Avalilação PREreview de Genome-wide chromatin reset at the onset of leg regeneration in pancrustaceans

Overall Assessment

The authors ask how chromatin is reorganized at the onset of arthropod leg regeneration, and whether this is conserved across pancrustaceans. Comparing regeneration-competent cricket nymphs with non-regenerating adults, and extending key analyses to Parhyale, they propose a two-phase model in which a genome-wide “enhancer shutdown” after amputation is followed by selective reopening associated with the pioneer factor Zelda.

I consider this an important study. The evidence supports the injury-associated chromatin changes and the identification of Zelda as a candidate regulator and somewhat supports Zelda-mediated reopening and its conservation across species.

To my knowledge, this is the first genome-wide chromatin profiling of appendage regeneration in a cricket. It extends work largely confined to genetically ablated Drosophila imaginal discs (Vizcaya-Molina et al., 2018; Harris et al., 2020; Llorens-Giralt et al., 2026) to amputation of a differentiated leg. Its finding of broad regulatory loss, rather than gains at damage-responsive elements, contrasts with that work. The nymph/adult comparison separates injury from regeneration responses, and the genome and epigenomes will be a valuable resource for the field. The Parhyale analyses were done carefully as depth downsampling (Supp. Table 1) shows the drop in FRiP score, indicating that it is not explained by sequencing depth. The authors also apply naked-DNA filtering and collapse same-individual samples.

My main point of feedback is that the evidence for the shutdown relies on a single normalization approach, so the agreement between assays does not by itself rule out technical or compositional explanations. A secondary point of feedback is matching the mechanistic and conservation claims to the data.

Major comments

1. The claim of “enhancer shutdown” may be confounded by signal quality and tissue composition. The shutdown is shown as flattened, depth-normalized (CPM/RPGC) profiles (Fig. 2B–E, 5B–D). Under this normalization, flattening is equivalent to a lower fraction of reads in peaks. That could reflect a real loss of accessibility, but also dying cells, debris or lower Tn5 efficiency at the site of the amputation. Tissue composition may also shift as suggested by the most downregulated genes being muscle structural genes (Fig. 3A), so the control peak landscape may be dominated by muscle regulatory elements. H3K27me3 is a helpful internal control for CUT&Tag. Supp. Fig. 5 shows that the peaks remaining after amputation are a subset of control peaks, and Supp. Table 1 rules out a depth artifact in Parhyale. The cricket ATAC data, however, have no equivalent QC. The authors could consider:

● reporting per-sample QC for the cricket data, analogous to Supp. Table 1 (FRiP, TSS enrichment, fragment sizes, library complexity)

● annotating the genes associated with lost peaks

● adding a scale to Supp. Fig. 5 and quantifying signal at those peaks across conditions

If lost peaks cluster near muscle genes or QC differs sharply between conditions, the shutdown may reflect tissue makeup rather than per-cell chromatin remodeling. If not, the two-phase model would be considerably stronger.

2. The ATAC methods list five samples for at least eight conditions, CUT&Tag replicate numbers aren't stated, and the antibodies used for H3K27ac and H3K4me1 aren't listed. The Hi-C comparison with published assemblies (Supp. Fig. 2) is reassuring, but annotation methods, assembly completeness metrics (e.g., BUSCO), data accessions and code are not yet described. A sample table (assay × condition × replicate × depth) with replicate concordance or explicitly disclosing the replicate numbers, together with data and code availability statements, would let readers judge whether differences exceed replicate variation and allow others to reuse the resource.

3. The text describes “High”, “Mid” and “Low” recovery peaks, whereas Fig. 4D shows “Common”, “Amp” and “Ctrl” peaks, and it isn't clear how "Amp peaks" fits with the report of virtually no amputation-specific peaks. The raw 6-mer counts are also presented without a statistical test or matched background, and retained peaks may be promoter-biased. The authors could consider:

● testing enrichment with AME (MEME suite) or HOMER against GC- and promoter-matched backgrounds

● running de novo motif discovery

● applying a Tn5-bias filter, as done for Parhyale

This would clarify whether Zelda motifs track recovery or simply promoter identity.

4. In regard to the scope of the claims, the Discussion describes reopening as "mediated by Zelda," and the Abstract describes the same sequence of events in Parhyale, although it was profiled at a single time point. GbZelda RNAi or Zelda CUT&Tag would make an exciting follow-up study. I don’t see them as necessary here because the paper’s main contribution is a genome-wide description of chromatin changes and the identification of Zelda as a candidate regulator, and testing Zelda’s function would require months of new experiments (knockdown, regeneration phenotyping and chromatin profiling). Rescoping the claims would serve the current study equally well. The authors could instead consider phrasing such as "associated with Zelda motif enrichment," and "consistent with" for the Parhyale results. Comparing their peaks with Zelda-bound regions in regenerating Drosophila discs (Bose et al., 2025) would also offer a cross-species reference without new experiments. Better aligning the claims to the evidence would make the conservation argument more persuasive.

5. The differential expression thresholds (padj < 0.15, ≥500 reads, 3-fold ratio) have no stated rationale. The regeneration metric divides by control-leg growth, which is close to zero in adults (Supp. Fig. 1), and Fig. 1B lacks sample sizes and a named statistical test. The authors could consider adding:

● a threshold sensitivity analysis

● the GbZelda interaction p-value

● an explanation of how adult values were computed

These would show whether GbZelda's candidacy rests on the data or on biological priors, and would make the decline in regenerative competence interpretable.

Minor comments

1. Fig. 3E uses different y-axis ranges across panels. Because H3K9me3 drops genome-wide, the change at GbZelda may simply mirror the global shift. Shared axes and values expressed relative to the global change would test the "double lock" interpretation.

2. Quantifying GbZelda signal across replicates, rather than showing single browser views (Fig. 3D), would strengthen what is arguably the paper's most distinctive finding.

3. In Fig. 3B, the GbZld label appears to point into the “Down in Regeneration” block, which seems inconsistent with its induction in Fig. 3A. The authors may want to check the label’s placement, so readers don’t misread the direction of change for the study’s central gene.

4. Adding a sentence to the Results section introducing the 8 dpa experiment, and to the Fig. 4A legend, clarifying whether the stage 8 animals were amputated at stage 7 and molted before collection would help readers relate the two time points.

5. Genome-wide studies of regenerating Drosophila discs (Vizcaya-Molina et al., 2018; Floc'hlay et al., 2023; Llorens-Giralt et al., 2026) report gains at damage-responsive elements. The authors could address this in the Discussion, for example by considering whether genetic ablation vs. amputation, or imaginal vs. differentiated tissue, explains why cricket legs behave differently. This would sharpen the novelty of the shutdown and place it more clearly within the field.

6. The title of Supp. Fig. 4 states that H3K9me3 “limits transcription,” whereas the main text appropriately notes that the data do not establish a causal role. Aligning the title with the text would avoid overstating this observation.

7. The text states that the six most C-terminal zinc fingers are strongly conserved, but Supp. Fig. 6 aligns four (ZF3–6), and the Drosophila sequence appears noticeably more divergent in ZF4. Reconciling the two, and perhaps reporting percent identity, would make the conservation argument easier to assess.

8. Several in-text citations (e.g., Kawamoto 2025, Worley 2022, Pashos 2025, Cevrim 2025) are missing from the reference list, and some listed references are uncited. Reconciling the in-text citations with the reference list would help readers locate the supporting literature.

9. A few figure labels and typos could be corrected to avoid confusion for readers:

   1. Fig. 2E reads "H3K34me1" rather than H3K4me1.

   2. Fig. 3D has two tracks labeled "H3K27ac"; the red one is presumably H3K9me3.

   3. Fig. 1D has overlapping scale labels, and the Supp. Fig. 5 color bar has no values.

   4. In the text, "Zelba," "Parrhyale," "analysis analysis," "and and," "memory that lock" and "decline … provide" are worth correcting.

Correcting these would ensure the figures accurately reflect the data and the text reads cleanly.

References cited

• Bose A, et al. (2025). Science Advances. doi:10.1126/sciadv.ads5743

• Floc'hlay S, et al. (2023). eLife. doi:10.7554/eLife.81173

• Harris RE, et al. (2020). eLife. doi:10.7554/eLife.58305

• Llorens-Giralt P, et al. (2026). Science Advances. doi:10.1126/sciadv.aea8281

• Vizcaya-Molina E, et al. (2018). Genome Research. doi:10.1101/gr.233098.117

Expertise disclosure and competing interests

My expertise is in chromatin profiling, 3D genome organization and computational genomics, and I am less positioned to evaluate arthropod husbandry or regeneration scoring. The author declares no competing interests.

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.