On 2025-10-09 07:16:25, user Ralf H. Adams wrote:
In a recent preprint at bioRxiv (Yang et al., https://doi.org/10.1101/2025.10.02.679940) , a group of authors led by Dr Anjali Kusumbe challenges an article from my lab published in 2024 (Koh et al. 2024, PMID: 39537918). The new preprint refers to Extended Data Figures and Supplementary videos that are unfortunately not provided at bioRxiv. Nevertheless, the manuscript text and the 5 main figures plus a proposed model contain a couple of major issues that I will cover here. A thread containing my comments plus useful illustrations is available on X ( https://x.com/ralfhadams/status/1976179960558539006 ).
But before I do so, let me give you a little background. Several papers over the recent years have proposed that the bone marrow (BM) in skull is specialized and acts as an immune cell reservoir for the brain parenchyma and the meninges, a multilayered tissue structure that encloses and protects the brain and spinal cord. Key evidence comes from landmark publications by the groups of Matthias Nahrendorf (Herisson et al. 2018, PMID: 30150661), Jonathan Kipnis (Cugurra et al. 2021, PMID: 34083447), Ali Ertürk (Kolabas et al. 2023, PMID: 37562402) and others. Kolabas et al. 2023, for example, provide compelling results showing that skull “has the most distinct transcriptomic profile compared with other bones in states of health and injury” with potential relevance for neurological pathologies. The same publication also reveals that skull shows a strong response to stroke, arguing for interactions between the brain and a nearby bone marrow compartment.
Our paper from 2024 shows that the BM of skull is undergoing substantial expansion in adult life but also during aging, which applies both to stromal cells (including vasculature) and a large range of hematopoietic cell subsets including stem and progenitor populations. Our paper makes use of well-established immunolabeling and imaging enabled by the injection of antibodies into living animals before they were sacrificed for tissue isolation. The removal of external tissues, including the meninges/dura mater on the brain side and dermal tissue on the skin side enabled undisturbed insights into the organization and expansion of bone marrow vessels, which exhibit the distinct dilated and irregular morphology typical for the sinusoidal vasculature of the BM. It is also noteworthy (and will be relevant later) that the expansion of the BM vasculature starts in frontal and parietal bones at the edges (near the sutures) and gradually progresses toward to center in young adult mice (10-14 weeks), whereas the interparietal bone is already filled with vessels (and hematopoietic cells) at this stage.
Apart from imaging, we provide an extensive set of flow cytometry data in a large (22 page) supplemental file accompanying the paper, which confirms the expansion of stromal, endothelial, total hematopoietic cells and various hematopoietic cell subsets during adulthood and aging but also in response to various challenges (such as pregnancy and stroke).
But even the skull of older mice is not spared from signs of aging. We observe that the expression of pro-inflammatory cytokines is elevated in old skull in comparison to samples from younger mice. However, this affects fewer cytokines and involves lower levels of upregulation relative to age-matched (young and old) femur samples.
Functionally, we show that shielding of either the head or the hindlimb allows the survival of young mice exposed to a lethal dose of irradiation (without BM transplantation). In irradiated old mice, however, only shielding of the head ensures survival, whereas animals with shielded legs gradually succumb over a period of 200 days, presumably reflecting the exhaustion of hematopoietic stem cells and early progenitor populations. These results and other evidence presented in the article led us to the conclusion that skull BM is more resilient to aging than its counterpart in femur.
Now, let’s take a look at some of the claims raised by Yang et al. in their recent bioRxiv preprint. In Figure 1, they have reanalyzed selected subsets of scRNA-seq data from our publication to show that samples from old skull show hallmarks of aging compared to young skull. It is not clear (and is not explained) why certain samples from our data (available at GEO under the accession number GSE275179) were included in their analysis whereas other, equally suitable samples were excluded. In any case, it is not surprising and also not controversial that cells from old skull show an increase of aging markers compared to samples from young mice. We have never claimed that the skull BM stays young forever and, instead, we have reported the upregulation of pro-inflammatory cytokines mentioned earlier together with other evidence such as myeloid-biased hematopoietic differentiation.
Figure 2 of the Yang et al. preprint shows a comparison of mass spectrometry data from young (10 weeks) and old skulls/calvaria (79 weeks), which also indicates that markers of DNA damage and inflammation are upregulated, whereas markers of angiogenesis, osteogenesis and mitochondrial activity are downregulated. One would have liked to see a more detailed description of the sample preparation (e.g., were the meninges and other adherent/external tissues removed from the skull samples?), but the findings themselves are not surprising or controversial. Of course, old tissue samples show evidence of aging and are different from young skull samples.
Things heat up a little in Figure 3 of the Yang et al. preprint. The authors claim that a dense, continuous network of vessels is found throughout skull samples from all ages and thereby challenge our observation that the calvarial vasculature and BM are dynamically expanding in adult mice.
The evidence provided by Yang et al. includes transversal sections through the skull roof. The low magnification overview images in Figure 3b are of poor quality/resolution so that it is not possible to see much detail. The DAPI nuclear staining in the higher magnification images, however, suggests that the enlarged areas primarily correspond to tissue near the sutures, which already contain vasculature and some BM even in young frontal and parietal bones. Clearly, one would like to see better quality overview images and an unbiased comparison of central vs. peripheral areas from frontal and parietal bones.
Things get more interesting and controversial in Figure 3c and d. Yang et al. use tissue clearing and light sheet microscopy to show the presence of a “continuous, dense vascular network throughout the skull frontal, parietal and interparietal regions in young mice”. The CD31 immunostaining in Fig. 3a, however, shows predominantly the meningeal vasculature, easily identifiable by hallmarks such as the sagittal and transverse sinuses consisting of large diameter blood and lymphatic vessels. Periosteal vessels might be also labeled, but it is not obvious that any sinusoidal vessels of the bone marrow are stained in this sample.
Higher magnification images in Fig. 3d show more details of the meningeal (and perhaps also periosteal) vasculature in mice from different age groups. Again, it is very obvious that sinusoidal vessels with their distinctive morphology and large caliber are not captured in these samples. This is not surprising because expression of CD31 is low in sinusoidal (also termed type L) endothelial cells, as previous work from my group has established (Kusumbe et al. 2014, PMID: 24646994).
Taken together, it is clear that Yang et al. have not carefully separated the vasculature of the skull and the adjacent meninges, leading to confusing results and wrong conclusions. Future research in this important field needs to clearly distinguish what is bone marrow, periosteal tissue, meninges or the surface of the adjacent brain.
There is another small but interesting nugget hidden in Fig. 3c of the Yang et al. preprint. The freshly dissected skull sample prior to clearing shows distinctive areas of red blood cells in the interparietal bone but also in parts of the frontal bone. Our work has revealed a similar pattern of red blood cell distribution, representing regions of bone marrow, in freshly dissected young skull. In samples from older mice, these areas increase substantially, confirming the expansion of calvarial BM even in the absence of any immunostaining. Kolabas et al. 2023 (PMID: 37562402) also show a strikingly similar distribution of Nr4a1+ and propidium iodide (PI) labeled immune cells in calvarial bone marrow of young adult mice. Note the absence of immune cells in central regions of the frontal and parietal bone, whereas labeled cells (blue and red) are abundant in the interparietal bone. This is totally consistent with our own findings.
Unfortunately, the sample showing aged skull is not included in the Yang et al. bioRxiv preprint so that a further comparison of young and old bone marrow will have to wait until the Extended Data Figures become publicly available.
Before I move on, I would briefly like to address a technical detail. In the Yang et al. bioRxiv preprint, the authors claim that our Endomucin staining shows “aberrant nuclear localization”, which, according to them, suggests “that the current staining is artefactual and calls into question the reliability of their vascular imaging data”.
This is just one of many strongly worded claims in the preprint. In reality, however, the Endomucin immunostaining shown by Yang et al. in the preprint is strongly overexposed, preventing any insight into the actual intracellular distribution of the antigen. The red (Endomucin) signal covers pretty much every part of the stained cells. The impact of different imaging and data presentation modalities, whole-mount vs. tissue sections and maximum intensity projection vs. isolated optical planes, also need to be considered.
Furthermore, I have already mentioned that our study has used an in vivo labeling approach involving the injection of antibodies before the animals were sacrificed. This might lead to a slightly different pattern in comparison to post-fixation staining. Nevertheless, it is unambiguously clear that the bone marrow vasculature is stained reliably in Koh et al. 2024 (PMID: 39537918), revealing striking details of vessel architecture and regional differences.
Before I continue with the next figure of the preprint, let’s have a quick look what other recent publications say about skull bone marrow.
Chang et al. 2025 (PMID: 40970910), using advanced tissue clearing and light sheet imaging, report bone marrow expansion in a comparison of 2-month-old and 2-year-old skull samples. The data shown in Figure 5 of their publication independently confirms that large parts of the young calvarium are devoid of VEGFR3+ sinusoidal vessels, whereas older samples show a profound expansion of vessels, consistent with our own findings.
A recent publication from the group of Warren Graham (Horenberg et al. 2025, PMID: 39984434) reports that “Emcn+ vessels demonstrated drastic morphological changes with aging, as bone marrow-resident sinusoidal blood vessel signal increases with age”.
A recent publication from the group of Shukri Habib (Reeves et al. 2024, PMID: 39692737) has evaluated calvaria from mouse of different age groups by micro-computed tomography and other methods. In Figure 1A+B and the corresponding text, they state that “overall thickness from outer surface to inner surface of the parietal bone significantly increased with age” and they also show a striking increase of cavities (which, as we know, represent bone marrow areas) compared to young and adult calvarial bone.
Overall, there is more and more emerging evidence supporting that the bone marrow in adult skull is expanding dynamically in adult mice, which involves changes in sinusoidal vessels but also in stromal and hematopoietic cell populations.
Let’s return to the preprint by Yang et al. with a focus on Figure 4, which is another reanalysis of scRNA-seq data from Koh et al. 2024 (PMID: 39537918). This time, the focus is initially on hematopoietic stem and progenitor cells (HSPCs) and CD45+ cells from young and old femur samples. According to the figure legend, this analysis is based on three samples – one from young and two from old femurs – that were subjected to the removal of hematopoietic cells prior to single cell sequencing. This depletion method is necessary to capture and enrich stromal cell populations, but these are obviously not samples that one would use for the analysis of hematopoietic cells. The remaining hematopoietic cells are residual contaminating cells that have escaped the depletion step, which is never totally complete.
I would dispute that the analysis of contaminating cells allows meaningful conclusions about the abundance/enrichment of hematopoietic cell populations in aging femurs, but this seems to be exactly what Yang et al. have done. Could they have selected the wrong samples from our deposited data or mixed up the sample numbers? Who knows? In this context, I would like to point out that we had provided a substantial amount of information about our scRNA-seq data, the pretreatment of the different samples and other aspects of our methodology per email to Dr Kusumbe in early June 2025.
Back to the preprint and Figure 5, which provides bulk mass spectrometry data comparing proteins from old vertebra and skull, which led the authors of the preprint to the conclusion that vertebrae are relatively protected from aging hallmarks in comparison to skull. The description of the methodology is unfortunately very brief. It is not clear which vertebrae were analyzed exactly (cervical, thoracic, lumbar, sacral, or all?). It is also not clear whether adherent muscle, fat, spinal cord and intervertebral discs were removed, all of which should influence substantially what is seen in the proteomic profile. In any case, it is obvious that this analysis is not confined to bone marrow and functional data for BM cells from vertebrae in comparison to other BM compartments is lacking. Obviously, future work will have to address how aging processes affect different BM compartments and thereby the function of the hematopoietic system.
We are almost done, but I want to highlight another controversial aspect in the Yang et al. preprint, namely the presence (or absence) of lymphatic vessels inside bone. Based on the reanalysis of our scRNA-seq data from Koh et al., it is claimed in Fig. 1 and the accompanying text that lymphangiogenic markers are downregulated in aged skull samples. I would dispute that the presence of Cdh5+ Pecam1+ Prox1+ Ptprc- Flt4+ endothelial cells in the data supports the presence of lymphatic endothelial cells inside the skull bone or marrow. Chang et al. 2025 (PMID: 40970910), a publication that was already mentioned earlier, have examined this question in detail and they state that lymphatic vessels are present outside the skull in the periosteum but not inside bone marrow.
The preparation of samples of skull and other tissues for scRNA-seq analysis is a race against time because it is essential to preserve RNA integrity as much as possible. This means that the removal of adherent, periosteal tissue residues including lymphatic vessels from the meningeal or the dermal side has to be done quickly, is most likely incomplete and might lead to residual contaminations. Essentially, the proposed presence of putative lymphatic endothelial cells in our data, which we have not validated ourselves, cannot resolve where these cells originate from.
I want to conclude here and thank all of you who have read this thread till the end. As I have not been able to cover all issues in the preprint, I am considering another detailed thread once the Extended Data becomes public.
I also hope that I have remained factual and sufficiently polite throughout my posts even though I perceive the language in the Yang et al. preprint as sometimes overly harsh and many claims unsubstantiated and wildly exaggerated.
Final full disclosure: I am not only the last author of Koh et al. 2024 (PMID: 39537918) but also the former postdoctoral supervisor of the two last authors of the Yang et al. preprint.