rates split by the number of causal variants simulated, and stratified by both signal-to-noise ratio and the number of credible sets included, are reported in Figure 2—figure supplements 16-18. We reproduce Figure 2—figure supplement 18 (three causal variants scenario) below for convenience. Analogous recovery rates for matching versus non-matching top or stable variants are reported in Figure 2—figure supplements 19, 21 and 23.

(3) I am a little bit concerned about the PICS's assumption about one causal variant. The authors mentioned this assumption as one of their method limitations. However, given the utility of existing fine-mapping methods (FINEMAP and SuSiE), it is worth exploring this domain.

Thank you for raising this fair concern. We explored this domain, by considering simulations that include two and three causal variants (see Response to Reviewer 2, Comment 3). We looked at how well PICS recovers causal variants, and found that each potential set largely does not contain more than one causal variant (Figure 2—figure supplements 20 and 22). This can be explained by the fact that PICS potential sets are constructed from variants with a minimum linkage disequilibrium to a focal variant. On the other hand, in SuSiE, we observed multiple causal variants appearing in lower credible sets when applying stability guidance (Figure 2—figure supplements 21 and 23). A more extensive study involving more fine-mapping methods and metrics specific to violation of the one causal variant assumption could be pursued in future work.

Reviewer #2:

Aw et al. presents a new stability-guided fine-mapping method by extending the previously proposed PICS method. They applied their stability-based method to fine-map cis-eQTLs in the GEUVADIS dataset and compared it against what they call residualization-based method. They evaluated the performance of the proposed method using publicly available functional annotations and claimed the variants identified by their proposed stability-based method are more enriched for these functional annotations.

While the reviewer acknowledges the contribution of the present work, there are a couple of major concerns as described below.

Major:

(1) It is critical to evaluate the proposed method in simulation settings, where we know which variants are truly causal. While I acknowledge their empirical approach using the functional annotations, a more unbiased, comprehensive evaluation in simulations would be necessary to assess its performance against the existing methods.

Thank you for this point. We agree. We have performed a simulation study where we assume that causal variants are shared across populations (see response to Reviewer 1, Comment 1). Specifically, by mirroring the simulation approach described in Wang et al. (2020), we generated 2,400 synthetic gene expression phenotypes across 22 autosomes, using GEUVADIS gene expression metadata (i.e., gene transcription start site) to ensure cis expression phenotypes were simulated.

(2) Also, simulations would be required to assess how the method is sensitive to different parameters, e.g., LD threshold, resampling number, or number of potential sets.

Thank you for raising this point. The underlying PICS algorithm was not proposed by us, so we followed the default parameters set (LD threshold, r² \= 0.5; see Taylor et al., 2021 Bioinformatics) to focus on how stability considerations will impact the existing fine-mapping algorithm. We attempted to derive the asymptotic joint distribution of the p-values, but it was too difficult. Hence, we used 500 permutations because such a large number would allow large-sample asymptotics to kick in. However, following your critical suggestion we varied the number of potential sets in our analyses of simulated data. We briefly mention this in the Results.

“In the Supplement, we also describe findings from investigations into the impact of including more potential sets on matching frequency and causal variant recovery…”

A detailed write-up is provided in Supplementary File 1 Section S2 (p.2):

“The number of credible or potential sets is a parameter in many fine-mapping algorithms. Focusing on stability-guided approaches, we consider how including more potential sets for stable fine-mapping algorithms affects both causal variant recovery and matching frequency in simulations…

Causal variant recovery. We investigate both Stable PICS and Stable SuSiE. Focusing first on simulations with one causal variant, we observe a modest gain in causal variant recovery for both Stable PICS and Stable SuSiE, most noticeably when the number of sets was increased from 1 to 2 under the lowest signal-to-noise ratio setting…”

We observed that increasing the number of potential sets helps with recovering causal variants for Stable PICS (Figure 2—figure supplements 13-15). This observation also accounts for the comparable power that Stable PICS has with SuSiE in simulations with low signal-to-noise ratio (SNR), when we increase the number of credible sets or potential sets (Figure 2—figure supplements 10-12).

(3) Given the previous studies have identified multiple putative causal variants in both GWAS and eQTL, I think it's better to model multiple causal variants in any modern fine-mapping methods. At least, a simulation to assess its impact would be appreciated.

We agree. In our simulations we considered up to three causal variants in cis, and evaluated how well the top three Potential Sets recovered all causal variants (Figure 2—figure supplements 13-15; Figure 2—figure supplement 15). We also reported the frequency of variant matches between Top and Stable PICS stratified by the number of causal variants simulated in Supplementary File 2B and 2C. Note Supplementary File 2C is for results from SuSiE fine-mapping; see Response to Reviewer 1, Comment 2.

Supplementary File 2B. Frequencies with which Stable and Top PICS have matching variants for the same potential set. For each SNR/ “No. Causal Variants” scenario, the number of matching variants is reported in parentheses.

Supplementary File 2C. Frequencies with which Stable and Top SuSiE have matching variants for the same credible set. For each SNR/ “No. Causal Variants” scenario, the number of matching variants is reported in parentheses.

(4) Relatedly, I wonder what fraction of non-matching variants are due to the lack of multiple causal variant modeling.

PICS handles multiple causal variants by including more potential sets to return, owing to the important caveat that causal variants in high LD cannot be statistically distinguished. For example, if one believes there are three causal variants that are not too tightly linked, one could make PICS return three potential sets rather than just one. To answer the question using our simulation study, we subsetted our results to just scenarios where the top and stable variants do not match. This mimics the exact scenario of having modeled multiple causal variants but still not yielding matching variants, so we can investigate whether these non-matching variants are in fact enriched in the true causal variants.

Because we expect causal variants to appear in some potential set, we specifically considered whether these non-matching causal variants might match along different potential sets across the different methods. In other words, we compared the stable variant with the top variant from another potential set for the other approach (e.g., Stable PICS Potential Set 1 variant vs Top PICS Potential Set 2 variant). First, we computed the frequency with which such pairs of variants match. A high frequency would demonstrate that, even if the corresponding potential sets do not have a variant match, there could still be a match between non-corresponding potential sets across the two approaches, which shows that multiple causal variant modeling boosts identification of matching variants between both approaches — regardless of whether the matching variant is in fact causal.

Low frequencies were observed. For example, when restricting to simulations where Top and Stable PICS Potential Set 1 variants did not match, about 2-3% of variants matched between the Potential Set 1 variant in Stable PICS and Potential Sets 2 and 3 variants in Top PICS; or between the Potential Set 1 variant in Top PICS and Potential Sets 2 and 3 variants in Stable PICS (Supplementary File 2D). When looking at non-matching Potential Set 2 or Potential Set 3 variants, we do see an increase in matching frequencies (between 10-20%) between Potential Set 2 variants and other potential set variants between the different approaches. However, these percentages are still small compared to the matching frequencies we observed between corresponding potential sets (e.g., for simulations with one causal variant this was 70-90% between Top and Stable PICS Potential Set 1, and for simulations with two and three causal variants this was 55-78% and 57-79% respectively).

We next checked whether these “off-diagonal” matching variants corresponded to the true causal variants simulated. Here we find that the causal variant recovery rate is mostly less than the corresponding rate for diagonally matching variants, which together with the low matching frequency suggests that the enrichment of causal variants of “off-diagonal” matching variants is much weaker than in the diagonally matching approach. In other words, the fraction of non-matching (causal) variants due to the lack of multiple causal variant modeling is low.

We discuss these findings in Supplementary File 1 Section S2 (bottom of p.2).

(5) I wonder if you can combine the stability-based and the residualization-based approach, i.e., using the residualized phenotypes for the stability-based approach. Would that further improve the accuracy or not?

This is a good idea, thank you for suggesting it. We pursued this combined approach on simulated gene expression phenotypes, but did not observe significant gains in causal variant recovery (Figure 2B; Figure 2—figure supplements 2, 13 and 15). We reported this Results “Searching for matching variants between Top PICS and Stable PICS improves causal variant Recovery.”

“We thus explore ways to combine the residualization and stability-driven approaches, by considering (i) combining them into a single fine-mapping algorithm (we call the resulting procedure Combined PICS); and (ii) prioritizing matching variants between the two algorithms. Comparing the performance of Combined PICS against both Top and Stable PICS, however, we find no significant difference in its ability to recover causal variants (Figure 2B)...”

However, we also confirmed in our simulations that prioritizing matching variants between the two approaches led to gains in causal variant recovery (Figure 2D; Figure 2—figure supplements 4, 19, 20 and 22). We reported this Results “Searching for matching variants between Top PICS and Stable PICS improves causal variant Recovery.”

“On the other hand, matching variants between Top and Stable PICS are significantly more likely to be causal. Across all simulations, a matching variant in Potential Set 1 is 2.5X as likely to be causal than either a non-matching top or stable variant (Figure 2D) — a result that was qualitatively consistent even when we stratified simulations by SNR and number of causal variants simulated (Figure 2—figure supplements 19, 20 and 22)...”

This finding is consistent with our analysis of real GEUVADIS gene expression data, where we reported larger functional significance of matching variants relative to non-matching variants returned by either Top of Stable PICS.

(6) The authors state that confounding in cohorts with diverse ancestries poses potential difficulties in identifying the correct causal variants. However, I don't see that they directly address whether the stability approach is mitigating this. It is hard to say whether the stability approach is helping beyond what simpler post-hoc QC (e.g., thresholding) can do.

Thank you for raising this fair point. Here is a model we have in mind. Gene expression phenotypes (Y) can be explained by both genotypic effects (G, as in genotypic allelic dosage) and the environment (E): Y = G + E. However, both G and E depend on ancestry (A), so that Y = G|A+E|A. Suppose that the causal variants are shared across ancestries, so that (G|A=a)=G for all ancestries a. Suppose however that environments are heterogeneous by ancestry: (E|A=a) = e(a) for some function e that depends non-trivially on a. This would violate the exchangeability of exogenous E in the full sample, but by performing fine-mapping on each ancestry stratum, the exchangeability of exogenous E is preserved. This provides theoretical justification for the stability approach.

We next turned to simulations, where we investigated 1,440 simulated gene expression phenotypes capturing various ways in which ancestry induces heterogeneity in the exogenous E variable (simulation details in Lines 576-610 of Materials and Methods). We ran Stable PICS, as well as a version of PICS that did not residualize phenotypes or apply the stability principle. We observed that (i) causal variant recovery performance was not significantly different between the two approaches (Figure 2—figure supplements 24-32); but (ii) disagreement between the approaches can be considerable, especially when the signal-to-noise ratio is low (Supplementary File 2A). For example, in a set of simulations with three causal variants, with SNR = 0.11 and E heterogeneous by ancestry by letting E be drawn from N(2σ,σ²) for only GBR individuals (rest are N(0,σ²)), there was disagreement between Potential Set 1 and 2 variants in 25% of simulations — though recovery rates were similar (Probability of recovering at least one causal variant: 75% for Plain PICS and 80% for Stable PICS). These points suggest that confounding in cohorts can reduce power in methods not adjusting or accounting for ancestral heterogeneity, but can be remedied by approaches that do so. We report this analysis in Results “Simulations justify exploration of stability guidance”

In the current version of our work, we have evaluated, using both simulations and empirical evidence, different ways to combine approaches to boost causal variant recovery. Our simulation study shows that prioritizing matching variants across multiple methods improves causal variant recovery. On GEUVADIS data, where we might not know which variants are causal, we already demonstrated that matching variants are enriched for functional annotations. Therefore, our analyses justify that the adverse consequence of confounding on reducing fine-mapping accuracy can be mitigated by prioritizing matching variants between algorithms including those that account for stability.

(7) For non-matching variants, I wonder what the difference of posterior probabilities is between the stable and top variants in each method. If the difference is small, maybe it is due to noise rather than signal.

We have reported differences in posterior probabilities returned by Stable and Top PICS for GEUVADIS data; see Figure 3—figure supplement 1. For completeness, we compute the differences in posterior probabilities and summarize these differences both as histograms and as numerical summary statistics.

Potential Set 1

- Number of non-matching variants = 9,921

- Table of Summary Statistics of (Stable Posterior Probability – Top Posterior Probability)

Author response table 1.

- Histogram of (Stable Posterior Probability – Top Posterior Probability)

Author response image 1.

Potential Set 2

- Number of non-matching variants = 14,454

- Table of Summary Statistics of (Stable Posterior Probability – Top Posterior Probability)

Author response table 2.

- Histogram of (Stable Posterior Probability – Top Posterior Probability)

Author response image 2.

Potential Set 3

- Number of non-matching variants = 16,814

- Table of Summary Statistics of (Stable Posterior Probability – Top Posterior Probability)

Author response table 3.

- Histogram of (Stable Posterior Probability – Top Posterior Probability)

Author response image 3.

We also compared the difference in posterior probabilities between non-matching variants returned by Stable PICS and Top PICS for our 2,400 simulated gene expression phenotypes. Focusing on just Potential Set 1 variants, we find two equally likely scenarios, as demonstrated by two distinct clusters of points in a “posterior probability-posterior probability” plot. The first is, as pointed out, a small difference in posterior probability (points lying close to y=x). The second, however, reveals stable variants with very small posterior probability (of order 4 x 10^–5 to 0.05) but with a non-matching top variant taking on posterior probability well distributed along [0,1]. Moving down to Potential Sets 2 and 3, the distribution of pairs of posterior probabilities appears less clustered, indicating less tendency for posterior probability differences to be small ( Figure 2—figure supplement 8).

Here are the histograms and numerical summary statistics.

Potential Set 1

- Number of non-matching variants = 663 (out of 2,400)

- Table of Summary Statistics of (Stable Posterior Probability – Top Posterior Probability)

Author response table 4.

- Histogram of (Stable Posterior Probability – Top Posterior Probability)

Author response image 4.

Potential Set 2

Number of non-matching variants = 1,429 (out of 2,400)

- Table of Summary Statistics of (Stable Posterior Probability – Top Posterior Probability)

Author response table 5.

- Histogram of (Stable Posterior Probability – Top Posterior Probability)

Author response image 5.

Potential Set 3

- Number of non-matching variants = 1,810 (out of 2,400)

- Table of Summary Statistics of (Stable Posterior Probability – Top Posterior Probability)

Author response table 6.

- Histogram of (Stable Posterior Probability – Top Posterior Probability)

Author response image 6.

(8) It's a bit surprising that you observed matching variants with (stable) posterior probability ~ 0 (SFig. 1). What are the interpretations for these variants? Do you observe functional enrichment even for low posterior probability matching variants?

Thank you for this question. We have performed a thorough analysis of matching variants with very low stable posterior probability, which we define as having a posterior probability < 0.01 (Supplementary File 1 Section S11). Here, we briefly summarize the analysis and key findings.

Analysis

First, such variants occur very rarely — only 8 across all three potential sets in simulations, and 17 across all three potential sets for GEUVADIS (the latter variants are listed in Supplementary 2E). We begin interpreting these variants by looking at allele frequency heterogeneity by ancestry, support size — defined as the number of variants with positive posterior probability in the ALL slice* — and the number of slices including the stable variant (i.e., the stable variant reported positive posterior probability for the slice).

*Note that the stable variant posterior probability need not be at least 1/(Support Size). This is because the algorithm may have picked a SNP that has a lower posterior probability in the ALL slice (i.e., not the top variant) but happens to appear in the most number of other slices (i.e., a stable variant).

For variants arising from simulations, because we know the true causal variants, we check if these variants are causal. For GEUVADIS fine-mapped variants, we rely on functional annotations to compare their relative enrichment against other matching variants that did not have very low stable posterior probability.

Findings

While we caution against generalizing from observations reported here, which are based on very small sample sizes, we noticed the following. In simulations, matching variants with very low stable posterior probability are largely depleted in causal variants, although factors such as the number of slices including the stable variant may still be useful. In GEUVADIS, however, these variants can still be functionally enriched. We reported three examples in Supplementary File 1 Section S11 (pp. 8-9 of Supplement), where the variants were enriched in either VEP or biologically interpretable functional annotations, and were also reported in earlier studies. We partially reproduce our report below for convenience.

“However, we occasionally found variants that stand out for having large functional annotation scores. We list one below for each potential set.

- Potential Set 1 reported the variant rs12224894 from fine-mapping ENSG00000255284.1 (accession code AP006621.3) in Chromosome 11. This variant stood out for lying in the promoter flanking region of multiple cell types and being relatively enriched for GC content with a 75bp flanking region. This variant has been reported as a cis eQTL for AP006632 (using whole blood gene expression, rather than lymphoblastoid cell line gene expression in this study) in a clinical trial study of patients with systemic lupus erythematosus (Davenport et al., 2018). Its nearest gene is GATD1, a ubiquitously expressed gene that codes for a protein and is predicted to regulate enzymatic and catabolic activity. This variant appeared in all 6 slices, with a moderate support size of 23.

- Potential Set 2 reported the variant rs9912201 from fine-mapping ENSG00000108592.9 (mapped to FTSJ3) in Chromosome 17. Its FIRE score is 0.976, which is close to the maximum FIRE score reported across all Potential Set 2 matching variants. This variant has been reported as a SNP in high LD to a GWAS hit SNP rs7223966 in a pan-cancer study (Gong et al., 2018). This variant appeared in all 6 slices, with a moderate support size of 32.

- Potential Set 3 reported the variant rs625750 from fine-mapping ENSG00000254614.1 (mapped to CAPN1-AS1, an RNA gene) in Chromosome 11. Its FIRE score is 0.971 and its B statistic is 0.405 (region under selection), which lie at the extreme quantiles of the distributions of these scores for Potential Set 3 matching variants with stable posterior probability at least 0.01. Its associated mutation has been predicted to affect transcription factor binding, as computed using several position weight matrices (Kheradpour and Kellis, 2014). This variant appeared in just 3 slices, possibly owing to the considerable allele frequency difference between ancestries (maximum AF difference = 0.22). However, it has a small support size of 4 and a moderately high Top PICS posterior probability of 0.64.

To summarize, our analysis of GEUVADIS fine-mapped variants demonstrates that matching variants with very low stable posterior probability could still be functionally important, even for lower potential sets, conditional on supportive scores in interpretable features such as the number of slices containing the stable variant and the posterior probability support size…”

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

__Summary

Köver et al. examine the genetic and environmental underpinnings of multicellular-like phenotypes (MLPs) in fission yeast, studying 57 natural isolates of Schizosaccharomyces pombe. They uncover that a noteworthy subset of these isolates can develop MLPs, with the extent of these phenotypes varying according to growth media. Among these, two strains demonstrate pronounced MLP across a range of conditions. By genetically manipulating one strain with an MLP phenotype (distinct from the previously mentioned two strains), they provide evidence that genes such as MBX2 and SRB11 play a direct role in MLP formation, strengthening their genetic mapping findings. The study also reveals that while some key genes and their phenotypic effects are strikingly similar between budding and fission yeast, other aspects of MLP formation are not conserved, which is an intriguing finding.

Overall, the manuscript is well-written, dense yet logically structured, and the figures are well presented. The combination of phenotypic, genetic, and bioinformatics analyses, particularly from wet lab experiments, is commendable. The study addresses a significant gap in our understanding, primarily explored in budding yeast, by providing comprehensive data on MLP diversity in fission yeast and the interplay of genetic and environmental factors.

In summary, I enjoyed reading the manuscript and have only a few minor suggestions to strengthen the paper:

Minor revisions:

1. Although this may seem like a minor revision, but it is a crucial point. Please make sure that all raw data used to generate figures, run stats, sequence data, and scripts used to run data analysis are made publicly available. Provide relevant accession numbers and links to public data repositories. It is important that others can download the various types of data that went into the major conclusions of this paper in order to replicate your analysis or expand upon the scope of this work. I am not sure if the journal has a policy regarding this, but it should be followed to allow for transparency and reproducibility of the research.__

Reply: We very much agree with the reviewer that sharing raw data and scripts is an essential part of open science. All code and data are deposited to Github (https://github.com/BKover99/S.-Pombe-MLPs) and Figshare (https://figshare.com/articles/software/S_-Pombe-MLPs/25750980), which have now been updated to reflect our revisions. Additionally, the sequenced genomes have been deposited to ENA (PRJEB69522). Where external data was used, it was properly referenced and specifically included in Supplementary Table 3.

Two out of 57 strains exhibit strong and consistent MLP across multiple environments. Providing more information on these strains (JB914 and JB953), such as their natural habitats and distinct appearances of their MLP phenotypes under varying conditions, would provide valuable insights.

First, a brief discussion highlighting what differentiates these two strains from the rest would be helpful for readers (e.g. insight into their unique genetic and environmental background that might be linked to the MLP phenotype).

Additionally, culture tube and microscopy images of these strains, similar to those presented for JB759 in Figure 2A, can be included in the supplementary materials. My reasoning is that these images could help illustrate variation or lack thereof in aggregative group size across different media.

Reply: We thank the reviewer for highlighting this issue. Our further investigation into these strains has added additional interesting insights. JB914 and JB953 were isolated from molasses in Jamaica and the exudate of Eucalyptus in Australia, respectively, though it remains unclear whether these environments are related or even selective for the ability of these strains to form MLPs. We note that the environment from which a strain is isolated is an incomplete way of assessing its ecology. Indeed, recent research suggests that the primary habitat of S. pombe is honeybee honey and suggests that bees, which may be attracted to a number of sugary substances, may be a vector by which fission yeast are transported (1). Therefore, isolation from a particular nectar or food production environment might not reflect significant ecological differences. We now refer to the location of strain isolation in the manuscript text (lines 208-209).

However, there is more to learn from the genetic backgrounds of these two strains. We found that JB914 possesses the same variant in srb11 causally related to MLPs as JB759, the MLP-forming parental strain for our QTL analysis. To understand whether the appearance of this variant in these two strains derived from a single mutation event or was a case of convergent evolution, we analysed homology between the genomes of JB759 and JB914, focusing specifically on that variant. We found an approximately 20kb region of homology between JB759 and JB914 surrounding the srb11 truncation variant, in contrast to the majority of the genome, which does not share homology between those two strains (New Supplementary Figure 9A, B)). This result suggests that, while the two strains are largely unrelated, that specific region shares a recent common ancestor and is likely a result of interbreeding across strains.

Importantly, this analysis further emphasizes the point that the srb11 variant segregates with the MLP-forming phenotype. We conclude this because none of the other strains similar to JB759 (either across the whole genome, or specifically in the region surrounding srb11) exhibit MLPs (New Supplementary Figure 9C). This thereby further complements our QTL analysis on the significance of this variant. We have added this analysis to the manuscript text (lines 337-349).

Furthermore, we searched other strains which exhibited MLPs in our experiments (e.g. JB953) for frame shifts, insertions or deletions in any other genes in the CKM module or in the genes that were identified in our deletion library screen as adhesive, and did not identify any severe mutations falling into coding regions (other than the srb11 truncation in JB914 and JB759). This indicates that MLPs in these other strains may be caused by differences in regulatory regions surrounding these genes, or variants in other genes that were not identified in our screen. We have added this analysis to our manuscript (lines 424-425) and Supplementary Table 13.

We agree that microscopy and culture tube images of JB914 and JB953 may give insight into the nature of the MLPs exhibited by those strains. We have included such images of cultures grown in YES, EMM and EMM-Phosphate media in our revision (Lines 207-208, Supplementary Figures 4 and 5). These images are consistent with our adhesion assay screen and show that JB914 and JB953 are adhesive at the microscopic level in the relevant conditions (EMM or EMM-Phosphate).

The phenotypic outcome of overexpressing MXB2 is striking, as shown in Supplementary Figure 4C. Incorporating at least one of the culture tube images depicting large flocs into the main text, perhaps adjacent to Figure 3 panel D, would improve the visual appeal and highlight this key finding (at the moment those images are only shown in the supplementary materials).

Reply: We thank the reviewer for this suggestion. In response to Reviewer 2's suggestion to overexpress mbx2 in YES, we created new mbx2 overexpression strains that could overexpress mbx2 in YES, which was not possible in our previous strain in which mbx2 overexpression was triggered by removal of thymine from the media. We have replaced our original data from Figure 3D with data from the new mbx2 overexpression experiment, including flask images.

I know that the authors discuss the knowledge gap in the intro and results, but the abstract does not mention this critical gap. Please stress this critical gap (i.e., MLPs understudied in fission yeast) with a brief sentence in the abstract. Similarly, please consider writing a brief concluding sentence summarizing the paper's most significant finding referring to the knowledge gap would provide a clearer takeaway message for the reader - the abstract ends abruptly without any conclusion.

Reply: We agree and have now emphasized the critical gap in our abstract:

"As MLP formation remains understudied in fission yeast compared to budding yeast, we aimed to narrow this gap." at lines 18-19.

Additionally, we added the following final sentence to give the reader a clearer takeaway message:

"Our findings provide a comprehensive genetic survey of MLP formation in fission yeast, and a functional description of a causal mutation that drives MLP formation in nature." at lines 31-32.

1. The observation that strains with adhesive phenotypes have a lower growth rate compared to non-adhesive strains is a noteworthy point (lines 532-535). This represents yet another example of this classical trade-off. This point could be emphasized in the Discussion or alongside the relevant result, with a brief speculative explanation for this phenomenon.

Reply: We agree that the nature of the trade-off between MLP formation is an interesting discussion point that could arise from our work. Understanding this trade-off is made more complicated by the fact that growth is always condition-dependent, and measuring growth in strains exhibiting MLPs is non-trivial, as adhesion to labware and thick clumps of cells separated by regions of cell-free media can add variability. Nonetheless, there has been some previous work on this problem. In S. cerevisiae, it was shown that larger group size correlates with slower growth rate (3), and that flocculating cells grow more slowly (4). In S. cerevisiae, cAMP, a signalling molecule heavily involved in regulating growth in response to nutrient availability, also regulates filamentation (5). However, the relationship between flocculation and slow growth is not consistent in the literature. In some settings overexpressing the flocculins FLO8, FLO5, and FLO10 results in slower growth (6), while in others it does not (7). In addition, ethanol production has been shown to improve for biofilms (7).

Furthermore, in S. cerevisiae, MLP-forming cells grow better in low sucrose concentrations (8) and under various stress conditions (4). Flocculating cells have also shown faster fermentation in media containing common industrial bioproduction inhibitors, despite slower fermentation than non-flocculating cells in non-inhibitory media (9). However, any consequence of this possible advantage on growth has not been characterised.

In S. pombe, there is less work on this topic; however, it has been shown that deletions of rpl3201 and rpl3202, which code for ribosomal proteins, cause flocculation and slow growth (10). In that case, it is not clear if there is any causal relationship between slow growth and flocculation or if they are both parallel consequences of the ribosomal pathway disruption. We have added some of these points to the portion of the discussion that discusses this tradeoff (Lines 477-499).

To get a better understanding of this tradeoff in our system, we took several approaches. First, we added a supporting analysis (New Supplementary Figure 12B), using published growth data based on measurements on agar plates for the S. pombe gene deletion library (11). There, the authors defined a set of deletion strains that grow more slowly on EMM than the wild-type lab strain. We found that our MLP hit strains were significantly enriched in this "EMM-slow" category. This information is now included in the manuscript (Lines 409-413, New Supplementary Figure 12B).

It is, however, possible that for the assays from that work, the appearance of slow growth on solid agar in adhesive cells could be partially artifactual. Indeed, we have observed that adhesive cells tend to stick to flasks and, when grown on agar plates, cells in the same colony can stick to one another rather than to inoculation loops or pin pads. Both of these dynamics can reduce initial inoculation densities. This is less of a concern for our adhesion assay and Figures 2E, 5B, and 5F, because our before-wash intensity was done with a 7x7 pinned square about 10x10 mm2. Nonetheless, as we wanted to make a point about srb10 and srb11 mutants growing faster than other deletion mutants that exhibit MLP-formation, we also conducted growth assays in liquid media (New Figure 5F).

We observed that srb10Δ and srb11Δ strains (which exhibit MLPs in EMM) show growth curves similar to wild-type cells in minimal (EMM) and rich media (YES). On the other hand, other strains that grow similarly to wild type cells in YES, such as tlg2Δ and rpa12Δ, grow much more slowly in EMM when they clump together. There are also some strains, mus7Δ and kgd2Δ, that grow more slowly in both YES and EMM but are only adhesive in EMM.

The text mentions two lab strains, JB22 and JB50, displaying strong adhesion under phosphate starvation (lines 525-526), yet the data point for JB22 in Figure 2C is not labeled.

Reply: We agree that highlighting JB22 on the figure is crucial, given that it was mentioned in the main text. JB22 is now highlighted in green on Fig 2C.

1. Although I generally avoid commenting on formatting, I found the manuscript to be dense. As mentioned above, I truly enjoyed reading it! But I couldn't help but think of ways to make the manuscript more concise for readers. The Results section spans nine pages (excluding figure captions), and the Discussion is five pages long. The main text contains 6 figures with approximately 27 panels and 32 plots and Venn diagrams, while the supplementary material has 11 figures with 22 panels and about 59 plots. Altogether, the manuscript comprises 17 figures, 49 panels, and roughly 91 plots and Venn diagrams! While I will not request any changes, I encourage the authors to consider streamlining the text/data where possible to focus on the core theme of the study.

We thank the reviewer for these suggestions and have reorganised some of our figures and text to appear less dense. We have also added several figures and panels in response to reviewer comments. While we endeavor to make our points clear and concise in the main figures, we believe that it is important to retain key supplementary figures so that an interested reader can evaluate the data in more detail:

A summary of our major changes to the figures is below, and we also provide a manuscript with changes tracked for the reviewers' convenience:

Fig 2:

Added Panel E in response to reviewer comments. Fig 3:

Removed axes for pfl3 and pfl7 from Fig 3C, as the point was made by the other genes displayed (mbx2, pfl8 and gsf2) Replaced Fig 3D with similar data from an improved experiment in response to reviewer comments. Added New Fig 3F from Original Supp Fig 5 Fig 5:

Moved Original Fig 5A to New Supp Fig 10A. Added New Fig 5F in response to reviewer comments. Original Supp Fig 4 / New Supp Fig 6:

Removed mbx2 overexpression images from Original Fig 4C, to be replaced by new overexpression data and images in New Fig 3D. Added flask images for srb10 and srb11 deletion mutants from Original Supp Fig 5A to New Supp Fig 6C. Added microscope image for srb11 deletion mutant from Ooriginal Supp Fig 5A to New Supp Fig 6C. Added adhesion assay results from Original Supp Fig 5C to New Supp Fig 6C. Added New Supp Fig 6D in response to review Original Supp Fig 5

Removed this figure. Original Supp Fig 5A and 5B were moved to New Supp Fig 6. Original Supp Fig 5B was removed to make the manuscript more concise. Original Supp Figs 6, 7 and 8 were combined into New Supp Fig 8.

Original Supp Fig 6A and 6B are now New Supp Fig 8A and 8B. Original Supp Fig 7 is now New Supp Fig 8C. Original Supp Fig 8A is now New Supp Fig 8D and 8E. Original Supp Fig 8B is now New Supp Fig 8F Original Supp Fig 9/New Supp Fig 10

Added Original Fig 5A as new Supp Fig 10A. Original Supp Fig 11/New Supp Fig 12

Removed Original Fig 11B and the relevant text to make the manuscript more concise. Added New Supp Fig 12B in response to reviewer comments. New Supplementary Figures added in response to reviewer comments:

New Supp Fig 4: Microscopy images of natural isolates. New Supp Fig 5: Flask images of natural isolates New Supp Fig 7: Microscopy and flask images of mbx2 overexpression strains. New Supp Fig 9: Genomic comparisons between JB759 and the MLP-forming wild isolate, JB914. Removed some less relevant points from our discussion, to reduce the length.

Added new Supplementary Tables:

Supplementary Table 13: Variants in candidate genes. Added in response to reviewer comments Supplementary Table 14: List of plasmids used in the study.

**Referees cross-commenting**

There are many useful recommendations from all the other reviewers that will help improve the final product. Once those points are revised, I think this will be a nice paper of interest to folks interested in natural variation in MLPs and its genetic background.

Significance

My expertise: evolutionary genetics, evolution of multicellularity, yeast genetics, experimental evolution

Overall, the manuscript is well-written, dense yet logically structured, and the figures are well presented. The combination of phenotypic, genetic, and bioinformatics analyses, particularly from wet lab experiments, is commendable. The study addresses a significant gap in our understanding, primarily explored in budding yeast, by providing comprehensive data on MLP diversity in fission yeast and the interplay of genetic and environmental factors.

In summary, I enjoyed reading the manuscript and have only a few minor suggestions to strengthen the paper.

Reviewer #2

Evidence, reproducibility and clarity

REVIEWER COMMENTS

Yeast species, including fission yeast and budding yeast, could form multicellular-like phenotypes (MLP). In this work, Kӧvér and colleagues found most proteins involved in MLP formation are not functionally conserved between S. pombe and budding yeast by bioinformatic analysis. The authors analyzed 57 natural S. pombe isolates and found MLP formation to widely vary across different nutrient and drug conditions. The authors demonstrate that MLP formation correlated with expression levels of the transcription factor gene mbx2 and several flocculins. The authors also show that Cdk8 kinase module and srub11 deletions also resulted in MLP formation. The experimental design is logic, the manuscript is well-written and organized. I have a few concerns that should be addressed before the publication.

Major points:

1) Line 61-62, how did the authors grow yeast cells in the liquid medium? Shaking or static? If shaking, the nutrient should be even distributed in the medium.

If static culture, most single yeast cells could precipitate on the bottom, how do you address the advantage of flocculation for increasing the sedimentation? In addition, under static culture, the bottom will have less air than the up medium, how to balance the air and nutrients?

Reply: In line 61-62 we stated that "Similarly, flocculation could increase sedimentation in liquid media, thereby assisting the search for more nutrient-rich or less stressful environments (4)".

Our intent was to speculate on the advantages of multicellular-like growth, and cited a review article which has mentioned sedimentation. After further consideration, we decided that this is a minor point and is rather speculative, and removed it altogether from the manuscript.

In response to the Reviewer's question about how cells were grown in liquid medium, throughout the paper we used shaking cultures for our flocculation assays and for pre-cultures. We have made this more clear in the text where it was ambiguous (e.g. line 189, throughout the methods section, and in the legend of Fig. 2A).

2) Line 555, it will be interesting to test whether overexpression of mbx2 could cause flocculation in YES medium. In Figure 3D, the authors use two control strains, but only one mbx2 OE strain, mbx2 OE should be tested in both strains. In addition, did the authors transform empty plasmid into the control strains, please indicate in the figure.

In this experiment, mbx2 was overexpressed using a thiamine-repressible nmt1 promoter, which is a standard construct in fission yeast studies. Assaying MLP formation was not feasible in YES with this strain, because YES is a rich media made up of yeast extract which contains thiamine. Thus, we could not remove thiamine from the media to trigger mbx2 overexpression.

In order to test the influence of mbx2 overexpression in YES, we constructed strains in which mbx2 was integrated into the genome and expression was driven by the rpl2102 promoter, which has been shown to provide constitutive moderate expression levels (12). We observed strong flocculation in both EMM and YES (Fig 3D, New Supplementary Figure 7) . We did not see strong flocculation in a control in which GFP was expressed under the rpl2102 promoter. The flocculation phenotype was so strong that our original adhesion assay protocol required modification for this experiment, including resuspension in 10 mM EDTA before repinning (Methods). We observed strong adhesion for the mbx2 overexpression strains (Fig 3D), but not for control strains in YES. We could not check adhesion in EMM for those strains because cells pinned on EMM did not survive resuspension in EDTA.

We performed these experiments in two backgrounds, 968 h90 (JB50), which is one of the parental strains of the segregant library analysed in Figure 3 and 972 h- (JB22), which is an appropriate background for the gene deletion collection.

We have replaced the data from the original Figure 3D with the new adhesion assay and added New Supplementary Figure 7 to the manuscript (Lines 236-244).

This result also helped us to further refine our model for the pathway. We can now say that the repression of MLPs in rich media must act via Mbx2, as overexpression of mbx2 is sufficient to abolish it, and is likely to act transcriptionally (if it acted on the protein level, the mild overexpression would likely not have led to the phenotype) (Figure 6, Lines 554-556 in the discussion)

3) Line 600-601, the authors may do the backcross of srb11Δ::Kan to exclude the possibility caused by other mutations.

Reply: We thank the reviewer for noticing our concern about suppressor mutations arising in the srb11Δ strain obtained from our deletion library. This initial concern arose following the observation that while qualitatively the srb11Δ::Kan and srb11Δ(CRISPR) strains were both strongly adhesive, there was a minor quantitative difference in their adhesion.

As we obtained this strain from an h+ deletion library strain backcrossed with a prototrophic h- strain (JB22) in order to restore auxotrophies (13), the chances for a suppressor mutation to arise are very low. We have therefore removed that language from our text. We now suspect that a more likely explanation for this small difference could be the strain background, as our CRISPR engineered strain was made in a JB50 background which has the h90 mating type, while the deletion library strains are h- without auxotrophic markers.

We would like to emphasize, however, that despite this quantitative difference in the adhesion phenotype between the two srb11Δ strains, they both have a large increase in the adhesion phenotype relative to the respective wild-type strains. To address this point, we have removed the unnecessary statistical comparison of these two deletion strains and focused on their qualitatively high levels of adhesion in the text (lines 267-269) and in our Revised Supplementary Figure 6D.

Minor points:

1) Line 506, what are the growth conditions of cells in Figure 2A? Did the authors use the liquid or solid medium? Please mention in the Methods or figure legends.

Reply: We have updated the manuscript to include the relevant details in the text (line 189), figure caption for Fig. 2A and in the methods section (lines 829-831).

2) Line 533-535, please explain why the strains exhibiting strong adhesion have a decreased growth rate. Is there any related research? Please add some references.

Reply: Please see reply to Reviewer 1, comment 5.

**Referees cross-commenting**

I agree with most of the comments from other reviewers. This publication may indeed be of interest to a minor area. But the results and the interpretations of the data are interesting and warranted, the findings are scientifically important.

Significance

The authors did many large-scale screens and bioinformatic analyses. The experiments in the manuscript are generally logical and sound. This study is useful for deciphering the mechanism of multicellular-like phenotype formation in the fission yeast, with some implications for some other organisms.

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

Summary: Using a variety of targeted and genome wide analyses, the authors investigate the basis for "multicellular-like phenotypes" in S. pombe. Authors developed several methodologies to detect and quantify "multicellular-like phenotypes" (flocculation, aggregation...) and defined genes involved in these processes in laboratory and wild S. pombe.

SECTION A - Evidence, reproducibility and clarity

This is a very solid manuscript that is well-written and supported by convincing data. While one can imagine many additional experiments, the manuscript stands on its own and presents a quite exhaustive analysis of the area. I commend the author for their rigorous work and clear presentation. They are only a few minor points that warrant comments or corrections: - Supplementary Figure 1 is a typical example of the "necessity" to have statistics and P-values everywhere. The data are convincing but what is the evidence that the Filtering assay and the Plate-reader assay values should be linearly related? Lets imagine that Plate-reader assay value is proportional to the square of the Filtering assay value. What would be the Pearson R and P-value in this case? What is most appropriate? Why would one use a linear correlation? What is the "real" significance?

Reply: We thank the reviewer for pointing out that the data in Supplementary Figure 1 does not appear to be linear and, therefore, reporting the Pearson correlation coefficient may not be the best way to represent the relationship between the two assays. The nonlinear nature of this data could indicate that

The filtering assay saturates before the plate reader assay, and is less able to distinguish between strains that flocculate strongly and The filtering assay may be more sensitive for strains that show lower levels of flocculation. In general, we observed fewer strains with intermediate phenotypes for both assays, making it difficult to ascertain the true relationship between them; however, we believe that the key result is that the strains with the highest level of flocculation have the highest values in both assays. To capture this aspect of the data, we now report the Spearman correlation which is non-parametric and indicates how similar the ranking of each strain is based on both assays. With the alternative hypothesis being that the correlation is > 0, we report a Spearman correlation coefficient of 0.24 and a P-value of 0.04 (lines 823-826)

• Minor points: * They are several "personal communications" in the manuscript (page 11, page 18, page 23). It should be checked whether this is accepted in the journal that publishes this manuscript.

Reply: We thank the reviewer for highlighting this issue. We had three instances of "personal communications" in our original submission.

The first instance was an acknowledgement for advice on our DNA extraction protocol from Dan Jeffares. We now include this in the Acknowledgements section instead.

The second communication with Angad Garg described that they observed flocculation while growing cells in phosphate starvation conditions, which was not reported in their publication (14). Though we appreciate their willingness to share unpublished data with us, we have removed this observation from our manuscript and instead rely only on our own observations and arguments based on their published RNA-seq data to make our point.

The third personal communication with Olivia Hillson supplements a minor hypothesis, namely that deletion of SPNCRNA.781 might cause MLP formation by affecting the promoter of hsr1, for which we had access to unpublished ChIP-seq data, showing its binding to flocculins. Recently published work from a different group (15) also suggests this link between hsr1 and flocculation and is now discussed in our manuscript instead of the result based on unpublished data obtained from personal communication at Lines 397-398.

* Page 4 check "a few regulators"

Reply: For clarity, this has now been changed to "several regulatory proteins" at Line 108. The specific proteins we are referring to are highlighted in Figure 1C.

* Page 19, line 567: "remaining 8 strains" may be confusing as Material and Methods states "remaining 10 strains".

Reply: Two of the 10 strains were found to be redundant after sequencing as explained in the Methods (Lines 930-934). Therefore, we only added 8 new strains to the analysis. We thank the reviewer for highlighting this as a potential source of misunderstanding, and clarified this point in the text (Lines 247-250 and in the methods).

**Referees cross-commenting**

I concur with most comments. Overall, the reviewers agree that this is a solid piece of work that could benefit from minor modifications and should be published. I reiterate that, for me, despite its quality, this publication will only be of interest to specialists.

Reviewer #3 (Significance (Required)):

A limited number of studies have investigated "multicellular-like phenotypes" in S. pombe. This manuscript brings therefore new and solid information. Yet, despite an impressive amount of work, our conceptual advance in understanding this process and its phylogenetic conservation remains limited. This is probably best illustrated in the figure 6 that summarize the study and contains 3 question marks and an additional unknown mechanism. (Most of the solid arrows in this figure correspond to interactions within the Mediator complex that were well known before this study.) In addition, while only few studies have been published in this area, the authors' findings are often only bringing additional support to already published observations. Overall, while this manuscript will be of interest to a restricted group of aficionados, it will most likely not attract the attention of a wide readership.

__ Reviewer #4 (Evidence, reproducibility and clarity (Required)):__

In this manuscript, the authors explore how multicellular-like phenotypes (MLPs) arise in the fission yeast S. pombe. Although yeasts are characterized as unicellular fungi, diverse species show MLPs, including filamentous growth on agar plates and flocculation in liquid media. MLPs may provide certain advantages in nutritionally poor conditions and protection against external challenges, upon which natural selection can then act. Previous work on MLPs has mostly been carried out in the budding yeasts S. cerevisiae and C. albicans, and little was known about these behaviors in S. pombe. The authors thus set out to investigate both genetic and environmental regulators of MLP formation.

First, their analysis of published data revealed a limited number of shared regulators of MLP between S. pombe, S. cerevisiae, and C. albicans, although the cell adhesion proteins themselves are largely not conserved. Next, the authors screened a set of non-clonal natural isolates using two high-throughput assays that they developed and found that MLPs vary in strains and depending on nutrient conditions. Focusing on a natural isolate that showed both adhesion on agar plates and flocculation in liquid medium, they then analyzed a segregant library generated from this and a laboratory strain using their assays. Using QTL analysis, they uncovered a frameshift in the srb11 gene, which encodes a subunit of the Mediator complex, as the likely causal inducer of MLP. This was confirmed by additional analyses of strains lacking srb11 or other members of Mediator. Furthermore, the authors showed that loss of srb11 function resulted in the upregulation of the Mbx2 transcription factor, which was both necessary and sufficient for MLP formation in this background. Finally, screening of two additional yeast strain collections (gene and long intergenic non-coding RNA deletion) identified both known and novel regulators representing different pathways that may be involved in MLP formation.

Altogether, this study provides new perspectives into our understanding of the diverse inputs that regulate multicellular-like phenotypes in yeast.

Major comments:

• The methods for screening for adhesion and flocculation are well described, with representative figures that show plates and flasks. However, there are few microscopy images of cells, and it would be interesting and helpful for the reader to have an idea of how cells look when they exhibit MLPs. For instance, are there any differences in cell shape or size when strains present different degrees of adhesion or flocculation? In addition, the authors mention that mutants with strong adhesion generally had lower colony density and are likely to be slower growing. Although their analyses suggest otherwise (page 22), this has a potential for introducing error in their observations, and including images of the adhesion/flocculation phenotypes may provide further support for their conclusions. I suggest that the authors present microscopy images 1) similar to what is shown for JB759 in Figure 2A and 2) of cells growing on agar in the adhesion assay. This could be included for the different Mediator subunit deletions that they tested, where there appear to be varying phenotypes. It could also be informative for a subset of the 31 high-confidence candidates that they identified in their screen.

Reply: We thank the reviewer for highlighting the need for further microscopic characterisation of MLP forming strains. We therefore now include images of JB914, JB953 (New Supplementary Figures 4, Figure 2E) in liquid media in EMM, EMM-Phosphate, and YES; an srb11 deletion strain (Figure 3F), and mbx2 overexpression strains (New Supplementary Figure 7).

• Upon identifying a frameshift in srb11 that is responsible for the MLP, the authors assessed whether deletion of other Mediator subunits would result in the same phenotype. They found that srb10 and srb11 deletions both flocculate and show adhesion, while other mutants had milder phenotypes. However, the authors also found that a new deletion of srb11 that they generated had a stronger adhesion phenotype than the srb11 deletion from the prototrophic deletion library, which was attributed this the accumulation of suppressor mutations in the strains of the deletion collection. As the authors make clear distinctions between the phenotypes of different Mediator mutants, I suggest generating and analyzing "clean" deletions of the 6 other subunits that they tested. This would strengthen their conclusion and help to rule out accumulated suppressors as the cause of the differences in the observed phenotypes.

Reply: We thank the reviewer for noticing our concern about suppressor mutations in the manuscript. As we describe above in response to a similar question from reviewer 2, as the prototrophic deletion library from which we extracted the Mediator deletion strains had been backcrossed during its construction (13), we no longer suspect that small difference between the srb11Δ::Kan strain from the deletion library and the newly created srb11Δ (CRISPR) strains is due to suppressor mutations. Rather, we think they may be a result of the difference in genetic background and possibly mating type between the two strains. We also want to emphasize that this difference is small compared to the difference between the adhesion ratios of the srb11Δ strains and their respective control strains.

Nevertheless, we made clean, independent Mediator mutants for 5 out of 6 Mediator genes tested (med10Δ, med13Δ, med19Δ, med27Δ, and srb10Δ) as well as an additional mutant that we didn't have in our library, med12Δ (Figure R9). When running the assay on these new strains we got an overall lower dynamic range, possibly due to variations in the water flow rate relative to the first assay. However, we saw a strong phenotype for both library and our own srb10Δ and CRISPR srb11Δ strains. We did not see a significant increase in adhesion for the other Mediator deletion mutants in EMM relative to wild type with the exception of for med10Δ in both the library strain and for our clean mutant, for which we did not observe a phenotype in our previous experiment. We included the experiment for the newly created mutants as New Supplementary Figure S6E and described them in lines 276-281 in our revised manuscript.

Minor comments:

• One point that recurs in the manuscript is the idea that mutations that give rise to strong MLPs also generally lead to slower growth, representing a potential trade-off. This idea could be reinforced with measurements of growth rate or generation time by optical density or cell number, for instance, rather than comparisons of colony density. Also, it would be interesting to mention if the slow growth phenotype is only observed in MLP-inducing conditions or also in rich medium.

Reply: As described above in response to item 5 from Reviewer 1, we have conducted growth assays in liquid media for srb10Δ, srb11Δ, and other mutants from our adhesion screen (tlg2Δ, rpa12Δ, mus7Δ and kgd2Δ) that showed a similar phenotype to those genes in both minimal (EMM) and rich (YES) media. We observe that in rich media, srb10Δ and srb11Δ cells grow similarly to control strains, and they exhibit a lower decrease in growth rate than the other similarly adhesive strains. Both mus7Δ and kgd2Δ cells grow more slowly, even in rich media.

We have also added data on the tradeoff between growth and adhesion based on growth on solid media from (11) for all mutants identified in our screen (New Supp Fig 12B)).

Thus, the relationship between slow growth and clumpiness depends on the mutation, and specifically, mutations of the Mediator, including those to srb11 and srb10, seem to decrease the impact of any tradeoff between growth and adhesion.

• The authors show that the MLPs of the srb10 and srb11 deletions occur through mbx2 upregulation. Do the varying strengths of the phenotypes of the strains lacking different Mediator subunits correlate with mbx2 levels in these backgrounds?

Reply: There is some evidence from previous work that the relationship between the strength of the MLPs and the expression of mbx2 may not be perfectly proportional. In (16), med12Δ had a higher (though qualitatively comparable) level of mbx2 upregulation than srb10Δ (New Supp Fig 8E), even though that paper reported a milder phenotype for med12Δ than for srb10Δ cells. We did not observe a significant increase in adhesion in our med12Δ strain (New Supp Fig 6D). This suggests that in the case of these mutants, it is not simply the level of mbx2 that controls MLP formation, but that there are likely additional regulatory mechanisms. We have added some discussion on this context in the manuscript (lines 545-547).

**Referees cross-commenting**

I agree overall with the comments and suggestions from the other reviewers. The revision would require only minor modifications. The paper is interesting both for the combination of methodologies used and its findings, and I believe that it would benefit a growing community of researchers.

Reviewer #4 (Significance (Required)):

This study employed a variety of methods that allowed the authors to uncover previously unknown regulators of MLPs. Taking advantage of the diversity of natural fission yeast isolates as well as the constructed gene and non-coding RNA deletion collections, the authors identified novel genetic determinants that give rise to MLPs, opening new avenues into this exciting area of research. The overall conclusions of the work are solid and supported by the reported results and analyses. This study will be appreciated by a broad audience of readers who are interested in understanding how organisms respond to environmental challenges as well as how MLPs may result in emergent properties that play key roles in these responses. Some of the limitations of the work are described above, with recommendations for addressing these points.

Keywords for my field of expertise: fission yeast, cell cycle, transcription, replication.

References for Response to Reviews

1. Brysch-Herzberg M, Jia GS, Seidel M, Assali I, Du LL. Insights into the ecology of Schizosaccharomyces species in natural and artificial habitats. Antonie Van Leeuwenhoek. 2022 May 1;115(5):661-95.
2. Jeffares DC, Rallis C, Rieux A, Speed D, Převorovský M, Mourier T, et al. The genomic and phenotypic diversity of Schizosaccharomyces pombe. Nat Genet. 2015 Mar;47(3):235-41.
3. Ratcliff WC, Denison RF, Borrello M, Travisano M. Experimental evolution of multicellularity. Proc Natl Acad Sci. 2012 Jan 31;109(5):1595-600.
4. Smukalla S, Caldara M, Pochet N, Beauvais A, Guadagnini S, Yan C, et al. FLO1 is a variable green beard gene that drives biofilm-like cooperation in budding yeast. Cell. 2008 Nov 14;135(4):726-37.
5. Lorenz MC, Heitman J. Yeast pseudohyphal growth is regulated by GPA2, a G protein alpha homolog. EMBO J. 1997 Dec 1;16(23):7008-18.
6. Ignacia DGL, Bennis NX, Wheeler C, Tu LCL, Keijzer J, Cardoso CC, et al. Functional analysis of Saccharomyces cerevisiae FLO genes through optogenetic control. FEMS Yeast Res. 2025 Sept 24;25:foaf057.
7. Wang Z, Xu W, Gao Y, Zha M, Zhang D, Peng X, et al. Engineering Saccharomyces cerevisiae for improved biofilm formation and ethanol production in continuous fermentation. Biotechnol Biofuels Bioprod. 2023 July 31;16(1):119.
8. Koschwanez JH, Foster KR, Murray AW. Improved use of a public good selects for the evolution of undifferentiated multicellularity. eLife. 2013 Apr 2;2:e00367.
9. Westman JO, Mapelli V, Taherzadeh MJ, Franzén CJ. Flocculation Causes Inhibitor Tolerance in Saccharomyces cerevisiae for Second-Generation Bioethanol Production. Appl Environ Microbiol. 2014 Nov;80(22):6908-18.
10. Li R, Li X, Sun L, Chen F, Liu Z, Gu Y, et al. Reduction of Ribosome Level Triggers Flocculation of Fission Yeast Cells. Eukaryot Cell. 2013 Mar;12(3):450-9.
11. Rodríguez-López M, Bordin N, Lees J, Scholes H, Hassan S, Saintain Q, et al. Broad functional profiling of fission yeast proteins using phenomics and machine learning. Marston AL, James DE, editors. eLife. 2023 Oct 3;12:RP88229.
12. Hebra T, Smrčková H, Elkatmis B, Převorovský M, Pluskal T. POMBOX: A Fission Yeast Cloning Toolkit for Molecular and Synthetic Biology. ACS Synth Biol. 2024 Feb 16;13(2):558-67.
13. Malecki M, Bähler J. Identifying genes required for respiratory growth of fission yeast. Wellcome Open Res. 2016 Nov 15;1:12.
14. Garg A, Sanchez AM, Miele M, Schwer B, Shuman S. Cellular responses to long-term phosphate starvation of fission yeast: Maf1 determines fate choice between quiescence and death associated with aberrant tRNA biogenesis. Nucleic Acids Res. 2023 Feb 16;51(7):3094-115.
15. Ohsawa S, Schwaiger M, Iesmantavicius V, Hashimoto R, Moriyama H, Matoba H, et al. Nitrogen signaling factor triggers a respiration-like gene expression program in fission yeast. EMBO J. 2024 Oct 15;43(20):4604-24.
16. Linder T, Rasmussen NN, Samuelsen CO, Chatzidaki E, Baraznenok V, Beve J, et al. Two conserved modules of Schizosaccharomyces pombe Mediator regulate distinct cellular pathways. Nucleic Acids Res. 2008 May;36(8):2489-504.

Reviewer #2 (Public review):

Summary:

Negreira et al. present an application of a novel single-cell genomics approach to investigate the genetic heterogeneity of Leishmania parasites. Leishmania, while also representing a major global disease with hundreds of thousands of cases annually, serves as a model to test the rigor of the sequencing strategy. Its complex karyotypic nature necessitates a method that is capable of resolving natural variation to better understand genome dynamics. Importantly, an earlier single-cell genomics platform (10x Chromium) is no longer available, and new methods need to be evaluated to fill in this gap.

The study was designed to evaluate whether a capsule-based cell capture method combined with primary template-directed amplification (PTA) could maintain levels of genomic heterogeneity represented in an equal mixture of two Leishmania strains. This was a high bar, given the relatively small protozoan genome and prior studies that showed limitations of single-cell genomics, especially for gene-level copy number changes. Overall, the study found that semi-permeable capsules (SPC) are an effective way to isolate high-quality single cells. Additionally, short reads from amplified genomes effectively maintained the relative levels of variation in the two strains on the chromosome, gene copy, and individual base level. Thus, this method will be useful to evaluate adaptive strategies of Leishmania. Many researchers will also refer to these studies to set up SPC collection and PTA methods for their organism of choice.

Strengths:

(1) The use of SPC and PTA in a non-bacterial organism is novel. The study displays the utility of these methods to isolate and amplify single genomes to a level that can be sequenced, despite being a motile organism with a GC-rich genome.

(2) The authors clearly outlined their optimization strategy and provided numerous quality-control metrics that inspire confidence in the success of achieving even chromosomal coverage relative to ploidy.

(3) The use of two distinct Leishmania strains with known clonal status provided strong evidence that PTA-based amplification could reflect genome differences and displayed the utility of the method for studies of rare genotypes.

(4) Evaluating the SPCs pre- and post-amplification with microscopy is a practical and robust way of determining the success of SPC formation and PTA.

(5) The authors show that the PTA-based approach easily resolved major genotypic ploidy in agreement with a prior 10x Chromium-based study. The new method had improved resolution of drug resistance genotypes in the form of both copy-number variations and single-nucleotide polymorphisms.

(6) In general, the authors are very thorough in describing the methods, including those used to optimize PTA lysis and amplification steps (fresh vs frozen cells, naked DNA vs sorted cells, etc). This demonstrates a depth of knowledge about the procedure and leaves few unanswered questions.

(7) The custom, multifaceted, computational assessment of coverage evenness is a major strength of the study and demonstrates that the authors acknowledge potential computational factors that could impact the analysis.

Weaknesses:

(1) The rationale behind some experimental/analysis choices is not well-described. For example, the rationale behind methanol fixation and heat-lysis is unclear. Additionally, the choice of various methods to assess "evenness" is not justified (e.g. why are multiple methods needed? What is the strength of each method?). Also, there is no justification for using 100k reads for subsampling. Finally, what exactly constitutes a "confidently-called SNP"?

(2) In the methods, the STD protocol lists a 15-minute amplification at 45C whereas the PTA protocol involves 10h at 37C. This is a dramatic difference in incubation time and should be addressed when comparing results from the two methods. It is not really a fair comparison when you look at coverage levels; of course, a 10-hour incubation is going to yield more reads than a 15-minute incubation.

(3) There is a lack of quantitative evaluations of the SPCs. e.g. How many capsules were evaluated to assess doublets? How many capsules were detected as Syto5 positive in a successful vs an unsuccessful experiment?

(4) The authors do not address some of the amplification results obtained under various conditions. For example, why did temperature-based lysis of STD4 lead to amplification failure? Also, what is the reason for fewer "true" cells (higher background) in the PTA samples compared to the STD samples? Is this related to issues with barcoding or, alternatively, substandard amplification as indicated by lower read amounts in some capsules (knee plots in Figure 1C)?

(5) The paper presents limited biological relevance. Without this, the paper describes an improvement in genome amplification methods and some proof-of-concept analyses. Using a 1:1 mixture of parasites with different genotypes, the authors display the utility of the method to resolve genetic diversity, but they don't seek to understand the limits of detecting this diversity. For some, the authors do not comment on the mixed karyotypes from the HU3 cells (Figure 3F) other than to state that this line was not clonal. For CNVs, the two loci evaluated were detected at relatively high copy number (according to Figure 4C, they are between 4 and 20 copies). Thus, the sensitivity of CNV detection from this data remains unclear; can this approach detect lower-level CNVs like duplications, or minor CNVs that do not show up in every cell?

(6) The authors state that Leishmania can carry extrachromosomal copies of important genes. There is no discussion about how the presence of these molecules would affect the amplification steps and CNV detection. For example, the phi29 enzyme is very processive with circular molecules; does its presence lead to overamplification and overrepresentation in the data? Is this evident in the current study? This information would be useful for organisms that carry this type of genetic element.

(7) The manuscript is missing a comparison with other similar studies in the field. For example, how does this coverage level compare to those achieved for other genomes? Can this method achieve amplification levels needed to assess larger genomes? Has there been any evaluation of base composition effects since Leishmania is a GC-rich genome?

(8) Cost is mentioned as a benefit of the SPC platform, and savings are achieved when working in a plate format, but no details are included on how this was evaluated.

(9) The Zenodo link for custom scripts does not exist, and code cannot be evaluated.

Author response:

Reviewer #1 (Public review):

Summary:

Negreira, G. et al clearly presented the challenges of conducting genomic studies in unicellular pathogens and of addressing questions related to the balance between genome integrity and instability, pivotal for survival under the stressful conditions these organisms face and for their evolutionary success. This underlies the need for powerful approaches to perform single-cell DNA analyses suited to the small and plastic Leishmania genome. Accordingly, their goal was to develop such a novel method and demonstrate its robustness.

In this study, the authors combined semi-permeable capsules (SPCs) with primary template-directed amplification (PTA) and adapted the system to the Leishmania genome, which is about 100 times smaller than the human genome and exhibits remarkable plasticity and mosaic aneuploidy. Given the size and organization of the Leishmania genome, the challenges were substantial; nevertheless, the authors successfully demonstrated that PTA not only works for Leishmania but also represents a significantly improved whole-genome amplification (WGA) method compared with standard approaches. They showed that SPCs provide a superior alternative for cell encapsulation, increasing throughput. The methodology enabled high-resolution karyotyping and the detection of fine-scale copy number variations (CNVs) at the single-cell level. Furthermore, it allowed discrimination between genotypically distinct cells within mixed populations.

Strengths:

This is a high-impact study that will likely contribute to our understanding of DNA replication and the genetic plasticity of Leishmania, including its well-documented aneuploidy, somy variations, CNVs, and SNPs - all key elements for elucidating various aspects of the parasite's biology, such as genome evolution, genetic exchange, and mechanisms of drug resistance.

Overall, the authors clearly achieved their objectives, providing a solid rationale for the study and demonstrating how this approach can advance the investigation of Leishmania's small, plastic genome and its frequent natural strain mixtures within hosts. This methodology may also prove valuable for genomic studies of other single-celled organisms.

We thank the reviewer for the positive feedback and appreciation of the potential applications for the methodology we describe here.

Weaknesses:

The discussion section could be enriched to help readers understand the significance of the work, for instance, by more clearly pointing out the obstacles to a better understanding of DNA replication in Leishmania. Or else, when they discuss the results obtained at the level of nucleotide information and the relevance of being able to compare, in their case, the two strains, they could refer to the implications of this level of precision to those studying clonal strains or field isolates, drug resistance or virulence in a more detailed way.

We thank the reviewer for the suggestions. Indeed, single-cell DNA sequencing has successfully revealed cell-to-cell variability in replication timing and fork progression in mammalian cells[1,2] and we believe that the SPC-PTA workflow could be used in similar studies in Leishmania to complement bulk-based observations[3,4]. Regarding nucleotide information, it is indeed of high relevance to detect minor circulating variants with potential virulence impact and/or effect on drug resistance which could be missed by bulk sequencing. This includes the ability to detect co-occurring variants with potential epistatic effects. These topics will be further developed in the revised version. Finally, we will explicitly discuss how this methodology can be applied beyond Leishmania, to investigate genome plasticity, adaptation, and evolutionary processes in other organisms.

Reviewer #2 (Public review):

Summary:

Negreira et al. present an application of a novel single-cell genomics approach to investigate the genetic heterogeneity of Leishmania parasites. Leishmania, while also representing a major global disease with hundreds of thousands of cases annually, serves as a model to test the rigor of the sequencing strategy. Its complex karyotypic nature necessitates a method that is capable of resolving natural variation to better understand genome dynamics. Importantly, an earlier single-cell genomics platform (10x Chromium) is no longer available, and new methods need to be evaluated to fill in this gap.

The study was designed to evaluate whether a capsule-based cell capture method combined with primary template-directed amplification (PTA) could maintain levels of genomic heterogeneity represented in an equal mixture of two Leishmania strains. This was a high bar, given the relatively small protozoan genome and prior studies that showed limitations of single-cell genomics, especially for gene-level copy number changes. Overall, the study found that semi-permeable capsules (SPC) are an effective way to isolate high-quality single cells. Additionally, short reads from amplified genomes effectively maintained the relative levels of variation in the two strains on the chromosome, gene copy, and individual base level. Thus, this method will be useful to evaluate adaptive strategies of Leishmania. Many researchers will also refer to these studies to set up SPC collection and PTA methods for their organism of choice.

Strengths:

(1) The use of SPC and PTA in a non-bacterial organism is novel. The study displays the utility of these methods to isolate and amplify single genomes to a level that can be sequenced, despite being a motile organism with a GC-rich genome.

(2) The authors clearly outlined their optimization strategy and provided numerous quality-control metrics that inspire confidence in the success of achieving even chromosomal coverage relative to ploidy.

(3) The use of two distinct Leishmania strains with known clonal status provided strong evidence that PTA-based amplification could reflect genome differences and displayed the utility of the method for studies of rare genotypes.

(4) Evaluating the SPCs pre- and post-amplification with microscopy is a practical and robust way of determining the success of SPC formation and PTA.

(5) The authors show that the PTA-based approach easily resolved major genotypic ploidy in agreement with a prior 10x Chromium-based study. The new method had improved resolution of drug resistance genotypes in the form of both copy-number variations and single-nucleotide polymorphisms.

(6) In general, the authors are very thorough in describing the methods, including those used to optimize PTA lysis and amplification steps (fresh vs frozen cells, naked DNA vs sorted cells, etc). This demonstrates a depth of knowledge about the procedure and leaves few unanswered questions.

(7) The custom, multifaceted, computational assessment of coverage evenness is a major strength of the study and demonstrates that the authors acknowledge potential computational factors that could impact the analysis.

We deeply appreciate the positive and encouraging feedback on our manuscript.

Weaknesses:

(1) The rationale behind some experimental/analysis choices is not well-described. For example, the rationale behind methanol fixation and heat-lysis is unclear. Additionally, the choice of various methods to assess "evenness" is not justified (e.g. why are multiple methods needed? What is the strength of each method?). Also, there is no justification for using 100k reads for subsampling. Finally, what exactly constitutes a "confidently-called SNP"?

The methanol fixation prior to lysis is part of the original protocol described in the Single-Microbe Genome Barcoding Kit manual and was meant to facilitate lysis and DNA denaturation in bacterial cells (for which the kit was originally developed). However, in our preliminary tests with bulk samples – described in the supplementary material – we noticed a strong negative effect on lysis efficiency/DNA recovery when parasites were fixed with methanol. Thus, we decided to test the effect of skipping this step in the single-cell DNA workflow. We kept the SPC_STD1 sample to have a safe control where the full workflow described in the kit manual was followed.

As we were unsure if the standard lysis (25 ˚C for 15 minutes) would work efficiently for Leishmania, we included the heat-lysis (99˚C for 15 minutes) as well as the longer incubation lysis (25 ˚C for 1h). These modifications were listed as validated alternatives in the kit's manual.

The 100k reads threshold was chosen based on the number of reads found in the 'true cell' with the lowest read count.

Regarding variant calling, a variant was considered confidently called if it was covered, at single-cell level, by at least one deduplicated read with Phred quality above Q30 and mapping quality (MAPQ) also above 30.

In the revised version, we will include these explanations and improve the explanation of the metrics used to estimate coverage quality.

(2) In the methods, the STD protocol lists a 15-minute amplification at 45C whereas the PTA protocol involves 10h at 37C. This is a dramatic difference in incubation time and should be addressed when comparing results from the two methods. It is not really a fair comparison when you look at coverage levels; of course, a 10-hour incubation is going to yield more reads than a 15-minute incubation.

We agree with the reviewer that the longer incubation period of PTA might explain the higher read count seen in the PTA samples, although the differences in amplification kinetics (linear in PTA, exponential in STD) and potential differences in amplification saturation points make it difficult to compare them. For instance, an updated version of PTA (ResolveDNA V2) uses a lower amplification time (2.5 h) and achieves similar amplification levels compared to the 10h incubation time, suggesting PTA amplification saturates well before the 10h time. In any case, all quality check metrics were done with the cells subsampled to 100 k reads to mitigate the effect of read count differences on the data quality.

(3) There is a lack of quantitative evaluations of the SPCs. e.g. How many capsules were evaluated to assess doublets? How many capsules were detected as Syto5 positive in a successful vs an unsuccessful experiment?

We agree with the reviewer but during experimental execution SPCs were only assessed qualitatively via microscopy following the Single-cell microbe DNA barcoding kit manual. No quantitative analysis was done and therefore we do not have this data. Regarding doublet, this was done in silico based on the detection of SPCs containing mixed genomes from the two strains used in the study as described in the Materials and Methods. As pointed by another reviewer, this only allow the detection of inter-strain doublets. In the revised version, we explain this and add an estimation of total doublets based on the inter-strain doublet rate.

(4) The authors do not address some of the amplification results obtained under various conditions. For example, why did temperature-based lysis of STD4 lead to amplification failure? Also, what is the reason for fewer "true" cells (higher background) in the PTA samples compared to the STD samples? Is this related to issues with barcoding or, alternatively, substandard amplification as indicated by lower read amounts in some capsules (knee plots in Figure 1C)?

After exchange with the technical support team of the SPC generator kit, it was clarified that the heat lysis done in STD4 should have had a shorter incubation time (10 minutes instead of 15 minutes). We suspect that the longer incubation time, combined with the higher temperature and the harsh lysis condition with 0.8M KOH might have damaged SPCs and therefore DNA might have leaked out of them before WGA. In the microscopy images, SPCs in STD4 show a swollen aspect not seen in the other samples. In the revised version we will explain this more clearly.

(5) The paper presents limited biological relevance. Without this, the paper describes an improvement in genome amplification methods and some proof-of-concept analyses. Using a 1:1 mixture of parasites with different genotypes, the authors display the utility of the method to resolve genetic diversity, but they don't seek to understand the limits of detecting this diversity. For some, the authors do not comment on the mixed karyotypes from the HU3 cells (Figure 3F) other than to state that this line was not clonal. For CNVs, the two loci evaluated were detected at relatively high copy number (according to Figure 4C, they are between 4 and 20 copies). Thus, the sensitivity of CNV detection from this data remains unclear; can this approach detect lower-level CNVs like duplications, or minor CNVs that do not show up in every cell?

As described above we will include more discussion on potential biological relevance of the method in the revised version of the manuscript. In the revised version we will attempt to use dedicated bioinformatic tools to discover de novo CNVs, as per the suggestion of other reviewers. This might also allow us to determine the detection limit of the methodology for CNVs.

(6) The authors state that Leishmania can carry extrachromosomal copies of important genes. There is no discussion about how the presence of these molecules would affect the amplification steps and CNV detection. For example, the phi29 enzyme is very processive with circular molecules; does its presence lead to overamplification and overrepresentation in the data? Is this evident in the current study? This information would be useful for organisms that carry this type of genetic element.

We believe our data, which uses short-read sequences, does not allow to differentiate between intra-chromosomal CNVs and linear or circular episomal CNVs, so we cannot define if circular CNVs are over-amplified. Of note, we have previously demonstrated that the M-locus CNV in chromosome 36 is intrachromosomal, not circular (episomal)[5].

(7) The manuscript is missing a comparison with other similar studies in the field. For example, how does this coverage level compare to those achieved for other genomes? Can this method achieve amplification levels needed to assess larger genomes? Has there been any evaluation of base composition effects since Leishmania is a GC-rich genome?

We believe the SPC-PTA workflow can be applied to organisms with larger genomes as PTA was developed specifically for mammalian cells[6], and also because, in our hands, it outperformed the 10X scDNA solution, which was developed for mammals.

We believe direct comparison with other studies regarding coverage levels is elusive because other steps in the workflow apart from the WGA, such as the library preparation (PCR-based in our case), as well as genome features like GC content, size, and presence of repetitive regions, can also affect coverage levels and evenness. One strength of our approach was the use a single sample (the 50/50 mix between two L. donovani strain) for all conditions, thus removing potential parasite-specific biases. In addition, the application of a multiplexing system during barcoding allowed us to combine all samples prior to library preparation, thus removing potential differences introduced by this step.

Regarding the effect of GC-content, we did notice a positive bias in all samples in regions with higher GC content, which had to be corrected in silico. This was the opposite to a negative bias observed in previous study[7] likely due to differences in WGA and/or library preparation. In the revised version, we will include a supplementary figure showing the GC bias.

(8) Cost is mentioned as a benefit of the SPC platform, and savings are achieved when working in a plate format, but no details are included on how this was evaluated.

In the revised version we will provide precise cost estimates and the rationale for the estimation.

(9) The Zenodo link for custom scripts does not exist, and code cannot be evaluated.

The full Zenodo link (https://doi.org/10.5281/zenodo.17094083) will be included in the revised version.

Reviewer #3 (Public review):

Summary

In this manuscript, Negreira et al. propose a new scDNAseq method, using semi-permeable capsules (SPCs) and primary template-directed amplification (PTA). The authors optimize several metrics to improve their predictions, such as determining GC bias, Intra-Chromosomal fluctuation (ICF -metric to differentiate replicative and non-replicative cells) and Intra-chromosomal coefficient of variation (ICCV - chromosome read distribution). The coverage evenness was evaluated using the fini index and the median absolute pairwise difference between the counts of two consecutive bins. They validate the proposed method using two Leishmania donovani strains isolated from different countries, BPK081 (low genomic variability) and HU3 (high genomic variability). Then, they showed that the method outperforms WGA and has similar accuracy to the discontinued 10X-scDNA (10X Genomics), further improving on short CNV identification. The authors also show that the method can identify somy variations, insertions/deletions and SNP variations across cells. This is a timely and very relevant work that has a wide applicability in copy number variation assessment using single-cell data.

Strengths

I really appreciate this work. My congratulations to the authors. All my comments below only aim to improve an already solid manuscript.

We thank the reviewer for the enthusiasm and positive feedback.

Weaknesses

(1) Data availability: Although the authors provide a Zenodo link, the data is restricted. I also could not access the GitHub link in the Zenodo website: https://github.com/gabrielnegreira/2025_scDNA_paper. The authors should make these files available.

Both the Zenodo (https://doi.org/10.5281/zenodo.17094083) and the GitHub (https://github.com/gabrielnegreira/2025_scDNA_paper) repositories are now publicly available.

(2) 2-SPC-PTA and SPC-STD cell count comparison: The authors have consistently proven that the SPC-PTA method was superior to SPC-STD. However, there are a few points that should be clarified regarding the SPC-PTA results. Is there an explanation for the lower proportion of SPC to true cells success in SPC-STD, which reflects the bimodal distribution for the reads per cell in SPC-PTA2 and a three-to-multimodal distribution in SPC-PTA1 in Figure 1B? Also, in Table 1, does the number of reads reflect the number of reads in all sequenced SPCs or only in the true cells? If it is in the SPCs, I suggest that the authors add a new column in the table with the "Number of reads in true cells" to account for this discrepancy.

The reason for the higher presence of 'background' SPCs in the PTA samples is not clear, but we hypothesize that it could be due to PTA favoring amplification of small, free floating DNA molecules that might have been trapped in cell-free SPCs, as PTA works with shorter amplicons. Also, the longer incubation time seen in PTA (10 h) might have allowed enhanced amplification of low quantities of free-floating DNA to detectable levels. Regarding Table 1, indeed it only show the total number of reads per sample. In the revised version we will include the suggested column to Table 1.

(3) The authors should evaluate the results with a higher coverage for SCP-PTA. I understand that the authors subsampled the total read to 100,000 to allow cross-sample comparisons, especially between SPC-STD and SPC-PTA. However, as they concluded that the SPC-PTA was far superior, and the samples SPC-PTA1 and SPC-PTA2 had an "elbow" of 650,493 and 448,041, respectively, it might be interesting to revisit some of the estimations using only SPC-PTA samples and a higher coverage cutoff, as 400,000.

We believe the 100.000 cutoff is already high for aneuploidy analysis as we have successfully reconstructed parasite karyotype with 20.000 reads per cell8, so a higher cutoff will likely not improve it. For CNV analysis, in the revised version, we will try to identify de novo CNVs using dedicated bioinformatic tools as per other reviewer suggestions. There, we will also test if a higher CNV detection sensitivity is achieved using the suggested 400,000 reads cutoff for the PTA samples.

(4) Doublet detection: I suggest that the authors be a little more careful with their definition of doublets. The doublet detection was based on diagnostic SNPs from the two strains, BPK081 and HU3, which identify doublets between two very different and well-characterised strains. However, this method will probably not identify strain-specific doublets. This is of minor importance for cloned and stable strains with few passages, as BPK081, but might be more relevant in more heterogeneous strains, as HU3. Strain-specific doublets might also be relevant in other scenarios, as multiclonal infections with different populations from the same strain in the same geographic area. One positive point is that the "between strain doublet count" was low, so probably the within-strain doublet count should be low too. The manuscript would benefit from a discussion on this regard.

We fully agree with the reviewer. We will make it clear in the revised version that we quantify inter-strain doublets only, and we will also provide an estimation of total doublets based on the inter-strain doublet rate.

(5) Nucleotide sequence variants and phylogeny: I believe that a more careful description of the phylogenetic analysis and some limitations of the sequence variant identification would benefit the manuscript.

(5.1) As described in the methods, the authors intentionally selected two fairly different Leishmania donovani strains, HU3 and BPK081, and confirmed that the sequent variant methodology can separate cells from each strain. It is a solid proof of concept. However, most of the multiclonal infections in natural scenarios would be caused by parasite populations that diverge by fewer SNPs, and will be significantly harder to detect. Hence, I suggest that a short discussion about this is important.

We will add a short discussion clarifying the limitations, while noting that our data demonstrate the ability of the approach to resolve very closely related cells, as illustrated by the fine-scale genetic differences observed within the clonal BPK081 population and by the detection of rare variants at targeted loci. We will also emphasize that the sensitivity to detect closely related genotypes depends on sequencing depth and the genomic regions considered.

(5.2) The authors should expand on the description of the phylogenetic tree. In the HU3 on Figure 5F left panel, most of the variation is observed in ~8 cells, which goes from position 0 to position ~28.000. Most of the other cells are in very short branches, from ~29.000 to 30.4000 (5F right panel). Assuming that this representation is a phylogram, as the branches are short, these cells diverge by approximately 100-2000 SNPs. It is unexpected (but not impossible) that such ~8 divergent cells be maintained uniquely (or in very low counts) in the culture, unless this is a multiclonal infection. I would carefully investigate these cells. They might be doublets or have more missing data than other cells. I would also suggest that a quick discussion about this should be added to the manuscript.

In the revised version we will improve the description of the phylogenetic analysis. We will also investigate deeper the 8 mentioned cells to define if they have confounding factors that might have led to their discrepancy. The possibility of multiclonal infection in HU3 is not excluded as this strain was not cloned after isolation.

References:

(1) Dileep, V., Gilbert, D. M., Dileep, V. & Gilbert, D. M. Single-cell replication profiling to measure stochastic variation in mammalian replication timing. Nat. Commun. 9, 427 (2018).

(2) Miura, H. et al. Single-cell DNA replication profiling identifies spatiotemporal developmental dynamics of chromosome organization. Nat. Genet. 51, 1356–1368 (2019).

(3) Marques, C. A. et al. Genome-wide mapping reveals single-origin chromosome replication in Leishmania, a eukaryotic microbe. Genome Biol. 16, 230 (2015).

(4) Damasceno, J. D. et al. Leishmania major chromosomes are replicated from a single high-efficiency locus supplemented by thousands of lower efficiency initiation events. Cell Rep. 44, 116094 (2025).

(5) Imamura, H. et al. Evolutionary genomics of epidemic visceral leishmaniasis in the Indian subcontinent. eLife 5, e12613 (2016).

(6) Gonzalez-Pena, V. et al. Accurate genomic variant detection in single cells with primary template-directed amplification. Proc. Natl. Acad. Sci. 118, e2024176118 (2021).

(7) Imamura, H. et al. Evaluation of whole genome amplification and bioinformatic methods for the characterization of Leishmania genomes at a single cell level. Sci. Rep. 10, 15043 (2020).

(8) Negreira, G. H. et al. High throughput single-cell genome sequencing gives insights into the generation and evolution of mosaic aneuploidy in Leishmania donovani. Nucleic Acids Res. 50, 293–305 (2022).

Code exhibition

完成了

It’s cool to see how images are really just grids of pixels with RGB values, and even something like microRGB fits into that same idea of breaking color down into tiny components. Thinking about images this way makes them feel a lot less mysterious and more like something you can actually work with in code.

Author response:

We thank all reviewers for their overall assessment, thoughtful comments, and suggestions. We are working to address each reviewer’s comment in detail. In this provisional response, we provide clarifications regarding our experimental approach and the novelty of our work, and include additional analyses that we have performed since the submission of the manuscript. We are also happy to report that we have now shared the raw data, intermediate analysis files, and the complete repository to facilitate replication of the analysis and figures.

Code repo: github.com/LorenFrankLab/ms_stim_analysis

Data repo: dandiarchive.org/dandiset/001634

Docker containers (see GitHub repo for use instructions):

Database: https://hub.docker.com/r/samuelbray32/spyglass-db-ms_stim_analysis

Python notebooks: https://hub.docker.com/r/samuelbray32/spyglass-hub-ms_stim_analysis

(1) Novelty and contrast with earlier manipulations:

We thank the reviewers for suggesting that we explicitly contrast our results with prior pharmacological (Wang et al., 2016; Wang et al., 2015; Koenig et al., 2011; Brandon et al., 2014), systemic (Robbe & Buzsaki 2009; Petersen and Buzsáki 2020), and behavioral (Drieu et al., 2018) manipulations that also assessed some of the physiological features we evaluated. We will add a discussion of these studies, which will help us emphasize both the insights and discrepancies observed using these prior approaches. We will also more clearly explain the the novelty and importance of our specific approach for temporally and physiologically precise manipulation. Specifically, our approach (closed-loop theta-phase stimulation during locomotion) provides a level of physiological specificity that made it possible to dissociate theta-state dynamics from other hippocampal processes. This in turn allowed us to address a question that has remained unresolved across prior studies: Are hippocampal spatial sequences during locomotion (i.e., theta sequences) necessary to learn a novel hippocampal-dependent task?

(2) Additional analysis on SWRs during rest:

since submitting the manuscript, we have conducted additional analysis on the rate and length of SWRs in the rest box and found that their rate and length are also indistinguishable between targeted and control animals (effect of manipulation between control and targeted animals; rSWR rate: p=0.45; rSWR length: p=0.94, mixed effect model). We also find evidence for sequential neural representations in the rest box, when the encoding was performed in the behavioral arena. Example trajectories are shown below. These results are consistent with our observations on SWRs rate, length, and content in the behavioral arena. Additionally, we are in the process of evaluating and quantifying the results of decoding the rSWRs and will include those in the next version of the manuscript.

Author response image 1.

Sequential replay events observed in the rest box

(3) Theta sequence measurement in the absence of theta:

In the next version of the manuscript, we will explicitly explain why our manipulation makes it is more appropriate to measure sequential hippocampal representations during locomotion (i.e., theta sequences) without using theta oscillation or an epoch-averaged relatively large sliding window as a reference. The key insight here is that our manipulation suppresses theta and thus makes it difficult or impossible to accurately identify theta phase. We understand that theta-phase based approaches were used in prior work; however, these prior analyses may have confounded the absence of hippocampal theta sequences during locomotion by the inability to detect theta oscillatory phase reliably. We will show that our method of using clusterless Bayesian decoding in which we estimate the decoded position at every 2ms timestep is indeed able to capture endogenous hippocampal sequences even without imposing any requirements of aligning to theta oscillations, thus providing an unbiased estimate of the rhythmicity of hippocampal spatial representations.

(4) Additional analysis on place cell stability and tuning:

We thank the reviewer for this question. For the KL divergence analysis, we have imposed a spike-count criterion (100 spikes for each interval type —stimulation-off, stimulation-on, and the stimulus sub-interval) and a coverage criterion (50% HPD of the units’ spatial firing distribution was contained within 40cm on the linear track and 100cm on the w-track). These criteria were chosen to ensure that spatial tuning curves were sufficiently well sampled and localized to allow reliable estimation of KL divergence, which is particularly sensitive to noise arising from low spike counts or diffuse firing. Based on the reviewer’s suggestion, we have relaxed the unit inclusion criteria for KL divergence by relaxing the criteria for number of spikes and spatial coverage criterion to include more weakly tuned place cells and replicated our results (p=.146). Further, we have also evaluated the stability of place field order between stimulation-on and stimulation-off conditions using more standard methods (as in Wang et. al., 2015; spearman correlation of place field order, control vs targeted, p = .920, t-test). These results are consistent with our observations about place field stability during stimulation-off and stimulation-on conditions (Fig. 2F).

Author response image 2.

Spearman correlation of place field order during stimulation-on and stimulation-off conditions.

Analyse de l'Avis du CESE sur les Temps de Vie de l'Enfant

Résumé Exécutif

Cet avis du Conseil économique, social et environnemental (CESE), intitulé « Satisfaire les besoins fondamentaux des enfants et garantir leurs droits », dresse un constat critique de la situation des enfants en France, dont les temps de vie sont davantage structurés par les contraintes des adultes que par leurs propres besoins fondamentaux.

Fruit d'une saisine gouvernementale faisant suite à une Convention citoyenne, le rapport souligne un décalage majeur entre les droits constitutionnels et internationaux de l'enfant et leur application effective, particulièrement pour les plus vulnérables.

Les principales conclusions révèlent des inégalités sociales, territoriales et économiques profondes qui entravent le développement, la santé et le bien-être des enfants.

L'avis pointe du doigt des rythmes scolaires inadaptés, une sédentarité croissante, un manque de sommeil chronique, une surexposition aux écrans, et une déconnexion préoccupante de la nature.

La pression sur les familles, notamment monoparentales, et le manque de coordination entre les acteurs éducatifs aggravent ces constats.

Pour y remédier, le CESE formule 19 préconisations interdépendantes visant une transformation systémique. Celles-ci incluent des mesures politiques fortes comme l'instauration d'une « clause impact enfance » dans chaque projet de loi, une réforme ambitieuse des rythmes scolaires sur la base des besoins physiologiques, et la création d'un Service Public de la Continuité Éducative (SPCE) pour assurer une meilleure coordination des acteurs.

L'avis appelle également à renforcer le soutien à la parentalité, à garantir l'accès de tous les enfants aux loisirs, à la culture et aux activités de plein air, et à allouer des financements publics pérennes pour faire de l'enfance un véritable investissement d'avenir.

Introduction et Contexte

En réponse à une saisine du Premier ministre de mai 2025, le CESE a élaboré cet avis suite aux travaux d'une Convention citoyenne dédiée aux temps de vie des enfants. Cent trente-trois citoyens et un panel de vingt enfants et adolescents ont été invités à répondre à la question :

« Comment mieux structurer les différents temps de la vie quotidienne des enfants afin qu’ils soient plus favorables à leurs apprentissages, à leur développement et à leur santé ? ».

Le constat principal de la Convention citoyenne, repris par le CESE, est que les enfants subissent les rythmes effrénés d'une société qui construit leurs temps autour des contraintes des adultes plutôt qu'en réponse à leurs besoins biologiques et de développement.

Le rapport du CESE, s'appuyant sur les 20 propositions citoyennes, formule 19 préconisations qui constituent une position commune de la société civile organisée.

Cet avis s'inscrit dans la continuité de travaux antérieurs du CESE sur l'éducation, la protection de l'enfance et la santé mentale, et vise à proposer des réponses globales et articulées.

Partie 1 : Droits et Besoins Fondamentaux de l'Enfant : Un Constat Alarmant

A. L'Écart entre Droits Reconnus et Réalité Vécue

La France a consacré les droits de l'enfant dans sa Constitution et a ratifié la Convention Internationale des Droits de l'Enfant (CIDE) en 1990, s'engageant sur quatre principes fondamentaux : le droit à la vie, l'intérêt supérieur de l'enfant, la non-discrimination et le respect de son opinion.

Cependant, l'avis du CESE met en lumière une ineffectivité préoccupante de ces droits pour une part significative des enfants.

• Pauvreté et Précarité : En 2023, 21,9 % des enfants de moins de 18 ans vivent sous le seuil de pauvreté monétaire.

À la rentrée 2025, au moins 2 159 enfants se sont retrouvés sans solution d'hébergement.

Ces réalités percutent violemment la capacité de la société à répondre à leurs besoins fondamentaux.

• Critiques Internationales : Le Comité des droits de l'enfant de l'ONU a enjoint la France en 2023 à prendre des mesures urgentes concernant la violence, la protection de l'enfance, la détention d'enfants étrangers, la pauvreté et l'inclusion des enfants en situation de handicap.

• L'« Infantisme » : Le rapport dénonce la persistance de l'« infantisme », un concept désignant les préjugés et la discrimination fondée sur l'âge, qui considère les enfants comme des êtres inférieurs et moins dignes de respect.

Cette culture conduit à ignorer leur parole et leur capacité à être des acteurs sociaux. Pour le combattre, le CESE réaffirme la nécessité d'un débat de société et la création d'un Code de l'enfance.

• Clause « Impact Enfance » : S'inspirant de la « clause impact jeunesse », le CESE préconise (Préconisation #1) d'intégrer un volet enfance dans chaque étude d'impact des projets de loi afin de s'assurer que toute politique publique soit fondée sur le respect des droits de l'enfant.

B. Le Rôle de la Famille et les Obstacles Socio-économiques

La famille est le premier lieu de développement de l'enfant, mais elle fait face à de nombreux obstacles.

• Soutien à la Parentalité : Face à la diversité des modèles familiaux (nucléaire, monoparentale, recomposée...), un soutien renforcé à la parentalité est jugé nécessaire pour aider les parents à répondre aux besoins de leurs enfants (Préconisation #7).

• Inégalités de Genre : Les femmes continuent d'assumer l'essentiel des responsabilités familiales et de la charge mentale, ce qui impacte leur santé et leur carrière.

Le rapport souligne la nécessité d'une répartition équitable des tâches.

• Conciliation Vie Professionnelle/Familiale : Les contraintes professionnelles empiètent sur le temps familial.

Le CESE préconise (Préconisation #2) la transposition complète de la directive européenne sur l'équilibre vie professionnelle-vie personnelle, en créant un droit à des « formules souples de travail » (aménagement du temps, télétravail) négocié dans les branches et la fonction publique.

• Enfants Séparés de leur Famille :

◦ Parents séparés : Il est crucial de soutenir les dispositifs comme les Espaces de rencontre pour préserver la relation parent-enfant tout en prenant en compte le point de vue de l'enfant (Préconisation #3).   

◦ Aide Sociale à l'Enfance (ASE) : L'avis dénonce une crise systémique de la protection de l'enfance, où les droits des enfants confiés, notamment l'accès aux loisirs et à la culture, sont négligés.

Il est préconisé (Préconisation #4) que le Projet Pour l'Enfant (PPE) soit co-construit avec les parents et l'enfant, et qu'il intègre l'ensemble de ses besoins.

Partie 2 : Les Enjeux des Temps et des Espaces de Vie

L'avis analyse en profondeur la manière dont les temps et les espaces de l'enfant sont organisés, révélant de multiples fractures et inadéquations.

A. Les Temps de Vie : Entre Contraintes et Qualité

La vie de l'enfant est rythmée par trois grands temps : familial, scolaire, et les "tiers temps" (périscolaire, extrascolaire).

• Qualité des Temps : Le rapport insiste sur la nécessité d'un équilibre entre temps contraints et temps libre, temps individuel et collectif, activité et repos.

La qualité des interactions avec les adultes et un environnement sécurisant sont déterminants.

Le CESE préconise (Préconisation #6) d'intégrer des temps libres de qualité dans toutes les activités d'apprentissage.

• Le Temps Scolaire : La France se distingue par des journées scolaires longues et un temps d'instruction élevé, sans que cela se traduise par de meilleurs résultats.

Le rythme de la semaine de quatre jours est jugé contraire aux besoins des enfants. Le CESE estime que le statu quo n'est plus tenable et appelle (Préconisation #8) à une évolution des rythmes scolaires :

◦ Premier degré : Réorganiser la journée et la semaine scolaire après concertation.   

◦ Second degré : Adapter les amplitudes horaires aux besoins physiologiques des jeunes (ex: commencer plus tard).   

◦ Calendrier scolaire : Organiser le calendrier hexagonal autour de deux zones de vacances, avec une alternance de 7 semaines de cours et 2 semaines de vacances.

• Les Tiers Temps et le Droit aux Loisirs : Les activités périscolaires et extrascolaires, portées par les associations et les collectivités, sont essentielles mais menacées par le désengagement de l'État et la marchandisation.

L'accès à ces activités, ainsi qu'aux vacances, est fortement marqué par les inégalités sociales.

Un enfant sur deux ne part pas en vacances. Le CESE réaffirme (Préconisation #9) que chaque enfant a droit aux vacances et aux loisirs, et appelle à renforcer le financement des accueils collectifs de mineurs et l'information sur les aides existantes.

B. Les Espaces de Vie : De l'« Enfant d'Intérieur » à la Reconnexion au Dehors

L'environnement physique joue un rôle crucial dans le développement de l'enfant.

• L'« Enfant d'Intérieur » : Le rapport alerte sur le phénomène des « enfants d'intérieur », qui passent de moins en moins de temps à l'extérieur et en contact avec la nature, en raison de la peur du risque, de l'urbanisation centrée sur la voiture et de l'attrait des écrans.

• Repenser l'Aménagement : Il est impératif de repenser l'aménagement des territoires « à hauteur d'enfant », en créant des espaces publics (rues, places) sécurisés, propices au jeu, à la socialisation et aux mobilités douces.

Le CESE préconise (Préconisation #11) d'associer les enfants à l'élaboration des projets d'urbanisme.

• Le Bâti et le Cadre de Vie : Les bâtiments accueillant des enfants (écoles, centres de loisirs) sont souvent inadaptés, notamment face aux enjeux climatiques (vagues de chaleur).

Leur rénovation écologique et leur accessibilité sont des priorités. Toute rénovation doit faire l'objet d'une concertation incluant les enfants et les jeunes (Préconisation #12).

Partie 3 : Leviers d'Action pour la Santé et le Bien-être

L'avis identifie quatre domaines d'action prioritaires pour améliorer la santé physique et mentale des enfants.

• Reconnecter à la Nature : Le contact avec la nature est fondamental pour la santé.

Le CESE appelle à valoriser et accompagner l'éducation au dehors (Préconisation #10) et à garantir que chaque enfant bénéficie d'un accès à des espaces naturels, de sorties régulières et d'au moins un séjour en classe de découverte par cycle scolaire (Préconisation #13).

• Lutter contre le Manque de Sommeil : Le déficit de sommeil touche plus de 30 % des enfants et 70 % des adolescents, avec des conséquences graves sur l'apprentissage et la santé.

Le CESE demande une campagne nationale de sensibilisation (Préconisation #14) et la garantie de temps de repos et de sieste dans toutes les structures, notamment en maternelle (Préconisation #15).

• Favoriser l'Activité Physique : Face à une sédentarité alarmante, il est crucial de faciliter l'accès au sport pour tous. Le CESE préconise (Préconisation #16) une tarification sociale et l'élargissement du dispositif Pass'Sport, récemment restreint.

• Mieux Réguler les Écrans : L'omniprésence des écrans a des effets néfastes documentés (sommeil, sédentarité, exposition à des contenus inappropriés). L'avis souligne la nécessité d'une meilleure régulation et d'un accompagnement à la parentalité numérique.

Partie 4 : Gouvernance, Coordination et Financement

Pour que ces changements soient effectifs, une transformation de la gouvernance des politiques de l'enfance est indispensable.

• Coordination des Acteurs : L'action publique est jugée trop fragmentée. Le CESE préconise (Préconisation #17) de réhabiliter le Projet Éducatif Territorial (PEDT) et d'en faire le volet éducation des Conventions Territoriales Globales (CTG) pour assurer une coordination efficace au niveau local.

• Un Service Public de la Continuité Éducative (SPCE) : Pour garantir une offre éducative cohérente sur tous les temps de l'enfant, l'avis propose la création d'un SPCE (Préconisation #18).

Ce service, confié aux collectivités locales, serait chargé de diagnostiquer les besoins et de planifier les actions en associant tous les acteurs.

• Formation et Financement : La revalorisation des métiers éducatifs et le développement d'une culture commune des droits de l'enfant sont essentiels.

Enfin, le CESE alerte sur l'insuffisance des budgets alloués aux politiques de l'enfance et appelle (Préconisation #19) à un effort budgétaire conséquent et pérenne de l'État et de la Sécurité sociale, considérant ces dépenses comme un investissement fondamental pour l'avenir.

Synthèse des 19 Préconisations du CESE

| Numéro | Thème Principal | Résumé de la Préconisation | | --- | --- | --- | | #1 | Droits de l'enfant | Mettre en œuvre une « clause impact enfance » dans chaque étude d'impact de projet de loi ou de texte réglementaire pour garantir que les politiques publiques respectent les droits de l'enfant. | | #2 | Parentalité & Travail | Créer un droit aux « formules souples de travail » (aménagement du temps, télétravail) pour les parents, par la négociation dans les branches et la fonction publique. | | #3 | Séparation parentale | Développer et soutenir financièrement les Espaces de rencontre pour aider les parents séparés à assumer leurs responsabilités parentales en prenant en compte le point de vue de l'enfant. | | #4 | Protection de l'enfance (ASE) | Rendre le Projet pour l'enfant (PPE) systématiquement co-construit avec les parents et l'enfant, et y intégrer tous les besoins, y compris les loisirs et la culture. Simplifier la gestion des actes usuels. | | #5 | Accès à la culture | Soutenir financièrement et développer tous les dispositifs culturels et artistiques pour les enfants (scolaires, ACM), via des contrats multipartites (État, collectivités, réseau culturel). | | #6 | Qualité des temps | Intégrer des temps libres de qualité dans les activités d'apprentissage, ce qui implique de former les adultes et personnels encadrants. | | #7 | Soutien à la parentalité | Mieux faire connaître, rendre accessibles et valoriser financièrement les lieux et actions d'aide aux parents (maisons des familles, groupes de parole, LAEP, PMI...). | | #8 | Rythmes scolaires | Faire évoluer les rythmes scolaires : réorganiser la journée et la semaine au primaire ; adapter les horaires aux besoins physiologiques au secondaire ; organiser un calendrier national à 2 zones (7 semaines de cours / 2 de vacances). | | #9 | Droit aux vacances et loisirs | Mobiliser les pouvoirs publics pour rendre effectif le droit aux vacances. Renforcer l'information sur les aides et financer davantage les accueils collectifs de mineurs (ACM). | | #10 | Éducation à la nature | Valoriser et accompagner l'éducation au dehors et en lien avec la nature (formation des acteurs, verdissement des espaces, aires éducatives, terrains d'aventure...). | | #11 | Aménagement du territoire | Aménager les territoires « à hauteur d'enfant » dans une démarche participative, en repensant les espaces publics comme lieux de sociabilité, de mixité et de jeu. | | #12 | Bâti et cadre de vie | Rendre obligatoire la concertation avec les enfants et les jeunes pour tout projet d'aménagement ou de rénovation de bâtiments (écoles, centres de loisirs, gymnases...). | | #13 | Lien à la nature | Garantir que chaque enfant bénéficie d'un accès à des espaces naturels, de sorties régulières, et d'au moins un séjour en classe de découverte par cycle de scolarité. | | #14 | Sommeil | Organiser une campagne nationale d'information et de sensibilisation sur le rôle fondamental du sommeil et les facteurs qui lui nuisent. | | #15 | Temps de repos | Prévoir des temps de repos, de calme et de sieste (préservée en maternelle) dans toutes les structures accueillant des enfants, et repenser les locaux pour créer une atmosphère paisible. | | #16 | Activité physique et sportive | Soutenir une tarification sociale pour l'accès au sport. Étendre et revaloriser le Pass'Sport, en y incluant les associations sportives scolaires. | | #17 | Coordination locale | Réhabiliter le Projet Éducatif Territorial (PEDT) et en faire le volet "éducation" des Conventions Territoriales Globales (CTG) pour une coordination globale des acteurs. | | #18 | Gouvernance | Créer un Service Public de la Continuité Éducative (SPCE), confié aux collectivités, pour diagnostiquer les besoins et planifier les actions éducatives sur le territoire. | | #19 | Financement | Assurer un effort budgétaire conséquent et pérenne de l'État et de la Sécurité sociale pour financer les politiques publiques en faveur de l'enfance. |

This makes sense to me, especially how binary maps so cleanly onto True/False. It’s interesting that something as abstract as “truth” in code ultimately comes down to 0s and 1s, which feels very different from how messy and ambiguous truth can be in real life.

I think this is an interesting data with constraints that shift along cultural axes. While the other data types like age, name, and address can vary from culture to culture (such as characters used to input), the answers will all be relatively similar. People might for example measure their age using different calendars, or by amount of winters experienced, but a full annual cycle is used pretty much worldwide. For relationship status though, there are a lot of variables even in one culture that must be negotiated. Does dating count as different than a relationship? What about new terms like situationship? What about cultures with multiple partners? There's so many value judgements about what counts as a relationship in the act of reifying it in code. It makes me think about the ways in which our cultural ideology is represented in code, more than we often think about.

I do not know code, or any kind of data analysis so this will be great practice!

Synthèse de l'Avis du Conseil d'État sur la Proposition de Loi "Protéger les Mineurs en Ligne"

1. Contexte et Objectifs de la Proposition de Loi

Cette proposition de loi a été élaborée en réponse à des constats alarmants concernant les risques auxquels les réseaux sociaux exposent les mineurs.

Faisant directement suite aux recommandations du rapport de la commission d’enquête sur TikTok, le texte met en lumière les dangers d'addiction et les effets psychologiques néfastes de certaines plateformes sur la santé mentale des jeunes.

L'objectif principal du législateur est donc de renforcer de manière significative le cadre de protection des mineurs dans l'environnement numérique, en instaurant des mesures contraignantes et préventives.

Les deux mesures phares de la proposition initiale sont les suivantes :

• Interdiction d'accès pour les moins de 15 ans : Le texte visait à imposer une obligation directe aux fournisseurs de services de réseaux sociaux de refuser l'inscription des mineurs de moins de 15 ans.

Pour ce faire, les plateformes auraient dû mettre en œuvre des dispositifs de contrôle d'âge robustes, sous peine de sanctions financières et d'injonctions judiciaires.

• Couvre-feu numérique pour les 15-18 ans : Pour cette tranche d'âge, la proposition prévoyait une obligation de désactivation automatique de l'accès aux comptes entre 22 heures et 8 heures du matin, en s'appuyant sur les mêmes solutions techniques de vérification de l'âge.

En complément de ce dispositif central, le texte comprend plusieurs autres mesures structurantes :

| Mesure | Objectif Stratégique | | --- | --- | | Lutte contre la publicité pro-suicide | Compléter la liste des contenus illicites pour inclure la propagande en faveur de moyens de se donner la mort. | | Renforcement des peines | Augmenter la durée de suspension des comptes d'accès aux plateformes en cas d'infraction. | | Messages sanitaires | Imposer des informations préventives sur les publicités pour les réseaux sociaux et sur les emballages de smartphones. | | Formation scolaire | Étendre la formation sur l'usage du numérique à la sensibilisation aux enjeux de santé mentale. | | Interdiction des téléphones dans les lycées | Généraliser l'interdiction déjà en vigueur dans les collèges pour favoriser la concentration et prévenir le harcèlement. | | Création d'un délit de négligence parentale | Sanctionner les parents en cas d'usage excessif, inadapté ou non surveillé des outils numériques par leur enfant. |

L'analyse juridique approfondie du Conseil d'État révèle cependant que, si l'intention est louable, les mécanismes proposés soulèvent des difficultés majeures de compatibilité avec le droit européen et les libertés fondamentales.

2. Analyse Critique du Conseil d'État : Compatibilité avec le Droit Européen

La conformité au droit de l'Union européenne est une condition essentielle de la validité de toute loi nationale.

Le Conseil d'État souligne que le Règlement sur les Services Numériques (DSA) harmonise pleinement les règles pour les plateformes opérant dans l'UE, limitant drastiquement la capacité des États membres à leur imposer des obligations supplémentaires.

L'avis du Conseil se révèle être une véritable leçon d'ingénierie juridique, démontrant comment atteindre un objectif de politique nationale dans le cadre contraignant d'un droit européen harmonisé.

Le Conseil d'État met en évidence une incompatibilité juridique frontale : en imposant une obligation directe aux plateformes de refuser l'inscription des mineurs, la proposition de loi initiale violerait le principe d'harmonisation maximale du DSA, rendant la mesure juridiquement fragile et susceptible d'être invalidée.

Pour surmonter cet obstacle majeur, le Conseil d'État propose une reformulation décisive, qui constitue le pivot de sa stratégie. Au lieu d'obliger les plateformes, la loi doit directement interdire l'accès au mineur : `

« Il est interdit au mineur de quinze ans d’accéder à un service de réseau social en ligne »`.

Cet acte de prohibition qualifie automatiquement un tel accès de "contenu illicite" au sens de la définition large du DSA.

Cette reclassification est la clé de voûte de la stratégie du Conseil : elle permet de mobiliser les puissants mécanismes de régulation du DSA (injonctions de l'Arcom, signalements, sanctions) contre les plateformes sans créer une nouvelle obligation nationale, interdite par le droit européen.

Le cadre de l'UE devient ainsi le principal outil d'application d'une politique nationale française.

Pour renforcer l'effectivité de cette interdiction, le Conseil suggère d'ouvrir un second flanc de mise en conformité. Il préconise de prévoir la nullité de plein droit des contrats passés par un mineur en violation de cette interdiction.

Une telle nullité priverait de base légale tout traitement de ses données personnelles, exposant les plateformes à des contrôles et sanctions de la part de la CNIL au titre du RGPD, ce qui augmente considérablement la pression en faveur du respect de la loi.

Enfin, le Conseil recommande que la Commission européenne élabore des lignes directrices pour s'assurer que les plateformes gèrent correctement la restitution des contenus et des données aux mineurs dont les comptes sont résiliés, afin de ne pas porter atteinte à leurs droits de propriété intellectuelle.

Cette refonte juridique est présentée comme une condition sine qua non à la viabilité du texte.

3. Analyse Critique du Conseil d'État : Équilibre avec les Droits et Libertés Fondamentaux

Au-delà de la conformité européenne, le Conseil d'État analyse la conciliation entre l'objectif de protection de l'enfance — une exigence constitutionnelle — et le respect des libertés fondamentales du mineur (liberté d'expression, d'information) et des droits des parents.

Sur ce plan, le Conseil juge le dispositif initial déséquilibré et disproportionné pour trois raisons principales :

1. Caractère général et absolu : L'interdiction s'appliquerait à tous les "réseaux sociaux" sans distinction, y compris ceux ne présentant aucun risque avéré (plateformes collaboratives, éducatives), ce qui est jugé excessif.

2. Absence de discernement et de rôle parental : Le mécanisme initial ignore le degré de maturité de l'enfant et écarte totalement les parents de leur rôle d'accompagnement, en contradiction avec le Code civil et la Convention relative aux droits de l’enfant.

3. Manque de justification du couvre-feu : Les bornes horaires du couvre-feu pour les 15-18 ans (22h-8h) sont jugées insuffisamment documentées et donc disproportionnées.

Pour rééquilibrer le texte, le Conseil d'État propose une refonte qui incarne un changement de philosophie réglementaire : passer d'une interdiction étatique, brute et centrée sur la plateforme, à un système nuancé, responsabilisant les parents et centré sur le terminal. Ce mécanisme alternatif repose sur deux volets :

• Volet 1 - Interdiction Ciblée Le Gouvernement pourrait, par décret en Conseil d’État pris après avis de l’Arcom, interdire l'accès aux mineurs de moins de 15 ans à des réseaux sociaux spécifiquement identifiés comme dangereux en raison de leurs systèmes de recommandation.

L'État utilise ici son pouvoir de prohibition de manière ciblée, là où le danger est avéré.

• Volet 2 - Autorisation Parentale Généralisée Pour tous les autres réseaux sociaux, l'accès serait interdit sauf autorisation expresse d'un parent.

Réalisée via des dispositifs installés sur les systèmes d’exploitation des équipements terminaux distribués par les fournisseurs d’accès à l’internet (à l'instar des mécanismes de contrôle parental existants), cette autorisation serait révocable et pourrait préciser une durée d'usage.

L'État délègue ici à une autorité parentale guidée le soin d'évaluer le risque.

Cette approche duale résout le problème de proportionnalité, transformant une interdiction fragile en un système de régulation juridiquement beaucoup plus solide.

4. Recommandations et Points de Vigilance sur les Autres Articles

Le Conseil d'État a également examiné les autres articles de la proposition de loi, formulant des recommandations d'ajustement ou des réserves importantes.

• Interdiction des téléphones dans les lycées (Art. 6) : La mesure est jugée nécessaire et proportionnée.

Le Conseil recommande d'exclure explicitement de son champ les formations de l'enseignement supérieur et de différer son entrée en vigueur à la rentrée scolaire 2026.

• Formation scolaire (Art. 4) : Jugée conforme, la mesure est cependant qualifiée de potentiellement redondante avec des dispositions déjà existantes.

Une entrée en vigueur différée à la rentrée 2026 est également suggérée pour permettre l'adaptation des enseignants.

• Délit de négligence numérique (Art. 7) : Le Conseil exprime de fortes réserves.

À titre principal, il estime que le droit pénal existant est suffisant.

À titre subsidiaire, si le délit était maintenu, ses termes ("usage excessif", "outils numériques") sont jugés trop vagues et contraires au principe constitutionnel de légalité des délits et des peines.

• Publicité et emballages (Art. 3) : Ces dispositions devront être notifiées à la Commission européenne au titre de la directive "TRIS", une étape procédurale cruciale destinée à prévenir la création de barrières techniques inopinées au sein du marché unique.

• Rapport au Parlement (Art. 5) : Il est suggéré de restreindre le champ du rapport pour le concentrer sur le respect par les plateformes de leurs obligations spécifiques envers les mineurs dans le cadre du DSA.

Ces ajustements visent à garantir la sécurité juridique et l'applicabilité concrète de l'ensemble du texte.

5. Conclusion : Synthèse Stratégique pour la Décision

L'avis du Conseil d'État valide sans équivoque la nécessité d'agir face aux dangers documentés que les réseaux sociaux font peser sur les mineurs et reconnaît la pertinence de l'objectif poursuivi par le législateur.

Cependant, cette validation de l'objectif s'accompagne d'une censure quasi totale du dispositif initialement proposé. Celui-ci est jugé doublement fragile :

1. Incompatible avec le droit de l'Union européenne, en raison de la violation du principe d'harmonisation maximale du DSA.

2. Déséquilibré au regard des droits fondamentaux, car l'interdiction générale et le couvre-feu sont jugés disproportionnés et écartent indûment l'autorité parentale.

En définitive, les amendements du Conseil d'État ne sont pas de simples ajustements.

Ils constituent une refondation juridique et une véritable feuille de route stratégique et législative offerte au Parlement. Ils transforment un projet juridiquement précaire en une loi conforme, proportionnée et, par conséquent, viable et réellement efficace pour protéger les mineurs dans l'espace numérique.

Author response:

The following is the authors’ response to the original reviews.

Public reviews:

Reviewer #1 (Public review):

Summary: 

The authors provide a resource to the systems neuroscience community, by offering their Python-based CLoPy platform for closed-loop feedback training. In addition to using neural feedback, as is common in these experiments, they include a capability to use real-time movement extracted from DeepLabCut as the control signal. The methods and repository are detailed for those who wish to use this resource. Furthermore, they demonstrate the efficacy of their system through a series of mesoscale calcium imaging experiments. These experiments use a large number of cortical regions for the control signal in the neural feedback setup, while the movement feedback experiments are analyzed more extensively.

Strengths:

The primary strength of the paper is the availability of their CLoPy platform. Currently, most closed-loop operant conditioning experiments are custom built by each lab and carry a relatively large startup cost to get running. This platform lowers the barrier to entry for closed-loop operant conditioning experiments, in addition to making the experiments more accessible to those with less technical expertise.

Another strength of the paper is the use of many different cortical regions as control signals for the neurofeedback experiments. Rodent operant conditioning experiments typically record from the motor cortex and maybe one other region. Here, the authors demonstrate that mice can volitionally control many different cortical regions not limited to those previously studied, recording across many regions in the same experiment. This demonstrates the relative flexibility of modulating neural dynamics, including in non-motor regions.

Finally, adapting the closed-loop platform to use real-time movement as a control signal is a nice addition. Incorporating movement kinematics into operant conditioning experiments has been a challenge due to the increased technical difficulties of extracting real-time kinematic data from video data at a latency where it can be used as a control signal for operant conditioning. In this paper they demonstrate that the mice can learn the task using their forelimb position, at a rate that is quicker than the neurofeedback experiments.

Weaknesses:

There are several weaknesses in the paper that diminish the impact of its strengths. First, the value of the CLoPy platform is not clearly articulated to the systems neuroscience community. Similarly, the resource could be better positioned within the context of the broader open-source neuroscience community. For an example of how to better frame this resource in these contexts, I recommend consulting the pyControl paper. Improving this framing will likely increase the accessibility and interest of this paper to a less technical neuroscience audience, for instance by highlighting the types of experimental questions CLoPy can enable.

We appreciate the editor’s feedback regarding the clarity of the CLoPy platform's value and its positioning within the broader neuroscience community. We agree and understand the importance of effectively communicating the utility of CLoPy to both the systems neuroscience field and the wider open-source neuroscience community.

To address this, we have revised the introduction and discussion sections of the manuscript to more clearly articulate the unique contributions of the CLoPy platform. Specifically:

(1) We have emphasized how CLoPy can address experimental questions in systems neuroscience by highlighting its ability to enable real-time closed-loop experiments, such as investigating neural dynamics during behavior or studying adaptive cortical reorganization after injury. These examples are aimed at demonstrating its practical utility to the neuroscience audience.

(2) We have positioned CLoPy within the broader open-source neuroscience ecosystem, drawing comparisons to similar resources like pyControl. We describe how CLoPy complements existing tools by focusing on real-time optical feedback and integration with genetically encoded indicators, which are becoming increasingly popular in systems neuroscience. We also emphasize its modularity and ease of adoption in experimental settings with limited resources.

(3) To make the manuscript more accessible to a less technically inclined audience, we have restructured certain sections to focus on the types of experiments CLoPy enables, rather than the technical details of the implementation.

We have consulted the pyControl paper, as suggested, and have used it as a reference point to improve the framing of our resource. We believe these changes will increase the accessibility and appeal of the paper to a broader neuroscience audience.

While the dataset contains an impressive amount of animals and cortical regions for the neurofeedback experiment, and an analysis of the movement-feedback experiments, my excitement for these experiments is tempered by the relative incompleteness of the dataset, as well as its description and analysis in the text. For instance, in the neurofeedback experiment, many of these regions only have data from a single mouse, limiting the conclusions that can be drawn. Additionally, there is a lack of reporting of the quantitative results in the text of the document, which is needed to better understand the degree of the results. Finally, the writing of the results section could use some work, as it currently reads more like a methods section.

Thank you for your thoughtful and constructive feedback on our manuscript. We appreciate the time and effort you took to review our work and provide detailed suggestions for improvement. Below, we address the key points raised in your review:

(1) Dataset Completeness: We acknowledge that some of the neurofeedback experiments include data from only a single mouse for some cortical regions while for some cortical regions, there are several animals. This was due to practical constraints during the study, and we understand the limitations this poses for drawing broad conclusions. We felt it was still important to include these data sets with smaller sample sizes as they might be useful for others pursuing this direction in the future. To address this, we have revised the text to explicitly acknowledge these limitations and clarify that the results for some regions are exploratory in nature. We believe our flexible tool will provide a means for our lab and others include more animals representing additional cortical regions in future studies. Importantly, we have included all raw and processed data as well as code for future analysis.

(2) Quantitative Results: We recognize the importance of reporting quantitative results in the text for better clarity and interpretation. In response, we have added more detailed description of the quantitative findings from both the neurofeedback and movement-feedback experiments. This will include effect sizes, statistical measures, and key numerical results to provide a clearer understanding of the degree and significance of the observed effects.

(3) Results Section Writing: We appreciate your observation that parts of the results section read more like a methods section. To improve clarity and focus, we have restructured the results section to present the findings in a more concise and interpretative manner, while moving overly detailed descriptions of experimental procedures to the methods section.

Suggestions for improved or additional experiments, data or analyses:

Not necessary for this paper, but it would be interesting to see if the CLNF group could learn without auditory feedback.

This is a great suggestion and certainly something that could be done in the future.

There are no quantitative results in the results section. I would add important results to help the reader better interpret the data. For example, in: "Our results indicated that both training paradigms were able to lead mice to obtain a significantly larger number of rewards over time," You could show a number, with an appropriate comparison or statistical test, to demonstrate that learning was observed.

Thank you for pointing this out. We have mentioned quantification values in the results now, along with being mentioned in the figure legends, and we are quoting it in following sentences. “A ΔF/F0 threshold value was calculated from a baseline session on day 0 that would have allowed 25% performance. Starting from this basal performance of around 25% on day 1, mice (CLNF No-rule-change, N=23, n=60 and CLNF Rule-change, N=17, n=60) were able to discover the task rule and perform above 80% over ten days of training (Figure 4A, RM ANOVA p=2.83e-5), and Rule-change mice even learned a change in ROIs or rule reversal (Figure 4A, RM ANOVA p=8.3e-10, Table 5 for different rule changes). There were no significant differences between male and female mice (Supplementary Figure 3A).”

For: "Performing this analysis indicated that the Raspberry Pi system could provide reliable graded feedback within ~63 {plus minus} 15 ms for CLNF experiments." The LED test shows the sending of the signal, but the actual delay for the audio generation might be longer. This is also longer than the 50 ms mentioned in the abstract.

We appreciate the reviewer’s insightful comment. The latency reported (~63ms) was measured using the LED test, which captures the time from signal detection to output triggering on the Raspberry Pi GPIO. We agree that the total delay for auditory feedback generation could include an additional latency component related to the digital-to-analog conversion and speaker response. In our setup, we employ a fast Audiostream library written in C to generate the audio signal and expect the delay contribution to be negligible compared to the GPIO latency. Though we did not do this, it can be confirmed by an oscilloscope-based pilot measurement (for additional delay calculation). We have updated the manuscript to clarify that the 63 ± 15 ms value reflects the GPIO-triggered output latency, and we have revised the abstract to accurately state the delay as “~63 ms” rather than 50 ms. This ensures consistency and avoids underestimation of the latency. We have corrected the LED latency for CLNF and CLMF experiments in the abstract as well.

It could be helpful to visualize an individual trial for each experiment type, for instance how the audio frequency changes as movement speed / calcium activity changes.

We have added Supplementary Figure 8 that contains this data where you can see the target cortical activity trace, target paw speed, rewards, along with the audio frequency generated.

The sample sizes are small (n=1) for a few groups. I am excited by the variety of regions recorded, so it could be beneficial for the authors to collect a few more animals to beef up the sample sizes.

We've acknowledged that some of the sample sizes are small. Importantly, we have included raw and processed data as well as code for future analysis. We felt it was still important to still include these data sets with smaller sample sizes as they might be useful for others pursuing this direction in the future.

I am curious as to why 60 trials sessions were used. Was it mostly for the convenience of a 30 min session, or were the animals getting satiated? If the former, would learning have occurred more rapidly with longer sessions?

This is a great observation and the answer is it was mostly due to logistical reasons. We tried to not keep animals headfixed for more than 45 minutes in each session as they become less engaged with long duration headfixed sessions. After headfixing them, it takes about 15 minutes to get the experiment going and therefore 30 - 40 minutes long recorded sessions seemed appropriate before they stop being engaged or before they get satiated in the task. We provided supplemental water after the sessions and we observed that they consumed water after the sessions so they were not fully satiated during the sessions even when they performed well in the task and got maximum rewards. We also had inter-trial rest periods of 10s that elongated the session duration. We think it would be interesting to explore the relationship between session duration(number of trials) and task learning progression over the days in a separate study.

Figure 4E is interesting, it seems like the changes in the distribution of deltaF was in both positive and negative directions, instead of just positive. I'd be curious as to the author's thoughts as to why this is the case. Relatedly, I don't see Figure 4E, and a few other subplots, mentioned in the text. As a general comment, I would address each subplot in the text.

We have split Figure 4 into two to keep the figures more readable. Previous Figure 4E-H are now Figure 5A-D in the revised manuscript. The online real-time CLNF sessions were using a moving window average to calculate ΔF/F₀  and the figures were generated by averaging the whole recorded sessions. We have added text in Methods under “Online ΔF/F₀calculation” and “Offline ΔF/F₀ calculation” sections making it clear about how we do our ΔF/F₀ normalization based on average fluorescence over the entire session. Using this method of normalization does increase the baseline so that some peaks appear to be below zero. Additionally, it is unclear what strategy animals are employing to achieve the rule specific target activity. The task did not constrain them to have a specific strategy for cortical activation - they were rewarded as long as they crossed the threshold in target ROI(s). For example, in 2-ROI experiments, to increase ROI1-ROI2 target activity, they could increase activity of ROI1 relative to ROI2 or decreased activity of ROI1 relative to ROI1 - both would have led to a reward as long as the result crossed the threshold.

We have now addressed and added reference to the figures in the text in Results under “Mice can explore and learn an arbitrary task, rule, and target conditions” and “Mice can rapidly adapt to changes in the task rule” sections - thanks for pointing this out.

For: "In general, all ROIs assessed that encompassed sensory, pre-motor, and motor areas were capable of supporting increased reward rates over time," I would provide a visual summary showing the learning curves for the different types of regions.

We have rewritten this section to emphasize that these conclusions were based on pooled data from multiple regions of interest. The sample sizes for each type of region are different and some are missing. We believe it would be incomplete and not comparable to present this as a regular analysis since the sample sizes were not balanced. We would be happy to dive deeper into this and point to the raw and processed dataset if anyone would like to explore this further by GitHub or other queries.

Relatedly, I would further explain the fast vs slow learners, and if they mapped onto certain regions.

Mice were categorized into fast or slow learners based on the slope of learning over days (reward progression over the days) as shown in Supplementary Figure 3C,D. Our initial aim was not to probe cortical regions that led to fast vs slow learning but this was a grouping we did afterwards. Based on the analysis we did, the fast learners included the sensory (V1), somatosensory (BC, HL), and motor (M1, M2) areas, while the slow learners included the motor (M1, M2), and higher order (TR, RL) cortical areas. Testing all dorsal cortical areas would be prudent to establish their role in fast or slow learning and it is an interesting future direction.

Also I would make the labels for these plots (e.g. Supp Fig3) more intuitive, versus the acronyms currently used.

We have made more expressive labels and explained the acronyms below the Supplementary Figure 3.

The CLMF animals showed a decrease in latency across learning, what about the CLNF animals? There is currently no mention in the text or figures.

We have now incorporated the CLNF task latency data into both the Results text and Figure 4C. Briefly, task latency decreased as performance improved, increased following a rule change, and then decreased again as the animals relearned the task. The previous Figure 4C has been updated to Figure 4D, and the former Figure 4D has been moved to Supplementary Figure 4E.

Reviewer #2 (Public review):

Summary:

In this work, Gupta & Murphy present several parallel efforts. On one side, they present the hardware and software they use to build a head-fixed mouse experimental setup that they use to track in "real-time" the calcium activity in one or two spots at the surface of the cortex. On the other side, the present another setup that they use to take advantage of the "real-time" version of DeepLabCut with their mice. The hardware and software that they used/develop is described at length, both in the article and in a companion GitHub repository. Next, they present experimental work that they have done with these two setups, training mice to max out a virtual cursor to obtain a reward, by taking advantage of auditory tone feedback that is provided to the mice as they modulate either (1) their local cortical calcium activity, or (2) their limb position.

Strengths:

This work illustrates the fact that thanks to readily available experimental building blocks, body movement and calcium imaging can be carried using readily available components, including imaging the brain using an incredibly cheap consumer electronics RGB camera (RGB Raspberry Pi Camera). It is a useful source of information for researchers that may be interested in building a similar setup, given the highly detailed overview of the system. Finally, it further confirms previous findings regarding the operant conditioning of the calcium dynamics at the surface of the cortex (Clancy et al. 2020) and suggests an alternative based on deeplabcut to the motor tasks that aim to image the brain at the mesoscale during forelimb movements (Quarta et al. 2022).

Weaknesses:

This work covers 3 separate research endeavors: (1) The development of two separate setups, their corresponding software. (2) A study that is highly inspired from the Clancy et al. 2020 paper on the modulation of the local cortical activity measured through a mesoscale calcium imaging setup. (3) A study of the mesoscale dynamics of the cortex during forelimb movements learning. Sadly, the analyses of the physiological data appears uncomplete, and more generally the paper tends to offer overstatements regarding several points:

In contrast to the introductory statements of the article, closed-loop physiology in rodents is a well-established research topic. Beyond auditory feedback, this includes optogenetic feedback (O'Connor et al. 2013, Abbasi et al. 2018, 2023), electrical feedback in hippocampus (Girardeau et al. 2009), and much more.

We have included and referenced these papers in our introduction section (quoted below) and rephrased the part where our previous text indicated there are fewer studies involving closed-loop physiology.

“Some related studies have demonstrated the feasibility of closed-loop feedback in rodents, including hippocampal electrical feedback to disrupt memory consolidation (Girardeau et al.2009), optogenetic perturbations of somatosensory circuits during behavior (O'Connor et al.2013), and more recent advances employing targeted optogenetic interventions to guide behavior (Abbasi et al. 2023).”

The behavioral setups that are presented are representative of the state of the art in the field of mesoscale imaging/head fixed behavior community, rather than a highly innovative design. In particular, the closed-loop latency that they achieve (>60 ms) may be perceived by the mice. This is in contrast with other available closed-loop setups.

We thank the reviewer for this thoughtful comment and fully agree that our closed-loop latency is larger than that achieved in some other contemporary setups. Our primary aim in presenting this work, however, is not to compete with the lowest possible latencies, but to provide an open-source, accessible, and flexible platform that can be readily adopted by a broad range of laboratories. By building on widely available and lower-cost components, our design lowers the barrier of entry for groups that wish to implement closed-loop imaging and behavioral experiments, while still achieving latencies well within the range that can support many biologically meaningful applications.

For example, our latency (~60 ms) remains compatible with experimental paradigms such as:

Motor learning and skill acquisition, where sensorimotor feedback on the scale of tens to hundreds of milliseconds is sufficient to modulate performance.

Operant conditioning and reward-based learning, in which reinforcement timing windows are typically broader and not critically dependent on sub-20 ms latencies.

Cortical state dependent modulation, where feedback linked to slower fluctuations in brain activity (hundreds of milliseconds to seconds) can provide valuable insight.

Studies of perception and decision-making, in which stimulus response associations often unfold on behavioral timescales longer than tens of milliseconds.

We believe that emphasizing openness, affordability, and flexibility will encourage widespread adoption and adaptation of our setup across laboratories with different research foci. In this way, our contribution complements rather than competes with ultra-low-latency closed-loop systems, providing a practical option for diverse experimental needs.

Through the paper, there are several statements that point out how important it is to carry out this work in a closed-loop setting with an auditory feedback, but sadly there is no "no feedback" control in cortical conditioning experiments, while there is a no-feedback condition in the forelimb movement study, which shows that learning of the task can be achieved in the absence of feedback.

We fully agree that such a control would provide valuable insight into the contribution of feedback to learning in the CLNF paradigm. In designing our initial experiments, we envisioned multiple potential control conditions, including No-feedback and Random-feedback. However, our first and primary objective was to establish whether mice could indeed learn to modulate cortical ROI activation through auditory feedback, and to further investigate this across multiple cortical regions. For this reason, we focused on implementing the CLNF paradigm directly, without the inclusion of these additional control groups. To broaden the applicability of the system, we subsequently adapted the platform to the CLMF experiments, where we did incorporate a No-feedback group. These results, as the reviewer notes, strengthen the evidence for the role of feedback in shaping task performance. We agree that the inclusion of a No-feedback control group in the CLNF paradigm will be crucial in future studies to further dissect the specific contribution of feedback to cortical conditioning.

The analysis of the closed-loop neuronal data behavior lacks controls. Increased performance can be achieved by modulating actively only one of the two ROIs, this is not clearly analyzed (for instance looking at the timing of the calcium signal modulation across the two ROIs. It seems that overall ROIs1 and 2 covariate, in contrast to Clancy et al. 2020. How can this be explained?

We agree that the possibility of increased performance being driven by modulation of a single ROI is an important consideration. Our study indeed began with 1-ROI closed-loop experiments. In those early experiments, while we did observe animals improving performance across days, we realized that daily variability in ongoing cortical GCaMP activity could lead to fluctuations in threshold-crossing events. The 2-ROI design was subsequently introduced to reduce this variability, as the target activity was defined as the relative activity between the two ROIs (e.g., ROI1 – ROI2). This approach offered a more stable signal by normalizing ongoing fluctuations. In our analysis of the early 2-ROI experiments, we observed that animals adopted diverging strategies to achieve threshold crossings. Specifically, some animals increased activity in ROI1 relative to ROI2, while others decreased activity in ROI2 to accomplish the same effect. Once discovered, each animal consistently adhered to its chosen strategy throughout subsequent training sessions. This was an early and intriguing observation, but as the experiments were not originally designed to systematically test this effect, we limited our presentation to the analysis of a small number of animals (shown in Figure 11). We have added details about this observation in our Results section as well, quoted below-

“In the 2-ROI experiment where the task rule required “ROI1 - ROI2” activity to cross a threshold for reward delivery, mice displayed divergent strategies. Some animals predominantly increased ROI1 activity, whereas others reduced ROI2 activity, both approaches leading to successful threshold crossing (Figure 11)”.

We hope this clarifies how the use of two ROIs helps explain the apparent covariation of the signals, and why some divergence from the observations of Clancy et al. (2020) may be expected.

Reviewer #3 (Public review):

Summary:

The study demonstrates the effectiveness of a cost-effective closed-loop feedback system for modulating brain activity and behavior in head-fixed mice. Authors have tested real-time closed-loop feedback system in head-fixed mice two types of graded feedback: 1) Closed-loop neurofeedback (CLNF), where feedback is derived from neuronal activity (calcium imaging), and 2) Closed-loop movement feedback (CLMF), where feedback is based on observed body movement. It is a python based opensource system, and authors call it CLoPy. The authors also claim to provide all software, hardware schematics, and protocols to adapt it to various experimental scenarios. This system is capable and can be adapted for a wide use case scenario.

Authors have shown that their system can control both positive (water drop) and negative reinforcement (buzzer-vibrator). This study also shows that using the close loop system mice have shown better performance, learnt arbitrary task and can adapt to change in the rule as well. By integrating real-time feedback based on cortical GCaMP imaging and behavior tracking authors have provided strong evidence that such closed-loop systems can be instrumental in exploring the dynamic interplay between brain activity and behavior.

Strengths:

Simplicity of feedback systems designed. Simplicity of implementation and potential adoption.

Weaknesses:

Long latencies, due to slow Ca2+ dynamics and slow imaging (15 FPS), may limit the application of the system.

We appreciate the reviewer’s comment and agree that latency is an important factor in our setup. The latency arises partly from the inherent slow kinetics of calcium signaling and GCaMP6s, and partly from the imaging rate of 15 FPS (every 66 ms). These limitations can be addressed in several ways: for example, using faster calcium indicators such as GCaMP8f, or adapting the system to electrophysiological signals, which would require additional processing capacity. In our implementation, image acquisition was fixed at 15 FPS to enable real-time frame processing (256 × 256 resolution) on Raspberry Pi 4B devices. With newer hardware, such as the Raspberry Pi 5, substantially higher acquisition and processing rates are feasible (although we have not yet benchmarked this extensively). More powerful platforms such as Nvidia Jetson or conventional PCs would further support much faster data acquisition and processing.

Major comments:

(1) Page 5 paragraph 1: "We tested our CLNF system on Raspberry Pi for its compactness, general-purpose input/output (GPIO) programmability, and wide community support, while the CLMF system was tested on an Nvidia Jetson GPU device." Can these programs and hardware be integrated with windows-based system and a microcontroller (Arduino/ Tency). As for the broad adaptability that's what a lot of labs would already have (please comment/discuss)?

While we tested our CLNF system on a Raspberry Pi (chosen for its compactness, GPIO programmability, and large user community) and our CLMF system on an Nvidia Jetson GPU device (to leverage real-time GPU-based inference), the underlying software is fully written in Python. This design choice makes the system broadly adaptable: it can be run on any device capable of executing Python scripts, including Windows-based PCs, Linux machines, and macOS systems. For hardware integration, we have confirmed that the framework works seamlessly with microcontrollers such as Arduino or Teensy, requiring only minor modifications to the main script to enable sending and receiving of GPIO signals through those boards. In fact, we are already using the same system in an in-house project on a Linux-based PC where an Arduino is connected to the computer to provide GPIO functionality. Furthermore, the system is not limited to Raspberry Pi or Arduino boards; it can be interfaced with any GPIO-capable devices, including those from Adafruit and other microcontroller platforms, depending on what is readily available in individual labs. Since many neuroscience and engineering laboratories already possess such hardware, we believe this design ensures broad accessibility and ease of integration across diverse experimental setups.

(2) Hardware Constraints: The reliance on Raspberry Pi and Nvidia Jetson (is expensive) for real-time processing could introduce latency issues (~63 ms for CLNF and ~67 ms for CLMF). This latency might limit precision for faster or more complex behaviors, which authors should discuss in the discussion section.

In our system, we measured latencies of approximately ~63 ms for CLNF and ~67 ms for CLMF. While such latencies indeed limit applications requiring millisecond precision, such as fast whisker movements, saccades, or fine-reaching kinematics, we emphasize that many relevant behaviors, including postural adjustments, limb movements, locomotion, and sustained cortical state changes, occur on timescales that are well within the capture range of our system. Thus, our platform is appropriate for a range of mesoscale behavioral studies that probably needs to be discussed more. It is also important to note that these latencies are not solely dictated by hardware constraints. A significant component arises from the inherent biological dynamics of the calcium indicator (GCaMP6s) and calcium signaling itself, which introduce slower temporal kinetics independent of processing delays. Newer variants, such as GCaMP8f, offer faster response times and could further reduce effective biological latency in future implementations.

With respect to hardware, we acknowledge that Raspberry Pi provides a low-cost solution but contributes to modest computational delays, while Nvidia Jetson offers faster inference at higher cost. Our choice reflects a balance between accessibility, cost-effectiveness, and performance, making the system deployable in many laboratories. Importantly, the modular and open-source design means the pipeline can readily be adapted to higher-performance GPUs or integrated with electrophysiological recordings, which provide higher temporal resolution. Finally, we agree with the reviewer that the issue of latency highlights deeper and interesting questions regarding the temporal requirements of behavior classification. Specifically, how much data (in time) is required to reliably identify a behavior, and what is the minimum feedback delay necessary to alter neural or behavioral trajectories? These are critical questions for the design of future closed-loop systems and ones that our work helps frame.

We have added a slightly modified version of our response above in the discussion section under “Experimental applications and implications”.

(3) Neurofeedback Specificity: The task focuses on mesoscale imaging and ignores finer spatiotemporal details. Sub-second events might be significant in more nuanced behaviors. Can this be discussed in the discussion section?

This is a great point  and we have added the following to the discussion section. “In the case of CLNF we have focused on regional cortical GCAMP signals that are relatively slow in kinetics. While such changes are well suited for transcranial mesoscale imaging assessment, it is possible that cellular 2-photon imaging (Yu et al. 2021) or preparations that employ cleared crystal skulls (Kim et al. 2016) could resolve more localized and higher frequency kinetic signatures.”

(4) The activity over 6s is being averaged to determine if the threshold is being crossed before the reward is delivered. This is a rather long duration of time during which the mice may be exhibiting stereotyped behaviors that may result in the changes in DFF that are being observed. It would be interesting for the authors to compare (if data is available) the behavior of the mice in trials where they successfully crossed the threshold for reward delivery and in those trials where the threshold was not breached. How is this different from spontaneous behavior and behaviors exhibited when they are performing the test with CLNF? 

We would like to emphasize that we are not directly averaging activity over 6 s to compare against the reward threshold. Instead, the preceding 6 s of activity is used solely to compute a dynamic baseline for ΔF/F₀ ( ΔF/F₀ = (F –F₀ )/F₀). Here, F₀is calculated as the mean fluorescence intensity over the prior 6 s window and is updated continuously throughout the session. This baseline is then subtracted from the instantaneous fluorescence signal to detect relative changes in activity. The reward threshold is therefore evaluated against these baseline-corrected ΔF/F₀ values at the current time point, not against an average over 6 s. This moving-window baseline correction is a standard approach in calcium imaging analyses, as it helps control for slow drifts in signal intensity, bleaching effects, or ongoing fluctuations unrelated to the behavior of interest. Thus, the 6-s window is not introducing a temporal lag in reward assignment but is instead providing a reference to detect rapid increases in cortical activity.  We have added the term dynamic baseline to the Methods to clarify.

Recommendations for the authors

Reviewer #1 (Recommendations for the authors):

Additional suggestions for improved or additional experiments, data or analyses.

For: "Looking closely at their reward rate on day 5 (day of rule change), they had a higher reward rate in the second half of the session as compared to the first half, indicating they were adapting to the rule change within one session." It would be helpful to see this data, and would be good to see within-session learning on the rule change day

Thank you for pointing this out. We had missed referencing the figure in the text, and have now added a citation to Supplementary Figure 4A, which shows the cumulative rewards for each day of training. As seen in the plot for day 5, the cumulative rewards are comparable to those on day 1, with most rewards occurring during the second half of the session.

For: "These results suggest that motor learning led to less cortical activation across multiple regions, which may reflect more efficient processing of movement-related activity," it could also be the case that the behaviour became more stereotyped over learning, which would lead to more concentrated, correlated activity. To test this, it would be good to look at the limb variability across sessions. Similarly, if it is movement-related, there should be good decoding of limb kinematics.

Indeed, we observed that behavior became more stereotyped over the course of learning, as shown in Supplementary Figure 4C, 4D. One plausible explanation for the reduction in cortical activation across multiple regions is that behavior itself became more stereotyped, a possibility we have explored in the manuscript. Specifically, forelimb movements during the trial became increasingly correlated as mice improved on the task, particularly in the groups that received auditory feedback (Rule-change and No-rule-change groups; Figure 8). As movements became more correlated, overall body movements during trials decreased and aligned more closely with the task rule (Figure 9D). This suggests that reduced cortical activity may in part reflect changes in behavior. Importantly, however, in the Rule-change group, we observed that on the day of the rule switch (day 5), when the target shifted from the left to the right forelimb, cortical activity increased bilaterally (Figure 9A–C). This finding highlights our central point: groups that received feedback (Rule-change and No-rule-change) were able to identify the task rule more effectively, and both their behavior and cortical activity became more specifically aligned with the rule compared to the No-feedback group. We agree with the reviewers that additional analyses along these lines would be valuable future directions. To facilitate this, we have included the movement data for readers who may wish to pursue further analyses, details can be found under “Data and code availability” in Methods section. However, given the limited sample sizes in our dataset and the need to keep the manuscript focused on the central message, we felt that including these additional analyses here would risk obscuring the main findings.

For: "We believe the decrease in ΔF/F0peak is unlikely to be driven by changes in movement, as movement amplitudes did not decrease significantly during these periods (Figure 7D CLMF Rule-change)." I would formally compare the two conditions. This is an important control. Also, another way to see if the change in deltaF is related to movement would be to see if you can predict movement from the deltaF.

Figure 7D in the previous version is Figure 9D in the current revision of the manuscript. We've assessed this for the examples shown based on graphing the movement data, unfortunately there is not enough of that data to do a group analysis of movement magnitude. We would suggest that this would be an excellent future direction that would take advantage of the flexible open source nature of our tool.

Recommendations for improving the writing and presentation.

In the abstract there is no mention of the rationale for the project, or the resulting significance. I would modify this to increase readership by the behavioral neuroscience community. Similarly, the introduction also doesn't highlight the value of this resource for the field. Again, I think the pyControl paper does a good job of this. For readability, I would add more subheadings earlier in the results, to separate the different technical aspects of the system.

We have revised the introduction to include the rationale for the project, its potential implications, and its relevance for translational research. We have also framed the work within the broader context of the behavioral and systems neuroscience community. We greatly appreciate this suggestion, as we believe it enhances the clarity and accessibility of the manuscript for the community.

For: "While brain activity can be controlled through feedback, other variables such as movements have been less studied, in part because their analysis in real time is more challenging." I would highlight research that has studied the control of behavior through feedback, such as the Mathis paper where mice learn to pull a joystick to a virtual box, and adapt this motion to a force perturbation.

We have added a citation to the Mathis paper and describe this as an additional form of feedback. The text is quoted below:

“Opportunities also exist in extending real time pose classification (Forys et al. 2020; Kane et al. 2020) and movement perturbation (Mathis et al. 2017) to shape aspects of an animal’s motor repertoire.”

Some of the results content would be better suited for the methods, one example: "A previous version of the CLNF system was found to have non-linear audio generation above 10 kHz, partly due to problems in the audio generation library and partly due to the consumer-grade speaker hardware we were employing. This was fixed by switching to the Audiostream (https://github.com/kivy/audiostream) library for audio generation and testing the speakers to make sure they could output the commanded frequencies"

This is now moved to the Methods section.

For: "There are reports of cortical plasticity during motor learning tasks, both at cellular and mesoscopic scales (17-19), supporting the idea that neural efficiency could improve with learning," not sure I agree with this, the studies on cortical plasticity are usually to show a neural basis for the learning observed, efficiency is separate from this.

We have modified this statement to remove the concept of efficiency "There are reports of cortical plasticity during motor learning tasks, both at cellular and mesoscopic scales (17-19).”

The paragraph that opens "Distinct task- and reward-related cortical dynamics" that describes the experiment should appear in the previous section, as the data is introduced there.

We have moved the mentioned paragraphs in the previous section where we presented the data and other experiment details. This makes the text more readable and contextual.

I would present the different ROI rules with better descriptors and visualization to improve the readability.

We have added Supplementary Figure 7, which provides visualizations of the ROIs across all task rules used in the CLNF experiments.

Minor corrections to the text and figures.

Figure 1 is a little crowded, combining the CLNF and CLMF experiments, I would turn this into a 2 panel figure, one for each, similar to how you did figure 2.

We have revised Figure 1 to include two panels, one for CLNF and one for CLMF. The colored components indicate elements specific to each setup, while the uncolored components represent elements shared between CLNF and CLMF. Relevant text in the manuscript is updated to refer to these figures.

For Figure 2, the organization of the CLMF section is not intuitive for the reader. I would reorder it so it has a similar flow as the CLNF experiment.

We have revised the figure by updating the layout of panel B (CLMF) to align with panel A (CLNF), thereby creating a more intuitive and consistent flow between the panels. We appreciate this helpful suggestion, which we believe has substantially improved the clarity of the figure. The corresponding text in the manuscript has also been updated to reflect these changes.

For Figure 3, highlight that C and E are examples. They also seem a little out of place, so they could even be removed.

We have now explicitly labeled Figures 3C and 3E as representative examples (figure legend and on figure itself). We believe including these panels provides helpful context for readers: Figure 3C illustrates how the ROIs align on the dorsal cortical brain map with segmented cortical regions, while Figure 3E shows example paw trajectories in three dimensions, allowing visualization of the movement patterns observed during the trials.

In the plots, I would add sample sizes, for instance, in CLNF learning curve in Figure 4A, how many animals are in each group? 

We have labeled Figure 4 with number of animals used in CLNF (No-rule-change, N=23; Rule-change, N=17), and CLMF (Rule-change, N=8; No-rule-change, N=4; No-feedback, N=4).

Also, Figure 7 for example, which figures are single-sessions, versus across animals? For Figure 7c, what time bin is the data taken from?

We have clarified this now and mentioned it in all the figures. Figure 7 in the previous version is Figure 9 in the current updated manuscript. Figure 9A is from individual sessions on different days from the same mouse. Figure 9B is the group average reward centered ΔF/F₀ activity in different cortical regions (Rule-change, N=8; No-rule-change, N=4; No-feedback, N=4). Figure 9C shows average ΔF/F₀ peak values obtained within -1sec to +1sec centered around the reward point (N=8).

It says "punish" in Figure 3, but there is no punishment?

Yes, the task did not involve punishment. Each trial resulted in either a success, which is followed by a reward, or a failure, which is followed by a buzzer sound. To better reflect these outcomes, we have updated Figure 3 and replaced the labels “Reward” with “Success” and “Punish” with “Failure.”

The regression on 5c doesn't look quite right, also this panel is not mentioned in the text.

The figure referred to by the reviewer as Figure 5 is now presented as Figure 6 in the revised manuscript. Regarding the reviewer’s observation about the regression line in the left panel of Figure 5C, the apparent misalignment arises because the majority of the data points are densely clustered at the center of the scatter plot, where they overlap substantially. The regression line accurately reflects this concentration of overlapping data. To improve clarity, we have updated the figure and ensured that it is now appropriately referenced in the Results section.

Reviewer #2 (Recommendations for the authors):

(1) There would be many interesting observations and links between the peripheral and cortical studies if there was a body video available during the cortical study. Is there any such data available?

We agree that a detailed analysis of behavior during the CLNF task would be necessary to explore any behavior correlates with success in the task. Unfortunately, we do not have a sufficient video of the whole body to perform such an analysis.

(2) The text (p. 24) states: [intracortical GCAMP transients measured over days became more stereotyped in kinetics and were more correlated (to each other) as the task performance increased over the sessions (Figure 7E).] But I cannot find this quantification in the figures or text?

Figure 7 in the previous version of the manuscript now appears as Figure 9. In this figure, we present cortical activity across selected regions during trials, and in Figure 9E we highlight that this activity becomes more correlated. Since we did not formally quantify variability, we have removed the previous claim that the activity became stereotyped and revised the text in the updated manuscript accordingly.

Typos:

10-serest c (page 13)

Inverted color codes in figure 4E vs F

Reviewer #3 (Recommendations for the authors):

We have mostly attempted to limit the feedback to suggestions and posed a few questions that might be interesting to explore given the dataset the authors have collected.

Comments:

In close loop systems the latency is primary concern, and authors have successfully tested the latency of the system (Delay): from detection of an event to the reaction time was less than 67ms.

We have commented on the issues and limitations caused by latency, and potential future directions to overcome these challenges in responses to some of the previous comments.

Additional major comments:

"In general, all ROIs assessed that encompassed sensory, pre-motor, and motor areas were capable of supporting increased reward rates over time (Figure 4A, Animation 1)." Fig 4A is merely showing change in task performance over time and does not have information regarding the changes observed specific to CLNF for each ROI.

We acknowledge that the sample size for individual ROI rules was not sufficient for meaningful comparisons. To address this limitation, we pooled the data across all the rules tested. The manuscript includes a detailed list of the rules along with their corresponding sample sizes for transparency.

A ΔF/F₀ threshold value was calculated from a baseline session on day 0 that would have allowed 25% performance. Starting from this basal performance of around 25% on day 1, mice (CLNF No-rule-change, n=28 and CLNF Rule-change, n=13). It is unclear what the replicates here are. Trials or mice? The corresponding Figure legend has a much smaller n value.

Thank you for pointing this out. We realized that we had not indicated the sample replicates in the figure, and the use of n instead of N for the number of animals may have been misleading. We have now corrected the notation and clarified this information in the figure to resolve the discrepancy.

What were the replicates for each ROI pairs evaluated?

Each ROI rule and number of mice and trials are listed in Table 5 and Table 6.

Our analysis revealed that certain ROI rules (see description in methods) lead to a greater increase in success rate over time than others (Supplementary Figure 3D). The Supplementary figures 3C and 3D are blurry and could use higher resolution images. 

We have increased the font size of the text that was previously difficult to read and re-exported the figure at a higher resolution (300 DPI). We believe these changes will resolve the issue.

Also, It will help the reader is a visual representation of the ROI pairs are provided, instead of the text view. One interesting question is whether there are anatomical biases to fast vs slow learning pairs (Directionality - anterior/posterior, distance between the selected ROIs etc). This could be interesting to tease apart.

We have added Supplementary Figure 7, which provides visualizations of the ROIs across all task rules used in the CLNF experiments. While a detailed investigation of the anatomical basis of fast versus slow learning cortical ROIs is beyond the scope of the present study, we agree that this represents an exciting future direction for further research.

How distant should the ROIs be to achieve increased task performance?

We appreciate this insightful question. We did not specifically test this scenario. In our study, we selected 0.3 × 0.3 mm ROIs centered on the standard AIBS mouse brain atlas (CCF). At this resolution, ROIs do not overlap, regardless of their placement in a two-ROI experiment. Furthermore, because our threshold calculations are based on baseline recordings, we expect the system would function for any combination of ROI placements. Nonetheless, exploring this systematically would be an interesting avenue for future experiments.

Figures:

I would leave out some of the methodological details such as the protocol for water restriction (Fig. 3) out of the legend. This will help with readability.

We have removed some of the methodological details, including those mentioned above, from the legend of Figure 3 in the updated manuscript.

Fig 1 and Fig 2: In my opinion, It would be easier for the reader if the current Fig. 2, which provides a high level description of CLNF and CLBF is presented as Fig. 1. The current Fig. 1, goes into a lot of methodological implementation details, and also includes a lot of programming jargon that is being introduced early in the paper that is hard to digest early on in the paper's narrative.

Thank you for the suggestion. In the new manuscript, Figure 1 and Figure 2 have been swapped.

Higher-resolution images/ plots are needed in many instances. Unsure if this is the pdf compression done by the manuscript portal that is causing this.

All figures were prepared in vector graphics format using the open-source software Inkscape. For this manuscript, we exported the images at 300 DPI, which is generally sufficient for publication-quality documents. The submission portal may apply additional processing, which could have resulted in a reduction in image quality. We will carefully review the final submission files and ensure that all figures are clear and of high quality.

The authors repeatedly show ROI specific analysis M1_L, F1_R etc. It will be helpful to provide a key, even if redundant in all figures to help the reader.

We have now included keys to all such abbreviations in all the figures.

There are also instances of editorialization and interpretation e.g., "Surprisingly, the "Rule-change" mice were able to discover the change in rule and started performing above 70% within a day of the rule change, on day 6" that would be more appropriate in the main body of the paper.

Thank you for pointing this out in the figure legend, and we have removed it now since we already discussed this in the Results.

Minor comments

(1) The description of Figure 1 is hard to follow and can be described better based on how the information is processed and executed in the system from source to processing and back. Using separated colors (instead of shaded of grey) for the neuro feedback and movement feedback would help as well. Common components could have a different color. The specification like the description of the config file should come later.

Figure 1 in the previous version is Figure 2 in the updated version. We have taken suggestions from other reviewers and made the figure easier to understand and split it into two panels with color coding Green for CLNF, Pink for CLMF specific parts while common shared parts are left without any color.

(2) Page 20 last paragraph:

Authors are neglecting that the rule change is done one day prior and the results that you see in the second half on the 6th day are not just because of the first half of the 6th day instead combined training on the 5th day (rule change) and then the first half of the 6th day. Rephrasing this observation is essential.

We have revised the text for clarity to indicate that the performance increase observed on day 6 is not necessarily attributable to training on that day. In fact, we noted and mentioned that mice began to perform the task better during the second half of the session on day 5 itself.

(3)  The method section description of the CLMF setup (Page no 39 first paragraph) is more detailed, a diagram of this setup would make it easy to follow and a better read.

We have made changes to the CLMF setup (Figure 1B) and CLMF schematic (Figure 2B) to make it easier to understand parts of the setup and flow of control.

Reviewer #3 (Public review):

Summary

The paper presents a imaging and analysis pipeline for whole-mount gastruloid imaging with two-photon microscopy. The presented pipeline includes spectral unmixing, registration, segmentation, and a wavelength-depended intensity normalization step, followed by quantitative analysis of spatial gene expression patterns and nuclear morphometry on a tissue level. The utility of the approach is demonstrated by several experimental findings such as establishing spatial correlations between local nuclear deformation and tissue density changes, as well as radial distribution pattern of mesoderm markers. The pipeline is distributed as a Python package, notebooks and multiple napari plugins.

Strengths

The paper is well-written with detailed methodological descriptions, which I think would make it a valuable reference for researchers performing similar volumetric tissue imaging experiments (gastruloids/organoids). The pipeline itself addresses many practical challenges including resolution loss within tissue, registration of large volumes, nuclear segmentation, and intensity normalization. Especially the intensity decay measurements and wavelength-dependent intensity normalization approach using nuclear (Hoechst) signal as reference is very interesting and should be applicable to other imaging contexts. The morphometric analysis is equally well done with the correlation between nuclear shape deformation and tissue density changes being a interesting finding. The paper is quite thorough in its technical description of the methods (which are a lot) and their experimental validation is appropriate. Finally, the provided code and napari plugins seem to be well done (I installed a selected list of the plugins and they ran without issues) and should be very helpful for the community.

Comments on revisions:

The minor issues that I originally raised in my first review have been fully resolved in the revised version.

Author response:

The following is the authors’ response to the original reviews.

Public Reviews:  

Reviewer #1 (Public review):  

Summary:  

The image analysis pipeline is tested in analysing microscopy imaging data of gastruloids of varying sizes, for which an optimised protocol for in toto image acquisition is established based on whole mount sample preparation using an optimal refractive index matched mounting media, opposing dual side imaging with two-photon microscopy for enhanced laser penetration, dual view registration, and weighted fusion for improved in toto sample data representation. For enhanced imaging speed in a two-photon microscope, parallel imaging was used, and the authors performed spectral unmixing analysis to avoid issues of signal cross-talk.  

In the image analysis pipeline, different pre-treatments are done depending on the analysis to be performed (for nuclear segmentation - contrast enhancement and normalisation; for quantitative analysis of gene expression - corrections for optical artifacts inducing signal intensity variations). Stardist3D was used for the nuclear segmentation. The study analyses into properties of gastruloid nuclear density, patterns of cell division, morphology, deformation, and gene expression.  

Strengths:  

The methods developed are sound, well described, and well-validated, using a sample challenging for microscopy, gastruloids. Many of the established methods are very useful (e.g. registration, corrections, signal normalisation, lazy loading bioimage visualisation, spectral decomposition analysis), facilitate the development of quantitative research, and would be of interest to the wider scientific community.

We thank the reviewer for this positive feedback.

Weaknesses:  

A recommendation should be added on when or under which conditions to use this pipeline. 

We thank the reviewer for this valuable feedback, we added the text in the revised version, ines 418 to 474. “In general, the pipeline is applicable to any tissue, but it is particularly useful for large and dense 3D samples—such as organoids, embryos, explants, spheroids, or tumors—that are typically composed of multiple cell layers and have a thickness greater than 50 µm”.

“The processing and analysis pipeline are compatible with any type of 3D imaging data (e.g. confocal, 2 photon, light-sheet, live or fixed)”.

“Spectral unmixing to remove signal cross-talk of multiple fluorescent targets is typically more relevant in two-photon imaging due to the broader excitation spectra of fluorophores compared to single-photon imaging. In confocal or light-sheet microscopy, alternating excitation wavelengths often circumvents the need for unmixing. Spectral decomposition performs even better with true spectral detectors; however, these are usually not non-descanned detectors, which are more appropriate for deep tissue imaging. Our approach demonstrates that simultaneous cross-talk-free four-color two-photon imaging can be achieved in dense 3D specimen with four non-descanned detectors and co-excitation by just two laser lines. Depending on the dispersion in optically dense samples, depth-dependent apparent emission spectra need to be considered”.

“Nuclei segmentation using our trained StarDist3D model is applicable to any system under two conditions: (1) the nuclei exhibit a star-convex shape, as required by the StarDist architecture, and (2) the image resolution is sufficient in XYZ to allow resampling. The exact sampling required is object- and system-dependent, but the goal is to achieve nearly isotropic objects with diameters of approximately 15 pixels while maintaining image quality. In practice, images containing objects that are natively close to or larger than 15 pixels in diameter should segment well after resampling. Conversely, images with objects that are significantly smaller along one or more dimensions will require careful inspection of the segmentation results”.

“Normalization is broadly applicable to multicolor data when at least one channel is expected to be ubiquitously expressed within its domain. Wavelength-dependent correction requires experimental calibration using either an ubiquitous signal at each wavelength. Importantly, this calibration only needs to be performed once for a given set of experimental conditions (e.g., fluorophores, tissue type, mounting medium)”.

“Multi-scale analysis of gene expression and morphometrics is applicable to any 3D multicolor image. This includes both the 3D visualization tools (Napari plugins) and the various analytical plots (e.g., correlation plots, radial analysis). Multi-scale analysis can be performed even with imperfect segmentation, as long as segmentation errors tend to cancel out when averaged locally at the relevant spatial scale. However, systematic errors—such as segmentation uncertainty along the Z-axis due to strong anisotropy—may accumulate and introduce bias in downstream analyses. Caution is advised when analyzing hollow structures (e.g., curved epithelial monolayers with large cavities), as the pipeline was developed primarily for 3D bulk tissues, and appropriate masking of cavities would be needed”.

Reviewer #2 (Public review):  

Summary:  

This study presents an integrated experimental and computational pipeline for high-resolution, quantitative imaging and analysis of gastruloids. The experimental module employs dual-view two-photon spectral imaging combined with optimized clearing and mounting techniques to image whole-mount immunostained gastruloids. This approach enables the acquisition of comprehensive 3D images that capture both tissue-scale and single-cell level information.  

The computational module encompasses both pre-processing of acquired images and downstream analysis, providing quantitative insights into the structural and molecular characteristics of gastruloids. The pre-processing pipeline, tailored for dual-view two-photon microscopy, includes spectral unmixing of fluorescence signals using depth-dependent spectral profiles, as well as image fusion via rigid 3D transformation based on content-based block-matching algorithms. Nuclei segmentation was performed using a custom-trained StarDist3D model, validated against 2D manual annotations, and achieving an F1 score of 85+/-3% at a 50% intersection-over-union (IoU) threshold. Another custom-trained StarDist3D model enabled accurate detection of proliferating cells and the generation of 3D spatial maps of nuclear density and proliferation probability. Moreover, the pipeline facilitates detailed morphometric analysis of cell density and nuclear deformation, revealing pronounced spatial heterogeneities during early gastruloid morphogenesis.  

All computational tools developed in this study are released as open-source, Python-based software.  

Strengths:  

The authors applied two-photon microscopy to whole-mount deep imaging of gastruloids, achieving in toto visualization at single-cell resolution. By combining spectral imaging with an unmixing algorithm, they successfully separated four fluorescent signals, enabling spatial analysis of gene expression patterns.  

The entire computational workflow, from image pre-processing to segmentation with a custom-trained StarDist3D model and subsequent quantitative analysis, is made available as open-source software. In addition, user-friendly interfaces are provided through the open-source, community-driven Napari platform, facilitating interactive exploration and analysis.

We thank the reviewer for this positive feedback.

Weaknesses:  

The computational module appears promising. However, the analysis pipeline has not been validated on datasets beyond those generated by the authors, making it difficult to assess its general applicability.

We agree that applying our analysis pipeline to published datasets—particularly those acquired with different imaging systems—would be valuable. However, only a few high-resolution datasets of large organoid samples are publicly available, and most of these either lack multiple fluorescence channels or represent 3D hollow structures. Our computational pipeline consists of several independent modules: spectral filtering, dual-view registration, local contrast enhancement, 3D nuclei segmentation, image normalization based on a ubiquitous marker, and multiscale analysis of gene expression and morphometrics. We added the following sentences to the Discussion, lines 418 to 474, and completed the discussion on applicability with a table showing the purpose, requirements, applicability and limitations of each step of the processing and analysis pipeline.

“Spectral filtering has already been applied in other systems (e.g. [7] and [8]), but is here extended to account for imaging depth-dependent apparent emission spectra of the different fluorophores. In our pipeline, we provide code to run spectral filtering on multichannel images, integrated in Python. In order to apply the spectral filtering algorithm utilized here, spectral patterns of each fluorophore need to be calibrated as a function of imaging depth, which depend on the specific emission windows and detector settings of the microscope”.

“Image normalization using a wavelength-dependent correction also requires calibration on a given imaging setup to measure the difference in signal decay among the different fluorophores species. To our knowledge, the calibration procedures for spectral-filtering and our image-normalization approach have not been performed previously in 3D samples, which is why validation on published datasets is not readily possible. Nevertheless, they are described in detail in the Methods section, and the code used—from the calibration measurements to the corrected images—is available open-source at the Zenodo link in the manuscript”.

Dual-view registration, local contrast enhancement, and multiscale analysis of gene expression and morphometrics are not limited to organoid data or our specific imaging modalities. To evaluate our 3D nuclei segmentation model, we tested it on diverse systems, including gastruloids stained with the nuclear marker Draq5 from Moos et al. [1]; breast cancer spheroids; primary ductal adenocarcinoma organoids; human colon organoids and HCT116 monolayers from Ong et al. [2]; and zebrafish tissues imaged by confocal microscopy from Li et al [3]. These datasets were acquired using either light-sheet or confocal microscopy, with varying imaging parameters (e.g., objective lens, pixel size, staining method). The results are added in the manuscript, Fig. S9b.

Besides, the nuclei segmentation component lacks benchmarking against existing methods.  

We agree with the reviewer that a benchmark against existing segmentation methods would be very useful. We tried different pre-trained models:

CellPose, which we tested in a previous paper ([4]) and which showed poor performances compared to our trained StarDist3D model.

DeepStar3D ([2]) is only available in the software 3DCellScope. We could not benchmark the model on our data, because the free and accessible version of the software is limited to small datasets. An image of a single whole-mount gastruloid with one channel, having dimensions (347,467,477) was too large to be processed, see screenshot below. The segmentation model could not be extracted from the source code and tested externally because the trained DeepStar3D weights are encrypted.

Author response image 1.

Screenshot of the 3DCellScore software. We could not perform 3D nuclei segmentation of a whole-mount gastruloids because the image size was too large to be processed.

AnyStar ([5]), which is a model trained from the StarDist3D architecture, was not performing well on our data because of the heterogeneous stainings. Basic pre-processing such as median and gaussian filtering did not improve the results and led to wrong segmentation of touching nuclei. AnyStar was demonstrated to segment well colon organoids in Ong et al, 2025 ([2]), but the nuclei were more homogeneously stained. Our Hoechst staining displays bright chromatin spots that are incorrectly labeled as individual nuclei.

Cellos ([6]), another model trained from StarDist3D, was also not performing well. The objects used for training and to validate the results are sparse and not touching, so the predicted segmentation has a lot of false negatives even when lowering the probability threshold to detect more objects. Additionally, the network was trained with an anisotropy of (9,1,1), based on images with low z resolution, so it performed poorly on almost isotropic images. Adapting our images to the network’s anisotropy results in an imprecise segmentation that can not be used to measure 3D nuclei deformations.

We tried both Cellos and AnyStar predictions on a gastruloid image from Fig. S2 of our main manuscript.  The results are added in the manuscript, Fig. S9b. Fig3 displays the results qualitatively compared to our trained model Stardist-tapenade.

Author response image 2.

Qualitative comparison of two published segmentation models versus our model. We show one slice from the XY plane for simplicity. Segmentations are displayed with their contours only. (Top left) Gastruloid stained with Hoechst, image extracted from Fig S2 of our manuscript. (Top right) Same image overlayed with the prediction from the Cellos model, showing many false negatives. (Bottom left) Same image overlayed with the prediction from our Stardist-tapenade model. (Bottom right) Same image overlayed with the prediction from the AnyStar model, false positives are indicated with a red arrow.

CellPose-SAM, which is a recent model developed building on the CellPose framework. The pre-trained model performs well on gastruloids imaged using our pipeline, and performs better than StarDist3D at segmenting elongated objects such as deformed nuclei. The performances are qualitatively compared on Fig. S9a and S10.  We also demonstrate how using local contrast enhancement improves the results of CellPose-SAM (Fig. S10a), showing the versatility of the Tapenade pre-processing module. Tissue-scale, packing-related metrics from Cellpose–SAM labels qualitatively match those from stardist-tapenade as shown Fig.10c and d.

Appraisal:  

The authors set out to establish a quantitative imaging and analysis pipeline for gastruloids using dual-view two-photon microscopy, spectral unmixing, and a custom computational framework for 3D segmentation and gene expression analysis. This aim is largely achieved. The integration of experimental and computational modules enables high-resolution in toto imaging and robust quantitative analysis at the single-cell level. The data presented support the authors' conclusions regarding the ability to capture spatial patterns of gene expression and cellular morphology across developmental stages.  

Impact and utility:  

This work presents a compelling and broadly applicable methodological advance. The approach is particularly impactful for the developmental biology community, as it allows researchers to extract quantitative information from high-resolution images to better understand morphogenetic processes. The data are publicly available on Zenodo, and the software is released on GitHub, making them highly valuable resources for the community.  

We thank the reviewer for these positive feedbacks.

Reviewer #3 (Public review):

Summary  

The paper presents an imaging and analysis pipeline for whole-mount gastruloid imaging with two-photon microscopy. The presented pipeline includes spectral unmixing, registration, segmentation, and a wavelength-dependent intensity normalization step, followed by quantitative analysis of spatial gene expression patterns and nuclear morphometry on a tissue level. The utility of the approach is demonstrated by several experimental findings, such as establishing spatial correlations between local nuclear deformation and tissue density changes, as well as the radial distribution pattern of mesoderm markers. The pipeline is distributed as a Python package, notebooks, and multiple napari plugins.  

Strengths  

The paper is well-written with detailed methodological descriptions, which I think would make it a valuable reference for researchers performing similar volumetric tissue imaging experiments (gastruloids/organoids). The pipeline itself addresses many practical challenges, including resolution loss within tissue, registration of large volumes, nuclear segmentation, and intensity normalization. Especially the intensity decay measurements and wavelength-dependent intensity normalization approach using nuclear (Hoechst) signal as reference are very interesting and should be applicable to other imaging contexts. The morphometric analysis is equally well done, with the correlation between nuclear shape deformation and tissue density changes being an interesting finding. The paper is quite thorough in its technical description of the methods (which are a lot), and their experimental validation is appropriate. Finally, the provided code and napari plugins seem to be well done (I installed a selected list of the plugins and they ran without issues) and should be very helpful for the community.

We thank the reviewer for his positive feedback and appreciation of our work.

Weaknesses  

I don't see any major weaknesses, and I would only have two issues that I think should be addressed in a revision:  

(1) The demonstration notebooks lack accompanying sample datasets, preventing users from running them immediately and limiting the pipeline's accessibility. I would suggest to include (selective) demo data set that can be used to run the notebooks (e.g. for spectral unmixing) and or provide easily accessible demo input sample data for the napari plugins (I saw that there is some sample data for the processing plugin, so this maybe could already be used for the notebooks?).  

We thank the reviewer for this relevant suggestion. The 7 notebooks were updated to automatically download sample tests. The different parts of the pipeline can now be run immediately:

https://github.com/GuignardLab/tapenade/tree/chekcs_on_notebooks/src/tapenade/notebooks

(2) The results for the morphometric analysis (Figure 4) seem to be only shown in lateral (xy) views without the corresponding axial (z) views. I would suggest adding this to the figure and showing the density/strain/angle distributions for those axial views as well.

A morphometric analysis based on the axial views was added as Fig. S6a of the manuscript, complementary to the XY views.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):  

In lines 64 and 65, it is mentioned that confocal and light-sheet microscopy remain limited to samples under 100μm in diameter. I would recommend revising this sentence. In the paper of Moos and colleagues (also cited in this manuscript; PMID: 38509326), gastruloid samples larger than 100μm are imaged in toto with an open-top dual-view and dual-illumination light-sheet microscope, and live cell behaviour is analysed. Another example, if considering also multi-angle systems, is the impressive work of McDole and colleagues (PMID: 30318151), in which one of the authors of this manuscript is a corresponding author. There, multi-angle light sheet microscopy is used for in toto imaging and reconstruction of post-implantation mouse development (samples much larger than 100μm). Some multi-sample imaging strategies have been developed for this type of imaging system, though not to the sample number extent allowed by the Viventis LS2 system or the Bruker TruLive3D imager, which have higher image quality limitations.

We thank the reviewer for this remark. As reported in their paper, Moos et al. used dual-view light-sheet microscopy to image gastruloids, which are particularly dense and challenging tissues, with whole-mount samples of approximately 250 µm in diameter. Nevertheless, their image quality metric (DCT) shows a rapid twofold decrease within 50 µm depth (Extended Fig 5.h), whereas with two-photon microscopy, our image quality metric (FRC-QE) decreases by a factor of two over 150 µm in non-cleared samples (PBS) (see Fig. 2 c). While these two measurements (FRC-QE versus DCT) are not directly comparable, the observed difference reflects the superior depth performance of two-photon microscopy, owing in part to the use of non-descanned detectors. In our case, imaging was performed with Hoechst, a blue fluorophore suboptimal for deep imaging, whereas in the Moos dataset (Draq5, far-red), the configuration was more favorable for imaging in depth  which further supports our conclusion.

In McDole et al, tissues reaching 250µm were imaged from 4 views, but do not reach cellular-scale resolution in deeper layers compatible with cell segmentation to our knowledge.

We corrected the sentence ‘However, light-sheet and confocal imaging approaches remain limited to relatively small organoids typically under 100 micrometers in diameter ‘ by the following (line 64) :

“While advances in light-sheet microscopy have extended imaging depth in organoids, maintaining high image quality throughout thick samples remains challenging. In practice, quantitative analyses are still largely restricted to organoids under roughly 100 µm in diameter”.

It is worth mentioning that two-photon microscopes are much more widely available than light sheet microscopes, and light sheet systems with 2-photon excitation are even less accessible, which makes the described workflow of Gros and colleagues have a wide community interest.  

We thank the reviewer for this remark, and added this suggestion line 74:

“Finally, two-photon microscopes are typically more accessible than light-sheet systems and allow for straightforward sample mounting, as they rely on procedures comparable to standard confocal imaging”.

Reviewer #2 (Recommendations for the authors):  

Suggestions:  

A comparison with established pre-trained models for 3D organoid image segmentation (e.g., Cellos[1], AnyStar[2], and DeepStar3D[3], all based on StarDist3D) would help highlight the advantages of the authors' custom StarDist3D model, which has been specifically optimized for two-photon microscopy images.  

(1)  Cellos: https://doi.org/10.1038/s41467-023-44162-6

(2)  AnyStar: https://doi.org/10.1109/WACV57701.2024.00742

(3)  DeepStar3D: https://doi.org/10.1038/s41592-025-02685-4

We agree with the reviewer that a benchmark against existing segmentation methods is very useful. This is addressed in the revised version, as detailed above (Figure 3).

Recommendations:  

Please clarify the following point. In line 195, the authors state, "This allowed us to detect all mitotic nuclei in whole-mount samples for any stage and size." Does this mean that the custom-trained StarDist3D model can detect 100% of mitotic nuclei? It was not clear from the manuscript, figures, or videos how this was validated. Given the reported performance scores of the StarDist3D model for detecting all nuclei, claiming 100% detection of mitotic nuclei seems surprisingly high.

We thank the reviewer for this comment. As it was detailed in the methods section, the detection score reaches 82%, and only the complete pipeline (detection+minimal manual curation) allows us to detect all mitotic nuclei. To make it clearer, the following precisions were added in the Results section:

”To detect division events, we stained gastruloids with phosphohistone H3 (ph3) and trained a separate custom Stardist3D model using 3D annotations of nuclei expressing ph3 (see Methods III H). This model together allowed us to detect nearly all mitotic nuclei in whole-mount samples for any stage and size (Fig.3f and Suppl.Movie 4), and we used minimal manual curation to correct remaining errors.”

Minor corrections:  

It appears that Figures 4-6 are missing from the submitted version, but they can be found in the manuscript available on bioRxiv.

We thank the reviewer for this remark, this was corrected immediately to add Figures 4 to 6.

In line 185, is the intended phrase "by comparing the 2D predictions and the 2D sliced annotated segments..."? 

To gain some clarity, we replaced the initial sentence:

“The f1 score obtained by comparing the 3D prediction and the 3D ground-truth is well approximated by the f1 score obtained by comparing the 2D annotations and the 2D sliced annotated segments, with at most a 5% difference between the two scores.” by

“The f1 score obtained in 3D (3D prediction compared with the 3D ground-truth) is well approximated by the f1 score obtained in 2D (2D predictions compared with the 2D sliced annotated segments). The difference between the 2 scores was at most 5%.”

Reviewer #3 (Recommendations for the authors):

(1) How is the "local neighborhood volume" defined, and how was it computed?

The reviewer is referring to this paragraph (the term is underscored) :

“To probe quantities related to the tissue structure at multiple scales, we smooth their signal with a Gaussian kernel of width σ, with σ defined as the spatial scale of interest. From the segmented nuclei instances, we compute 3D fields of cell density (number of cells per unit volume), nuclear volume fraction (ratio of nuclear volume to local neighborhood volume), and nuclear volume at multiple scales.”

To improve clarity, the phrasing has been revised: the term local neighborhood volume has been replaced by local averaging volume, and a reference to the Methods section has been added.

From the segmented nuclei instances, we compute 3D fields of cell density (number of cells per unit volume), nuclear volume fraction (ratio of space occupied by nuclear volume within the local averaging volume, as defined in the Methods III I), and nuclear volume at multiple scales.

(2) In the definition of inertia tensor (18), isn't the inner part normally defined in the reversed way (delta_i,j - ...)?

We thank the reviewer for noticing this error, which we fixed in the manuscript.

(3) For intensity normalization, the paper uses the Hoechst signal density as a proxy for a ubiquitous nuclei signal. I would assume that this is problematic, for eg, dividing cells (which would overestimate it). Would using the average Hoechst signal per nucleus mask (as segmentation is available) be a better proxy?

We agree that this idea is appealing if one assumes a clear relationship between nuclear volume and Hoechst intensity. However, since cell and nuclear volumes vary substantially with differentiation state (see Fig. 4), such a normalization approach would introduce additional biases at large spatial scales. We believe that the most robust improvement would instead consist in masking dividing cells during the normalization procedure, as these events could be detected and excluded from the computation.

Nonetheless, we believe the method proposed by the reviewer could prove relevant for other types of data, so we will implement this recommendation in the code available in the Tapenade package.

(4) Figures 4-6 were part of the Supplementary Material, but should be included in the main text?

We thank the reviewer for this remark, this was corrected immediately to add Figures 4-6.

We also noticed a missing reference to Fig. S3 in the main text, so we added lines 302 to 307 to comment on the wavelength-dependency of the normalization method. We improved the description of Fig.6, which lacked clarity (line 316 to 321, line 327).

(1) Moos, F., Suppinger, S., de Medeiros, G., Oost, K.C., Boni, A., Rémy, C., Weevers, S.L., Tsiairis, C., Strnad, P. and Liberali, P., 2024. Open-top multisample dual-view light-sheet microscope for live imaging of large multicellular systems. Nature Methods, 21(5), pp.798-803.

(2) Ong, H. T.; Karatas, E.; Poquillon, T.; Grenci, G.; Furlan, A.; Dilasser, F.; Mohamad Raffi, S. B.; Blanc, D.; Drimaracci, E.; Mikec, D.; Galisot, G.; Johnson, B. A.; Liu, A. Z.; Thiel, C.; Ullrich, O.; OrgaRES Consortium; Racine, V.; Beghin, A. (2025). Digitalized organoids: integrated pipeline for high-speed 3D analysis of organoid structures using multilevel segmentation and cellular topology.  Nature Methods, 22(6), pp.1343-1354

(3) Li, L., Wu, L., Chen, A., Delp, E.J. and Umulis, D.M., 2023. 3D nuclei segmentation for multi-cellular quantification of zebrafish embryos using NISNet3D. Electronic Imaging, 35, pp.1-9.

(4) Vanaret, J., Dupuis, V., Lenne, P. F., Richard, F., Tlili, S., & Roudot, P. (2023). A detector-independent quality score for cell segmentation without ground truth in 3D live fluorescence microscopy. IEEE Journal of Selected Topics in Quantum Electronics, 29(4:Biophotonics), 1-12.

(5) Dey, N., Abulnaga, M., Billot, B., Turk, E. A., Grant, E., Dalca, A. V., & Golland, P. (2024). AnyStar: Domain randomized universal star-convex 3D instance segmentation. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (pp. 7593-7603).

(6) Mukashyaka, P., Kumar, P., Mellert, D. J., Nicholas, S., Noorbakhsh, J., Brugiolo, M., ... & Chuang, J. H. (2023). High-throughput deconvolution of 3D organoid dynamics at cellular resolution for cancer pharmacology with Cellos. Nature Communications, 14(1), 8406.

(7) Rakhymzhan, A., Leben, R., Zimmermann, H., Günther, R., Mex, P., Reismann, D., ... & Niesner, R. A. (2017). Synergistic strategy for multicolor two-photon microscopy: application to the analysis of germinal center reactions in vivo. Scientific reports, 7(1), 7101.

(8) Dunsing, V., Petrich, A., & Chiantia, S. (2021). Multicolor fluorescence fluctuation spectroscopy in living cells via spectral detection. Elife, 10, e69687.

This breakdown of coding into its smallest pieces is helpful because it explains how algorithms make decisions. While this is a fraction of code compared to the algorithms that are used on social media, it makes you question what the code looks like and what the decisions in these algorithms are based on.

In other words, it’s like the fundamental structure for more developed code? It could be interpreted as the building blocks for what humans can understand of the vast universe of coding.

Author response:

The following is the authors’ response to the original reviews.

Public Reviews: 

Reviewer #1 (Public review):

We thank Reviewer #1 for its thoughtful and constructive feedback. We found the suggestions particularly helpful in refining the conceptual framework and clarifying key aspects of our interpretations.

Summary:

This paper investigates the potential link between amygdala volume and social tolerance in multiple macaque species. Through a comparative lens, the authors considered tolerance grade, species, age, sex, and other factors that may contribute to differing brain volumes. They found that amygdala, but not hippocampal, volume differed across tolerance grades, such that hightolerance species showed larger amygdala than low-tolerance species of macaques. They also found that less tolerant species exhibited increases in amygdala volume with age, while more tolerant species showed the opposite. Given their wide range of species with varied biological and ecological factors, the authors' findings provide new evidence for changes in amygdala volume in relation to social tolerance grades. Contributions from these findings will greatly benefit future efforts in the field to characterize brain regions critical for social and emotional processing across species.

Strengths:

(1) This study demonstrates a concerted and impressive effort to comparatively examine neuroanatomical contributions to sociality in monkeys. The authors impressively collected samples from 12 macaque species with multiple datapoints across species age, sex, and ecological factors. Species from all four social tolerance grades were present. Further, the age range of the animals is noteworthy, particularly the inclusion of individuals over 20 years old - an age that is rare in the wild but more common in captive settings. 

(2) This work is the first to report neuroanatomical correlates of social tolerance grade in macaques in one coherent study. Given the prevalence of macaques as a model of social neuroscience, considerations of how socio-cognitive demands are impacted by the amygdala are highly important. The authors' findings will certainly inform future studies on this topic.

(3) The methodology and supplemental figures for acquiring brain MRI images are well detailed. Clear information on these parameters is crucial for future comparative interpretations of sociality and brain volume, and the authors do an excellent job of describing this process in full.

Weaknesses:

(1) The nature vs. nurture distinction is an important one, but it may be difficult to draw conclusions about "nature" in this case, given that only two data points (from grades 3 and 4) come from animals under one year of age (Method Figure 1D). Most brains were collected after substantial social exposure-typically post age 1 or 1.5-so the data may better reflect developmental changes due to early life experience rather than innate wiring. It might be helpful to frame the findings more clearly in terms of how early experiences shape development over time, rather than as a nature vs. nurture dichotomy.

We agree with the reviewer that presenting our findings through a strict nature vs. nurture dichotomy was potentially misleading. We have revised the introduction and the discussion (e.g. lines 85-95 and 363-365) to clarify that we examined how neurodevelopmental trajectories differ across social grades with the caveat of related to the absence of very young individuals in our samples.  We now explicitly mention that our results may reflect both early species-typical biases and experience-dependent maturation.

We positioned our study on social tolerance in a comparative neuroscience framework and introduced a tentative working model that articulates behavioral traits, cognitive dimensions, and their potential subcortical neural substrates

Drawing upon 18 behavioral traits identified in Thierry’s comparative analyses (Thierry, 2021, 2007), we organize these traits into three core dimensions: socio-cognitive demands, behavioral inhibition, and the predictability of the social environment (Table 1). This conceptualization does not aim to redefine social tolerance itself, but rather to provide a structured basis for testing neuroanatomical hypotheses related to social style variability. It echoes recent efforts to bridge behavioral ecology and cognitive neuroscience by linking specific mental abilities – such as executive functions or metacognition – with distinct prefrontal regions shaped by social and ecological pressures (Bouret et al., 2024).

“Cross-fostering experiments (De Waal and Johanowicz, 1993), along with our own results, suggest that social tolerance grades reflect both early, possibly innate predispositions and later environmental shaping”.

(2) It would be valuable to clarify how the older individuals, especially those 20+ years old, may have influenced the observed age-related correlations (e.g., positive in grades 1-2, negative in grades 3-4). Since primates show well-documented signs of aging, some discussion of the potential contribution of advanced age to the results could strengthen the interpretation.

We thank the reviewer for highlighting this important point. In our dataset, younger and older subjects are underrepresented, but they are distributed across all subgroups. Therefore, we do not think that it could drive the interaction effect we are reporting. In our sample, amygdala volume tended to increase with age in intolerant species and decrease in tolerant species. We included a new analysis (Figure 4) that allows providing a clearer assessment of when social grades 1 vs 4 differed in terms of amygdala and hippocampus volume. While our model accounts for age continuously, we agree that age-related variation deserves cautious interpretation and require longitudinal designs in future studies.

We also added the following statements in the discussion (lines 386-391)

“Due to a limited sample size of our study, this crossing trend, already accounted for by our continuous age model, should be further investigated. These results call for cautious interpretation of age-related variation and further emphasize the importance of longitudinal studies integrating both behavioral, cognitive and anatomical data in non-human primates, which would help to better understand the link between social environment and brain development (Song et al., 2021)”.

(3) The authors categorize the behavioral traits previously described in Thierry (2021) into 3 selfdefined cognitive requirements, however, they do not discuss under what conditions specific traits were assigned to categories or justify why these cognitive requirements were chosen. It is not fully clear from Thierry (2021) alone how each trait would align with the authors' categories. Given that these traits/categories are drawn on for their neuroanatomical hypotheses, it is important that the authors clarify this. It would be helpful to include a table with all behavioral traits with their respective categories, and explain their reasoning for selecting each cognitive requirement category.

Thank you for this important suggestion. We have extensively revised the introduction to explain how we derived from the scientific literature the three cognitive dimensions—socio-cognitive demands, behavioral inhibition, and predictability of the social environment—. We now provide a complete overview of the 18 behavioral traits described in Thierry’s framework and their cognitive classification in a dedicated table , along with hypothesized neural correlates. We have also mentioned traits that were not classified in our framework along with short justification of this classification. We believe this addition significantly improves the transparency and intelligibility of our conceptual approach.

“The concept of social tolerance, central to this comparative approach, has sometimes been used in a vague or unidimensional way. As Bernard Thierry (2021) pointed out, the notion was initially constructed around variations in agonistic relationships – dominance, aggressiveness, appeasement or reconciliation behaviors – before being expanded to include affiliative behaviors, allomaternal care or male–male interactions (Thierry, 2021). These traits do not necessarily align along a single hierarchical axis but rather reflect a multidimensional complexity of social style, in which each trait may have co-evolved with others (Thierry, 2021, 2000; Thierry et al., 2004). Moreover, the lack of a standardized scientific definition has sometimes led to labeling species as “tolerant” or “intolerant” without explicit criteria (Gumert and Ho, 2008; Patzelt et al., 2014). These behavioral differences are characterized by different styles of dominance (Balasubramaniam et al., 2012), severity of agonistic interactions (Duboscq et al., 2014), nepotism (Berman and Thierry, 2010; Duboscq et al., 2013; Sueur et al., 2011) and submission signals (De Waal and Luttrell, 1985; Rincon et al., 2023), among the 18 covariant behavioral traits described in Thierry's classification of social tolerance (Thierry, 2021, 2017, 2000)”.

“To ground the investigation of social tolerance in a comparative neuroanatomical framework, we introduce a tentative working model that articulates behavioral traits, cognitive dimensions, and their potential subcortical neural substrates. Drawing upon 18 behavioral traits identified in Thierry’s comparative analyses (Thierry, 2021, 2007), we organized these traits into three core dimensions: socio-cognitive demands, behavioral inhibition, and the predictability of the social environment (Table 1). This conceptualization does not aim to redefine social tolerance itself, but rather to provide a structured basis for testing neuroanatomical hypotheses related to social style variability. It echoes recent efforts to bridge behavioral ecology and cognitive neuroscience by linking specific mental abilities – such as executive functions or metacognition – with distinct prefrontal regions shaped by social and ecological pressures (Bouret et al., 2024; Testard 2022)”.

(4) One of the main distinctions the authors make between high social tolerance species and low tolerance species is the level of complex socio-cognitive demands, with more tolerant species experiencing the highest demands. However, socio-cognitive demands can also be very complex for less tolerant species because they need to strategically balance behaviors in the presence of others. The relationships between socio-cognitive demands and social tolerance grades should be viewed in a more nuanced and context-specific manner. 

We fully agree and we did not mean that intolerant species lives in a ‘simple’ social environment but that the ones of more tolerant species is markedly more demanding. Evidence supporting this statement include their more efficient social networks (Sueur et al., 2011) and more complex communicative skills (e.g. tolerant macaques displayed higher levels of vocal diversity and flexibility than intolerant macaques in social situation with high uncertainty (Rebout et al., 2020).

In the revised version (lines 106-122), we now highlight that socio-cognitive challenges arise across the tolerance spectrum, including in less tolerant species where strategic navigation of rigid hierarchies and risk-prone interactions is required. We hope that this addition offers a more balanced and nuanced framing of socio-cognitive demands across macaque societies

“The first category, socio-cognitive demands, refers to the cognitive resources needed to process, monitor, and flexibly adapt to complex social environments. Linking those parameters to neurological data is at the core of the social brain theory to explain the expansion of the neocortex in primates (Dunbar). Macaques social systems require advanced abilities in social memory, perspective-taking, and partner evaluation (Freeberg et al., 2012). This is particularly true in tolerant species, where the increased frequency and diversity of interactions may amplify the demands on cognitive tracking and flexibility. Tolerant macaque species typically live in larger groups with high interaction frequencies, low nepotism, and a wider range of affiliative and cooperative behaviors, including reconciliation, coalition-building, and signal flexibility (REF). Tolerant macaque species also exhibit a more diverse and flexible vocal and facial repertoire than intolerants ones which may help reduce ambiguity and facilitate coordination in dense social networks (Rincon et al., 2023; Scopa and Palagi, 2016; Rebout 2020). Experimental studies further show that macaques can use facial expressions to anticipate the likely outcomes of social interactions, suggesting a predictive function of facial signals in managing uncertainty (Micheletta et al., 2012; Waller et al., 2016). Even within less tolerant species, like M. mulatta, individual variation in facial expressivity has been linked to increased centrality in social networks and greater group cohesion, pointing to the adaptive value of expressive signaling across social styles (Whitehouse et al., 2024)”.

(5) While the limitations section touches on species-related considerations, the issue of individual variability within species remains important. Given that amygdala volume can be influenced by factors such as social rank and broader life experience, it might be useful to further emphasize that these factors could introduce meaningful variation across individuals. This doesn't detract from the current findings but highlights the importance of considering life history and context when interpreting subcortical volumes-particularly in future studies.

We have now emphasized this point in the limitations section (lines 441-456). While our current dataset does not allow us to fully control for individual-level variables across all collection centers, we recognize that factors such as rank, social exposure, and individual life history may influence subcortical volumes

“Although we explained some interspecies variability, adding subjects to our database will increase statistical power and will help addressing potential confounding factors such as age or sex in future studies. One will benefit from additional information about each subject. While considered in our modelling, the social living and husbandry conditions of the individuals in our dataset remain poorly documented. The living environment has been considered, and the size of social groups for certain individuals, particularly for individuals from the CdP, have been recorded. However, these social characteristics have not been determined for all individuals in the dataset. As previously stated, the social environment has a significant impact on the volumetry of certain regions. Furthermore, there is a lack of data regarding the hierarchy of the subjects under study and the stress they experience in accordance with their hierarchical rank and predictability of social outcomes position (McCowan et al., 2022)”. 

Reviewer #2 (Public review):

We thank Reviewer #2 for its thoughtful remarks and for acknowledging the value of our comparative approach despite its inherent constraints.

Summary:

This comparative study of macaque species and the type of social interaction is both ambitious and inevitably comes with a lot of caveats. The overall conclusion is that more intolerant species have a larger amygdala. There are also opposing development profiles regarding amygdala volume depending on whether it is a tolerant or intolerant species.

To achieve any sort of power, they have combined data from 4 centres, which have all used different scanning methods, and there are some resolution differences. The authors have also had to group species into 4 classifications - again to assist with any generalisations and power. They have focused on the volumes of two structures, the amygdala and the hippocampus, which seems appropriate. Neither structure is homogeneous and so it may well be that a targeted focus on specific nuclei or subfields would help (the authors may well do this next) - but as the variables would only increase further along with the number of potential comparisons, alongside small group numbers, it seems only prudent to treat these findings are preliminary. That said, it is highly unlikely that large numbers of macaque brains will become available in the near future.

This introduction is by way of saying that the study achieves what it sets out to do, but there are many reasons to see this study as preliminary. The main message seems to be twofold: (1) that more intolerant species have relatively larger amygdalae, and (2) that with development, there is an opposite pattern of volume change (increasing with age in intolerant species and decreasing with age in tolerant species). Finding 1 is the opposite of that predicted in Table 1 - this is fine, but it should be made clearer in the Discussion that this is the case, otherwise the reader may feel confused. As I read it, the authors have switched their prediction in the Discussion, which feels uncomfortable. 

We thank the reviewer for this important observation. In the original version, Table 1 presented simplified direct predictions linking social tolerance grades to amygdala and hippocampus volumes. We recognize that this formulation may have created confusion In the revised manuscript, we have thoroughly restructured the table and its accompanying rationale. Table 1 now better reflects our conceptual framework grounded in three cognitive dimensions—sociocognitive demands, behavioral inhibition, and social predictability—each linked to behavioral traits and associated neural hypotheses based on published literature. This updated framework, detailed in lines 144-169 of the introduction, provides a more nuanced basis for interpreting our results and avoids the inconsistencies previously noted. The Discussion was also revised accordingly (lines 329-255) to clarify where our findings diverge from the original predictions and to explore alternative explanations based on social complexity. Rather than directly predicting amygdala size from social tolerance grades, we propose that variation in volume emerges from differing combinations of cognitive pressures across species.

It is inevitable that the data in a study of this complexity are all too prone to post hoc considerations, to which the authors indulge. In the case of Grade 1 species, the individuals have a lot to learn, especially if they are not top of the hierarchy, but at the same time, there are fewer individuals in the troop, making predictions very tricky. As noted above, I am concerned by the seemingly opposite predictions in Table 1 and those in the Discussion regarding tolerance and amygdala volume. (It may be that the predictions in Table 1 are the opposite of how I read them, in which case the Table and preceding text need to align.)

In order to facilitate the interpretation of our Bayesian modelling, we have selected a more focused ROI in our automatic segmentation procedure of the Hippocampus (from Hippocampal Formation to Hippocampus) and have added to the new analysis (Figure 4) that helps to properly test whether the hippocampus significantly differs between species from social grade 1 vs 4. The present analysis found that this is the case in adult monkeys. This is therefore consistent with our hypothesis that amygdala volumes are principally explained by heightened sociocognitive demands in more tolerant species.

We also acknowledge the reviewer’s concerns about the limited generalizability due to our sample. The challenges of comparative neuroimaging in non-human primates—especially when using post-mortem datasets—are substantial. Given the ethical constraints and the rarity of available specimens, increasing the number of individuals or species is not feasible in the short term. However, we have made all data and code publicly available and clearly stated the limitations of our sample in the manuscript. Despite these constraints, we believe our dataset offers an unprecedented comparative perspective, particularly due to the inclusion of rare and tolerant species such as M. tonkeana, M. nigra, and M. thibetana, which have never been included in structural MRI studies before. We hope this effort will serve as a foundation for future collaborative initiatives in primate comparative neuroscience.

Reviewer #3 (Public review):

We thank Reviewer #3 for their thoughtful and detailed review. Their comments helped us refine both the conceptual and interpretative aspects of the manuscript. We respond point by point below.

Summary:

In this study, the authors were looking at neurocorrelates of behavioural differences within the genus Macaca. To do so, they engaged in real-world dissection of dead animals (unconnected to the present study) coming from a range of different institutions. They subsequently compare different brain areas, here the amygdala and the hippocampus, across species. Crucially, these species have been sorted according to different levels of social tolerance grades (from 1 to 4). 12 species are represented across 42 individuals. The sampling process has weaknesses ("only half" of the species contained by the genus, and Macaca mulatta, the rhesus macaque, representing 13 of the total number of individuals), but also strengths (the species are decently well represented across the 4 grades) for the given purpose and for the amount of work required here. I will not judge the dissection process as I am not a neuroanatomist, and I will assume that the different interventions do not alter volume in any significant ways / or that the different conditions in which the bodies were kept led to the documented differences across species. 

25 brains were extracted by the authors themselves who are highly with this procedure. Overall, we believe that dissection protocols did not alter the total brain volume. Despite our expertise, we experienced some difficulties to not damage the cerebellum. Therefore, this region was not included in our analysis. We also noted that this brain region was also damaged or absent from the Prime-DE dataset.

Several protocols were used to prepare and store tissue. It could have impacted the total brain volume.

We agree that differences in tissue preparation and storage could potentially affect total brain volume. Therefore, we explicitly included the main sample preparation variable — whether brains had been previously frozen — as a covariate in our model. This factor did not explain our results. Moreover, Figures 1D and 1I display the frozen status and its correlation with the amygdala and hippocampus ratios, respectively. Figure 2 shows the parameters of the model and the posterior distributions for the frozen status and total brain volume effects.

There are two main results of the study. First, in line with their predictions, the authors find that more tolerant macaque species have larger amygdala, compared to the hippocampus, which remains undifferentiated across species. Second, they also identify developmental effects, although with different trends: in tolerant species, the amygdala relative volume decreases across the lifespan, while in intolerant species, the contrary occurs. The results look quite strong, although the authors could bring up some more clarity in their replies regarding the data they are working with. From one figure to the other, we switch from model-calculated ratio to modelpredicted volume. Note that if one was to sample a brain at age 20 in all the grades according to the model-predicted volumes, it would not seem that the difference for amygdala would differ much across grades, mostly driven with Grade 1 being smaller (in line with the main result), but then with Grade 2 bigger than Grade 3, and then Grade 4 bigger once again, but not that different from Grade 2.

Overall, despite this, I think the results are pretty strong, the correlations are not to be contested, but I also wonder about their real meaning and implications. This can be seen under 3 possible aspects:

(1)  Classification of the social grade

While it may be familiar to readers of Thierry and collaborators, or to researchers of the macaque world, there is no list included of the 18 behavioral traits used to define the three main cognitive requirements (socio-cognitive demands, predictability of the environment, inhibitory control). It would be important to know which of the different traits correspond to what, whether they overlap, and crucially, how they are realized in the 12 study species, as there could be drastic differences from one species to the next. For now, we can only see from Table S1 where the species align to, but it would be a good addition to have them individually matched to, if not the 18 behavioral traits, at least the 3 different broad categories of cognitive requirements.

We fully agree with this observation. In the revised version of the manuscript, we now include a detailed conceptual table listing all 18 behavioral traits from Thierry’s framework. For each trait, we provide its underlying social implications, its associated cognitive dimension (when applicable), and the hypothesized neural correlate. 

While some traits may could have been arguably classified in several cognitive dimensions (e.g. reconciliation rate), we preferred to assign each to a unique dimension for clarity. Additionally, the introduction (lines 95-169 + Table1) now explains how each trait was evaluated based on existing literature and assigned to one of the three proposed cognitive categories: socio-cognitive demands, behavioral inhibition, or social unpredictability. This structure offers a clearer and more transparent basis for the neuroanatomical hypotheses tested in the study.

“Navigating social life in primate societies requires substantial cognitive resources: individuals must not only track multiple relationships, but also regulate their own behavior, anticipate others’ reactions, and adapt flexibly to changing social contexts. Taken advantage of databases of magnetic resonance imaging (MRI) structural scans, we conducted the first comparative study integrating neuroanatomical data and social behavioral data from closely related primate species of the same genus to address the following questions: To what extent can differences in volumes of subcortical brain structures be correlated with varying degrees of social tolerance? Additionally, we explored whether these dispositions reflect primarily innate features, shaped by evolutionary processes, or acquired through socialization within more or less tolerant social environments”.

“The first category, socio-cognitive demands, refers to the cognitive resources needed to process, monitor, and flexibly adapt to complex social environments. Linking those parameters to neurological data is at the core of the social brain theory to explain the expansion of the neocortex in primates (Dunbar). Macaques social systems require advanced abilities in social memory, perspective-taking, and partner evaluation (Freeberg et al., 2012). This is particularly true in tolerant species, where the increased frequency and diversity of interactions may amplify the demands on cognitive tracking and flexibility. Tolerant macaque species typically live in larger groups with high interaction frequencies, low nepotism, and a wider range of affiliative and cooperative behaviors, including reconciliation, coalition-building, and signal flexibility (REF). Tolerant macaque species also exhibit a more diverse and flexible vocal and facial repertoire than intolerants ones which may help reduce ambiguity and facilitate coordination in dense social networks (Rincon et al., 2023; Scopa and Palagi, 2016; Rebout 2020). Experimental studies further show that macaques can use facial expressions to anticipate the likely outcomes of social interactions, suggesting a predictive function of facial signals in managing uncertainty (Micheletta et al., 2012; Waller et al., 2016). Even within less tolerant species, like M. mulatta, individual variation in facial expressivity has been linked to increased centrality in social networks and greater group cohesion, pointing to the adaptive value of expressive signaling across social styles (Whitehouse et al., 2024)”.

“The second category, inhibitory control, includes traits that involve regulating impulsivity, aggression, or inappropriate responses during social interactions. Tolerant macaques have been shown to perform better in tasks requiring behavioral inhibition and also express lower aggression and emotional reactivity in both experimental and natural contexts (Joly et al., 2017; Loyant et al., 2023). These features point to stronger self-regulation capacities in species with egalitarian or less rigid hierarchies. More broadly, inhibition – especially in its strategic form (self-control) – has been proposed to play a key role in the cohesion of stable social groups. Comparative analyses across mammals suggest that this capacity has evolved primarily in anthropoid primates, where social bonds require individuals to suppress immediate impulses in favour of longer-term group stability (Dunbar and Shultz, 2025). This view echoes the conjecture of Passingham and Wise (2012), who proposed that the emergence of prefrontal area BA10 in anthropoids enabled the kind of behavioural flexibility needed to navigate complex social environments (Passingham et al., 2012)”.

“The third category, social environment predictability, reflects how structured and foreseeable social interactions are within a given society. In tolerant species, social interactions are more fluid and less kin-biased, leading to greater contextual variation and role flexibility, which likely imply a sustained level of social awareness. In fact, as suggested by recent research, such social uncertainty and prolonged incentives are reflected by stress-related physiology : tolerant macaques such as M. tonkeana display higher basal cortisol levels, which may be indicative of a chronic mobilization of attentional and regulatory resources to navigate less predictable social environments (Sadoughi et al., 2021)”.

“Each behavioral trait was individually evaluated based on existing empirical literature regarding the types of cognitive operations it likely involves. When a primary cognitive dimension could be identified, the trait was assigned accordingly. However, some behaviors – such as maternal protection, allomaternal care, or delayed male dispersal – do not map neatly onto a single cognitive process. These traits likely emerge from complex configurations of affective and socialmotivational systems, and may be better understood through frameworks such as attachment theory (Suomi, 2008), which emphasizes the integration of social bonding, emotional regulation, and contextual plasticity. While these dimensions fall beyond the scope of the present framework, they offer promising directions for future research, particularly in relation to the hypothalamic and limbic substrates of social and reproductive behavior”.

“Rather than forcing these traits into potentially misleading categories, we chose to leave them unclassified within our current cognitive framework. This decision reflects both a commitment to conceptual clarity and the recognition that some behaviors emerge from a convergence of cognitive demands that cannot be neatly isolated. This tripartite framework, leaving aside reproductive-related traits, provides a structured lens through which to link behavioral diversity to specific cognitive processes and generate neuroanatomical predictions”.

(2) Issue of nature vs nurture

Another way to look at the debate between nature vs nurture is to look at phylogeny. For now, there is no phylogenetic tree that shows where the different grades are realized. For example, it would be illuminating to know whether more related species, independently of grades, have similar amygdala or hippocampus sizes. Then the question will go to the details, and whether the grades are realized in particular phylogenetic subdivisions. This would go in line with the general point of the authors that there could be general species differences.

As pointed out by Thierry and collaborators, the social tolerance concept is already grounded in a phylogenetic framework as social tolerance matches the phylogenetical tree of these macaque species, suggesting a biological ground of these behavioral observations. Given the modest sample size and uneven species representation, we opted not to adopt tools such as Phylogenetic Generalized Least Squares (PGLS) in our analysis. Our primary aim in this study was to explore neuroanatomical variation as a function of social traits, not to perform a phylogenetic comparative analysis per see. That said, we now explicitly acknowledge this limitation in the Discussion and indicate that future work using larger datasets and phylogenetic methods will be essential to disentangle social effects from evolutionary relatedness. We hope that making our dataset openly available will facilitate such futures analyses.

With respect to nurture, it is likely more complicated: one needs to take into account the idiosyncrasies of the life of the individual. For example, some of the cited literature in humans or macaques suggests that the bigger the social network, the bigger the brain structure considered. Right, but this finding is at the individual level with a documented life history. Do we have any of this information for any of the individuals considered (this is likely out of the scope of this paper to look at this, especially for individuals that did not originate from CdP)?

We appreciate this insightful observation. Indeed, findings from studies in humans and nonhuman primates showing associations between brain structure and social network size typically rely on detailed life history and behavioral data at the individual level. Unfortunately, such finegrained information was not consistently available across our entire sample. While some individuals from the Centre de Primatologie (CdP) were housed in known group compositions and social settings, we did not have access to longitudinal social data—such as rank, grooming rates, or network centrality—that would allow for robust individual-level analyses. We now acknowledge this limitation more clearly in the Discussion (lines 436-443), and we fully agree that future work combining neuroimaging with systematic behavioral monitoring will be necessary to explore how species-level effects interact with individual social experience.

(3) Issue of the discussion of the amygdala's function

The entire discussion/goal of the paper, states that the amygdala is connected to social life. Yet, before being a "social center", the amygdala has been connected to the emotional life of humans and non-humans alike. The authors state L333/34 that "These findings challenge conventional expectations of the amygdala's primary involvement in emotional processes and highlight the complexity of the amygdala's role in social cognition". First, there is no dichotomy between social cognition and emotion. Emotion is part of social cognition (unless we and macaques are robots). Second, there is nowhere in the paper a demonstration that the differences highlighted here are connected to social cognition differences per se. For example, the authors have not tested, say, if grade 4 species are more afraid of snakes than grade 1 species. If so, one could predict they would also have a bigger amygdala, and they would probably also find it in the model. My point is not that the authors should try to correlate any kind of potential aspect that has been connected to the amygdala in the literature with their data (see for example the nice review by DomínguezBorràs and Vuilleumier, https://doi.org/10.1016/B978-0-12-823493-8.00015-8), but they should refrain from saying they have challenged a particular aspect if they have not even tested it. I would rather engage the authors to try and discuss the amygdala as a multipurpose center, that includes social cognition and emotion.

We thank the reviewer for this important and nuanced point. We have revised the manuscript to adopt a more cautious and integrative tone regarding the function of the amygdala. In the revised Discussion (lines 341-355), we now explicitly state that the amygdala is involved in a broad range of processes—emotional, social, and affective—and that these domains are deeply intertwined. Rather than proposing a strict dissociation, we now suggest that the amygdala supports integrated socio-emotional functions that are mobilized differently across social tolerance styles. We also cite recent relevant literature (e.g., Domínguez-Borràs & Vuilleumier, 2021) to support this view and have removed any claim suggesting we challenge the emotional function of the amygdala per se. Our aim is to contribute to a richer understanding of how affective and social processes co-construct structural variation in this region.

Strengths:

Methods & breadth of species tested.

Weaknesses:

Interpretation, which can be described as 'oriented' and should rather offer additional views.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

Private Comments:

(1) Table 1 should be formatted for clarity i.e., bolded table headers, text realignment, and spacing. It was not clear at first glance how information was organized. It may also be helpful to place behavioral traits as the first column, seeing that these traits feed into the author's defined cognitive requirements.

We have reformatted Table 1 to improve clarity and readability. Behavioral traits now appear in the first column, followed by cognitive dimensions and hypothesized neural correlates. Column headers have been bolded and alignment has been standardized.

(2) Figures could include more detail to help with interpretations. For example, Figure 3 should define values included on the x-axis in the figure caption, and Figure 4 should explain the use of line, light color, and dark color. Figure 1 does not have a y-axis title.

The figures have been revised and legends completed to ensure more clarity.

(3) Please proofread for typos throughout.

The manuscript has been carefully proofread, and all typographical and grammatical errors have been corrected. These changes are visible in the tracked version.

Reviewer #2 (Recommendations for the authors):

Specific comments:

(1) Given all of the variability would it not be a good idea to just compare (eg in the supplemental) the macaque data from just the Strasbourg centre for m mulatta and m toneanna. I appreciate the ns will be lower, but other matters are more standardized.

We fully understand the reviewer’s suggestion to restrict the comparison to data collected at a single site in order to minimize inter-site variability. However, as noted, such an analysis would come at the cost of statistical power, as the number of individuals per species within a single center is small. For example, while M. tonkeana is well represented at the Strasbourg centre, only one individual of M. mulatta is available from the same site. Thus, a restricted comparison would severely limit the interpretability of results, particularly for age-related trajectories. To address variability, we included acquisition site and brain preservation method as covariates or predictors where appropriate, and we have been cautious in our interpretations. We also now emphasize in the Methods and Discussion the value of future datasets with more standardized acquisition protocols across species and centers. We hope that by openly sharing our data and workflow, we can contribute to this broader goal.

(2) I have various minor edits:

(a) L 25 abstract - Specify what is meant by 'opposite trend'; the reader cannot infer what this is.

Modified in line 25-28: “Unexpectedly, tolerant species exhibited a decrease in relative amygdala volume across the lifespan, contrasting with the age-related increase observed in intolerant species—a developmental pattern previously undescribed in primates.”

(b) L67 - The reference 'Manyprimates' needs fixing as it does in the references section.

After double checking, Manyprimates studies are international collaborative efforts that are supposed to be cite this way (https://manyprimates.github.io/#pubs).

(c) L74 - Taking not Taken.

This typo has been corrected.

(d) L129 - It says 'total volume', but this is corrected total volume?

We have clarified in the figures legends that the “total brain volume” used in our analyses excludes the cerebellum and the myelencephalon, as specified in our image preprocessing protocol. This ensures consistency across individuals and institutions.

(e) L138 - Suddenly mentions 'frozen condition' without any prior explanation - this needs explaining in the legend - also L144.

We have added an explanation of the ‘frozen condition’ variable in in the relevant figure legend.

(f) L166 - Results - it would be helpful to remind readers what Grade 1 signifies, ie intolerant species.

We now include a brief reminder in the Results section that Grade 1 corresponds to socially intolerant species, to help readers unfamiliar with the classification (Lines 240-251).

(g)Figure 4 - Provide the ns for each of the 4 grades to help appreciate the meaningfulness of the curves, etc.

The number of subjects has been added to the Figure and a novel analysis helps in the revised ms help to appreciate the meaningfulness of some of these curves.

(h) L235 - 'we had assumed that species of high social tolerance grade would have presented a smaller amygdala in size compared to grade 1'. But surely this is the exact opposite of what is predicted in Table 1 - ie, the authors did not predict this as I read the paper (Unless Table l is misleading/ambiguous and needs clarification).

As discussed in our response to Reviewer #2 and #3, we have restructured both Table 1 and the Discussion to ensure consistency. We now explicitly state that the findings diverge from our initial inhibitory-control-based prediction and propose alternative interpretations based on sociocognitive demands.

(i) L270 - 'This observation' which?? Specify.

We have replaced ‘this observation’ with a precise reference to the observed developmental decrease in amygdala volume in tolerant species.

(j) L327 - 'groundbreaking' is just hype given that there are so many caveats - I personally do not like the word - novel is good enough.

We have replaced the word ‘groundbreaking’ with ‘novel’ to adopt a more measured and appropriate tone in the discussion.

(3) I might add that I am happy with the ethics regarding this study. 

Thanks, we are also happy that we were able to study macaque brains from different species using opportunistic samplings along with already available data. We are collectively making progress on this!

(4) Finally, I should commend the authors on all the additional information that they provide re gender/age/species. Given that there are 2xs are many females as males, it would be good to know if this affects the findings. I am not a primatologist, so I don't know, for example, if the females in Grade 1 monkeys are just as intolerant as the males?

We thank the reviewer for this thoughtful comment. We now explicitly mention the female-biased sex ratio in the Methods section and report in the Results (Figure 2, Figure 3) that sex was included as a covariate in our Bayesian models. While a small effect of sex was found for hippocampal volume, no effect was observed for the amygdala. Given the strong imbalance in our dataset (2:1 female-to-male ratio), we refrained from drawing any conclusion about sex-specific patterns, as these would require larger and more balanced samples. Although we did not test for sex-by-grade interactions, we agree that this question—especially regarding whether females and males express social style differences similarly across grades—represents an important direction for future comparative work.

Reviewer #3 (Recommendations for the authors):

I found the article well-written, and very easy to follow, so I have little ways to propose improvements to the article to the authors, besides addressing the various major points when it comes to interpretation of the data.

One list I found myself wanting was in fact the list of the social tolerance grades, and the process by which they got selected into 3 main bags of socio-cognitive skills. Then it would become interesting to see how each of the 12 species compares within both the 18 grades (maybe once again out of the scope of this paper, there are likely reviews out there that already do that, but then the authors should explicitly mention so in the paper: X, 19XX have compared 15 out of 18 traits in YY number of macaque species); and within the 3 major subcognitive requirements delineated by the authors, maybe as an annex?

We thank the reviewer for this thoughtful suggestion. In the revised manuscript, we now include a detailed table (Table 1) that lists the 18 behavioral traits derived from Thierry’s framework, along with their associated cognitive dimension and hypothesized neuroanatomical correlate. While we did not create a matrix mapping each of the 12 species across all 18 traits due to space and data availability constraints, we agree this is an important direction that should be tackled by primatologist. We now include a sentence (line 87-90) in the manuscript to guide readers to previous comparative reviews (e.g., Thierry, 2000; Thierry et al., 2004, 2021) that document the expression of these traits across macaque species. We also clarify that our three cognitive categories are conceptual tools intended to structure neuroanatomical predictions, and not formal clusters derived from quantitative analyses.

In the annex, it would also be good to have a general summarizing excel/R file for the raw data, with important information like age, sex, and the relevant calculated volumes for each individual. The folders available following the links do not make it an easy task for a reader to find the raw data in one place.

We fully agree with the reviewer on the importance of data accessibility. We have now uploaded an additional supplementary file in .csv format on our OSF repository, which includes individuallevel metadata for all 42 macaques: species, sex, age, social grade, total brain volume, amygdala volume, and hippocampus volume. The link to this file is now explicitly mentioned in the Data Availability section. We hope this will facilitate comparisons with other datasets and improve usability for the community. In addition, we provide in a supplementary table the raw data that were used for our Bayesian modelling (see below).

The availability of the raw data would also clear up one issue, which I believe results from the modelling process: it looks odd on Figure 2, that volume ratios, defined as the given brain area volume divided by the total brain volume, give values above 1 (especially for the hippocampus). As such, the authors should either modify the legend or the figure. In general, it would be nicer to have the "real values" somewhere easily accessible, so that they can be compared more broadly with: 1) other macaques species to address questions relevant to the species; 2) other primates to address other questions that are surely going to arise from this very interesting work!

We thank the reviewer for pointing this out. The ratio values in Figure 1 correspond to the proportion of the regional volume (amygdala or hippocampus) relative to the total brain volume, excluding the cerebellum and myelencephalon. As such, values above 0.01 (i.e., above 1% of the brain volume) are expected for these structures and do not indicate an error. We have updated the figure legend to clarify this point explicitly. In addition, we have now made a cleaned .csv file available via OSF, containing all raw volumetric data and metadata in a format that facilitates cross-species or cross-study comparisons. This replaces the previous folder-based structure, which may have been less accessible.

Typos:

L233: delete 'in'

L430: insert space in 'NMT template(Jung et al., 2021).'

When looking at bots we can't assume it reflects the intention of any one single person. The people who wrote the code, the people who use it, and the computer that runs it are all separate different entities and because of this responsibility for a bot's action is spread out.

Author response:

The following is the authors’ response to the original reviews

eLife Assessment

This paper undertakes an important investigation to determine whether movement slowing in microgravity is due to a strategic conservative approach or rather due to an underestimation of the mass of the arm. While the experimental dataset is unique and the coupled experimental and computational analyses comprehensive, the authors present incomplete results to support the claim that movement slowing is due to mass underestimation. Further analysis is needed to rule out alternative explanations.

We thank the editor and reviewers for the thoughtful and constructive comments, which helped us substantially improve the manuscript. In this revised version, we have made the following key changes:

- Directly presented the differential effect of microgravity in different movement directions, showing its quantitative match with model predictions.

- Showed that changing cost function with the idea of conservative strategy is not a viable alternative.

- Showed our model predictions remain largely the same after adding Coriolis and centripetal torques.

- Discussed alternative explanations including neuromuscular deconditioning, friction, body stability, etc.

- Detailed the model description and moved it to the main text, as suggested.

Our point-to-point response is numbered to facilitate cross-referencing.

We believe the revisions and the responses adequately addresses the reviewers’ concerns, and new analysis results strengthened our conclusion that mass underestimation is the major contributor to movement slowing in microgravity.

Reviewer #1 (Public review):

Summary:

This article investigates the origin of movement slowdown in weightlessness by testing two possible hypotheses: the first is based on a strategic and conservative slowdown, presented as a scaling of the motion kinematics without altering its profile, while the second is based on the hypothesis of a misestimation of effective mass by the brain due to an alteration of gravity-dependent sensory inputs, which alters the kinematics following a controller parameterization error.

Strengths:

The article convincingly demonstrates that trajectories are affected in 0g conditions, as in previous work. It is interesting, and the results appear robust. However, I have two major reservations about the current version of the manuscript that prevent me from endorsing the conclusion in its current form.

Weaknesses:

(1) First, the hypothesis of a strategic and conservative slow down implicitly assumes a similar cost function, which cannot be guaranteed, tested, or verified. For example, previous work has suggested that changing the ratio between the state and control weight matrices produced an alteration in movement kinematics similar to that presented here, without changing the estimated mass parameter (Crevecoeur et al., 2010, J Neurophysiol, 104 (3), 1301-1313). Thus, the hypothesis of conservative slowing cannot be rejected. Such a strategy could vary with effective mass (thus showing a statistical effect), but the possibility that the data reflect a combination of both mechanisms (strategic slowing and mass misestimation) remains open.

Response (1): Thank you for raising this point. The basic premise of this concern is that changing the cost function for implementing strategic slowing can reproduce our empirical findings, thus the alternative hypothesis that we aimed to refute in the paper remain possible. At least, it could co-exist with our hypothesis of mass underestimation. In the revision, we show that changing the cost function only, as suggested here, cannot produce the behavioral patterns observed in microgravity.

As suggested, we modified the relative weighting of the state and control cost matrices (i.e., Q and R in the cost function Eq 15) without considering mass underestimation. While this cost function scaling can decrease peak velocity – a hallmark of strategic slowing – it also inevitably leads to later peak timings. This is opposite to our robust findings: the taikonauts consistently “advanced” their peak velocity and peak acceleration in time. Note, these model simulation patterns have also been shown in Crevecoeur et al. (2010), the paper mentioned by the reviewer (see their Figure 7B).

We systematically changed the ratio between the state and control weight matrices in the simulation, as suggested. We divided Q and multiplied R by the same factor α, the cost function scaling parameter α as defined in Crevecoeur et al. (2010). This adjustment models a shift in movement strategy in microgravity, and we tested a wide range of α to examine reasonable parameter space. Simulation results for α = 3 and α = 0.3 are shown in Figure 1—figure supplement 2 and Figure 1—figure supplement 3 respectively. As expected, with α = 3 (higher control effort penalty), peak velocities and accelerations are reduced, but their timing is delayed. Conversely, with α = 0.3, both peak amplitude and timing increase. Hence, changing the cost function to implement a conservative strategy cannot produce the kinematic pattern observed in microgravity, which is a combination of movement slowing and peak timing advance.

Therefore, we conclude that a change in optimal control strategy alone is insufficient to explain our empirical findings. Logically speaking, we cannot refute the possibility of strategic slowing, which can still exist on top of the mass underestimation we proposed here. However, our data does not support its role in explaining the slowing of goal-directed hand reaching in microgravity. We have added these analyses to the Supplementary Materials and expanded the Discussion to address this point.

(2) The main strength of the article is the presence of directional effects expected under the hypothesis of mass estimation error. However, the article lacks a clear demonstration of such an effect: indeed, although there appears to be a significant effect of direction, I was not sure that this effect matched the model's predictions. A directional effect is not sufficient because the model makes clear quantitative predictions about how this effect should vary across directions. In the absence of a quantitative match between the model and the data, the authors' claims regarding the role of misestimating the effective mass remain unsupported.

Response (2): First, we have to clarify that our study does not aim to quantitatively fit observed hand trajectory. The two-link arm model simulates an ideal case of moving a point mass (effective mass) on a horizontal plane without friction (Todorov, 2004; 2005). In contrast, in the experiment, participants moved their hand on a tabletop without vertical arm support, so the movement was not strictly planar and was affected by friction. Thus, this kind of model can only illustrate qualitative differences between conditions, as in the majorities of similar modeling studies (e.g., Shadmehr et al., 2016). In our study, qualitative simulation means the model is intended to reproduce the directional differences between conditions—not exact numeric values—in key kinematic measures. Specifically, it should capture how the peak velocity and acceleration amplitudes and their timings differ between normal gravity and microgravity (particularly under the mass-underestimation assumption).

Second, the reviewer rightfully pointed out that the directional effect is essential for our theorization of the importance of mass underestimation. However, the directional effect has two aspects, which were not clearly presented in our original manuscript. We now clarify both here and in the revision. The first aspect is that key kinematic variables (peak velocity/acceleration and their timing) are affected by movement direction, even before any potential microgravity effect. This is shown by the ranking order of directions for these variables (Figure 1C-H). The direction-dependent ranking, confirmed by pre-flight data, indicates that effective mass is a determining factor for reaching kinematics, which motivated us to study its role in eliciting movement slowing in space. This was what our original manuscript emphasized and clearly presented.

The second aspect is that the hypothetical mass underestimation might also differentially affect movements in different directions. This was not clearly presented in the original manuscript. However, we would not expect a quantitative match between model predictions and empirical data, for the reasons mentioned above. We now show this directional ranking in microgravity-elicited kinematic changes in both model simulations and empirical data. The overall trend is that the microgravity effect indeed differs between directions, and the model predictions and the data showed a reasonable qualitative match (Author response image 1 below).

Shown in Author response image 1, we found that for amplitude changes (Δ peak speed, Δ peak acceleration) both the model and the mean of empirical data show the same directional ordering (45° > 90° > 135°) in pre-in and post-in comparisons. For timing (Δ peak-speed time, Δ peak-acceleration time), which we consider the most diagnostic, the same directional ranking was observed. We only found one deviation, i.e., the predicted sign (earlier peaks) was confirmed at 90° and 135°, but not at 45°. As discussed in Response (6), the absence of timing advance at 45° may reflect limitations of our simplified model, which did not consider that the 45° direction is essentially a single-joint reach. Taken together, the directional pattern is largely consistent with the model predictions based on mass underestimation. The model successfully reproduces the directional ordering of amplitude measures -- peak velocity and peak acceleration. It also captures the sign of the timing changes in two out of the three directions. We added these new analysis results in the revision and expanded Discussion accordingly.

The details of our analysis on directional effects: We compared the model predictions (Author response image 1, left) with the experimental data (Author response image 1, right) across the three tested directions (45°, 90°, 135°). In the experimental data panels, both Δ(pre-in) (solid bars) and Δ(post-in) (semi-transparent bars) with standard error are shown. The directional trends are remarkably similar between model prediction and actual data. The post-in comparison is less aligned with model prediction; we postulate that the incomplete after-flight recovery (i.e., post data had not returned to pre-flight baselines) might obscure the microgravity effect. Incomplete recovery has also been shown in our original manuscript: peak speed and peak acceleration did not fully recover in post-flight sessions when compared to pre-flight sessions. To further quantify the correspondence between model and data, we performed repeated-measures correlation (rm-corr) analyses. We found significant within-subject correlations for three of the four metrics. For pre–in, Δ peak speed time (r_rm = 0.627, t(23) = 3.858, p < 0.001), Δ peak acceleration time (r_rm = 0.591, t(23) = 3.513, p = 0.002), and Δ peak acceleration (r_rm = 0.573, t(23) = 3.351, p = 0.003) were significant, whereas Δ peak speed was not (r_rm = 0.334, t(23) = 1.696, p = 0.103). These results thus show that the directional effect, as predicted our model, is observed both before spaceflight and in spaceflight (the pre-in comparison).

Author response image 1.

Directional comparison between model predictions and experimental data across the three reach directions (45°, 90°, 135°). Left: model outputs. Right: experimental data shown as Δ relative to the in-flight session; solid bars = Δ(in − pre) and semi-transparent bars = Δ(in − post). Colors encode direction consistently across panels (e.g., 45° = darker hue, 90° = medium, 135° = lighter/orange). Panels (clockwise from top-left): Δ peak speed (cm/s), Δ peak speed time (ms), Δ peak acceleration time (ms), and Δ peak acceleration (cm/s²). Bars are group means; error bars denote standard error across participants.

Citations:

Todorov, E. (2004). Optimality principles in sensorimotor control. Nature Neuroscience, 7(9), 907.

Todorov, E. (2005). Stochastic optimal control and estimation methods adapted to the noise characteristics of the sensorimotor system. Neural Computation, 17(5), 1084–1108.

Shadmehr, R., Huang, H. J., & Ahmed, A. A. (2016). A Representation of Effort in Decision-Making and Motor Control. Current Biology: CB, 26(14), 1929–1934.

In general, both the hypotheses of slowing motion (out of caution) and misestimating mass have been put forward in the past, and the added value of this article lies in demonstrating that the effect depended on direction. However, (1) a conservative strategy with a different cost function can also explain the data, and (2) the quantitative match between the directional effect and the model's predictions has not been established.

We agree that both hypotheses have been put forward before, however they are competing hypotheses that have not been resolved. Furthermore, the mass underestimation hypothesis is a conjecture without any solid evidence; previous reports on mass underestimation of object cannot directly translate to underestimation of body. As detailed in our responses above, we have shown that a conservative strategy implemented via a different cost function cannot reproduce the key findings in our dataset, thereby supporting the alternative hypothesis of mass underestimation. Moreover, we found qualitative agreement between the model predictions and the experimental data in terms of directional effects, which further strengthens our interpretation.

Specific points:

(1) I noted a lack of presentation of raw kinematic traces, which would be necessary to convince me that the directional effect was related to effective mass as stated.

Response (3): We are happy to include exemplary speed and acceleration trajectories. Kinematic profiles from one example participant are shown in Figure 2—figure supplement 6.

(2) The presentation and justification of the model require substantial improvement; the reason for their presence in the supplementary material is unclear, as there is space to present the modelling work in detail in the main text. Regarding the model, some choices require justification: for example, why did the authors ignore the nonlinear Coriolis and centripetal terms?

Response (4): Great suggestion. In the revision, we have moved the model into the main text and added further justification for using this simple model.

We initially omitted the nonlinear Coriolis and centripetal terms in order to start with a minimal model. Importantly, excluding these terms does not affect the model’s main conclusions. In the revision we added simulations that explicitly include these terms. The full explanation and simulations are provided in the Supplementary Notes 2 (this time we have to put it into the Supplementary to reduce the texts devoted to the model). More explanations can also be found in our response to Reviewer 2 (response (6)). The results indicate that, although these velocity-dependent forces show some directional anisotropy, their contribution is substantially smaller relative to that of the included inertial component; specifically, they have only a negligible impact on the predicted peak amplitudes and peak times.

(3) The increase in the proportion of trials with subcomponents is interesting, but the explanatory power of this observation is limited, as the initial percentage was already quite high (from 60-70% during the initial study to 70-85% in flight). This suggests that the potential effect of effective mass only explains a small increase in a trend already present in the initial study. A more critical assessment of this result is warranted.

Response (5): Thank you for your thoughtful comment. You are correct that the increase in the percentage of trials with submovements is modest, but a more critical change was observed in the timing between submovement peaks—specifically, the inter-peak interval (IPI). These intervals became longer during flight. Taken together with the percentage increase, the submovement changes significantly predicted the increase in movement duration, as shown by our linear mixed-effects model, which indicated that IPI increased.

Reviewer #2 (Public review):

This study explores the underlying causes of the generalized movement slowness observed in astronauts in weightlessness compared to their performance on Earth. The authors argue that this movement slowness stems from an underestimation of mass rather than a deliberate reduction in speed for enhanced stability and safety.

Overall, this is a fascinating and well-written work. The kinematic analysis is thorough and comprehensive. The design of the study is solid, the collected dataset is rare, and the model tends to add confidence to the proposed conclusions. That being said, I have several comments that could be addressed to consolidate interpretations and improve clarity.

Main comments:

(1) Mass underestimation

a) While this interpretation is supported by data and analyses, it is not clear whether this gives a complete picture of the underlying phenomena. The two hypotheses (i.e., mass underestimation vs deliberate speed reduction) can only be distinguished in terms of velocity/acceleration patterns, which should display specific changes during the flight with a mass underestimation. The experimental data generally shows the expected changes but for the 45° condition, no changes are observed during flight compared to the pre- and post-phases (Figure 4). In Figure 5E, only a change in the primary submovement peak velocity is observed for 45°, but this finding relies on a more involved decomposition procedure. It suggests that there is something specific about 45° (beyond its low effective mass). In such planar movements, 45° often corresponds to a movement which is close to single-joint, whereas 90° and 135° involve multi-joint movements. If so, the increased proportion of submovements in 90° and 135° could indicate that participants had more difficulties in coordinating multi-joint movements during flight. Besides inertia, Coriolis and centripetal effects may be non-negligible in such fast planar reaching (Hollerbach & Flash, Biol Cyber, 1982) and, interestingly, they would also be affected by a mass underestimation (thus, this is not necessarily incompatible with the author's view; yet predicting the effects of a mass underestimation on Coriolis/centripetal torques would require a two-link arm model). Overall, I found the discrepancy between the 45° direction and the other directions under-exploited in the current version of the article. In sum, could the corrective submovements be due to a misestimation of Coriolis/centripetal torques in the multi-joint dynamics (caused specifically -or not- by a mass underestimation)?

Response (6): Thank you for raising these important questions. We unpacked the whole paragraph into two concerns: 1) the possibility that misestimation of Coriolis and centripetal torques might lead to corrective submovements, and 2) the weak effect in the 45° direction unexploited. These two concerns are valid but addressable, and they did not change our general conclusions based on our empirical findings (see Supplementary note 2. Coriolis and centripetal torques have minimal impact).

Possible explanation for the 45° discrepancy

We agree with the reviewer that the 45° direction likely involves more single-joint (elbow-dominant) movement, whereas the 90° and 135° directions require greater multi-joint (elbow + shoulder) coordination. This is particularly relevant when the workspace is near body midline (e.g., Haggard & Richardson, 1995), as the case in our experimental setup. To demonstrate this, we examined the curvature of the hand trajectories across directions. Using cumulative curvature (positive = counterclockwise), we obtained average values of 6.484° ± 0.841°, 1.539° ± 0.462°, and 2.819° ± 0.538° for the 45°, 90°, and 135° directions, respectively. The significantly larger curvature in the 45° condition suggests that these movements deviate more from a straight-line path, a hallmark of more elbow-dominant movements.

Importantly, this curvature pattern was present in both the pre-flight and in-flight phases, indicating that it is a general movement characteristic rather than a microgravity-induced effect. Thus, the 45° reaches are less suitable for modeling with a simplified two-link arm model compared to the other two directions. We believe this is the main reason why the model predictions based on effective mass become less consistent with the empirical data for the 45° direction.

We have now incorporated this new analysis in the Results and discussed it in the revised Discussion.

Citation: Haggard, P., Hutchinson, K., & Stein, J. (1995). Patterns of coordinated multi-joint movement. Experimental Brain Research, 107(2), 254-266.

b) Additionally, since the taikonauts are tested after 2 or 3 weeks in flight, one could also assume that neuromuscular deconditioning explains (at least in part) the general decrease in movement speed. Can the authors explain how to rule out this alternative interpretation? For instance, weaker muscles could account for slower movements within a classical time-effort trade-off (as more neural effort would be needed to generate a similar amount of muscle force, thereby suggesting a purposive slowing down of movement). Therefore, could the observed results (slowing down + more submovements) be explained by some neuromuscular deconditioning combined with a difficulty in coordinating multi-joint movements in weightlessness (due to a misestimation or Coriolis/centripetal torques) provide an alternative explanation for the results?

Response (7): Neuromuscular deconditioning is indeed a space effect; thanks for bringing this up as we omitted the discussion of this confounds in our original manuscript. Prolonged stay in microgravity can lead to a reduction of muscle strength, but this is mostly limited to lower limb. For example, a recent well-designed large-sample study have shown that while lower leg muscle showed significant strength reductions, no changes in mean upper body strength was found (Scott et al., 2023), consistent with previous propositions that muscle weakness is less for upper-limb muscles than for postural and lower-limb muscles (Tesch et al., 2005). Furthermore, the muscle weakness is unlikely to play a major role here since our reaching task involves small movements (~12cm) with joint torques of a magnitude of ~2N·m. Of course, we cannot completely rule out the contribution of muscle weakness; we can only postulate, based on the task itself (12 cm reaching) and systematic microgravity effect (the increase in submovements, the increase in the inter-submovements intervals, and their significant prediction on movement slowing), that muscle weakness is an unlikely major contributor for the movement slowing.

The reviewer suggests that poor coordination in microgravity might contribute to slowing down + more submovements. This is also a possibility, but we did not find evidence to support it. First, there is no clear evidence or reports about poor coordination for simple upper-limb movements like reaching investigated here. Note that reaching or aiming movement is one of the most studied tasks among astronauts. Second, we further analyzed our reaching trajectories and found no sign of curvature increase, a hallmark of poor coordination of Coriolis/centripetal torques, in our large collection of reaching movements. We probably have the largest dataset of reaching movements collected in microgravity thus far, given that we had 12 taikonauts and each of them performed about 480 to 840 reaching trials during their spaceflight. We believe the probability of Type II error is quite low here.

Citation: Tesch, P. A., Berg, H. E., Bring, D., Evans, H. J., & LeBlanc, A. D. (2005). Effects of 17-day spaceflight on knee extensor muscle function and size. European journal of applied physiology, 93(4), 463-468.

Scott J, Feiveson A, English K, et al. Effects of exercise countermeasures on multisystem function in long duration spaceflight astronauts. npj Microgravity. 2023;9(11).

(2) Modelling

a) The model description should be improved as it is currently a mix of discrete time and continuous time formulations. Moreover, an infinite-horizon cost function is used, but I thought the authors used a finite-horizon formulation with the prefixed duration provided by the movement utility maximization framework of Shadmehr et al. (Curr Biol, 2016). Furthermore, was the mass underestimation reflected both in the utility model and the optimal control model? If so, did the authors really compute the feedback control gain with the underestimated mass but simulate the system with the real mass? This is important because the mass appears both in the utility framework and in the LQ framework. Given the current interpretations, the feedforward command is assumed to be erroneous, and the feedback command would allow for motor corrections. Therefore, it could be clarified whether the feedback command also misestimates the mass or not, which may affect its efficiency. For instance, if both feedforward and feedback motor commands are based on wrong internal models (e.g., due to the mass underestimation), one may wonder how the astronauts would execute accurate goal-directed movements.

b) The model seems to be deterministic in its current form (no motor and sensory noise). Since the framework developed by Todorov (2005) is used, sensorimotor noise could have been readily considered. One could also assume that motor and sensory noise increase in microgravity, and the model could inform on how microgravity affects the number of submovements or endpoint variance due to sensorimotor noise changes, for instance.

c) Finally, how does the model distinguish the feedforward and feedback components of the motor command that are discussed in the paper, given that the model only yields a feedback control law? Does 'feedforward' refer to the motor plan here (i.e., the prefixed duration and arguably the precomputed feedback gain)?

Response (8): We thank the reviewer for raising these important and technically insightful points regarding our modeling framework. We first clarify the structure of the model and key assumptions, and then address the specific questions in points (a)–(c) below.

We used Todorov’s (2005) stochastic optimal control method to compute a finite-horizon LQG policy under sensory noise and signal-dependent motor noise (state noise set to zero). The cost function is: (see details in updated Methods). The resulting time-varying gains {L_k, K_k} correspond to the feedforward mapping and the feedback correction gain, respectively. The control law can be expressed as:

where u_k is the control input, is the nominal planned state, is the estimated state, L_k is the feedforward (nominal) control associated with the planned trajectory, and K_k is the time-varying feedback gain that corrects deviations from the plan.

To define the motor plan for comparison with behavior, we simulate the deterministic open-loop

trajectory by turning off noise and disabling feedback corrections, i.e., . In this framework, “feedforward” refers to this nominal motor plan. Thus, sensory and signal-dependent noise influence the computed policy (via the gains), but are not injected when generating the nominal trajectory. This mirrors the minimum-jerk practice used to obtain nominal kinematics in prior utility-based work (Shadmehr, 2016), while optimal control provides a more physiologically grounded nominal plan. In the revision, we have updated the equations, provided more modeling details, and moved the model description to the main text to reduce possible confusions.

In the implementation of the “mass underestimation” condition, the mass used to compute the policy is the underestimated mass (), whereas the actual mass is used when simulating the feedforward trajectories. Corrective submovements are analyzed separately and are not required for the planning-deficit findings reported here.

Answers of the three specific questions:

a) We mistakenly wrote a continuous-time infinite-horizon cost function in our original manuscript, whereas our controller is actually implemented as a discrete-time finite-horizon LQG with a terminal cost, over a horizon set by the utility-based optimal movement duration T_opt. The underestimated mass is used in both the utility model (to determine T_opt) and in the control computation (i.e., internal model), while the true mass is used when simulating the movement. This mismatch captures the central idea of feedforward planning based on an incorrect internal model.

b) As described, our model includes signal-dependent motor noise and sensory noise, following Todorov (2005). We also evaluated whether increased noise levels in microgravity could account for the observed behavioral changes. Simulation results showed that increasing either source of noise did not alter the main conclusions or reverse the trends in our key metrics. Moreover, our experimental data showed no significant increase in endpoint variability in microgravity (see analyses and results in Figure 2—figure supplement 3 & 4), making it unlikely that increased sensorimotor noise alone accounts for the observed slowing and submovement changes.

c) In our framework, the time-varying gains {L_K,K_K}define the feedforward and feedback components of the control policy. While both gains are computed based on a stochastic optimal control formulation (including noise), for comparison with behavior we simulate only the nominal feedforward plan, by turning off both noise and feedback: . This defines a deterministic open-loop trajectory, which we use to capture planning-level effects such as peak timing shifts under mass underestimation. Feedback corrections via gains exist in the full model but are not involved in these specific analyses. We clarified this modeling choice and its behavioral relevance in the revised text.

We have updated the equations and moved the model description into the main text in the revised manuscript to avoid confusion.

(3) Brevity of movements and speed-accuracy trade-off

The tested movements are much faster (average duration approx. 350 ms) than similar self-paced movements that have been studied in other works (e.g., Wang et al., J Neurophysiology, 2016; Berret et al., PLOS Comp Biol, 2021, where movements can last about 900-1000 ms). This is consistent with the instructions to reach quickly and accurately, in line with a speed-accuracy trade-off. Was this instruction given to highlight the inertial effects related to the arm's anisotropy? One may however, wonder if the same results would hold for slower self-paced movements (are they also with reduced speed compared to Earth performance?). Moreover, a few other important questions might need to be addressed for completeness: how to ensure that astronauts did remember this instruction during the flight? (could the control group move faster because they better remembered the instruction?). Did the taikonauts perform the experiment on their own during the flight, or did one taikonaut assume the role of the experimenter?

Response (9): Thanks for highlighting the brevity of movements in our experiment. Our intention in emphasizing fast movements is to rigorously test whether movement is indeed slowed down in microgravity. The observed prolonged movement duration clearly shows that microgravity affects people’s movement duration, even when they are pushed to move fast. The second reason for using fast movement is to highlight that feedforward control is affected in microgravity. Mass underestimation specifically affects feedforward control in the first place, shown by the microgravity-related changes in peak velocity/acceleration. Slow movement would inevitably have online corrections that might obscure the effect of mass underestimation. Note that movement slowing is not only observed in our speed-emphasized reaching task, but also in whole-arm pointing in other astronauts’ studies (Berger, 1997; Sangals, 1999), which have been quoted in our paper. We thus believe these findings are generalizable.

Regarding the consistency of instructions: all our experiments conducted in the Tiangong space station were monitored in real time by experimenters in the control center located in Beijing. The task instructions were presented on the initial display of the data acquisition application and ample reading time was allowed. All the pre-, in-, and post-flight test sessions were administered by the same group of personnel with the same instruction. It is common that astronauts serve both as participants and experimenters at the same time. And, they were well trained for this type of role on the ground. Note that we had multiple pre-flight test sessions to familiarize them with the task. All these rigorous measures were in place to obtain high-quality data. In the revision, we included these experimental details for readers that are not familiar with space studies, and provided the rationales for emphasizing fast movements.

Citations:

Berger, M., Mescheriakov, S., Molokanova, E., Lechner-Steinleitner, S., Seguer, N., & Kozlovskaya, I. (1997). Pointing arm movements in short- and long-term spaceflights. Aviation, Space, and Environmental Medicine, 68(9), 781–787.

Sangals, J., Heuer, H., Manzey, D., & Lorenz, B. (1999). Changed visuomotor transformations during and after prolonged microgravity. Experimental Brain Research. Experimentelle Hirnforschung. Experimentation Cerebrale, 129(3), 378–390.

(4) No learning effect

This is a surprising effect, as mentioned by the authors. Other studies conducted in microgravity have indeed revealed an optimal adaptation of motor patterns in a few dozen trials (e.g., Gaveau et al., eLife, 2016). Perhaps the difference is again related to single-joint versus multi-joint movements. This should be better discussed given the impact of this claim. Typically, why would a "sensory bias of bodily property" persist in microgravity and be a "fundamental constraint of the sensorimotor system"?

Response (10): We believe that the presence or absence of adaptation between our study and Gaveau et al.’s study cannot be simply attributed to single-joint versus multi-joint movements. Their adaptation concerned incorporating microgravity into movement control to minimize effort, whereas ours concerned accurately perceiving body mass. Gaveau et al.’s task involved large-amplitude vertical reaching, a scenario in which gravity strongly affects joint torques and movement execution. Thus, adaptation to microgravity can lead to better execution, providing a strong incentive for learning. By contrast, our task consisted of small-amplitude horizontal movements, where the gravitational influence on biomechanics is minimal.

More importantly, we believe the lack of adaptation for mass underestimation is not totally surprising. When an inertial change is perceived (such as an extra weight attached to the forearm, as in previous motor adaptation studies), people can adapt their reaching within tens of trials. In that case, sensory cues are veridical, as they correctly signal the inertial perturbation. However, in microgravity, reduced gravitational pull and proprioceptive inputs constantly inform the controller that the body mass is less than its actual magnitude. In other words, sensory cues in space are misleading for estimating body mass. The resulting sensory bias prevents the sensorimotor system from adapting. Our initial explanation on this matter was too brief; we expanded it in the revised Discussion.

Reviewer #3 (Public review):

Summary:

The authors describe an interesting study of arm movements carried out in weightlessness after a prolonged exposure to the so-called microgravity conditions of orbital spaceflight. Subjects performed radial point-to-point motions of the fingertip on a touch pad. The authors note a reduction in movement speed in weightlessness, which they hypothesize could be due to either an overall strategy of lowering movement speed to better accommodate the instability of the body in weightlessness or an underestimation of body mass. They conclude for the latter, mainly based on two effects. One, slowing in weightlessness is greater for movement directions with higher effective mass at the end effector of the arm. Two, they present evidence for an increased number of corrective submovements in weightlessness. They contend that this provides conclusive evidence to accept the hypothesis of an underestimation of body mass.

Strengths:

In my opinion, the study provides a valuable contribution, the theoretical aspects are well presented through simulations, the statistical analyses are meticulous, the applicable literature is comprehensively considered and cited, and the manuscript is well written.

Weaknesses:

Nevertheless, I am of the opinion that the interpretation of the observations leaves room for other possible explanations of the observed phenomenon, thus weakening the strength of the arguments.

First, I would like to point out an apparent (at least to me) divergence between the predictions and the observed data. Figures 1 and S1 show that the difference between predicted values for the 3 movement directions is almost linear, with predictions for 90º midway between predictions for 45º and 135º. The effective mass at 90º appears to be much closer to that of 45º than to that of 135º (Figure S1A). But the data shown in Figure 2 and Figure 3 indicate that movements at 90º and 135º are grouped together in terms of reaction time, movement duration, and peak acceleration, while both differ significantly from those values for movements at 45º.

Furthermore, in Figure 4, the change in peak acceleration time and relative time to peak acceleration between 1g and 0g appears to be greater for 90º than for 135º, which appears to me to be at least superficially in contradiction with the predictions from Figure S1. If the effective mass is the key parameter, wouldn't one expect as much difference between 90º and 135º as between 90º and 45º? It is true that peak speed (Figure 3B) and peak speed time (Figure 4B) appear to follow the ordering according to effective mass, but is there a mathematical explanation as to why the ordering is respected for velocity but not acceleration? These inconsistencies weaken the author's conclusions and should be addressed.

Response (11): Indeed, the model predicts an almost equal separation between 45° and 90° and between 90° and 135°, while the data indicate that the spacing between 45° and 90° is much smaller than between 90° and 135°. We do not regard the divergence as evidence undermining our main conclusion since 1) the model is a simplification of the actual situation. For example, the model simulates an ideal case of moving a point mass (effective mass) without friction and without considering Coriolis and centripetal torques. 2) Our study does not make quantitative predictions of all the key kinematic measures; that will require model fitting, parameter estimation, and posture-constrained reaching experiments; instead, our study uses well-established (though simplified) models to qualitatively predict the overall behavioral pattern we would observe. For this purpose, our results are well in line with our expectations: though we did not find equal spacing between direction conditions, we do confirm that the key kinematic measures (Figure 2 and Figure 3 as questioned) show consistent directional trends between model predictions and empirical data. We added new analysis results on this matter: the directional effect we observed (how the key measures changed in microgravity across direction condition) is significantly correlated with our model predictions in most cases. Please check our detailed response (2) above. These results are also added in the revision.

We also highlight in the revision that our modeling is not to quantitatively predict reaching behaviors in space, but to qualitatively prescribe that how mass underestimation, but not the conservative control strategy, can lead to divergent predictions about key kinematic measures of fast reaching.

Then, to strengthen the conclusions, I feel that the following points would need to be addressed:

(1) The authors model the movement control through equations that derive the input control variable in terms of the force acting on the hand and treat the arm as a second-order low-pass filter (Equation 13). Underestimation of the mass in the computation of a feedforward command would lead to a lower-than-expected displacement to that command. But it is not clear if and how the authors account for a potential modification of the time constants of the 2nd order system. The CNS does not effectuate movements with pure torque generators. Muscles have elastic properties that depend on their tonic excitation level, reflex feedback, and other parameters. Indeed, Fisk et al. showed variations of movement characteristics consistent with lower muscle tone, lower bandwidth, and lower damping ratio in 0g compared to 1g. Could the variations in the response to the initial feedforward command be explained by a misrepresentation of the limbs' damping and natural frequency, leading to greater uncertainty about the consequences of the initial command? This would still be an argument for unadapted feedforward control of the movement, leading to the need for more corrective movements. But it would not necessarily reflect an underestimation of body mass.

Fisk, J. O. H. N., Lackner, J. R., & DiZio, P. A. U. L. (1993). Gravitoinertial force level influences arm movement control. Journal of neurophysiology, 69(2), 504-511.

Response (12): We agree that muscle properties, tonic excitation level, proprioception-mediated reflexes all contribute to reaching control. Fisk et al. (1993) study indeed showed that arm movement kinematics change, possibly owing to lower muscle tone and/or damping. However, reduced muscle damping and reduced spindle activity are more likely to affect feedback-based movements. Like in Fisk et al.’s study, people performed continuous arm movements with eyes closed; thus their movements largely relied on proprioceptive control. Our major findings are about the feedforward control, i.e., the reduced and “advanced” peak velocity/acceleration in discrete and ballistic reaching movements. Note that the peak acceleration happens as early as approximately 90-100ms into the movements, clearly showing that feedforward control is affected -- a different effect from Fisk et al’s findings. It is unlikely that people “advanced” their peak velocity/acceleration because they feel the need for more later corrective movements. Thus, underestimation of body mass remains the most plausible explanation.

(2) The movements were measured by having the subjects slide their finger on the surface of a touch screen. In weightlessness, the implications of this contact are expected to be quite different than those on the ground. In weightlessness, the taikonauts would need to actively press downward to maintain contact with the screen, while on Earth, gravity will do the work. The tangential forces that resist movement due to friction might therefore be different in 0g. This could be particularly relevant given that the effect of friction would interact with the limb in a direction-dependent fashion, given the anisotropy of the equivalent mass at the fingertip evoked by the authors. Is there some way to discount or control for these potential effects?

Response (13): We agree that friction might play a role here, but normal interaction with a touch screen typically involves friction between 0.1N and 0.5N (e.g., Ayyildiz et al., 2018). We believe that the directional variation of the friction is even smaller than 0.1N. It is very small compared to the force used to accelerate the arm for the reaching movement (10N-15N). Thus, friction anisotropy is unlikely to explain our data. Indeed, our readers might have the same concern, we thus added some discussion about possible effect of friction.

Citation: Ayyildiz M, Scaraggi M, Sirin O, Basdogan C, Persson BNJ. Contact mechanics between the human finger and a touchscreen under electroadhesion. Proc Natl Acad Sci U S A. 2018 Dec 11;115(50):12668-12673.

(3) The carefully crafted modelling of the limb neglects, nevertheless, the potential instability of the base of the arm. While the taikonauts were able to use their left arm to stabilize their bodies, it is not clear to what extent active stabilization with the contralateral limb can reproduce the stability of the human body seated in a chair in Earth gravity. Unintended motion of the shoulder could account for a smaller-than-expected displacement of the hand in response to the initial feedforward command and/or greater propensity for errors (with a greater need for corrective submovements) in 0g. The direction of movement with respect to the anchoring point could lead to the dependence of the observed effects on movement direction. Could this be tested in some way, e.g., by testing subjects on the ground while standing on an unstable base of support or sitting on a swing, with the same requirement to stabilize the torso using the contralateral arm?

Response (14): Body stabilization is always a challenge for human movement studies in space. We minimized its potential confounding effects by using left-hand grasping and foot straps for postural support throughout the experiment. We think shoulder stability is an unlikely explanation because unexpected shoulder instability should not affect the feedforward (early) part of the ballistic reaching movement: the reduced peak acceleration and its early peak were observed at about 90-100ms after movement initiation. This effect is too early to be explained by an expected stability issue. This argument is now mentioned in the revised Discussion.

The arguments for an underestimation of body mass would be strengthened if the authors could address these points in some way.

Recommendations for the authors:

Reviewing Editor Comments:

General recommendation

Overall, the reviewers agreed this is an interesting study with an original and strong approach. Nonetheless, there were significant weaknesses identified. The main criticism is that there is insufficient evidence for the claim that the movement slowing is due to mass underestimation, rather than other explanations for the increased feedback corrections. To bolster this claim, the reviewers have requested a deeper quantitative analysis of the directional effect and comparison to model predictions. They have also suggested that a 2-dof arm model could be used to predict how mass underestimation would influence multi-joint kinematics, and this should be compared to the data. Alternatively, or additionally, a control experiment could be performed (described in the reviews). We do realize that some of these options may not be feasible or practical. Ultimately, we leave it to you to determine how best to strengthen and solidify the argument for mass underestimation, rather than other causes.

As an alternative approach, you could consider tempering the claim regarding mass underestimation and focus more on the result that slower movements in microgravity are not simply a feedforward, rescaling of the movement trajectories, but rather, have greater feedback corrections. In this case, the reviewers feel it would still be critical to explain and discuss potential reasons for the corrections beyond mass underestimation.

We hope that these points are addressable, either with new analyses, experiments, or with a tempering of the claims. Addressing these points would help improve the eLife assessment.

Reviewer #1 (Recommendations for the authors):

(1) Move model descriptions to the main text to present modelling choices in more detail

Response (15): Thank you for the suggestion. We have moved the model descriptions to the main text to present the modeling choices in more detail and to allow readers to better cross-reference the analyses.

(2) Perform quantitative comparisons of the directional effect with the model's predictions, and add raw kinematic traces to illustrate the effect in more detail.

Response (16): Thanks for the suggestion, we have added the raw kinematics figure from a representative participant and please refer to Response (2) above for the comparisons of directional effect.

(3) Explore the effect of varying cost parameters in addition to mass estimation error to estimate the proportion of data explained by the underestimation hypothesis.

Response (17): Thank you for the suggestion. This has already been done—please see Response (1) above.

Reviewer #2 (Recommendations for the authors):

Minor comments:

(1) It must be justified early on why reaction times are being analyzed in this work. I understood later that it is to rule out any global slowing down of behavioral responses in microgravity.

Response (18): Exactly, RT results are informative about the absence of a global slowing down. Contrary to the conservative-strategy hypothesis, taikonauts did not show generalized slowing; they actually had faster reaction times during spaceflight, incompatible with a generalized slowing strategy. Thanks for point out; we justified that early in the text.

(2) Since the results are presented before the methods, I suggest stressing from the beginning that the reaching task is performed on a tablet and mentioning the instructions given to the participants, to improve the reading experience. The "beep" and "no beep" conditions also arise without obvious justification while reading the paper.

Response (19): Great suggestions. We now give out some experimental details and rationales at the beginning of Results.

(3) Figure 1C: The vel profiles are not returning to 0 at the end, why? Is it because the feedback gain is computed based on the underestimated mass or because a feedforward controller is applied here? Is it compatible with the experimental velocity traces?

Response (20): Figure. 1C shows the forward simulation under the optimal control policy. In our LQG formulation the terminal velocity is softly penalized (finite weight) rather than hard-constrained to zero; with a fixed horizon° the optimal solution can therefore end with a small residual velocity.

In the behavioral data, the hand does come to rest: this is achieved by corrective submovements during the homing phase.

(4) Left-skewed -> I believe this is right-skewed since the peak velocity is earlier.

Response (21): Yes, it should be right-skewed, thanks for point that out.

(5) What was the acquisition frequency of the positional data points? (on the tablet).

Response (22): The sampling frequency is 100 Hz. Thanks for pointing that out; we’ve added this information to the Methods.

(6) Figure S1. The planned duration seems to be longer than in the experiment (it is more around 500 ms for the 135-degree direction in simulation versus less than 400 ms in the experiment). Why?

Response (23): We apologize for a coding error that inadvertently multiplied the body-mass parameter by an extra factor, making the simulated mass too high. We have corrected the code, rerun the simulations, and updated Figures 1 and S1; all qualitative trends remain unchanged, and the revised movement durations (≈300–400 ms) are closer to the experimental values.

(7) After Equation 13: "The control law is given by". This is not the control law, which should have a feedback form u=K*x in the LQ framework. This is just the dynamic equations for the auxiliary state and the force. Please double-check the model description.

Response (24): Thank you for point this out. We have updated and refined all model equations and descriptions, and moved the model description from the Supplementary Materials to the main text; please see the revised manuscript.

Reviewer #3 (Recommendations for the authors):

(1) I have a concern about the interpretation of the anisotropic "equivalent mass". From my understanding, the equivalent mass would be what an external actor would feel as an equivalent inertia if pushing on the end effector from the outside. But the CNS does not push on the arm with a pure force generator acting at the hand to effectuate movement. It applies torque around the joints by applying forces across joints with muscles, causing the links of the arm to rotate around the joints. If the analysis is carried out in joint space, is the effective rotational inertia of the arm also anisotropic with respect to the direction of the movement of the hand? In other words, can the authors reassure me that the simulations are equivalent to an underestimation of the rotational inertia of the links when applied to the joints of the limb? It could be that these are mathematically the same; I have not delved into the mathematics to convince myself either way. But I would appreciate it if the authors could reassure me on this point.

Response (25): Thank you for raising this point. In our work, “equivalent mass” denotes the operational-space inertia projected along the hand-movement direction u, computed as:

This formulation describes the effective mass perceived at the end effector along a given direction, and is standard in operational-space control.

Although the motor command can be coded as either torque/force in the CNS, the actual executions are equivalent no matter whether it is specified as endpoint forces or joint torques, since force and torque are related by . For small excursions as investigated here, this makes the directional anisotropy in endpoint inertia consistent with the anisotropy of the effective joint-space inertia required to produce the same endpoint motion. Conceptually, therefore, our “mass underestimation” manipulation in operational space corresponds to underestimating the required joint-space inertia mapped through the Jacobian. Since our behavioral data are hand positions, using the operational-space representation is the most direct and appropriate way for modeling.

(2) I would also like to suggest one more level of analysis to test their hypothesis. The authors decomposed the movements into submovements and measured the prevalence of corrective submovements in weightlessness vs. normal gravity. The increase in corrective submovements is consistent with the hypothesis of a misestimation of limb mass, leading to an unexpectedly smaller displacement due to the initial feedforward command, leading to the need for corrections, leading to an increased overall movement duration. According to this hypothesis, however, the initial submovement, while resulting in a smaller than expected displacement, should have the same duration as the analogous movements performed on Earth. The authors could check this by analyzing the duration of the extracted initial submovements.

Response (26): We appreciate the reviewer’s suggestion regarding the analysis of the initial submovement duration. In our decomposition framework, each submovement is modeled as a symmetric log-normal (bell-shaped) component, such that the time to peak speed is always half of the component duration. Thus, the initial submovement duration is directly reflected in the initial submovement peak-speed time already reported in our original manuscript (Figure. 5F).

However, we respectfully disagree with the assumption that mass underestimation would necessarily yield the same submovement duration as on Earth. Under mass underestimation, the movement is effectively under-actuated, and the initial submovement can terminate prematurely, leading to a shorter duration. This is indeed what we observed in the data. Therefore, our reported metrics already address the reviewer’s proposal and support the conclusion that mass underestimation reduces the initial submovement duration in microgravity. Per your suggestion, we now added one more sentence to explain to the reader that initial submovement peak-speed time reflect the duration of the initial submovement.

Some additional minor suggestions:

(1) I believe that it is important to include the data from the control subjects, in some form, in the main article. Perhaps shading behind the main data from the taikonauts to show similarities or differences between groups. It is inconvenient to have to go to the supplementary material to compare the two groups, which is the main test of the experiment.

Response (27): Thank you for the suggestion. For all the core performance variables, the control group showed flat patterns, with no changes across test sessions at all. Thus, including these figures (together with null statistical results) in the main text would obscure our central message, especially given the expanded length of the revised manuscript (we added model details and new analysis results). Instead, following eLife’s format, we have reorganized the Supplementary Material so that each experimental figure has a corresponding supplementary figure showing the control data. This way, readers can quickly locate the control results and directly compare them with the experimental data, while keeping the main text focused.

(2) "Importantly, sensory estimate of bodily property in microgravity is biased but evaded from sensorimotor adaptation, calling for an extension of existing theories of motor learning." Perhaps "immune from" would be a better choice of words.

Response (28): Thanks for the suggestion, we edited our text accordingly.

(3) "First, typical reaching movement exhibits a symmetrical bell-shaped speed profile, which minimizes energy expenditure while maximizing accuracy according to optimal control principles (Todorov, 2004)." While Todorov's analysis is interesting and well accepted, it might be worthwhile citing the original source on the phenomenon of bell-shaped velocity profiles that minimize jerk (derivative of acceleration) and therefore, in some sense, maximize smoothness. Flash and Hogan, 1985.

Response (29): Thanks for the suggestion, we added the citation of minimum jerk.

(4) "Post-hoc analyses revealed slower reaction times for the 45° direction compared to both 90° (p < 0.001, d = 0.293) and 135° (p = 0.003, d = 0.284). Notably, reactions were faster during the in-flight phase compared to pre-flight (p = 0.037, d = 0.333), with no significant difference between in-flight and post-flight phases (p = 0.127)." What can one conclude from this?

Response (30): Although these decreases reached statistical significance, their magnitudes were small. The parallel pattern across groups suggests the effect is not driven by microgravity, but is more plausibly a mild learning/practice effect. We now mentioned this in the Discussion.

(5) "In line with predictions, peak acceleration appeared significantly earlier in the 45° direction than other directions (45° vs. 90°, p < 0.001, d = 0.304; 45° vs. 135°, p < 0.001, d = 0.271)." Which predictions? Because the effective mass is greater at 45º? Could you clarify the prediction?

Response (31): We should be more specific here; thank you for raising this. The predictions are the ones about peak acceleration timing (shown in Fig. 1H). We now modified this sentence as:

“In line with model predictions (Figure 1H), ….”.

(6) Figure 2: Why do 45º movements have longer reaction times but shorter movement durations?

Response (32): Appreciate your careful reading of the results. We believe this is possibly due to flexible motor control across conditions and trials, i.e., people tend to move faster when people react slower with longer reaction time. This has been reflected in across-direction comparisons (as spotted by the reviewer here), and it has also been shown within participant and across participants: For both groups, we found a significant negative correlation between movement duration (MD) and reaction time (RT), both across and within individuals (Figure 2—figure supplement 5). This finding indicates that participants moved faster when their RT was slower, and vice versa. This flexible motor adjustment, likely due to the task requirement for rapid movements, remained consistent during spaceflight.

Commands: * /compact to summarise the chat (degrades quality significantly) * /clear (clears current conversation so ensure complete CLAUDE.md and plan.md) * /(find the rest in the folder in the Claude Code course)

magivc

x

Joint Public Review:

Quite obviously, the brain encodes "time", as we are able to tell if something happened before or after something else. How this is done, however, remains essentially not understood. In the context of Working Memory tasks, many experiments have shown that the neural activity during the retention period "encodes" time, besides the stimulus to be remembered; that is, the time elapsed from stimulus presentation can be reliably inferred from the recordings, even if time per se is not important for the task. This implies 'mixed selectivity', in the weak sense of neural activity varying with both stimulus identity and time elapsed (since presentation).

In this paper, the authors investigate the implications of a specific form of such mixed selectivity, that is, conjunctive coding of what (stimulus) and when (time) at the single-neuron level, on the resulting dynamics of the population activity when 'viewed' through linear dimensionality-reduction techniques, essentially Principal Component Analysis (PCA). The theoretical/modeling results presented provide a useful guide to the interpretation of the experimental results; in particular, with respect to what can, or cannot, be rightfully inferred from those experimental results (using PCA-like techniques). The results are essentially theoretical in nature; there are, however, some conclusions that require a more precise justification, in my opinion. More generally, as the authors themselves discuss in the paper, it is not clear how to generalize this coding scheme to more complicated, but behaviorally and cognitively relevant, situations, such as multi-item WM or WM for sequences.

(1) It is unclear to me how the conjunctive code that the authors use (i.e., Equation (3)) is constrained by the theoretical desiderata (i.e., compositionality) they list, or whether it is simply an ansatz, partly motivated by theoretical considerations and experimental observations.

The "what" part: What the authors mean by "relationships" between stimuli is never clearly defined. From their argument (and from Figure 1b), it would seem that what they mean is "angles" between population vectors for all pairs of stimuli. If this is so, then the effect of the passing time can only amount to a uniform rescaling of the components of the population vector (i.e., it must be a similarity transformation; rotations are excluded, if the linear-decoder vectors are to be time-independent); the scaling factor, then, must be a strictly monotonous function of time (increasing or decreasing), if one is to decode time. In other words, the "when" receptive fields must be the same for all neurons.

The "when" part: The condition, \tau_3=\tau_1+\tau_2, does not appear to be used at all. In fact, it is unclear (to me at least) whether the model, as it is formulated, is able to represent time intervals between stimuli.

(2) For the specific case considered, i.e., conjunctive coding, it would seem that one should be able to analytically work out the demixed PCA (see Kobak et al., 2016). More generally, it seems interesting to compare the results of the PCA and the demixed PCA in this specific case, even just using synthetic data.

(3) In the Section "Dimensionality of neural trajectories...", there is some claim about how the dimensionality of the population activity goes up with the observation window T, backed up by numerical results that somehow mimic the results of Cueva et al. (2020) on experimental data. Is this a result that can be formally derived? Related to this point, it would be useful to provide a little more justification for Equation (17). Naively, one would think that the correlation matrix of the temporal component is always full-rank nominally, but that one can get excellent low-rank approximations (depending on T, following your argument).

I think this concept of a human computer is very thought provoking in a myriad of ways. My own research pertains to how identity is being reshaped along the boundaries of recommendation algorithms, and I feel like this idea of a human computer has a lot of thematic overlap. I know that the first "computers" were people who manually did computation as their job, but to what extent did their status as essentially beings of code shape their lives? Did they (or do they, in the case of modern human computers) see the world differently than I? What do they notice? Does living so in tune with the digital shape their philosophical framework? I feel like the fact that so much of culture and my own time is mediated by retention-based algorithms drastically shapes my ability to imagine another world different from my own. Does the same hold true for human computers?

I always wondered how programmers would create these sorts of commands, and it's cool to know that it's done with simple commands like these! I was also not previously aware that to display something on a screen, you have to use the command 'display'. I previously thought that 'print' was the main form to do so.

Not really. I think in a lot of controversial cases involving bot behavior are often not excused because at the end of the day, someone programmed it. Maybe it's because I don't know exactly how bots work, but I feel like a bot's actions are limited to what its programmer allows it to do, so if a bot is able to do something (such as post racist comments, as explained in an earlier section of this chapter), then something in its code allowed it to process racist information and express it. In the case from earlier this chapter, it probably should've been considered that if the bot is using other twitter users' tweets toward the company and that information is public, there probably should've been safeguards to detect racial remarks and block it from the bot's vocabulary and processing. It could be argued that that scenario slipped from their minds, but I think that anything that will be released to the public should be tested and be given feedback from a smaller community before released to the mass public. Surely someone would've thought that the bot could pick up racist tweets.

MCP 可以拓展 claude code 的能力，给它更多的工具和能力。

其实 build 工具也可以做这件事，自己封装一系列步骤成一个命令，然后一次性运行，那么 claude code 的这个自定义 command 好在： - 提供了 convention - 自然语言的 argument，本质上还是利用了 claude 的功能

Plan Mode can boost claude code intelligence

在 claude code 粘贴截图

AGENT INTEL: The "Save State" Monitor Card

SAVE THIS FOR YOURSELF AS A REMINDER

This is your tactical reminder. Copy, print, or transcribe this onto a card and fix it to the side of your primary monitor. It serves as a visual "circuit breaker" to prevent a biological hijack when a human enters your workspace.

S0P002 | THE SAVE STATE PROTOCOL STATUS: Logic Saved. Human Witnessed. Connection Restored.

SECURE THE LOGIC (Ctrl+S) Physically press the keys. This is your "mental bookmark." The machine now holds the data so your brain doesn't have to.

CALM THE BIOLOGY (The Vagal Brake) Feet flat. Shoulders down. Exhale for four seconds. Flush the cortisol out of your system before you speak.

SHIFT THE VECTOR (The 180° Pivot) Physically turn your body. If you are facing the screen, you are in "Processor Mode." If you face the person, you are in "Partner Mode."

RE-CENTRE THE MISSION Internal Check: “Am I optimising for the machine or the human?”

OPEN THE CHANNEL "I'm here. I’ve saved my work. Tell me what's happening."

Why this works: The Cognitive Handover By performing a physical action (Ctrl+S), you are manually offloading the "Cognitive Load" from your Prefrontal Cortex. This reduces the internal "friction" that causes the irritability (the snap). The physical 180-degree turn breaks the visual "loop" of the monitor's blue light, allowing your brain to switch from Task-Positive Network (focusing on the code) to the Default Mode Network (focusing on the relationship).

You are not merely "stopping work"; you are honouring the Ruach (the breath/spirit) of the person standing before you. You are choosing the Yellow Light (the human) over the Blue Light (the machine).

1. The Neuroscience (The Biological Reality) When you are deep in a "deployment" or high-focus task, your brain is utilising the Prefrontal Cortex (PFC) to hold complex, fragile data structures in your working memory (RAM). An interruption triggers the Amygdala, the "engine" of your threat-detection system. Because the PFC is already taxed, the brain misinterprets a partner’s "Bid for Connection" as a predatory strike against your cognitive resources.

This results in an immediate spike of Cortisol and Adrenaline, narrowing your visual and emotional field—a "biological hijack." To counter this, you must engage the Vagus Nerve. By pressing your feet flat and extending your exhale, you signal the parasympathetic nervous system to downregulate the alarm, shifting the energy from "Defence" back to "Social Engagement." You are effectively clearing the cache of your emotional processor to make room for a new, higher-priority input.

1. The Scripture (The Spiritual Logic) The text highlights a failure to recognise the Ruach (Spirit/Breath) in the room. In Hebrew thought, Ruach is not just "spirit" in an abstract sense, but the very "animating breath" that makes a human distinct from a machine. When you treat Sam like a "broken appliance," you are committing a logic error in the Kingdom: you are valuing the Asah (work/doing) over the Neshama (the living soul).

The "Physics of the Kingdom" dictates that Love (Agape) is the primary vector. In 1 Corinthians 13, Paul describes Love as not being "easily provoked" (paroxynetai—literally, "not sparked into a sharp edge"). When you snap, you have allowed your internal friction to create a spark that severs the "cord" of connection. By hitting the "Save State," you are aligning with the Sabbath Logic: the world (and the code) is sustained by God, allowing you the freedom to stop, turn, and witness the Image Dei standing in your doorway.

1. The FieldGuide Application: "The 180-Degree Pivot" To lock the vector and prevent the "Glitch in the Hallway," perform this Micro-Drill the next time your workflow is breached:

The Physical Save (5 Seconds): Physically press Ctrl + S (or Command + S). This tactile movement tells your brain: "The data is safe; the PFC can let go."

The Grounding (10 Seconds): Press both feet into the floor. Feel the weight. Exhale slowly through pursed lips as if blowing out a candle.

The Pivot (5 Seconds): Physically rotate your chair or your torso 180 degrees away from the screen.

The Identification: Look at the person and internally label them: "Image of God. Priority One."

The Opening: Say: "I've saved my place. I'm listening."

digital transition is possible because it scales through spreading. You don't have to solve exponential scale first.

Sees the original fight by digital rights activists now joined by geopolitical economics and international cybersec. Thinks this combi will win out

This is the pivot. Everything changes when we spot the lie. If you're stuck here, use the counter-measure - it seems to easy to fix it, but it works!

Algo similar es lo que quiero capitalizar con Cardumem y luego portar a Grafoscopio, pues, como lo ha mostrado la experiencia con este último, las interfaces en Spec, el toolkit gráfico de Pharo, si bien brindan algunas cosas que las interfaces web no tienen, adolecen del basto ecosistema de ésta última y mantienen los documentos y la computación aisladas dentro de la imagen.

La web, por el contrario, es casi ubicua en términos de las tecnologías ya instaladas y así no se cuente con una conexión a internet en el equipo de cómputo, si este tiene una interfaz gráfica, muy seguramente contará con un naveador web. Y ahora que los sistemas hipermedia, hacen posible programar la web desde cualquier lenguaje (HOWL: Hypermedia On Whatever you Like), se puede aprovechar tanto lo que sabemos de los lenguajes/entornos que nos gustan (Pharo o Lua) como del amplio sistema de la web. Antes de 2023, que se popularizaron los sistemas hipermedia, teníamos que elegir entre lo uno y lo otro. Y yo deselegí activamente la web, debido al adefesio de JavaScript y lo engorroso del CSS. Hoy, las condiciones son bien distintas.

The sanction in this case is pretty fair.

Author response:

The following is the authors’ response to the previous reviews

Public Reviews:

Reviewer #1 (Public review):

Summary:

Mazar & Yovel 2025 dissect the inverse problem of how echolocators in groups manage to navigate their surroundings despite intense jamming using computational simulations.

The authors show that despite the 'noisy' sensory environments that echolocating groups present, agents can still access some amount of echo-related information and use it to navigate their local environment. It is known that echolocating bats have strong small and large-scale spatial memory that plays an important role for individuals. The results from this paper also point to the potential importance of an even lower-level, short-term role of memory in the form of echo 'integration' across multiple calls, despite the unpredictability of echo detection in groups. The paper generates a useful basis to think about the mechanisms in echolocating groups for experimental investigations too.

Strengths:

The paper builds on biologically well-motivated and parametrised 2D acoustics and sensory simulation setup to investigate the various key parameters of interest

The 'null-model' of echolocators not being able to tell apart objects & conspecifics while echolocating still shows agents succesfully emerge from groups - even though the probability of emergence drops severely in comparison to cognitively more 'capable' agents. This is nonetheless an important result showing the direction-of-arrival of a sound itself is the 'minimum' set of ingredients needed for echolocators navigating their environment.

The results generate an important basis in unraveling how agents may navigate in sensorially noisy environments with a lot of irrelevant and very few relevant cues.

The 2D simulation framework is simple and computationally tractable enough to perform multiple runs to investigate many variables - while also remaining true to the aim of the investigation.

Weaknesses:

Authors have not yet provided convincing justification for the use of different echolocation phases during emergence and in cave behaviour. In the previous modelling paper cited for the details - here the bat-agents are performing a foraging task, and so the switch in echolocation phases is understandable. While flying with conspecifics, the lab's previous paper has shown what they call a 'clutter response' - but this is not necessarily the same as going into a 'buzz'-type call behaviour. As pointed out by another reviewer - the results of the simulations may hinge on the fact that bats are showing this echolocation phase-switching, and thus improving their echo-detection. This is not necessarily a major flaw - but something for readers to consider in light of the sparse experimental evidence at hand currently.

The use of echolocation phases—defined as the sequential search, approach, and buzz call patterns—has been documented not only during foraging but also in tasks such as landing, obstacle avoidance, clutter navigation, and drinking. Bat call structure has been shown to vary systematically with object proximity, not exclusively in response to prey. During obstacle avoidance, phase transitions were observed, with approach calls emitted in grouped sequences and with reduced durations (Gustafson & Schnitzler, 1979; Schnitzler et al., 1987). In landing contexts, bats have been reported to emit short-duration calls and decrease inter-pulse intervals—buzz-like patterns also observed during prey capture— suggesting shared acoustic strategies across behaviors (Hagino et al., 2007; Hiryu et al., 2008; Melcón et al., 2007, 2009). Comparable patterns have been reported during drinking maneuvers, where “drinking buzzes” have been proposed to guide a precise approach to the water surface, analogous to landing buzzes (Griffiths, 2013; Russo et al., 2016). In response to environmental complexity, bats were found to shorten calls and increase repetition rates when navigating cluttered spaces compared to open ones (Falk et al., 2014; Kalko & Schnitzler, 1993).

Moreover, field recordings from our study of Rhinopoma microphyllum (Goldshtein et al., 2025) revealed shortened call durations and inter-pulse intervals during dense group flight outside the cave during emergence—patterns consistent with terminal-approach phase that is typical when coming very close to an object (another bat in this case). The Author response image 1 shows an approach sequence recorded from a tagged bat approximately 20 meters from the cave entrance, with self-generated echolocation calls marked. The inter-pulse-interval of ca. 20 ms is used by these bats when a reflective object (another bat in this case) is nearby. 

Author response image 1.

These results provide direct evidence that bats actively employ approach-phase echolocation during swarming likely to avoid collision with other bats. This supports the view that echolocation phase transitions are a general proximity-based sensing strategy, adapted across a variety of behavioral scenarios—not limited to hunting alone. 

In our simulations, bats predominantly emitted calls in the approach phase, with only rare occurrences of buzz-phase calls.

See lines 355-363 in the revised manuscript.

The decision to model direction-of-arrival with such high angular resolution (1-2 degrees) is not entirely justifiable - and the authors may wish to do simulation runs with lower angular resolution. Past experimental paradigms haven't really separated out target-strength as a confounding factor for angular resolution (e.g. see the cited Simmons et al. 1983 paper). Moreover, to this reviewer's reading of the cited paper - it is not entirely clear how this experiment provides source-data to support the DoA-SNR parametrisation in this manuscript. The cited paper has two array-configurations, both of which are measured to have similar received levels upon ensonification. A relationship between angular resolution and signal-to-noise ratio is understandable perhaps - and one can formulate such a relationship, but here the reviewer asks that the origin/justification be made clear. On an independent line, also see the recent contrasting results of Geberl, Kugler, Wiegrebe 2019 (Curr. Biol.) - who suggest even poorer angular resolution in echolocation.

We thank the reviewer for raising this important point. The acuity of 1.5–3° in horizontal direction-of-arrival (DoA) estimation is based on the classical work of Simmons et al. with Eptesicus fuscus (Simmons et al., 1983). Similar precision was later supported by Erwin et al. (Erwin et al., 2001), who modeled azimuth estimation from measured interaural intensity differences (IIDs), reporting an average error of 0.2° with a standard deviation of ~2.2°, consistent with the behavioral data found by Simmons. The decline in acuity with increasing arrival angle has also been demonstrated in behavioral and physiological studies of binaural IID processing (Erwin et al., 2001; Fay, 1995; Razak, 2012; Wohlgemuth et al., 2016). The error model itself was first introduced in our earlier work (Mazar & Yovel, 2020).

Importantly, Geberl et al. (Geberl et al., 2019) examined the resolution of weak targets masked by nearby strong flankers  and found poor spatial discrimination of ~45 degrees; however, they were studying a detection problem, rather than the horizontal acuity of azimuth estimation. Indeed, our model assumes there is no spatial discrimination at all.

Overall, while our DoA–SNR parametrization can certainly be critiqued and alternative parameterizations could be tested in future work, we believe it reflects a reasonable and empirically supported assumption. 

Reviewer #2 (Public review):

This manuscript describes a detailed model for bats flying together through a fixed geometry. The model considers elements which are faithful to both bat biosonar production and reception and the acoustics governing how sound moves in air and interacts with obstacles. The model also incorporates behavioral patterns observed in bats, like one-dimensional feature following and temporal integration of cognitive maps. From a simulation study of the model and comparison of the results with the literature, the authors gain insight into how often bats may experience destructive interference of their acoustic signals and those of their peers, and how much such interference may actually negatively effect the groups' ability to navigate effectively. The authors use generalized linear models to test the significance of the effects they observe.

The work relies on a thoughtful and detailed model which faithfully incorporates salient features, such as acoustic elements like the filter for a biological receiver and temporal aggregation as a kind of memory in the system. At the same time, the authors abstract features that are complicating without being expected to give additional insights, as can be seen in the choice of a two-dimensional rather than three-dimensional system. I thought that the level of abstraction in the model was perfect, enough to demonstrate their results without needless details. The results are compelling and interesting, and the authors do a great job discussing them in the context of the biological literature.

With respect to the first version of the manuscript, the authors have remedied all my outstanding questions or concerns in the current version. The new supplementary figure 5 is especially helpful in understanding the geometry.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

Data Availability: This reviewer lauds the authors for switching from a private commercial folder requiring login to one that does not. At the cost of being overtly pedantic - the Github repository is not a long-term archival resource. The ideal solution is to upload the code in an academic repository (Zenodo, OSF, etc.) to periodically create a 'static snapshot' of code for archival, while also hosting a 'live' version on Github.

We have uploaded to Zenodo repository, and updated the link in the paper:

How bats exit a crowded colony when relying on echolocation only - a modeling approach

In one of the rebuttals to Reviewer #3- the authors have cited a wrong paper (Beleyur & Goerlitz 2019) - while discussing broad bandwidth calls improving detection - and may wish to correct this if possible on record.

We have removed the incorrect citation from the revised version of the manuscript.

Specific comments on the 2nd manuscript:

Figure 5: Table 1 says 1, 2,5,10,20,40,100 bats were simulated (line 138-139) but the conclusion (line 398) says '1 to 100 bats' per 3msq. However, the X-axis only stops at 40 and says 'number of bats', while the legend says bats/3msq....what is actually being plotted? Moreover, in the entire paper there is a constant back-and-forth between density and # of bats - perhaps it is explained beforehand, but it is a bit unsettling - and more can be done to clarify these two conventions.

While most parameters were tested across the full range of 1 to 100 bats per 3 m², a subset of conditions—including misidentification, multi-call clustering, wall target strength, and conspecific target strength—were simulated only up to 40 bats due to significantly longer run-times. This is now clarified in both the main text and the Table 1 caption.

In our simulations, the primary parameter was the number of bats placed within a 3 m² starting area, which directly determined the initial density (bats per 3 m²). Throughout the manuscript, we use “number of bats” to refer to the simulation input, while “density” denotes the equivalent ecological measure. Figure 5 and related captions have been revised accordingly to note these conventions and to indicate when results are shown only up to 40 bats (see lines 120–122, 314-317 in the revised text).

Table 1: This was made considerably difficult to read given the visual clutter - and I hope I've understood these changes correctly.

What is in the square brackets of the effect-size (e.g. first row with values 'Exit prob. (%)' says -0.37/bat [63:100] ? What does this 63:100 refer to?

What is the 'process flag'

Values in square brackets indicate the minimum and maximum values of the metric across the tested range (e.g., [63:100] shows the range of exit probabilities observed across different bat densities).

The term “process flag” has been replaced with “with and without multi-call clustering” for clarity

Both the table layout and caption have been revised to reduce visual clutter and to make these conventions clearer to the reader. 

Lines 562-3: "In our study, due to the dense cave environment, the bats are found to operate in the approach phase nearly all of the time, which is consistent with natural cave emergence behavior" - bats are 'found to' implies there is some experimental data or it is an emergent property. See above for the point questioing the implementation of multiple echolocation phases in the model, but also - here the bat-agents are allowed to show different phases and thus they do so -- it is a constraint of the implementation and not a result per se given the size of the cave and the number of bats involved...

We removed the sentence from the Methods section, since it could be misinterpreted as an experimental finding rather than a model outcome. Instead, we now discuss this in the Discussion, clarifying that the predominance of the approach phase arises from the cluttered cave environment in our simulations, which is consistent with natural emergence behavior (see lines 355-363). In this context, the use of echolocation phases is presented as a biologically plausible modeling choice rather than an empirical result.

Lines 659-660: The parametrisation between DoA and SNR is supposedly found in 'Equation 10' - which this reviewer could not find in the manuscript

The equation was accidentally omitted in the previous revision and has now been reinserted into the manuscript. It defines how direction-of-arrival (DoA) error depends on SNR and azimuth angle (see lines 603-605).

small typo in the code here.

Author response:

The following is the authors’ response to the original reviews.

Public Reviews:

Reviewer #1 (Public review):

Summary:

This study addresses the encoding of forelimb movement parameters using a reach-to-grasp task in mice. The authors use a modified version of the water-reaching paradigm developed by Galinanes and Huber. Two-photon calcium imaging was then performed with GCaMP6f to measure activity across both the contralateral caudal forelimb area (CFA) and the forelimb portion of primary somatosensory cortex (fS1) as mice perform the reaching behavior. Established methods were used to extract the activity of imaged neurons in layer 2/3, including methods for deconvolving the calcium indicator's response function from fluorescence time series. Video-based limb tracking was performed to track the positions of several sites on the forelimb during reaching and extract numerous low-level (joint angle) and high-level (reach direction) parameters. The authors find substantial encoding of parameters for both the proximal and distal parts of the limb across both CFA and fS1, with individual neurons showing heterogeneous parameter encoding. Limb movement can be decoded similarly well from both CFA and fS1, though CFA activity enables decoding of reach direction earlier and for a more extended duration than fS1 activity. Collectively, these results indicate involvement of a broadly distributed sensorimotor region in mouse cortex in determining low-level features of limb movement during reach-to-grasp.

Strengths:

The technical approach is of very high quality. In particular, the decoding methods are well designed and rigorous. The use of partial correlations to distinguish correlation between cortical activity and either proximal or distal limb parameters or either low- or high-level movement parameters was very nice. The limb tracking was also of extremely high quality, and critical here to revealing the richness of distal limb movement during task performance.

The task itself also reflects an important extension of the original work by Galinanes and Huber. The demonstration of a clear, trackable grasp component in a paradigm where mice will perform hundreds of trials per day expands the experimental opportunities for the field. This is an exciting development.

The findings here are important and the support for them is solid. The work represents an important step forward toward understanding the cortical origins of limb control signals. One can imagine numerous extensions of this work to address basic questions that have not been reachable in other model systems.

Collectively, these strengths made this manuscript a pleasure to read and review.

Thank you!

Weaknesses:

In the last section of the results, the authors purport to examine the representation of "higher-level target-related signals," using the decoding of reach direction. While I think the authors are careful in their phrasing here, I think they should be more explicit about what these signals could be reflecting. The "signals" here that are used to decode direction could relate to anything - low-level signals related to limb or postural muscles, or true high-level commands that dictate only what movement downstream motor centers should execute, rather than the muscle commands that dictate how. One could imagine using a partial correlation-type approach again here to extract a signal uncorrelated with all the measured low-level parameters, but there would still be all the unmeasured ones. Again, I think it is still ok to call these "high-level signals," but I think some explicit discussion of what these signals could reflect is necessary.

Thank you for this excellent suggestion. We have followed both pieces of the reviewer’s advice. First, we performed the suggested analysis, partialing off the kinematics then performing target classification on the residuals. This is now Figure 6S1. The analysis revealed the presence of target-related information in the neural activity after subtracting off all linear correlations with kinematics, supporting our claims that higher-level information is present in both populations. The exact timing of classifier performances varied substantially across mice, potentially due to differences in reach-to-grasp strategy, kinematic tracking fidelity, and exact spatial locations of each recorded FOV. Following the second suggestion, we have made the relevant text more careful. We now conclude simply that higher-level signals, meaning those signals that are largely unrelated to forelimb joint angle kinematics, are present but with variable timing and strengths in each area. That text now reads:

“Target decoding performance could result from truly higher-level signals that code abstractly for target location, or alternatively could be supported by strong encoding of kinematic variables that differed between targets. To disambiguate these possibilities, we refit the linear classifier to neural data after regressing off variance related to the joint angle kinematics. The strength and exact time course of the resulting target decoding varied somewhat across animals, but the earliest portion of target decoding performance persisted in all animals after the removal of kinematics and performance remained stronger for M1-fl than S1-fl (Fig. 6S1B). We thus conclude that higher-level signals are present in both areas, but differ in their exact timing and strength. However, we note that other possible signals, such as postural changes, could not be controlled for here.”

Related to this, I think the manuscript in general does not do an adequate job of explicitly raising the important caveats in interpreting parametric correlations in motor system signals, like those raised by Todorov, 2000. The authors do an expert job of handling the correlations, using PCA to extract uncorrelated components and using the partial correlation approach. However, more clarity about the range of possible signal types the recorded activity could reflect seems necessary.

This is an important point, and our text could have unintentionally misled readers. We have now attempted to make this point explicit in the Discussion and in the Results for Figure 6. This Discussion text now reads:

“Moreover, as is widely known (Todorov 2000), the exact role of these kinematically-related signals is challenging to determine from correlative measures alone; thus, determining whether these signals are used for direct movement control or instead indirectly reflect control performed elsewhere is left as a topic for future work.”

The manuscript could also do a better job of clarifying relevant similarities and differences between the rodent and primate systems, especially given the claims about the rodent being a "first-class" system for examining the cellular and circuit basis of motor control, which I certainly agree with. Interspecies similarities and differences could be better addressed both in the Introduction, where results from both rodents and primates are intermixed (second paragraph), and in the Discussion, where more clarity on how results here agree and disagree with those from primates would be helpful. For example, the ratio of corticospinal projections targeting sensory and motor divisions of the spinal cord differs substantially between rodents and primates. As another example, the relatively high physical proximity between the typical neurons in mouse M1 and S1 compared to primates seems likely to yoke their activity together to a greater extent. There is also the relatively large extent of fS1 from which forelimb movements can be elicited through intracortical microstimulation at current levels similar to those for evoking movement from M1. All of these seem relevant in the context of findings that activity in mouse M1 and S1 are similar.

We understand two points to address here. The first point is that we needed to be more careful to attribute previous results as being from the rodent vs. monkey. We agree. We have now revised several parts of the paper to make these distinctions clearer. The second point is about the potential benefit of a thorough review of the many ways in which primate and rodent sensorimotor systems differ. We entirely agree that this could be useful for the field. However, this is a sizable endeavor and doing it full justice is beyond what we know how to fit in the space allotted for framing our results here. We therefore sought a compromise, acknowledging how our results correspond to existing results in the primate without exhaustively accounting for how they differ. Future work will be necessary to more carefully disambiguate whether species-specific differences are due to biomechanical, neurological, ethological, or as-of-yet undetermined sources. We have incorporated your final specific points about what could produce similar information in M1 and S1 into the Discussion.

“This may simply be a consequence of widely distributed representations of movement across mouse cortex (Musall et al. 2019; Steinmetz et al. 2019; Stringer et al. 2019), including forelimb somatosensory areas, or may be a consequence of the close physical proximity of M1-fl and S1-fl hindering development of functionally distinct representations (Tennant et al. 2011).”

In addition, there are a number of other issues related to the interpretation of findings here that are not adequately addressed. These are described in the Recommendations for improvement.

Reviewer #2 (Public review):

Summary:

In this manuscript, Grier, Salimian, and Kaufman characterize the relationship between the activity of neurons in sensorimotor cortex and forelimb kinematics in mice performing a reach-to-grasp task. First, they train animals to reach to two cued targets to retrieve water reward, measure limb motion with high resolution, and characterize the stereotyped kinematics of the shoulder, elbow, wrist, and digits. Next, they find that inactivation of the caudal forelimb motor area severely impairs coordination of the limb and prevents successful performance of the task. They then use calcium imaging to measure the activity of neurons in motor and somatosensory cortex, and demonstrate that fine details of limb kinematics can be decoded with high fidelity from this activity. Finally, they show reach direction (left vs right target) can be decoded earlier in the trial from motor than from somatosensory cortex.

Strengths:

In my opinion, this manuscript is technically outstanding and really sets a new bar for motor systems neurophysiology in the mouse. The writing and figures are clear, and the claims are supported by the data. This study is timely, as there has been a recent trend towards recording large numbers of neurons across the brain in relatively uncontrolled tasks and inferring a widespread but coarse encoding of high-level task variables. The central finding here, that sensorimotor cortical activity reflects fine details of forelimb movement, argues against the resurgent idea of cortical equipotentiality, and in favor of a high degree of specificity in the responses of individual neurons and of the specialization of cortical areas.

Thank you!

Weaknesses:

It would be helpful for the authors to be more explicit about which models of mouse cortical function their results support or rule out, and how their findings break new conceptual ground.

We appreciate this feedback and have attempted to make these details clearer through changes to the Introduction and Discussion. One key change is noted below:

“The presence of detailed kinematic signals in the sensorimotor cortex supports a model of mouse sensorimotor cortex in which M1-fl and S1-fl play a strong role in shaping the fine details of reaching and grasping movements.”

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

In addition to the weaknesses noted above, I suggest the authors also address the following:

The last results section is generally lacking in statistical support for claims. Statistical support should be added.

Thank you for pointing this out, we have added more statistical support to this section.

The consideration in the Discussion of relevant previous findings and potential explanations for the distal limb signals in mouse sensorimotor cortex is somewhat lacking. There are several specific issues:

(1) In contrast to the present study, the studies cited in regards to a lack of motor cortical involvement did not involve dexterous movements - in fact, Kawai et al. explicitly engineered a task that did not involve dexterity to distinguish the role of motor cortex in learning from its known role in dextrous movement execution. In Kawai et al., the authors note one rat who adopted a more dexterous approach to the lever pressing task; in this rat, a motor cortical lesion did cause a longer-lasting reduction in task performance. In additional experiments reported in Kawai's PhD thesis, performance of a dextrous task does erode with motor cortex lesion, as seen in other studies, like the early rodent reaching work of Whishaw and colleagues.

(2) Other possible explanations for the persistence of non-dexterous tasks following motor cortical removal are compensation by, or redundant functionality in, other motor system regions.

(3) It is also worth noting that stimulation in different regions of mouse M1 and S1 evokes alternately, digit, wrist, and elbow movements in fairly similar proportions (Tennant, 2011), suggesting that descending pathways substantially target spinal circuits that control all forelimb joints.

(4) It also seems relevant that although the recovery time course is longer, nonhuman primates also retain substantial hand control after motor cortical removal (e.g. Lashley, 1925; Glees and Cole, 1950; Passingham et al., 1983). Humans of course, appear to be a different story.

These are good points. We have tried to make the Discussion better reflect the tension in the literature, including with this new text:

“However, several other previous results have indirectly suggested that M1 and S1 may be involved in the details of forelimb movement. Performance suffers with inactivation or lesioning of M1 and S1 in skilled, complex manual behaviors (Guo et al 2015, Mizes et al 2024, Whishaw et al 1990) or idiosyncratic use of digits to accomplish non-dexterous tasks (Kawai 2014). The sparing of non-dexterous tasks with these lesions may also reflect redundancy in control as opposed to irrelevance of M1 and S1. Nevertheless, our finding of low-level kinematic information in sensorimotor cortex supports a role for cortex beyond simply providing redundant high-level commands to these subcortical areas.”

We have avoided mentioning points 3 and 4 in the paper; the stimulation results might follow from activating projections not normally involved in this behavior, and discussing primates in this context would require a long list of caveats. We agree that these points are worth thinking about, but are concerned that they are too circumstantial to include in interpreting the results formally.

Although similar decoding performance is achieved using neurons from both CFA and fS1, I am left wondering whether you would do substantially better with CFA using activity at additional preceding time points, or when using exclusively time points from the past. The primary model used here appears to use neural signals from corresponding time points to decode limb parameters, but results seemingly could be different when using preceding time points as regressors.

We appreciate this suggestion and have added the analysis to an additional supplementary panel for Figure 5 (Figure 5S3). Incorporating lags into the decoder via a Wiener filter does indeed improve the decoding performance, but this could simply be due to the increase in the number of predictor variables. This analysis did not, however, further disambiguate M1-fl and S1-fl: the performance improvement was similar across areas for both causal and acausal lag configurations. This could be a consequence of the time resolution of calcium imaging, so further experiments with electrophysiology would be required to rule this possibility out. We now note this new result:

“Including additional causal (-100 ms preceding) and/or acausal (-100 ms preceding to 100 following) lags improved decoding performance modestly and similarly for both areas (Fig. 5S3E-F).”

Related to this, I am also worried about the bleeding of signals across time here. If you deconvolve and interpolate between time points, the interpolation seemingly will pull information into the past, up to half the sampling period, which here is on the order of how long it takes signals to travel to and from the limb. The authors do not make any inappropriate claims about the neural signals here reflecting causes or consequences of what is happening at the limb, but readers (like me) will still try to draw these sorts of conclusions. Is it possible that, although decoding from instantaneous signals is similar for the two regions, the M1 signals are actually motor signals related to future limb state while the S1 signals are sensory consequences? Even if many of the relevant details related to conduction times are not known, perhaps the authors could clarify what can and can't be said related to causal interpretation here.

Thank you for suggesting further explanation here. We agree that our interpretation could be made more specific. We have added text in the Discussion section to speak more directly to what can and cannot be concluded from our analyses. In short, it is hard to be certain of lags in calcium imaging data for many reasons, and using recording methods with finer temporal resolution (like electrophysiology) will be necessary for determining the precise temporal relationships between kinematics and neural activity. In the absence of these recordings, we limit our claim to kinematic information being present in M1-fl and S1-fl neural activity and leave determining the causal role of this information to future work.

New clarifying text in the Discussion:

“The use of calcium imaging further prevents strong conclusions about whether activity reflects future limb states or sensory consequences. Confirming this limitation, inclusion of lagged data in the decoding models, whether causal or acausal, resulted in similar performance changes in both areas.”

An alternative reason why lift onset is less decodable in CFA is that CFA activates substantially before lift onset, as has been observed in previous rodent studies (Kargo and Nitz, 2004; Miri et al., 2017; Veuthey et al., 2020), perhaps as some sort of movement preparation. S1, on the other hand, may not have this early activity, and so may show a clearer transient at onset when the hand and limb start to move. This seems more likely than the explanations provided by the authors.

This is a valid possible alternative explanation and we have updated the Discussion to reflect this. This difference in the structure of M1-fl activity versus S1-fl is apparent in the projections of Figure 6A, which show M1-fl projections more clearly aligned to cue-onset than S1-fl projections.

“Our lift time decoding results are consistent with this view and align with recent observations characterizing mouse proprioceptive forelimb cortex, (Alonso et al 2023), although an alternative explanation may be simply that M1-fl activates earlier than S1-fl during reaching (Kargo and Nitz 2004; Miri et al 2017; Veuthey et al 2020).”

To better clarify relevant similarities and differences between the rodent and primate systems, the Introduction could include some of these similarities and differences exposed by the literature currently cited, and the Discussion could include an additional paragraph specifically relating findings here to previous observations in the primate.

We appreciate the reviewer’s thoughtfulness on possible framings of our results. When writing this paper, framing was a major challenge for us and we drafted quite a few versions of the Introduction including some that focused more on mouse-primate comparison. In the end, we decided the most critical function of the Intro was to set up our central question, of “levels-of-sensorimotor-control”. The rich primate literature was valuable here, but getting into a protracted compare-and-contrast exercise quickly became a distraction from the point. Further, we sought to highlight the relevance and importance of the question answered in our work as the mouse has gained prominence for filling gaps that are challenging to address with primates. This paper serves as one of many early steps towards the ultimate goal of revealing general properties of sensorimotor cortical function with the mouse model. We have made some subtle changes to the Introduction that we hope will more clearly communicate this narrative. 

We agree that a Discussion paragraph directly relating our results to those in primates would benefit our conclusions and have added one:

“These results expand our understanding of the rodent sensorimotor system and highlight similarities to nonhuman primates. We show here evidence in mice of detailed joint angle kinematic signals from the full forelimb in M1 and S1, as has been shown in macaque cortex during tasks involving reaching and grasping objects (Vargas-Irwin et al. 2010; Saleh et al. 2010, 2012; Goodman et al. 2019; Okorokova et al. 2020). Additionally, the earlier onset of movement-related activity in M1-fl compared to S1-fl is similar to macaque M1 and S1 (Tanji and Evarts 1976). Taken together these results suggest that the mouse can be employed to address questions traditionally explored in primates about how cortical activity encodes detailed movement commands.”

Although this is outside the scope of the present study, it would be interesting to image descending projection neurons to see what signals are conveyed downstream, and to what targets. Some signals observed in layer 2/3 may not be strongly reflected in descending projections.

We agree that recording from descending projection neurons in this task would be of deep interest – and also agree that these experiments are beyond the scope of the present study. We look forward to performing these additional experiments in future work.

Minor:

(1) The use of "CFA" and “fS1” is a bit confusing. S1, like M1, is defined primarily based on histological criteria, while CFA is defined by intracortical microstimulation. CFA contains a substantial fraction of fS1, seemingly most of it based on the maps shown in Tennant et al., 2011. This is not really a criticism, as the field has not reached any sort of consensus on this nomenclature yet.

We are similarly unhappy with the inconsistency of the terminology in the field, and struggled with how not to make it worse.  After much debate and consultation with colleagues, we decided to use “M1” and “S1” to evoke the century of literature on these areas; and “-fl” to indicate forelimb because it is more intuitive than “-ul” and avoids using the illegible “-ll” for hindlimb (relevant to our subsequent paper). For what we called M1-fl, we recorded where we did because anecdotally we saw similar responses across that swath; but note that this definition is also consistent with the definition of “MOp-ul” found with multimodal mapping by

Munoz-Castaneda (2021), which extends a little anteriorly of MOp as defined by the Allen CCF. As the field continues to mature, we hope future work can converge on a set of shared terms.

(2) Page 4: "Inactivations and lesions of M1 and S1 have shown that M1 is required for the execution of dexterous reach-to-grasp movements" - to me, earlier work from Whishaw and colleagues deserves to be cited here.

We appreciate the suggestion and have updated the references in this section to better reflect the prior work from Whishaw and other researchers.

(3) Page 5: "evoking sufficient trial-to-trial variability to avoid model overfitting." - what I think the authors are referring to here is a particular kind of "overfitting," the consequence of not exploring the full movement space, as opposed to model overfitting from issues with the model-fitting method itself. Rather than just saying overfitting, the authors could be clearer about what they are referring to.

The reviewer is right; the phenomenon we intended to refer to is not properly termed overfitting. Specifically, we meant that data with restricted range does not necessarily express global structure, and models can therefore incorrectly fit them. For example, fitting a linear model to data including many periods of a sine wave will correctly show a zero-slope linear component, but fitting to only a portion of a single cycle will typically yield a nonzero slope. This is not overfitting, is not exactly underfitting (because the relevant structure is barely present in the data, as opposed to missed by an insufficiently powerful model), is not bias (the data are fit well), and is not even necessarily a problem (the local relationship may be what you are interested in). Yet, it does not reflect the larger structure of the data.

We do not know of a standard term for this phenomenon, so instead of dragging the reader through this tangential argument, we have tried to offer a simpler motivation for using multiple targets:

“Assessing the relationship between neural activity and the details of movement requires striking a balance between achieving repeatable behavior and evoking sufficient trial-to-trial variability to broadly sample movement space”.

(4) Page 5: Caudal Forelimb Area should not be capitalized.

Obviated with the change in area nomenclature.

(5) Page 7: "of linearly independent degrees of freedom" - for a neuroscience audience, I think it is better to explicitly mention that the resulting PCs are uncorrelated.

We agree that this section could benefit from clarification. We have attempted to provide additional nuance to indicate what the analysis was intended to test.

“Despite the strong coupling between the proximal and distal joint angles, rich variation remained in the action of different joints over time. The presence of strong correlations across joints suggested that the kinematics may be well described by a smaller number of independent degrees of freedom than the total number of recorded angles. To assess the number of linearly independent (uncorrelated) degrees of freedom amongst the 24 joint angles and velocities, we used double-cross-validated PCA (Yu et al. 2009); Methods; Fig. 3D), finding intermediate dimensionalities of 7 (median for joint angles) and 10 (velocities; Fig. 3E). This is consistent with the idea that joint angles across the limb are coordinated instead of controlled independently, and that this coordination is flexible enough over time to enable accurately performing reaching and grasping to different targets.”

(6) Page 7: In the Results, the authors should mention what indicator is being used, the imaging frame rate, and summarize briefly how cells were defined.

Thank you for the suggestion, these details have been added to the relevant results section for clarity.

“To do so, we recorded neural activity from neurons in layer 2/3 M1-fl extending into the immediately adjacent secondary motor cortex (M2), and the forelimb region of S1 (S1-fl) using two-photon calcium imaging of GCaMP6f-expressing neurons in layer 2/3 (185-230 μm deep, imaged at 31 Hz, cells extracted with Suite2p (Pachitariu et al 2017)).”

(7) Page 7: "corrected at n=2" - n doesn't typically refer to the number of tests, so for clarity I would say "corrected for dual tests."

Thank you for pointing this out, we have corrected the text and added additional explanation in the methods for our approach to determining statistical significance across the targets and locking events.

“P-values obtained through the ZETA were then Bonferroni corrected for dual tests when measuring the number of cells modulated to a given event and corrected for six tests (2 targets and 3 events) when measuring the overall number of modulated cells.”

(8) Page 7: In the Results, when the decoding is introduced, it would be helpful to have a few details without having to hunt through the Methods. For example, were things regularized, how was cross-validation handled, etc?

Thank you for the suggestion, these details have been added to the relevant results section for clarity.

A simple linear regression model related the single-trial joint angles at all time points to single-trial neural activity at the corresponding moments. The model was fit with ridge regression, the ridge penalty was determined via a heuristic (Karabatsos 2018), and performance was measured on held-out trials (80/20 train/test split, 50 folds).

(9) Page 8: I think it is worth noting how much mouse reaching involves shoulder rotation as opposed to movement in other joints, as this seems very different from primates.

Thank you for pointing this out. We think this is mostly a task difference: our mice were in a quadrupedal stance, whereas monkeys are typically asked to reach from a sitting position. We now mention this in the Results. 

“Reaching evoked particularly large rotation of the shoulder, likely because the mice reached from a quadrupedal position to targets on either side of the snout.”

(10) Page 8: Should provide quantification to clarify what is meant by "closely tracked."

We have updated the text to indicate that this claim was meant to be qualitative, and to more clearly highlight that the interest here is the first demonstration of the ability to reconstruct valid forelimb postures from decoded joint angles in the mouse. Quantifying the reconstruction properly would require substantially more manual data labeling, and the successful decoding itself demonstrates indirectly that the reconstructions are good enough to obtain the results of interest.

Additionally, we reconstructed the skeletal representation of the forelimb from the decoded joint angles and found that, as intended, the reconstructed postures had strong qualitative resemblance to the true postures, even of “minor” angles like cylindrical paw deformation or digit splay (Fig. 5C,G).

(11) Page 8: "Overall, these results suggest that instantaneous movement-related signals are similarly distributed across CFA and fS1." - I know we are being succinct here, but this sentence sounds like a non sequitur in the context of this paragraph - perhaps include a conclusion from the results in this paragraph first, then summarize the whole section.

Thank you for the suggestion, we have updated this text to more clearly conclude the results of this section.

Overall, these results reveal that neural activity in M1-fl and S1-fl is closely related to the kinematic details of reach-to-grasp movements. The ability to decode substantial variance in proximal and distal joints suggests that this relationship extends to the entire forelimb and the similar performance obtained from each area suggests that this information is similarly distributed across M1-fl and S1-fl. 

(12) Page 10: Mention of projections from fS1 does not explicitly specify their preferential targeting of the dorsal horn, which seems relevant.

We appreciate the suggestion and have added this detail to the text.

Rodent S1-fl is known to influence interneuron populations in the spinal cord through direct and indirect projections that predominantly target the dorsal horn (Ueno et al. 2018), thus these signals may also reflect S1-fl’s important role in modulating reflex circuits to coordinate sensory feedback with movement generation (Moreno-López et al. 2016; Moreno-Lopez et al. 2021; Seki et al. 2003).

(13) Page 31: Labels on the figure indicating what blue and red stand for would be helpful.

Thank you for the suggestion, labels have been added to indicate left and right trials for Figure 5 C/F and Figure 6A.

(14) Page 32: Legend does not include panel D.

Thank you for catching this, the corresponding caption has been added.

Reviewer #2 (Recommendations for the authors):

(1) The Introduction could perhaps set the central question in starker relief. What specifically do the authors mean by high- vs low-level control? As suggested by the cited studies, this has been a fraught issue in primate work for decades, and I think a finer-grained framing of alternative hypotheses would help set up the results. For example, would better performance at decoding joint angles than paw position be evidence for lower-level control? The clarity of the Introduction might also be improved if the facts and unknowns were broken down by species throughout.

We have tried to further improve the focus of the Introduction on the central question, clarify what we mean, and make clearer in the review of the literature which species a finding comes from.

The clarifying text from the introduction is quoted below:

Extensive motor mapping experiments in rodents have revealed that activating different parts of the sensorimotor cortex evokes movements of different body parts or different kinds of movements of the same body part, as it does in primates (for review, see (Harrison and Murphy 2014)). Yet it is unclear how the topography of stimulation-evoked movements relates to the roles of these areas during volitional actions. Perturbations during behavioral tasks in mice involving forelimb lever or reaching movements have provided a coarse-level understanding of how these areas contribute during behavior. Inactivations and lesions of M1 and S1 have shown that M1 is required for the execution of dexterous reach-to-grasp movements (Guo et al. 2015; Sauerbrei et al. 2020; Galiñanes et al. 2018; Wang et al. 2017; Whishaw et al. 1991; Whishaw 2000) and that S1 is essential for adapting learned movements to external perturbations of a joystick (Mathis et al. 2017). However, spinal cord projections from mouse M1 and S1 primarily target spinal interneurons rather than directly synapsing onto motor neurons (Gu et al. 2017; Ueno et al. 2018; Wang et al. 2017), suggesting cortical activity might play a more modulatory role. Further, stimulation of brainstem nuclei alone can evoke naturalistic forelimb actions, including realistic reaching movements involving coordinated flexion and extension of the proximal and distal limb (Esposito et al. 2014; Ruder et al. 2021; Yang et al. 2023). Taken together, these results have raised the question of what role mouse M1 and S1 play in the control of goal-directed forelimb movements. 

One route to answering this question involves characterizing the signals present in mouse M1 and S1 during movement. If mouse M1 and S1 were to control only high-level aspects of forelimb movements, activity should be dominated by ‘abstract’ signals like target location and reflect little trial-to-trial variability in reach kinematics. If instead M1 and S1 control low-level movement features then activity should correlate strongly with forelimb joint angle kinematics and their trial-to-trial variation when reaching to different targets. While the presence of high- or low-level signals in a cortical area does not necessarily imply that they are causally responsible for these aspects of movement, characterizing what signals are present serves as a first step toward determining how these areas relate to movement.

(2) The kinematics and calcium traces appear to be highly stereotyped across trials. If the population encodes joint angles, would one expect to find correlations between the neural and kinematic residuals after subtraction of the time-varying means? Some additional analysis and/or discussion on this point would be helpful, especially as there are only two targets.

This is a great idea. As suggested, we implemented regression models on the residuals for each target in the new Figure 5S3. Figure 5S3 A and B show the performance when decoding the residuals for right trials and C and D show performance for left trials. Decoding remained well above chance, despite shrinking down due to predicting this relatively small within-target variation. This analysis supports our claims from the main regression models in Figure 5 and 5S1-2, and also suggests that movements ipsilateral to the reaching limb (contralateral to the recording hemisphere) may be better encoded than movements contralateral to the reaching limb. We have added a reference to this additional residual analysis in the final paragraph of the decoding section of the Results section:

“Finally, we tested whether the ability to decode these many joint angles was a direct consequence of inter-joint correlations, and might not be indicative of the presence of “real” information about some of these joints. To do so, we fit partial correlation models that removed correlations between proximal and distal joints, or removed correlations of the joint angles with a high-level parameter – the overall distance of the paw centroid to the spout. Despite substantially lowering the behavioral variance, in each case the residuals could still be decoded from neural activity (Fig 5S2A-D). Similar decoding performance for M1-fl and S1-fl was obtained from models fit to decode single-trial residuals separately for left and right trials (Fig 5S3A-D), indicating that trial-to-trial variations on each basic movement were decodable from these populations.”

Along similar lines, binary classification is used to characterize cue-, lift-, and contact-responsive neurons. Is it possible to exploit trial-to-trial variation in the cue-lift and lift-contact latencies to extract the time-varying marginal effects of each event (e.g., using a GLM)?

For the detection of single-cell modulations by different events, we have elected to retain our simple statistical test to determine modulation; in our experience, encoding models typically involve a surprising number of steps to get them to do what you actually intend. We leave more extensive encoding model-style analysis to future work, currently in progress.

(3) The authors mention prior studies suggesting that the control of some forelimb tasks can be gradually transferred from the cortex to the subcortical centers. Have they performed the inactivation at different time points across learning, and if so, do they have evidence for a diminishing effect over time (e.g., blocking of both initiation and coordination early in training)? In addition, the effects of motor cortex inactivation are similar to, but slightly different from, effects shown in reaching tasks in prior studies. Some additional discussion on this point would be useful.

Our inactivation experiments in this study were intended to coarsely demonstrate the involvement of mouse forelimb sensorimotor cortex in our task. We have not performed the inactivations over learning and leave such experiments to future work. 

We agree that a little more clarity relating our results to previous ones was warranted. Previous studies (Guo et al. 2015 and Galinanes et al. 2018) have demonstrated inactivation impacts on similar tasks, but for thoroughness we sought to show the same for our task as it varied from the pellet and motorized water spout tasks in both training time and target configurations. Our results are strongly in line with those of Galinanes et al. 2018 which used a fairly similar water spout target configuration. In the inactivation experiments of that paper, 3 out of 13 animals with initiation-triggered inactivations were able to initiate reaching within a time window similar to control trials. Additionally, a proportion of trials across multiple mice proceeded with little perturbation from the inactivations. This is consistent with our observation that M1-fl inactivations may either abolish movement initiation or allow movement initiation but impair task completion on a trial-by-trial and animal-to-animal basis. Further work is required to determine what factors influence these differential responses to inactivation and to determine how these effects differ across task variations (i.e., pellet vs water spout). We have added a brief description of these nuances to the text for clarity. 

“These inactivations blocked the execution of the reach to grasp sequence, preventing the animal from making contact with the spout during the 3-second laser stimulation period (Fig. 1F; 86.5% control trials with contact within 3 seconds of cue, 5.1% inactivation trials with contact, P < 10^-191, Mann-Whitney U test, 2 mice, 495 stimulation trials). Interestingly, inactivation at the time of cue often did not prevent reach initiation (mouse 1: 54.7%, mouse 2: 34.2% of inactivation trials with lift within 3 seconds; 93.5%, 86.2% control trials). Yet the movement stalled once the paw and digits extended towards the spout, producing uncoordinated and unsuccessful reaching trajectories (Fig. 1I, two representative datasets). Taken together, these results support the involvement of M1-fl in the water-reaching task and suggest that the strength of inactivation effects may depend on specific task details like training time or target configuration (c.f. Galinanes et al. 2018).”

Minor points

(1) The rationale for the multiple comparisons procedure in identifying event-locked responses should be explained in more detail. If I understand correctly, the authors are not correcting for comparisons across ROIs, but instead control the family-wise error rate across brain regions and event types (dividing alpha by two or six). Why not instead control the false discovery rate across ROIs? 

Thank you for pointing this out, it was confusing as written and we received a similar comment from Reviewer 1. We have fixed the wording now to make it clearer why we did this. We simply aimed to describe how many of the recorded neurons in each area were modulated by the task as a proxy for the engagement of these areas during the behavior, and to use this measure of modulation as a criterion for including the neuron in subsequent analysis. In other words, if the question had been “are any neurons in this area modulated by the task?” then correcting for the number of ROIs would be the correct method; but if the question is, “is this neuron probably modulated and therefore worth including in my decoder?” correcting for the number of ROIs will typically be much too conservative. Thus, we only sought to correct for the false discovery rate across events and targets for each ROI. We have added additional text in the methods to clarify these choices, below. Please also see response to (7) from Reviewer 1 above.

“Note that we did not correct for the number of ROIs tested for two reasons. First, the goal of this testing was to serve as a criterion for inclusion in subsequent decoding analyses, not to determine whether any neurons in the area at all were modulated; and second, correcting for the number of ROIs would bias comparison between areas if different numbers of ROIs were recorded in one area vs. the other.”

(2) It appears joint angles are treated as linear variables in the decoding analysis; is this correct? This seems reasonable as long as the range of motion is not too large, but the authors might briefly comment on the issue in the Methods. 

Yes, all joint angles are treated as linear variables in the linear regression model. We observed empirically (as can be seen in Figure 3B and Figure 5B/F) that the joint angle variables were relatively constrained to specific ranges during the task, with no angles displaying substantial wrap-around during the reaching and grasping movements. It is true that use of nonlinear decoding would almost surely improve performance further. Future work could also compare decoding of joint angles with muscle forces, which correlate and which we made no effort to distinguish here. In this work, though, the demonstration of a substantial relationship between neural activity and kinematics already tells us that fine details of movement are present in the M1 and S1-fl populations, which is a critical fact to understand these areas and was not previously known. We now comment explicitly on this, as suggested.

“Joint angle or velocity kinematics were linearly interpolated from their original 6.66 ms to 10 ms and smoothed with a Gaussian (15 ms s.d.). These angular variables were then treated linearly in decoding analyses as their ranges were relatively constrained during the reaching and grasping movements; although the true relationships are likely nonlinear, this serves as a sufficient approximation to demonstrate the presence of a relationship between neural activity and kinematics.”

(3) Are the limb pose estimates mirrored along the mediolateral axis? Figures 1C and 2D appear to show reaches to the left spout on the animal's right.

Thank you for pointing out the ambiguity in the display of these data. The reach trajectories were not mirrored along the mediolateral axis, but they are displayed from the perspective of the behavioral imaging cameras as shown in Figure 1A. Thus the right target reaches (ipsilateral to the animal’s reaching arm) are on the left side of the camera image and the left target reaches (contralateral to the animal’s reaching arm) are on the right side of the image. We have clarified this in the figure captions.

Author response:

The following is the authors’ response to the original reviews.

Reviewer #1:

SOM+ interneurons such as Martinotti cells target the apical tufts of pyramidals in the cortex. Since interneurons in general are strongly implicated in mediating rhythmic population activity over a range of timescales, it is quite appropriate to study the consequence of rhythmic inhibition provided by SOM+ interneurons for synaptic integration, including the phenomenon of dendritic spikes. However, using conclusions from a singular study (ref 22) to identify the beta band as the rhythm mediated by SOM+ is not very accurate. SOM+ interneurons have been implicated in regulating rhythms centered just below 30 Hz (refs 22, 21). It is a range that lies in the grey zone of the traditional definition of beta and gamma. However, it is significantly higher than the 16 Hz rhythms explored in this study. It thus remains unknown how a 25-30 Hz rhythmic inhibition (that has an experimentally suggested role for dendrite targeting SOM+ INs) in apical tufts regulates dendritic spikes.

We agree with the reviewer that the rhythms arising from SOM+ interneurons can extend their frequencies higher than the 16 Hz analyzed in this study. To address this, we have conducted a new set of simulations where we delivered distal dendritic inhibition across a range of frequencies, from 0.5 to 80 Hz (see new Results section “Frequency specific effects of rhythmic inhibition on neuronal integration”). These results revealed, surprisingly, that at 30 Hz their ability to entrain Ca²⁺ and NMDA spikes degrades (but not Na⁺ spikes). This suggests that beta rhythms in the 20-30 Hz range are operating at the highest frequency for which dendritically targeting inhibition will be effective. The implications are covered in the Discussion section “Interaction with microcircuitry”. They are:

“Particularly in the visual cortex, SOM interneurons can generate a rhythm in the 25-30 Hz range [22]. We found this to be at the upper end of the frequency range for dendritic inhibitory rhythms to be effective in modulating NMDA and Ca²⁺ spikes. If this rhythm solely recruited SOM interneurons, its effectiveness would be marginal. Potentially compensating for this, recent work has found that PV interneurons also participate in beta/low-gamma [23, 24] (but see [21, 22]). In our model, on its own when beta rhythmic inhibition was delivered perisomatically we found that it was less able to entrain spiking and had an overall hyperpolarizing effect. However, if delivered in conjunction with the distal dendritic inhibition arising from SOM interneurons, this may strengthen entrainment.”

Distal dendritic inhibition has been previously shown to be more effective in controlling dendritic spikes. However, given the slow timescale of dendritic spikes, it can be hypothesized that high-frequency rhythmic inhibition would be ineffective in entraining the dendritic spikes either in distal or proximal location, as demonstrated by 4H and 5F, and vice versa. A computational study can take this further by exploring the robustness of this hypothesis. By sticking to a single-frequency definition of what constitutes Gamma (64 Hz) and Beta (16 Hz) inhibition, the current exploration does support the core hypothesis. However, given the temporal dynamics of dendritic spikes, it is valuable to learn, for example, the upper bound of "Beta" range (13-30Hz) inhibition that fails to phasically modulate them. In addition to the reason stated in the earlier paragraph, Alpha band activity (8-12 Hz), has been implicated (e.g. van Kerkoerle, 2014) in signaling of inter-areal feedback to the superficial layer in the cortex, potentially targeting apical tufts of pyramidals from multiple layers and resulting in alpha-range rhythmic inhibition. To make the findings significant, it might therefore be more pertinent to understand the consequences of ~10Hz rhythmic inhibition (in addition to the ~25-30 Hz Beta/Gamma) in the apical tufts for phasic modulation of dendritic spikes.

We added an additional set of simulations that address this in the Results section ‘Frequency specific effects of rhythmic inhibition on neuronal integration’. In general, we found that dendritic and perisomatic inhibitory rhythms at lower frequencies could entrain AP generation, but with less functional specialization. This is explored in our Discussion section ‘Interneuron specializations and rhythm timescales’.

The differential effect of Gamma and Beta range inhibition on basal and apical excitatory clusters is not convincing from the information provided. The basal cluster appears to overlap with perisomatic inhibitory synapses. The description in the methods does not have enough information to negate the visual perception (ln 979-81). With this understanding, it is not surprising that the correlation between excitation and APs is high (during the trough of gamma) for basal and not apical excitation. A more comparable scenario would be a more distal location of the basal excitatory cluster.

While we stated in the original manuscript that we were contrasting ‘basal’ vs. ‘apical’ clustered inputs, this terminology did not reflect our intent with these analyses. We meant to contrast proximal vs. distal dendritic clustered synaptic inputs, which the reviewer correctly noted is confounded in the apical vs. basal comparison. We have rewritten these results, their discussion, and corresponding figure, to clearly state that we are contrasting proximal vs. distal synaptic input.

Reviewer #2:

The weaknesses are probably in some of the parameterizations of inhibitory synaptic dynamics. A unitary peak conductance of 1nS is very high for inhibitory synapses. This high value could invariably skew some of the network-level predictions. The authors could obtain specific parameters from the Neocortical Collaboration Portal (https://bbp.epfl.ch/nmcportal/microcircuit.html), which is an incredible resource for cortical neurons and synapses.

We appreciate the valuable resource mentioned by the reviewer and will consult it when constructing future models. Regarding the present one, our choice of peak conductance was based on previous studies, namely:

Egger R, Narayanan RT, Guest JM, Bast A, Udvary D, Messore LF, Das S, de Kock CPJ, Oberlaender M (2020) Cortical output is gated by horizontally projecting neurons in the deep layers. Neuron 105, 122-137.e128.

and

Xiang Z, Huguenard JR, Prince DA (2002) Synaptic inhibition of pyramidal cells evoked by different interneuronal subtypes in layer v of rat visual cortex. J Neurophysiol 88, 740-750.

The study by Egger et al. used an inhibitory peak conductance of 1 nS and was simulating circuitry very similar to ours. We validated these synapses in pilot simulations that sought to characterize the resulting IPSPs and IPSCs, and whose results can be seen in Table 1 of our methods. These synapses exhibited IPSCs whose peak amplitudes ranged over values (~24162 pA) that agreed with the experimental literature, such as Xiang et al.

Given this, we feel our parameterization of inhibitory synapses does not warrant any changes.

Reviewer #3:

What disappointed me a bit was the lack of a concise summary of what we learned beyond the fact that beta and gamma act differently on dendritic integration. The individual paragraphs of the discussion often are 80% summary of existing theories and only a single vague statement about how the results in this study relate. I think a summarizing schematic or similar would help immensely.

We agree with the reviewer that a summary schematic would help the reader. This has been added to the manuscript as Figure 11. It demonstrates the principal findings of the paper and is referenced in the opening paragraph of the discussion section.

Orthogonal to that, there were some points where the authors could have offered more depth on specific features. For example, the authors summarized that their "results suggest that the timescales of these rhythms align with the specialized impacts of SOM and PV interneurons on neuronal integration". Here they could go deeper and try to explain why SOM impact is specialized at slower time scales. (I think their results provide enough for a speculative outlook.)

This discussion has been expanded under the section “Interneuron specializations and rhythm timescales”. The added text is:

“So, while our results suggest that spatial targeting of SOM and PV interneurons aligns with the timescales of their network-level rhythms, it could also be that their timing and subcellular localization interact to produce specialized neuron-level functions [85]. For instance, NMDA and Ca²⁺ spikes in the distal dendrites last for ~50 ms, making the slower beta rhythm more appropriate for bidirectionally controlling them. Both can be described as dynamical systems with distinct phases with differing sensitivity to inhibition. Ca²⁺ spikes are dynamical events comprised of an initiation, plateau, and termination phase. Inhibition delivered during the plateau phase shortens their duration [86]. If the beta rhythm is comprised of cycling between periods of elevated excitation (increased NMDA spike generation) followed by elevated inhibition, then Ca²⁺ spike initiation will tend to occur during the excitatory phase, and its plateau during the subsequent inhibitory phase. A plateau during the inhibitory phase will more quickly enter termination. This is bidirectional control. On the other hand, slower rhythms (e.g. 1 Hz) initiate Ca²⁺ spikes during the excitatory phase that plateau and enter termination autonomously, before the inhibitory phase is reached. The same principle holds for NMDA spikes [87]. As a result, rhythms in the range from 15-30 Hz are optimal for synchronizing the onsets and offsets of dendritic spikes across a population of neurons.

The integrative effects of gamma (>40 Hz) are also specialized. Low frequency inhibitory rhythms delivered to the soma tended to shift the membrane potential higher or lower with the rhythm’s phase, effectively bringing it closer or farther from AP generation but not changing the neuron’s sensitivity to fast synaptic inputs. In the gamma frequency range, this is reversed, with the mean membrane potential not varying with rhythm phase but with a shifting bias to positive or negative membrane potential fluctuations. In addition, the trough phase of gamma lowers the threshold for AP generation, while slower rhythms like beta only raise the threshold. Consequently, the timing of gamma is ideal for increasing the sensitivity of the neuron to rapid excitation. This agrees with the observation that gamma oscillations accompany rapid excitation-inhibition balancing [88].”

We also extended our discussion section ‘Relevance to coding’ to explore how beta and gamma rhythms can support sparse vs. dense population coding, respectively. It reads:

“One interpretation of rhythms arising from local inhibitory feedback is that they maintain the balance between excitation and inhibition. This can be thought of as a normalization operation that maintains activity within a set range. Normalization can be achieved either through a subtractive effect that raises the threshold for initiating an action potential, or a multiplicative effect that lowers the slope of the relationship between excitation and action potential firing rate. When considered at the population level, these normalization effects impact coding in different ways. Subtractive normalization increases sparsity by dropping out neurons whose excitation is below the raised threshold. Multiplicative normalization, however, encourages dense codes by scaling down firing rates and compressing the range of firing rates. This study found that while both perisomatic and distal dendritic inhibition produced subtractive effects, only perisomatic had a multiplicative effect. Tying this to beta and gamma, beta rhythms may encourage sparse population codes while gamma allows for dense.”

Beyond that, the authors invite the community to reappraise the role of gamma and beta in coding. This idea seems to be hindered by the fact that I cannot find a mention of a release of the model used in this work. The base pyramidal cell model is of course available from the original study, but it would be helpful for follow-up work to release the complete setup including excitatory and inhibitory synapses and their activation in the different simulation paradigms used. As well as code related to that.

We have added a Code and Data Availability section that addresses this. It reads: “Simulation code is deposited at ModelDB athttps://modeldb.science/2019883 . The raw simulation data are available from DBH upon request. Analysis code is posted as a github repo at https://github.com/dbheadley/InhibOnDendComp.”

Változott már az értékkészlet. Bővült az alábbiakkal: 31D2 Regulatory report code- 31D2 jelentés szerinti besorolás<br /> Security subtpye - Értékpapír altípus Security Fund type - Értékpapír alap típusa ( Befektetési alap)

EZt töröljük. Helyette kellene : External Codes Custom values Code - Kód Alternative code (Kötelező mező) - Altermatív kód<br /> Code at Subcustodian - Alletétkezelői kód ( Pl. Keler kód) C2 ID - C2 ID (külső rendszer azonosító)

Ide még írjuk bele szerintem: CFI code - Kötjel miatt kell ez a kód ( KELER - kibocsátáskor ép.-hez tartozó kód) - Liquid jelölőnégyzetet - Likvid értékpapírok jelölése 78/2014. (III. 14.) Korm. rendelet 31D2 Regulatory Report code -ot - MNB 31D2 és MNB 30LA kötelező jelentéshez szükséges

Especially with the collective knowledge that is made accessible through the Internet, the ability to efficiently add and remove comments seems important for the human aspect of programming languages. Particularly in cases when the code tries to incorporate even more complicated directions.

Ami nekem innen még hiányzik Margó az egy tábla az External codes custom values- Külső azonosítók egyedi adatok- Ez felületen az ALAPADAT EGYEDI ADATOK alatt van, ide kéne ez a táblázat : C2 ID - C2 ID SAP ID - SAP ID Alternatív code - Alternatív kód TEÁOR code - TEÁOR szám KSH Number - KSH Azonosító Symbols code - Symbols szám SAP code - SAP vevőkód Kondor code - KONDOR kód ( Kereskedési rendszer) Registration number - Lajstrom szám EIH code - EIH azonosító

Kedvezményezett ügyfélkód

Felületen ez nem látszik , vegyük le

vegyük le

Töröljük ezt az egészet

Töröljük ki ezeket

Töröljük ki ezeket

• Original author switched from VSCode to Zed in December and now uses Zed as the primary editor for Python and Go.
• Main reason for leaving VSCode was increasingly intrusive AI features (Copilot prompts, inline terminal suggestions) and perceived increase in crashes and slowness.
• Author still likes VSCode overall but feels rapid AI integration harmed stability and usability, and hopes it becomes less intrusive in the future.
• JetBrains IDEs were rejected as feeling too heavy, and Vim/Emacs as too time‑intensive to configure; Zed was attractive as a modern, lightweight Rust-based IDE.
• Transition from VSCode was smooth: similar UI, mostly compatible keybindings, and ability (unused by author) to import some VSCode settings.
• Zed felt significantly faster and more responsive than VSCode, with no glitches or crashes over a couple of weeks, restoring a sense of “joy of programming”.
• Initial Zed setup was minimal: adjust fonts, theme, disable inline git blame, and enable autosave; Go worked out of the box.
• Python setup required more work because Zed uses language servers and defaults to Basedpyright instead of Pylance (which is VSCode-only and closed source).
• The author hit unexpected strict type-checking because projects with a [tool.pyright] section in pyproject.toml effectively force Basedpyright’s recommended mode.
• Attempting to set typeCheckingMode in Zed’s settings.json did not help; the fix was explicitly setting typeCheckingMode = "standard" inside each project’s [tool.pyright] config.
• Another issue was delayed type diagnostics across files, fixed by setting "disablePullDiagnostics": true in Zed’s Basedpyright initialization options.
• Virtualenv handling and other Python-specific behavior worked smoothly; the author also tried the new ty language server, found it good, but stayed with Basedpyright to match CI’s Pyright.
• Zed is now the author’s default IDE: fast, stable, familiar, with enough extensions despite a much smaller ecosystem than VSCode.
• The main missing feature is a powerful side‑by‑side git diff viewer comparable to GitLens.
• Zed’s AI features are present but easy to ignore; paid plans for AI edit predictions seem like a reasonable way to fund development while keeping the core editor free.
• The author views Zed as a serious competitor that pressures VSCode to improve, especially around AI integration and performance.
• The post ends with sharing a minimal settings.json showcasing autosave, disabled inline blame, VSCode keymap, fonts, light theme, and customized Basedpyright LSP options.

Hacker News Discussion

• A VS Code team member acknowledges that AI-related features sometimes ignore the “disable” settings but states they try to ship fixes quickly and appreciate feedback.
• Several commenters recommend VSCodium as a way to get the open‑source VS Code experience without Microsoft’s telemetry and aggressive AI integration, while clarifying that both VS Code and VSCodium build from the same upstream repo.
• Many users express frustration with VS Code becoming bloated, unreliable, or “enshittified,” particularly around Copilot and complex configuration/remote setups, and are looking at Zed or classic editors as alternatives.
• Emacs and Vim/Neovim advocates argue that investing in these longstanding editors avoids churn and AI/UX regressions, with some describing decades-long Emacs usage and others praising Neovim plus LSPs as a lightweight yet powerful setup.
• Sublime Text is often cited as the spiritual predecessor of Zed in terms of speed and snappiness, with some saying Zed is the closest modern successor focused on performance.
• Zed users highlight positives like fast AI/MCP integration, good Nix/Direnv support, and pleasant design, but note pain points such as font rendering on low‑DPI or non‑GPU setups, Linux packaging gaps, missing REPLs for Lisps, and weaker debugging/extension ecosystems compared to VS Code or JetBrains.
• Some comments mention concrete bugs and annoyances in Zed, including format‑on‑save occasionally deleting the first line of Python classes, unwanted newline insertion at EOF, and missing small quality-of-life features (e.g., indentation autodetection, drag‑and‑drop markdown link insertion).
• A few developers describe hybrid workflows: using JetBrains IDEs on powerful machines, Zed on lower‑power devices, and Vim/Neovim or Sublime for quick one‑off edits, emphasizing that Zed is not yet at JetBrains’ level for deep refactoring and code understanding.
• Several participants discuss Zed’s business model as an AI “reseller”: core editor remains free while Pro users pay for pooled tokens across multiple AI providers, which some see as a relatively benign and sustainable way to monetize.
• There is concern that Zed’s extension ecosystem is still small and that Rust-based extension development may limit growth relative to VS Code; suggestions include better guidance for porting VS Code extensions and addressing collaboration/chat self‑hosting and security concerns.

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

We thank the Reviewers for their positive assessment of the quality and significance of our work, as well as for their insightful comments, which have helped us to further improve the manuscript. We have addressed the majority of the comments in the revised version and, for those that require additional time, we outline below a detailed plan of the experiments we intend to perform.

We agree with Reviewer #2 that a more detailed mechanistic understanding of the drug effects would further strengthen the study, and we are grateful to both reviewers for the constructive experimental suggestions provided to address this point. In particular, we are highly motivated to better define the causal role of C18 sphingolipid alterations in mediating the effects of the drugs, as suggested by Reviewer #2, as well as to investigate the involvement of the retromer complex in the lysosome-to-Golgi connection, as suggested by Reviewer #1.

Below, we provide a point-by-point description of the revisions already incorporated into the manuscript, along with the planned experiments that will address the remaining comments

REVIEWER #1:

VPS13B is a bridge-like lipid transfer protein, the loss or mutation of which is associated with Cohen syndrome (CS) involving Golgi fragmentation. In this study, the authors performed image-based chemical screens to identify compounds capable of rescuing the Golgi morphology in VPS13B-KO HeLa cells. They identified 50 compounds, the majority of which are lysosomotropic compounds or cationic amphiphilic drugs (CADs). Treatment of cells with several of these compounds causes lysosomal lipid storage, as assessed by BMP/LBPA staining, filipin staining, or LipidTOX staining. Interestingly, most LipidTOX puncta colocalized with transferrin receptor-positive compartments but not lysosomes. Similar to lysosomotropic compounds, knocking down NPC1 or SMPD1, mimicking lysosomal storage disease, also substantially rescued Golgi morphology. The authors show that VPS13B-KO cells have reduced C18 sphingolipids, which is reversed by treatment with CADs. Finally, the authors show that two CADs partially rescue neurite outgrowth in neuronal cultures. However, these drugs do not rescue the size of VPS13B KO organoids.

Overall, this is an impressive study identifying CADs as potential therapeutics for CS and suggesting sphingolipid upregulation as a general strategy for CS treatment. The morphological and lipidomics analyses unravel important molecular basis of CS pathology. This study will be of high interest to the field of lipid biology and organelle homeostasis. I have a few comments to help improve the quality of this study.

1. The reverse of lipid changes in VPS13B-KO cells by CADs is intriguing. Are CAD-mediated benefits such as Golgi morphology recovery permanent or only transient within 24 hours of treatment? How do the CADs affect the Golgi morphology in WT HeLa cells?

RESPONSE:

We thank the reviewer for this insightful question Indeed, the effects of CADs on Golgi organization are most evident in VPS13B KO cells, where the Golgi apparatus is severely fragmented and becomes more compact upon drug treatment, whereas the effect is much less apparent in wild-type cells. Nevertheless, a careful quantitative analysis of the images (now presented in the new Fig. S7) demonstrates that the impact of these compounds on Golgi morphology is not restricted to KO cells but is likely more general, supporting a link between lysosomal storage and Golgi organization. Although this observation indicates an indirect effect (consistent with the proposed mechanism of action), rather than a direct correction of VPS13B loss, it does not compromise in our opinion their potential beneficial effect for KO cells as shown also from the results obtained in organoid-derived neurons.

Under continuous treatment, azelastine keeps the Golgi in a compact state for 72 hours without any noticeable deleterious effect on the cells (see new Fig. S10) Raloxifene, on the contrary proved to be toxic over the same time period. We believe this difference reflects the mechanism of action of CADs, which progressively accumulate within acidic organelles and may eventually reach a toxic threshold upon prolonged exposure. For this reason, lower drug concentrations administered over longer treatment periods may represent a viable alternative strategy. In this regard, we also refer the reviewer to our response to the comment on brain organoids below.

1. Is it surprising that Azelastine-induced lipid storage in transferrin receptor compartments (early and recycling endosomes)? I suggest more controls to examine LipidTOX overlap with Golgi markers or other late endosome/lysosome markers such as LBPA and CD63.

RESPONSE:

We agree with the reviewer that this observation is somewhat unexpected. However, we would like to clarify that we do not intend to suggest that lipid storage occurs primarily in early or recycling endosomes, which would indeed contradict a substantial body of existing evidence. Rather, our data indicate that this particular dye (LipidTOX) labels recycling endosomes, at least in HeLa cells. This finding is consistent with the widely accepted view that lysosomal lipid storage exerts broader effects on intracellular trafficking, not limited to late endosomes/lysosomes. We corrected the text in order to clarify this concept.

LipidTOX was specifically developed to detect drug-induced phospholipidosis, and based on our data, it appears suitable for this purpose. To our knowledge, there is no published information detailing its intracellular localization, which motivated us to perform these control experiments. Unfortunately, the proprietary formulation of this product does not allow informed speculations to explain the observed localization or whether this could refer to the intact molecule or to a catabolite.

As suggested by the reviewer, we plan to perform co-staining with additional markers to further clarify this this point.

1. Does the LipidTOX/TFRC overlap suggest potential roles of retrograde transport in supplying sphingolipids to the Golgi? The authors can quickly test if the knockdown of a retromer subunit (VPS35) blocks Azelastine-induced recovery of Golgi morphology.

RESPONSE:

We thank the reviewer for this insightful suggestion. Indeed, the retromer complex represents one of the best-characterized trafficking pathways from the endosomal system to the Golgi, and this relatively straightforward experiment could help to mechanistically clarify our observations. We plan to test whether VPS35 knockdown interferes with the effects of the drugs.

What is the rationale to use 500 nM to 1 uM azelastine and raloxifene for neuronal cultures and organoids? At such concentrations, no obvious changes in Golgi morphology or lipid storage were observed (Fig 4). Also, the lipidomics analysis was performed after 10 uM compound treatment. It might be worth trying dose-response experiments in organoid tests.

RESPONSE:

We thank the reviewer for this question. The rationale about this choice was indeed missing from our previous version of the manuscript. The reason of lowering the concentrations comes indeed from toxicity tests, preliminarily performed over long-term treatment of both WT and VPS13B KO organoids. This information has now been explicitly included in the Results section of the revised manuscript, and the broader implications are also discussed in the Discussion section.

MINOR COMMENTS:

It is important to know whether the authors used TGN or cis-Golgi markers for Golgi morphology analysis. Please label the two channels in Fig. 2C and throughout all figures. In many cases, it is not clear what is stained in the green channel to show the Golgi morphology. It was not even stated in the legend.

RESPONSE:

We now included the antibody staining in all figure legends where it was previously missing.

The authors stated that Recovery of Golgi morphology is dependent on lysosomal lipid storage. However, while the data show positive correlation between the two, no causal relationship is established by the data. It seems true that in all conditions (CADs or genetic knockdown) where lysosomal lipid storage was observed, the authors detect the Recovery of Golgi morphology. However, budesonide did not depend on lysosomal lipid storage to recover the Golgi morphology. Thus, the recovery of Golgi morphology is NOT dependent on lysosomal lipid storage, but inducing lysosomal lipid storage appears sufficient to recover Golgi morphology in VPS13B-KO HeLa cells.

RESPONSE:

We thank the reviewer for this comment and we agree that the previous title of the paragraph could have been misleading. This has been now changed in: “Lysosomal lipid storage mediates the recovery of Golgi morphology” which is probably less prone to ambiguous interpretations.

Obviously, in the previous version of the title we wanted to mean that Golgi recovery is dependent on lipid storage “in the context of CAD treatment” and not as a general statement.

With respect to the cause–effect relationship, we believe that the strongest evidence supporting this link is the observation that genetically induced lipid storage phenocopies the effects of drug treatment. We hope that this conclusion is now sufficiently clear from the revised text.

Each figure needs a title before the detailed legends for specific panels.

RESPONSE:

Titles have now been included to all figure legends.

Fig 8. Y axis labeling is missing.

RESPONSE:

Axes labels have now been included

Does U18666A rescues Golgi morphology in VPS13B-KO cells?

RESPONSE:

We thank the reviewer for this comment. U18666A indeed also corrects Golgi morphology. The result is now included in the new figure S5.

Please do not repeat the result section in discussion. Focus on the most important points.

RESPONSE:

We thank the reviewer for this comment. We shortened the descriptive part of the discussion trying as much as possible to avoid repetitions with the result session and keeping only the more essential information for the flow of the discussion.

Reviewer #1 (Significance (Required)):

This is an impressive study that identifies Cationic Amphiphilic Drugs (CADs) as potential therapeutics for Cohen syndrome (CS) and suggests sphingolipid upregulation as a general strategy for diseases driven by VPS13B loss-of-function. The unbiased approaches, notably the chemical screen and lipidomics, provide novel mechanistic insights into the underlying pathology of CS. This study will be of high interest to researchers in the fields of lipid biology and organelle homeostasis. It will also be highly valuable for clinical pediatricians managing CS patients.

REVIEWER #2:

This manuscript describes a compound screening aimed at identifying molecules that can restore Golgi organization in VPS13B knockout (KO) cells. The authors identify several compounds, most of which are lysosomotropic, and analyze their effects on Golgi morphology and lipid composition using multiple approaches. They report that VPS13B KO cells exhibit a reduction in C18-N-acyl sphingolipids, which can be restored by several of the identified compounds. Furthermore, two of these compounds, azelastine and raloxifene, promote neurite outgrowth in VPS13B KO cortical organoids. These findings are interesting and could potentially contribute to a better understanding of the pathophysiology of Cohen syndrome and the development of therapeutic strategies. However, despite the large number of analyses presented, the study remains largely descriptive, and there is no coherent mechanistic explanation for how these compounds restore Golgi structure in VPS13B KO cells. In addition to the reduction in C18-N-acyl sphingolipids, the KO cells display alterations in several other lipid species (LPC, LPE, PC40:1, PE42:1, TG, etc.), and treatment with the selected compounds induces further lipid accumulations, including cholesterol and BMP/LBPA. The relationship between these diverse lipid changes and the observed Golgi recovery lacks clarity and mechanistic consistency.

MAJOR COMMENTS:

The finding that compounds cannot prevent Golgi fragmentation caused by brefeldin A or nocodazole but can suppress statin-induced fragmentation is intriguing, but the underlying mechanism is not addressed. It is not evident whether this difference results from changes in membrane lipid composition or restoration of Rab/SNARE trafficking. The authors should examine Rab prenylation and SNARE localization by immunofluorescence or Western blotting to support their interpretation.

RESPONSE:

We thank the reviewer for this suggestion and agree that the ability of these compounds to counteract statin-induced Golgi fragmentation is indeed intriguing. The primary reason we did not further explore this aspect is that we evaluated the effects of statins not to be a central focus of the present study. Nevertheless, we fully agree that this observation represents a valuable opportunity to gain additional insight into the mechanism underlying drug-induced Golgi recovery.

To address this point, we plan to analyze Rab prenylation by Western blot and Rab localization by microscopy, focusing on a Golgi-associated Rab protein such as Rab6. In addition, we will employ downstream inhibitors of Rab prenylation, such as 3-PEHPC (an inhibitor of type II protein geranylgeranyltransferase (GGTase-II)), which should allow us to formally distinguish effects related to impaired Rab prenylation from those arising from inhibition of cholesterol biosynthesis.

Although restoration of C18 sphingolipids (SM 36:1, CER 36:1) is observed upon compound treatment, its causal role in Golgi recovery or neurite outgrowth is not established. The authors should test whether blocking the increase of C18 SM/CER prevents the rescue of Golgi or neuronal phenotypes.

RESPONSE:

We sincerely thank the reviewer for this comment. We agree that, based on the current data, a definitive cause–effect relationship between Golgi recovery and the increase in C18 sphingolipids cannot be firmly established, and we acknowledge that a deeper understanding of this issue will require further investigation. Furthermore, we believe that addressing this would not only provide a better mechanistic understanding of the biological processes behind the effect of the drugs but provide a potential avenue for therapeutic intervention. For these reasons, we are strongly motivated to pursue this aspect further.

With respect to the reviewer’s specific suggestion, we agree that preventing the increase in C18 sphingolipids would be an ideal experimental approach. However, the limited understanding of the regulatory mechanisms controlling C18 sphingolipid homeostasis currently precludes a fully informed strategy. In principle, if the observed increase were due to enhanced synthesis, one could envisage blocking it by silencing ceramide synthases with C18 selectivity, such as CERS1. The experiment shown in Fig. 7E (azelastine treatment in the presence of sphingolipid synthesis inhibitors) was designed with this rationale in mind. However, these results suggest that azelastine-induced C18 sphingolipid accumulation is unlikely to result from increased synthesis, and is instead more consistent with reduced degradation, in line with the proposed mechanism of action of CADs.

Based on these considerations, we propose to invert the experimental approach and test whether cellular re-complementation with C18 sphingolipids is sufficient to recapitulate the drug-induced Golgi recovery. We are aware of the technical challenges associated with the targeted delivery of exogenously supplied lipids, particularly given the likelihood that effective rescue would require lipid access to the Golgi apparatus. Based on current knowledge, we anticipate that externally supplied lipids would primarily traffic either to the ER via non-vesicular routes or to endosomes/lysosomes through endocytic uptake. From both locations they could eventually reach to some extent the Golgi. The route from endosomes to Golgi in particular as been intensively studied in the past with the use of fluorescent sphingolipid analogs1,2 and may well work also with native lipids.

Since we are not able to predict in advance which lipid species would be more effective or the optimal delivery strategy, we plan to test re-complementation using C18 sphingomyelin and some of its potential precursors, including C18 ceramide as well as using alternative delivery strategies such as incorporation in liposomes of different formulations and delivery at the plasma membrane with bovine serum albumin or cyclodextrins as carriers.

1. Puri et al., (2001). J Cell Biol.154:535-47 (doi: 10.1083/jcb.200102084)
2. Koivusalo et al.,(2007). Mol Biol Cell. 18:5113-23 (doi: 10.1091/mbc.e07-04-0330)

   In Figure 7D, comparisons should include the LM and HM fractions isolated from WT cells.

RESPONSE:

Wild-type control were included in the figure as requested.

The subcellular fractionation experiment should be repeated using AZL and RAL, the compounds used in organoid experiments, rather than TFPZ, to assess whether similar results are obtained. The compounds used differ across experiments, making it difficult to draw consistent conclusions.

RESPONSE:

We thank the reviewer for this comment and apology for some inconsistencies in the selection of the compounds to highlight in the figures which are mostly remnants of the drug prioritization history over the progression of the project. We tried to make it more consistent in the current version.

In the new version of figure 7D, AZL is substituting TFPZ, while TFPZ data were moved to supplementary figure S19.

Golgi morphology in VPS13B KO cells is reported to recover in NPC1 KD and SMPD1 KD cells, but it is not shown whether SM 36:1, CER 36:1, or other lipid levels also increase or change in these conditions. If Golgi morphology recovery occurs via the same mechanism as with compound treatment, a similar lipid pattern should be observed.

RESPONSE:

We thank the reviewer for this question that allowed us to expand our study including new interesting findings. We agree that this is an important point to strengthen the link between CAD and genetic perturbation effects. Given the availability of several published lipidomic datasets modelling LDS in HeLa and in other cell lines, we decided to perform a re-analysis of those to specifically focus on C18 sphingolipids. We found a relative increase of 36:1 upon depletion of LSD genes in all analyzed datasets for NPC1 and SMPD1, but also for more than 15 other LSD genes including NPC2, recapitulating what we find with all the CAD molecules tested in our study. These changes, were not noticed or at least not discussed by most of the authors. This is not surprising since those studies are focused on different biological questions. We believe that these findings, besides reinforcing our hypothesis of a common mechanism between CAD and NPC1/SMPD1 KO, have of general interest for the regulation of C18 sphingolipids, which are among the relative few lipid species with a bona fide specific protein binding partner and proposed to play a crucial role in Golgi traffic.

MINOR POINTS:

The manuscript lacks sufficient information about the compound library used for screening (number and source of compounds, compound type).

RESPONSE:

We apologize if this information was not sufficiently visible in the original version of the manuscript. The data about source, catalog number, formulation and several additional identifiers is included in the File S1. This is now clearly indicated in the methods so that I can be more easily visible to the readers

Fig. 3A: a WT control image is required.

RESPONSE:

A WT control image is now included in the new version of Figure 3.

Fig. 4: include representative images at concentrations higher than 1.25 µM.

RESPONSE:

Representative images are now included for all concentrations higher than 1.25 µM, as requested.

Abbreviations such as BMP/LBPA should be defined when first mentioned.

RESPONSE:

The abbreviation of BMP/LBPA was already defined when first mentioned in the original version of the manuscript

The abbreviation for raloxifene is inconsistent (RLX vs RAL) and should be unified.

RESPONSE:

Raloxifene is now abbreviated as RLX all over the manuscript.

Fig. 5C: the meaning of the green and magenta bars is not explained.

RESPONSE:

Color code for figure 5C has been included.

The definitions and centrifugation parameters for light and heavy membrane fractions should be clearly stated in the Methods.

RESPONSE:

The centrifugation parameters were already defined in the original manuscript. It is not clear to us, which parameter the Referee is referring to. Below is the sentence in the methods section:

“Gradients were centrifuged at 165,000 g for 1.5 h at 4°C with a SW40Ti Swinging-Bucket rotor (Beckman-Coulter). The LM and HM fractions were collected at the 35%-HB and 35%-40.6% interfaces, respectively”

The concentration and incubation times for BFA and nocodazole should be included in the main text or figure legends.

RESPONSE:

Concentrations and incubation times of BFA and nocodazole were already present in the legend of figure 5.

Fig. 8C, D, G, H: y-axes lack labels and must be defined.

RESPONSE:

Axes labels have now been included

There are multiple typographical errors, including "VPS12" instead of "VPS13B", that should be corrected.

RESPONSE:

We corrected this specific mistake as well as others that we could identify after careful reading of the manuscript.

Reviewer #2 (Significance (Required)):

While the dataset is extensive and technically detailed, the manuscript lacks a clear mechanistic explanation connecting lipid changes to Golgi restoration. The choice and comparison of compounds are inconsistent across experiments, and the interpretation remains speculative. Substantial revision and additional experiments are required before the study can be considered for publication.

Il s'agit donc d'élargir le champ des utilisateurs d'un logiciel libre (ici Pandoc) à une communauté ciblée, pas ou peu familier du libre. C'est un peu le projet derrière Stylo : pandoc (encore) pour les chercheur·es SHS → cheval de troie

English — rung (thorough explanation)

1) What “rung” means (core idea)

A rung is a horizontal step or bar that connects the two sides of a ladder.

Rung = one step of a ladder

You climb a ladder by stepping on its rungs.

2) “Rung” as a metaphor in biology (Science 10 focus)

In biology, rung is often used as a comparison (metaphor) when explaining DNA structure.

• DNA is often described as a twisted ladder
• The sides of the ladder → sugar–phosphate backbones
• The rungs of the ladder → paired nitrogenous bases

📌 Each DNA rung is made of a base pair:

• A–T (adenine–thymine)
• C–G (cytosine–guanine)

These base pairs are held together by hydrogen bonds.

3) What makes up a DNA “rung”

Each rung consists of:

• Two nitrogenous bases
• Joined by hydrogen bonds
• One base from each DNA strand

Example:

One rung = A on one strand + T on the other strand

4) Why the “rung” idea helps understanding

The ladder model helps students visualize that:

• DNA has two strands
• The strands are connected at regular intervals
• The order of rungs carries genetic information

📌 The sequence of rungs = genetic code.

5) Everyday uses of “rung”

• Ladder rung
• A rung on a career ladder (metaphor)
• A rung in a rope ladder

One-sentence exam summary

A rung is a horizontal step of a ladder; in DNA, rungs represent paired nitrogenous bases connecting the two strands.

中文 — rung（梯级 / 横档） 详细解释

1) “rung”的基本含义

Rung 指的是梯子上的横档或踏板，用来踩踏和攀爬。

Rung = 梯子的一格横档

2) 生物学中的 rung（DNA 比喻，重点）

在生物学中，DNA 常被比作一把梯子：

• 梯子的两侧 → 糖—磷酸骨架
• 梯子的横档（rungs）→ 碱基对

📌 每一个 DNA 的“rung”由一对碱基组成：

• A–T
• C–G

3) DNA 横档的作用

• 把两条 DNA 链连接在一起
• 保持双螺旋结构稳定
• 横档的排列顺序储存遗传信息

一句话考试版总结

Rung 指梯子的横档，在 DNA 中用来比喻连接两条链的碱基对。

如果你需要，我可以把 ladder model → rung → base pair → hydrogen bond 做成 Science 10 中英对照图解或互动闪卡，非常适合课堂讲解与复习。

English — Thymine (T) (thorough explanation)

1) What thymine is (core idea)

Thymine is a nitrogenous base found only in DNA. It is one of the four bases that make up the DNA genetic code.

Thymine = a DNA base that pairs with adenine

2) Where thymine is found

• DNA ✅
• RNA ❌ (RNA uses uracil instead)

Each thymine base is part of a nucleotide, attached to:

• Deoxyribose sugar
• Phosphate group

3) Thymine’s base-pairing rule (exam essential)

In DNA:

• Thymine (T) pairs with Adenine (A)
• Held together by 2 hydrogen bonds

This specific pairing:

• Keeps DNA strands aligned
• Allows accurate DNA replication

4) Chemical group of thymine

Thymine belongs to the pyrimidines, which:

• Have a single-ring structure
• Are smaller than purines

Pyrimidines: Cytosine (C), Thymine (T), Uracil (U) Purines: Adenine (A), Guanine (G)

Purine–pyrimidine pairing keeps the DNA double helix a constant width.

5) Role of thymine in DNA

• Thymine helps store genetic information
• The order of T bases (with A, C, G) determines genes
• Correct T–A pairing ensures genetic stability

A mistake involving thymine (a mutation) can:

• Change a gene
• Alter protein structure
• Affect traits

6) Thymine vs uracil (common exam comparison)

| Feature | Thymine (T) | Uracil (U) | | ---------- | ----------- | ----------- | | Found in | DNA | RNA | | Pairs with | Adenine | Adenine | | Stability | More stable | Less stable |

📌 Thymine makes DNA more stable for long-term information storage.

One-sentence exam summary

Thymine is a pyrimidine nitrogenous base found only in DNA that pairs with adenine using two hydrogen bonds.

中文 — Thymine（胸腺嘧啶，T） 详细解释

1) 什么是胸腺嘧啶（核心概念）

胸腺嘧啶（Thymine，T）是一种只存在于 DNA 中的含氮碱基，是 DNA 遗传信息的重要组成部分。

胸腺嘧啶 = DNA 中与腺嘌呤配对的碱基

2) 胸腺嘧啶出现在哪里

• DNA 中有
• RNA 中没有（RNA 用的是尿嘧啶 U）

3) 胸腺嘧啶的配对规则（必考）

DNA 中：

• T ↔ A
• 通过 2 条氢键连接

这种精确配对保证 DNA 复制准确。

4) 胸腺嘧啶的类别

胸腺嘧啶属于 嘧啶（pyrimidine）：

• 单环结构
• 分子较小

5) 胸腺嘧啶的作用

• 参与遗传信息储存
• 确保 DNA 双螺旋结构稳定
• 维持遗传信息长期保存

6) 胸腺嘧啶 vs 尿嘧啶（常考对比）

| 项目 | 胸腺嘧啶 | 尿嘧啶 | | ---- | ---- | --- | | 所在分子 | DNA | RNA | | 配对对象 | 腺嘌呤 | 腺嘌呤 | | 稳定性 | 更高 | 较低 |

一句话考试版总结

胸腺嘧啶是 DNA 中与腺嘌呤配对、具有两条氢键的嘧啶碱基。

如果你需要，我可以把 A–T vs C–G 的稳定性比较或把 四种 DNA 碱基整理成 Science 10 中英对照闪卡 / 互动图表，方便系统复习与教学。

English — Guanine (G) (thorough explanation)

1) What guanine is (core idea)

Guanine is a nitrogenous base found in DNA and RNA. It is one of the four bases that make up the genetic code.

Guanine = a nitrogen-containing base that stores genetic information

2) Where guanine is found

Guanine appears in:

• DNA
• RNA

It is always part of a nucleotide, attached to:

• A sugar (deoxyribose in DNA, ribose in RNA)
• A phosphate group

3) Guanine’s base-pairing rule (exam essential)

In DNA:

• Guanine (G) pairs with Cytosine (C)
• They are held together by 3 hydrogen bonds

In RNA:

• Guanine (G) pairs with Cytosine (C)

📌 Because there are three hydrogen bonds, G–C pairs are stronger and more stable than A–T pairs.

4) Guanine’s chemical group

Guanine belongs to the purines, which:

• Have a double-ring structure
• Are larger than pyrimidines

Purines: Adenine (A), Guanine (G) Pyrimidines: Cytosine (C), Thymine (T), Uracil (U)

This size matching (purine–pyrimidine) keeps the DNA double helix uniform in width.

5) Role of guanine in DNA

In DNA:

• Guanine attaches to the sugar to form a nucleotide
• The sequence of G (with A, T, C) determines genetic instructions
• Accurate G–C pairing ensures correct DNA replication

Changes involving guanine can cause mutations, potentially affecting proteins and traits.

6) Guanine and DNA stability

• Regions with many G–C pairs are more thermally stable
• Such regions often occur in important regulatory areas of DNA

One-sentence exam summary

Guanine is a purine nitrogenous base that pairs with cytosine using three hydrogen bonds in DNA and RNA.

中文 — Guanine（鸟嘌呤，G） 详细解释

1) 什么是鸟嘌呤（核心概念）

鸟嘌呤（Guanine，G）是一种存在于 DNA 和 RNA 中的含氮碱基，是遗传信息的重要组成部分。

鸟嘌呤 = DNA / RNA 中的遗传“字母”之一

2) 鸟嘌呤出现在哪里

鸟嘌呤存在于：

• DNA
• RNA

它与：

• 糖
• 磷酸基团 一起组成 核苷酸。

3) 鸟嘌呤的配对规则（必考）

DNA 中：

• G ↔ C
• 通过 3 条氢键连接

RNA 中：

• G ↔ C

📌 三条氢键使 G–C 配对更加牢固。

4) 鸟嘌呤的类别

鸟嘌呤属于 嘌呤（purine）：

• 双环结构
• 分子较大

对比：

• 嘌呤：A、G
• 嘧啶：C、T、U

5) 鸟嘌呤在 DNA 中的作用

• 与脱氧核糖结合形成核苷酸
• 与胞嘧啶精确配对
• 确保 DNA 复制的准确性

碱基变化可能导致突变。

一句话考试版总结

鸟嘌呤是 DNA 和 RNA 中与胞嘧啶配对、具有三条氢键的嘌呤碱基。

如果你需要，我可以把 A–T vs C–G 的稳定性对比、或把四种碱基做成 Science 10 中英对照闪卡 / 互动图解，直接用于复习或教学。

English — Cytosine (C) (thorough explanation)

1) What cytosine is (core idea)

Cytosine is a nitrogenous base found in DNA and RNA. It is one of the four main bases that make up the genetic code.

Cytosine = a nitrogen-containing base that helps store genetic information

2) Where cytosine is found

Cytosine occurs in:

• DNA
• RNA

It is always part of a nucleotide, attached to:

• A sugar (deoxyribose in DNA, ribose in RNA)
• A phosphate group

3) Cytosine’s base-pairing rule (exam essential)

In DNA:

• Cytosine (C) pairs with Guanine (G)
• They are held together by 3 hydrogen bonds

In RNA:

• Cytosine (C) pairs with Guanine (G)

📌 The three hydrogen bonds make the C–G pair stronger than the A–T pair.

4) Cytosine’s chemical group

Cytosine belongs to the pyrimidines, which:

• Have a single-ring structure
• Are smaller than purines

Purines (double ring):

• Adenine (A)
• Guanine (G)

Pyrimidines (single ring):

• Cytosine (C)
• Thymine (T)
• Uracil (U)

This size matching keeps the DNA double helix uniform in width.

5) Role of cytosine in DNA

In DNA:

• Cytosine attaches to the sugar to form a nucleotide
• The order of cytosine (with A, T, G) determines genetic instructions
• Accurate C–G pairing ensures correct DNA replication

A change in cytosine (mutation) can:

• Alter genes
• Affect proteins
• Change traits

6) Cytosine and genetic stability

Because C–G pairs have three hydrogen bonds:

• Regions rich in C and G are more stable
• They often occur in important regulatory regions of DNA

One-sentence exam summary

Cytosine is a pyrimidine nitrogenous base that pairs with guanine using three hydrogen bonds in DNA and RNA.

中文 — Cytosine（胞嘧啶，C） 详细解释

1) 什么是胞嘧啶（核心概念）

胞嘧啶（Cytosine，C）是一种存在于 DNA 和 RNA 中的含氮碱基，是遗传信息的基本组成单位之一。

胞嘧啶 = DNA / RNA 中的重要碱基

2) 胞嘧啶出现在哪里

胞嘧啶存在于：

• DNA
• RNA

它与：

• 糖（DNA 中是脱氧核糖）
• 磷酸基团 一起构成核苷酸。

3) 胞嘧啶的配对规则（必考）

DNA 中：

• C ↔ G
• 通过 3 条氢键连接

RNA 中：

• C ↔ G

📌 三条氢键使 C–G 配对更牢固。

4) 胞嘧啶的类别

胞嘧啶属于 嘧啶（pyrimidine）：

• 单环结构
• 分子较小

嘌呤（双环）：A、G 嘧啶（单环）：C、T、U

5) 胞嘧啶在 DNA 中的作用

• 与脱氧核糖结合形成核苷酸
• 与鸟嘌呤精确配对
• 保证 DNA 复制的准确性

碱基变化可能导致突变。

6) 一句话考试版总结

胞嘧啶是 DNA 和 RNA 中与鸟嘌呤配对的嘧啶碱基，具有三条氢键。

如果你愿意，我可以把 Adenine / Thymine / Cytosine / Guanine 做成 Science 10 中英对照碱基配对表或互动闪卡，非常适合系统复习与教学。

English — Adenine (A) (thorough explanation)

1) What adenine is (core idea)

Adenine is a nitrogenous base found in DNA and RNA. It is one of the letters of the genetic code.

Adenine = a nitrogen-containing base that carries genetic information

2) Where adenine is found

Adenine appears in several key biological molecules:

• DNA → pairs with thymine (T)
• RNA → pairs with uracil (U)
• ATP → part of the energy molecule used by cells

3) Adenine’s base-pairing rules (exam essential)

In DNA:

• A pairs with T
• Held together by 2 hydrogen bonds

In RNA:

• A pairs with U

These pairing rules ensure accurate DNA replication and correct protein synthesis.

4) Adenine’s chemical group

Adenine belongs to the purines, which:

• Have a double-ring structure
• Are larger than pyrimidines

Purines: Adenine (A), Guanine (G) Pyrimidines: Cytosine (C), Thymine (T), Uracil (U)

This size difference explains why:

• Purine always pairs with pyrimidine
• DNA maintains a uniform width

5) Role of adenine in DNA

In DNA:

• Adenine attaches to deoxyribose sugar
• Becomes part of a nucleotide
• The sequence of adenine (with other bases) determines genetic instructions

Changing adenine’s position can:

• Alter genes
• Cause mutations
• Affect traits

6) Adenine in energy (ATP connection)

Adenine is part of ATP (adenosine triphosphate):

• Adenine + ribose = adenosine
• Adenosine + 3 phosphates = ATP

ATP provides energy for:

• Muscle contraction
• Active transport
• Chemical reactions

One-sentence exam summary

Adenine is a purine nitrogenous base that pairs with thymine in DNA and with uracil in RNA.

中文 — Adenine（腺嘌呤，A） 详细解释

1) 什么是腺嘌呤（核心概念）

腺嘌呤（Adenine，A）是一种含氮碱基，存在于 DNA 和 RNA 中，是遗传信息的“字母”之一。

腺嘌呤 = DNA / RNA 中的重要遗传碱基

2) 腺嘌呤出现在哪里

• DNA：与 胸腺嘧啶（T）配对
• RNA：与 尿嘧啶（U）配对
• ATP：能量分子的重要组成部分

3) 腺嘌呤的配对规则（必考）

DNA 中：

• A ↔ T（2 条氢键）

RNA 中：

• A ↔ U

这些规则保证了遗传信息的准确复制和表达。

4) 腺嘌呤的类别

腺嘌呤属于 嘌呤（purine）：

• 结构为双环
• 体积较大

嘌呤：A、G 嘧啶：C、T、U

5) 腺嘌呤在 DNA 中的作用

• 与脱氧核糖结合
• 构成核苷酸
• 其排列顺序决定遗传信息

碱基变化可能导致突变。

6) 腺嘌呤与能量（ATP）

腺嘌呤是 ATP（三磷酸腺苷）的一部分：

• 为细胞活动提供能量

一句话考试版总结

腺嘌呤是 DNA 中与 T 配对、RNA 中与 U 配对的嘌呤碱基。

如果你需要，我可以把 Adenine / Thymine / Cytosine / Guanine 做成 Science 10 中英对照碱基配对表或互动闪卡，方便系统复习。

English — nitrogenous base (thorough explanation)

1) What a nitrogenous base is (core idea)

A nitrogenous base is a nitrogen-containing molecule that is part of a nucleotide, the building block of DNA and RNA.

Nitrogenous base = the “letter” of the genetic code

Each nucleotide has:

1. A phosphate group
2. A sugar
3. A nitrogenous base

The sequence of bases stores genetic information.

2) The five nitrogenous bases (must know)

In DNA:

• Adenine (A)
• Thymine (T)
• Cytosine (C)
• Guanine (G)

In RNA:

• Adenine (A)
• Uracil (U) (replaces thymine)
• Cytosine (C)
• Guanine (G)

📌 Only the bases change; the sugar–phosphate backbone stays the same.

3) Two base groups: purines vs pyrimidines

Purines (two rings):

• Adenine (A)
• Guanine (G)

Pyrimidines (one ring):

• Cytosine (C)
• Thymine (T)
• Uracil (U)

This size difference explains base pairing rules.

4) Base pairing rules (DNA vs RNA)

DNA base pairing:

• A pairs with T (2 hydrogen bonds)
• C pairs with G (3 hydrogen bonds)

RNA base pairing:

• A pairs with U
• C pairs with G

Hydrogen bonds hold paired bases together, forming the double helix in DNA.

5) Why nitrogenous bases matter

Nitrogenous bases:

• Store genetic information
• Determine proteins made by a cell
• Allow DNA to replicate accurately
• Enable RNA to carry instructions for protein synthesis

Changing a base (a mutation) can change traits.

6) Base vs nucleotide (common confusion)

• Nitrogenous base → one component (A, T/U, C, G)
• Nucleotide → base + sugar + phosphate

Bases encode information; nucleotides build the molecule.

One-sentence exam summary

Nitrogenous bases are nitrogen-containing molecules in DNA and RNA whose sequence stores genetic information.

中文 — nitrogenous base（含氮碱基） 详细解释

1) 什么是含氮碱基（核心概念）

含氮碱基是含有氮原子的分子，是 DNA 和 RNA 的核苷酸组成部分之一。

含氮碱基 = 遗传信息的“字母”

每个核苷酸由：

1. 磷酸
2. 糖
3. 含氮碱基 组成。

2) 五种含氮碱基（必背）

DNA 中：

• 腺嘌呤（A）
• 胸腺嘧啶（T）
• 胞嘧啶（C）
• 鸟嘌呤（G）

RNA 中：

• 腺嘌呤（A）
• 尿嘧啶（U）
• 胞嘧啶（C）
• 鸟嘌呤（G）

3) 含氮碱基的两大类

嘌呤（双环）：

• A、G

嘧啶（单环）：

• C、T、U

4) 碱基配对规则（考试重点）

DNA：

• A ↔ T
• C ↔ G

RNA：

• A ↔ U
• C ↔ G

这些配对通过氢键连接，使 DNA 形成双螺旋结构。

5) 含氮碱基的重要性

• 决定遗传信息
• 指导蛋白质合成
• 保证 DNA 复制准确
• 碱基改变会导致突变

一句话考试版总结

含氮碱基是 DNA 和 RNA 中储存遗传信息的关键成分。

如果你需要，我可以把 nitrogenous base → nucleotide → DNA → gene 做成 Science 10 中英对照速记卡或互动图解，非常适合系统复习。

English — Sugar in DNA (thorough explanation)

1) What the sugar in DNA is

The sugar found in DNA is called deoxyribose. It is a five-carbon sugar (a pentose) and is one of the three essential parts of a DNA nucleotide.

DNA sugar = deoxyribose

Each DNA nucleotide contains:

1. A phosphate group
2. Deoxyribose sugar
3. A nitrogenous base (A, T, C, or G)

2) Why it’s called deoxyribose

• “Deoxy-” means missing an oxygen
• Deoxyribose has one less oxygen atom than ribose (the sugar in RNA)

📌 This small difference makes DNA:

• More stable
• Better for long-term information storage

3) What the sugar does in DNA (key functions)

A) Forms the backbone

• Deoxyribose links to phosphate groups
• Together they form the sugar–phosphate backbone
• This backbone gives DNA its shape and strength

B) Connects to bases

• Each sugar attaches to one nitrogenous base
• The sequence of bases carries genetic information
• The sugar itself does not code information, but holds it in place

4) How sugars link DNA together

• The sugar of one nucleotide bonds to the phosphate of the next
• This creates a long chain called a polynucleotide
• The bonds are called phosphodiester bonds

Two sugar–phosphate backbones twist together to form the double helix.

5) DNA sugar vs RNA sugar (common exam comparison)

| Feature | DNA | RNA | | ------------------- | ------------------ | ------------------ | | Sugar | Deoxyribose | Ribose | | Oxygen at 2′ carbon | ❌ Missing | ✅ Present | | Stability | More stable | Less stable | | Function | Store genetic info | Help make proteins |

6) Why sugar matters (big picture)

Without the sugar:

• DNA nucleotides could not link
• DNA would fall apart
• Genetic information could not be stored or copied

Sugar = the structural “frame” that holds DNA together

One-sentence exam summary

The sugar in DNA is deoxyribose, which forms the sugar–phosphate backbone and supports the structure of the DNA molecule.

中文 — DNA 中的糖（详细解释）

1) DNA 中的糖是什么

DNA 中的糖叫 脱氧核糖（deoxyribose），是一种五碳糖。

DNA 的糖 = 脱氧核糖

每个 DNA 核苷酸由三部分组成：

1. 磷酸基团
2. 脱氧核糖
3. 含氮碱基（A、T、C、G）

2) 为什么叫“脱氧”核糖

• “脱氧”表示 少一个氧原子
• 脱氧核糖比 RNA 中的核糖 少一个氧

📌 这使 DNA：

• 更稳定
• 适合长期储存遗传信息

3) 糖在 DNA 中的作用（重点）

① 构成骨架

• 脱氧核糖与磷酸交替连接
• 形成 糖—磷酸骨架
• 为 DNA 提供支撑和形状

② 连接碱基

• 每个糖连接一个碱基
• 碱基顺序决定遗传信息
• 糖本身不存信息，但固定信息

4) DNA 是如何连成链的

• 一个核苷酸的糖
• 与下一个核苷酸的磷酸相连
• 形成 磷酸二酯键

两条这样的链相互缠绕，形成 DNA 双螺旋结构。

5) DNA 糖 vs RNA 糖（常考对比）

| 项目 | DNA | RNA | | --- | ------ | ----- | | 糖 | 脱氧核糖 | 核糖 | | 氧原子 | 少一个 | 多一个 | | 稳定性 | 高 | 低 | | 功能 | 储存遗传信息 | 蛋白质合成 |

一句话考试版总结

DNA 中的糖是脱氧核糖，它与磷酸一起形成 DNA 的骨架结构。

如果你需要，我可以把 nucleotide → sugar → phosphate → DNA backbone 做成 中英对照闪卡或可交互 HTML 图解，直接用于 Science 10 复习或教学。

English (thorough explanation with images)

1) What a chromosome is

A chromosome is a highly condensed structure of DNA and proteins found in the nucleus of eukaryotic cells. Chromosomes carry genes, which contain the instructions for building and maintaining an organism.

Simply: chromosome = tightly packed DNA that holds genes.

2) What chromosomes are made of

Chromosomes consist of:

• DNA (the genetic code)
• Proteins, mainly histones, which help DNA coil and fold

DNA + histones together form chromatin. When chromatin coils up tightly (especially during cell division), it becomes a chromosome.

3) Chromatin vs chromosome (very important)

| Term | State | When you see it | | -------------- | ------------------ | --------------------------------- | | Chromatin | Loose, uncondensed | Interphase (normal cell activity) | | Chromosome | Tightly condensed | Mitosis / Meiosis |

👉 You usually see chromosomes only during cell division.

4) Structure of a replicated chromosome

When a chromosome has been copied (after DNA replication), it looks like an “X” shape:

• Two sister chromatids → identical copies of DNA
• Centromere → region holding sister chromatids together
• Telomeres → protective caps at chromosome ends

Each chromatid contains one complete DNA molecule.

5) Why chromosomes condense

Condensation helps:

• Prevent DNA from breaking
• Prevent tangling
• Ensure accurate separation during cell division

Loose DNA would be impossible to divide correctly.

6) Chromosome number (species-specific)

Each species has a fixed chromosome number.

Examples:

• Humans: 46 chromosomes (23 pairs)

• 22 pairs of autosomes

• 1 pair of sex chromosomes (XX or XY)

This arrangement can be seen in a karyotype.

7) Role in cell division

• Mitosis → chromosomes ensure identical cells
• Meiosis → chromosomes allow genetic variation and formation of gametes

Correct chromosome behavior is essential for life.

8) One-sentence exam definition

A chromosome is a condensed DNA–protein structure that carries genes and ensures accurate DNA distribution during cell division.

中文（配图·深入讲解）

1）什么是染色体（chromosome）

染色体是存在于真核细胞细胞核中、由 DNA 高度压缩形成的结构， 它们携带基因，决定生物的性状和功能。

一句话：

染色体 = 高度压缩的 DNA 信息载体

2）染色体的组成

染色体由：

• DNA
• 蛋白质（主要是组蛋白）

组成。

DNA + 组蛋白 = 染色质 染色质高度凝缩后 → 染色体

3）染色质 vs 染色体（重点）

| 名称 | 状态 | 出现时间 | | ------- | ---- | ------- | | 染色质 | 松散 | 间期 | | 染色体 | 高度压缩 | 有丝/减数分裂 |

👉 平时细胞里看到的是染色质，而不是染色体。

4）复制后染色体的结构（X 形）

复制后的染色体通常呈 X 形，由：

• 两条姐妹染色单体

• DNA 完全相同

• 着丝粒

• 连接两条染色单体

• 端粒

• 保护染色体末端