Ir para a Avaliação PREreview

Avalilação PREreview de Mammalian spermatogenesis at single-cell resolution downstream of genetic perturbations in meiotic double-strand break repair

Publicado
DOI
10.5281/zenodo.24092782
Licença
CC BY 4.0

The following is a peer review written by a trainee enrolled in HHMI’s Transparent & Accountable Peer Review (TAP) training. As part of the workshop, I have selected and reviewed a preprint with support from my research mentor, who has provided supervision and feedback.

Preprint citation:

Mammalian spermatogenesis at single-cell resolution downstream of genetic perturbations in meiotic double-strand break repair

I. Agarwal, B. Myers, M. Houlard, A. Hinch, E. Bitoun, S. R. Myers

bioRxiv 2026.05.14.724910; doi: https://doi.org/10.64898/2026.05.14.724910

Disclosure:

The contents and conclusions of this preprint rely heavily on single-cell ATAC- and RNA-sequencing. The reviewer is not an expert in the technical details or bioinformatic strategies appropriate for this type of data, and therefore this review should not be considered a rigorous evaluation of the methodological strategies used.

Overall assessment:

Agarwal et al., uses established inbred mice models of mixed genetic backgrounds (B6, CAST, and PWD) to evaluate the effect of PRDM9 binding asymmetry on spermatogenesis using combined single-cell ATAC- and RNA-seq as a functional readout of chromatin and transcription. PRDM9 targets DNA double strand break (DSB) formation in meiosis to allele-specific motifs. When PRDM9 alleles from one species (or subspecies of inbred mice in this case) are exposed to an evolutionarily divergent genome which has not been subject to allele-specific motif erosion, allele-specific PRDM9 binding prefers to bind to the “naïve” sequences. It is within this context that the authors devise a system to further evaluate the effects asynapsis, PRDM9 asymmetric binding, and genetic variation outside of hotspot motifs on spermatogenesis. This is achieved through the generation of a 23-mouse cohort that possess a PWD maternal genome and a random mixture of CAST/B6 paternal genomes. Crucially, this cohort all possess the same two PRDM9 alleles: PRDM9B6/PWD. Symmetric PRDM9 binding in this system will increase with higher paternal CAST genome composition specifically via PRDM9B6. Thus, different degrees of PRDM9 asymmetry over broad genomic scales is achieved via the generation of varying degrees of CAST paternal genome inheritance.

This reviewer finds this genetic system and the evaluation of these mice via dual scATAC- and scRNA-seq remarkably clever, and the resulting findings to be highly informative and novel for the reasons outlined below. Overall, this paper represents a significant advance forward in the field of PRDM9 biology and meiosis. It makes important contributions both methodologically and conceptually for the following reasons:

1. Intermediate phenotypes associated with varying levels of CAST/B6 genome composition has never been explored previously. The titration of CAST genome into the otherwise sterile PWD x B6 F1 hybrid confirms and expands on previous understanding regarding hybrid sterility and the contribution of PRDM9 binding asymmetry therein.

2. scRNA-seq data shows clear signatures of transcriptional silencing associated with unsynapsed chromatin in meiosis which strongly supports higher rates of asynapsis with increased PRDM9 asymmetric binding (at least on short chromosomes – see major point #1).

3. Deriving ploidy from scATAC-seq data, scRNA-seq data stages cells as in a more mature transcriptional program relative to the physical status of their ploidy.

4. A particular locus on chromosome 15 which contains Dmc1 and Mei1 predicts asynapsis status in opposition of the asymmetric PRDM9 model. This represents an interesting area for future study, as the implication from these observations is that MEI1 may preform different non-substitutable molecular roles in B6 vs CAST genetic backgrounds.

5. Mapping crossover rates, the data supports the conclusion that crossovers track with PRDM9 symmetry. Crossovers are enriched in CAST containing genome segments, and higher asynapsis rates leads to higher crossover enrichment in CAST vs B6 backgrounds. Confounding factors such as chromosome size may complicate these interpretations (see Major point #2).

The genetic system used (PWD x (CAST x B6)) is particularly innovative and the single cell methods are well justified within the framework of the study design. Importantly, previous aspects of PRDM9 binding asymmetry are well recapitulated within this data, providing a rigorous internal control for the subsequent experiments. The only weakness I can note is to look at the contribution of chromosome size and its influence on the proposed models and some minor textual/figure clarity points. Below I have outlined major and minor points but do not ask for any additional experiments, rather additional analysis and data interpretations to be explored. This reviewer hopes these suggestions strengthen the claims of this work. Recommended cosmetic changes are italicized for clarity.

Major points:

  1. Transcriptional silencing is used as an indirect measurement of chromosome asynapsis. Using this method, figure 3E shows that higher B6 paternal genome composition leads to higher inferred asynapsis rates and this claim is generalized across all chromosomes. However, looking at this figure, it appears that the longer chromosomes (specifically 1-14) show negligible differences in inferred asynapsis compared with shorter chromosomes (15-19). This is important, as it supports previous work demonstrating that shorter chromosomes have outsized sensitivity to asynapsis as a result of asymmetric hotspots (1. Kauppi et al, 2013, Genes & Development, PMID: 23599345. 2. Gregorova et al, 2018, eLife, https://doi.org/10.7554/eLife.34282). As no evidence currently suggests that symmetric hotspots are more likely to experience a DSB than asymmetric hotspots (beyond chromatin compatibility considerations), perhaps an alternative explanation could be that DSB alterations (either formation or repair) have an outsized impact on shorter chromosomes and result in fewer chances for homolog engagement (via the assumption that symmetric PRDM9 binding promotes homolog engagement) as a product of chromosome length. It will be important to generate a statistical model evaluating or simulating the contribution of hotspot symmetry/asymmetry specifically in the context of chromosome length to clarify if the interpretation of this data should be that symmetric PRDM9 binding matters specifically to short chromosomes or to all chromosomes equally. This will clarify if the authors current claims are influenced by chromosome size and may also lead to new discovery.

  2. A model is proposed about crossovers, specifically that crossover designation is likely to occur during early pachytene as this would explain why CAST crossovers are preferentially enriched on chromosomes which have higher rates of asynapsis. This model assumes that partial synapsis is made in early meiosis via symmetric PRDM9 binding on the PWDxCAST homologs and that these DSBs are funneled into a crossover decision pathway which occurs early (presumably the authors are implying at late zygotene/early pachytene). Similar to Major point #1, in figure 5K there appears to be little to no difference between the crossover rate in the CAST background between chromosomes 1-14, with a jump from 15-19. Incorporating the variable of chromosome size via statistical modeling and simulations would greatly improve the rigor of the analysis and allow inferences to be made regarding length-dependent/-independent effects. In turn, this would allow extrapolation to biology such as crossover timing as suggested by the authors.

Minor points:

  1. In the introduction, 2n replication results in 4c not 4n chromosomes. Correction will ensure technical preciseness, as 4n refers to ploidy and 4c refers to DNA content.

  2. Can the ATAC-seq reads be shown from PWD x CAST homologs (similar to 2C)? This would nicely visualize the symmetric nature of the data. It is useful to compare these side by side such as in Figure S6 which will emphasize the authors point about asymmetric binding being confined to PWDxB6 genomic loci.

  3. It is not clear from Figure 1 or the abstract that the B6 x CAST mouse is bringing in only the PRDM9B6/B6 allele. This detail is key to emphasize in texts and figures early, as it is central to the design of the study.

  4. Indicating the PRDM9 allele present in the experimental genotype as a legend in corresponding figures throughout the paper would improve interpretation and clarity.

  5. What is the correlation between the fertility measurements (TW, sperm count) and CAST genome % composition? Can this simple correlation (Spearman, Pearson, etc) between genome composition and fertility measurements be measured and justified in the context of the proposed model? This offers a sanity check on the assumption that fertility tracks with CAST genome composition.

  6. The justification for calling class differences between “likely replicating” and other pre-division cells is not clear, or why this distinction matters for the claims. It may be more straightforward to remove the “likely replicating” cells from the analysis unless there is a specific claim tied to them.

  7. ATAC-seq data at hotspots looks great (2B), but this data is binned by pseudotime (likely L/Z stage). Can the authors show the chromatin accessibility across more pseudotime points? PreL, L, Z, P, etc. This would be interesting for readers to compare with other published datasets to see if similar conclusions can be drawn about chromatin accessibility and timing (e.g. Lam et al., 2019, Nat Comm, PMID: 31444359 shows peak hotspot chromatin modifications in zygotene which may track with hotspot chromatin architecture).

  8. Even though asynapsis predicts MSUC, it may or may not be true that MSUC predicts asynapsis. Cytologically fully synapsed chromosomes have been reported to be in the sex body in pachytene-like cells (see Kim et al, 2025, Nature Communications, https://doi.org/10.1038/s41467-025-57928-x) which could also result in transcriptional silencing. Could this limitation be addressed in the discussion section of the text?

Competing interests

The author declares that they have no competing interests.

Use of Artificial Intelligence (AI)

The author declares that they did not use generative AI to come up with new ideas for their review.