Escrever um comentário

The study asks whether DNA replication in the early mouse embryo serves only to produce cells, or whether it also helps establish the chromatin states that govern stage-specific gene expression during zygotic genome activation (ZGA) and the first lineage decisions.

The authors block S phase with aphidicolin at three windows: zygote to E2C, E2C to L2C, and 4C to 8C. Embryos stay at the same cell number but still transcribe (EU incorporation, Fig. 2), compact and polarize (Fig. 3), and express lineage markers TEAD4 and GATA3 at 8C levels (Fig. 3). In RNA-seq, aphidicolin-treated embryos resulted in a subset of genes strongly upregulated after replication arrest (Fig. 4, 5). These genes are low-expressed, depleted of accessible chromatin, enriched in late-replicating domains, and carry Polycomb-associated features in published datasets (Fig. 5). Although aphidicolin-treated L2C embryos show reduced H3K27me3 and H3K9me3 immunofluorescence (Fig. 5), embryos after washout blocked at 4C recover to blastocysts with normal ICM/TE ratios, while embryos blocked at the 2-cell stage mostly fail to survive (Fig. 6).

The manuscript is clearly written, and the figures follow the logic of the text. The question is important. Classical work showed that early developmental events can proceed without cell division, but the idea that replication itself builds repressive chromatin at ZGA has had little direct testing in mammals. The genome-wide RNA-seq in both windows, the careful treatment of the common aphidicolin response, and the integration with replication-timing, ATAC and ChIP-seq datasets are real strengths. The finding that aphidicolin-treated 2-cell embryos derepress a distinct set of Polycomb-associated developmental genes, and that this is stage-specific, is a valuable result for the ZGA and heterochromatin-establishment fields. However, some conclusions go further than the data currently allow. Addressing the points below would substantially strengthen the conclusions of this manuscript.

Major comments

1. Aphidicolin specificity: separating the requirement for replication from replication stress and DNA damage. Lines 218–232 and 507–518 argue that aphidicolin does not cause a uniform stress response, based on only 43 shared DEGs across the four comparisons (Fig. S2I) and on the absence of DNA-damage-repair transcript signatures in E2C embryos. Lines 699–705 then propose that persistent ATR signaling may explain the poor survival of 2-cell embryos. Transcript signatures are a weak readout in 2-cell embryos, where the authors themselves note that ATM-dependent responses are not yet active (lines 516–518). Therefore, an undetectable RNA signature does not definitively exclude replication stress. Furthermore, using a single inhibitor and concentration (5 µg/mL) leaves open the possibility that the changes in cluster #1 gene expression and histone modifications stem from stalled-fork toxicity or off-target consequences rather than a strict biological requirement for replication. To rule out these effects, consider the following: (1) Add immunostaining for γH2AX and a replication-stress readout (pCHK1 or pRPA) in control versus aphidicolin-treated E2C-to-L2C embryos, and report it in the Results next to Fig. 2B. (2) Reproduce the key readouts (cluster #1 genes by RT-qPCR, and the H3K27me3/H3K9me3 immunofluorescence of Fig. 5D,E) with an orthogonal replication block, such as a CDC7 inhibitor (for example XL413 or PHA-767491) or a lower aphidicolin dose. (3) Because the Discussion proposes a 2-cell-like or totipotent state (lines 703–705), report the expression of 2C-like markers (for example Zscan4 family and MERVL-driven transcripts) in the existing RNA-seq data.

2. Replication versus cell-cycle arrest and loss of mitosis. The title and abstract attribute the chromatin phenotype to DNA replication (lines 28–34, 706–712). Aphidicolin, however, also prevents mitosis, and the “age-matched” treated embryos are arrested in S/G2 while controls have passed through division. Mitosis and nuclear envelope breakdown are themselves relevant to chromatin reorganization and heterochromatin re-establishment, so the current design cannot attribute the phenotype to replication rather than to the absence of division. To address this, consider adding a control that blocks mitotic entry after replication is complete, for example a CDK1 inhibitor (RO-3306) applied from late S/G2 of the E2C stage. And then quantify H3K27me3 and H3K9me3 as in Fig. 5D,E and test cluster #1 gene expression. If a G2 block does not reproduce the phenotype, this directly supports the replication-specific interpretation. If it does, the title and Discussion could refer to “replication and cell-cycle progression” to accurately reflect the data shown.

3. The evidence for altered “heterochromatin organization” (Fig. 5D,E) could be further strengthened. Lines 480–486 state that the signals of both marks are reduced, which “further support the functional link between replication and heterochromatin.” However, mean nuclear intensity is normalized to controls, and aphidicolin-arrested nuclei have unreplicated, half-doubled DNA content and have not deposited new nucleosomes, so a lower per-nucleus signal could reflect dilution or lower histone content rather than a defect in establishing or maintaining the marks. Additionally, the effect sizes are modest (P<0.05 for H3K27me3 and P<0.01 for H3K9me3) and are shown as one experiment with nuclei as units (legend, line 424).

To address these consider the following options, 1) Co-stain total histone H3 (or normalize to DAPI) and report H3K27me3/H3K9me3 relative to it. 2) Add a locus-specific measurement of H3K27me3 and H3K9me3 at cluster #1 genes and at an expression-matched unaffected set, using low-input CUT&Tag or CUT&RUN in control versus aphidicolin-treated L2C embryos. If this is not feasible, another idea is to add a morphological analysis of heterochromatin (H3K9me3/DAPI foci). 3) The authors could also consider updating the title and abstract from “heterochromatin organization” to “repressive histone modification levels.”

4. The link between cluster #1 and a Polycomb/late-replicating/closed-chromatin signature would benefit from stronger controls and quantification. Lines 489–501 and 641–653 conclude that aphidicolin disrupts repression of Polycomb-regulated genes, based on enrichment in Fig. 5A–C and Fig. S5A,B. However, Fig. 5A,B show Z-scores without effect sizes or permutation statistics in the main figure. The comparison group is cluster #0 and the other clusters, which contain expressed genes. Any gene silent at the 2-cell stage will tend to be late-replicating, ATAC-poor and H3K27me3-marked, so the signature may reflect “silent developmental genes” in general rather than genes specifically sensitive to replication arrest. Cluster #1 genes are lowly expressed (with a filter of only 30 total reads), so the upregulation is vulnerable to low-count noise in pooled samples. The cluster-based analysis is categorical; no relationship between replication timing and aphidicolin sensitivity is tested across all genes.

To address these consider the following options, 1) add a control gene set matched to cluster #1 for E2C/L2C expression level that is not changed by aphidicolin, and show whether Polycomb, replication-timing and ATAC features still differ, with permutation P-values and effect sizes in Fig. 5. 2) Add a continuous analysis, such as aphidicolin-induced log2 fold change versus replication-timing value or H3K27me3 signal across all low-expressed genes. 3) Validate several cluster #1 genes (for example Irx5, Nefm, Foxd3, Gsc, Tbx2) by RT-qPCR or single-embryo RNA FISH, which would also show whether derepression occurs in all blastomeres. 4) Compare the cluster #1 gene list with published datasets of genes derepressed by EZH2 or PRC2 inhibition in 2-cell embryos or ESCs.

5. The claims about developmental progression, ZGA and lineage specification would benefit from more support. The title, abstract (lines 22–27) and Discussion (lines 706–712) state that preimplantation development proceeds without replication and that ZGA is not prevented. Fig. 2E shows a significant decrease in EU incorporation at MZGA, and the 555 DEGs between aphidicolin-treated and L2C embryos include cell-cycle genes (clusters #3 and #5). ZGA is therefore partly affected. Fig. 6C,F shows delay and, at E3.5, ~90% death of embryos blocked at the 2-cell stage. The conclusion of “comparable” TEAD4/GATA3 (Fig. 3E, lines 188–189) rests on a non-significant t-test, and the plot suggests lower and more variable TEAD4 in aphidicolin-treated embryos. Lineage specification is assessed through two TE factors at 8C. The PCA-based statement that developmental time is “largely” independent of replication (lines 202–206) is not quantified, and maternal RNA decay and polyadenylation changes, which occur independently of replication, could drive the age-matched clustering. To address these consider the following ideas, 1) Temper related claims. 2) Replace the “comparable” claim in Fig. 3E with an equivalence test or describe the difference explicitly. 2) Revise the title, abstract and Discussion to state the limits: the block is transient (16 h), 2-cell blocks are poorly tolerated, and blastocyst formation after a 4C block requires washout.

6. The metabolic analysis relies on indirect predictions rather than direct metabolic measurements. Lines 351–378 and 677–686 infer metabolic “rewiring,” enhanced membrane lipid remodeling, and a possible role for lipid droplets, using GSEA/KEGG and METAFlux. METAFlux was developed for human cancer data and applied here with a human blood metabolite list converted to mouse annotation (Methods, lines 840–843). It estimates fluxes from bulk mRNA of pooled embryos, and no metabolic measurement is provided. Validation with an embryo-compatible readout, for example BODIPY or Nile Red staining for neutral lipids/lipid droplets in aphidicolin-treated versus control embryos at 4C-to-8C would better support the claims. Alternatively, the authors could consider moving the analysis to the supplement and rephrasing the text as a transcriptional prediction, removing the lipid-droplet interpretation from the Discussion.

Minor comments

1. Confirmation of replication block at 4C. EdU is shown only for E2C (Fig. 2B). Consider showing EdU incorporation for the 4C-to-8C treatment (Fig. 3A) and for the zygote window. This would show that the 4C results reflect a true replication block rather than partial inhibition, rule out differences in drug efficacy as a confound in the 2-cell versus 4C comparison.

2. RNA-seq normalization. Global transcription changes at ZGA (and the reduction of EU in Fig. 2E) violate the assumption behind DESeq2 size-factor normalization that most genes are unchanged. To address this consider stating this limitation in the Methods or Results, and noting whether spike-ins were used or how conclusions depend on normalization.

3. Nuclear volume. Line 131 reports no change in nuclear volume at E2C-to-L2C, while 4C-to-8C (Fig. S1C) and E4.5 (Fig. 6K) show enlarged nuclei. Consider adding one sentence interpreting this stage-specific difference, since it bears on the replication-independent growth of nuclei.

4. Figure/legend inconsistencies: 1) Line 494 cites “Fig. 4C” for cluster enrichment of cluster #1 (Table S3 is correct, but the figure should be 4E). 2) METAFlux panels are cited as Fig. S5B,C (lines 358, 361, 370) while chromatin states and histone marks are also called Fig. S5A–D. To improve clarity, the authors could check and update supplementary panel labels as needed (S4 versus S5).

I reviewed this preprint as part of the Transparent and Accountable Peer Review training program. This report was written with the support of my research mentor.

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.