BMP signaling promotes zebrafish heart regeneration via alleviation of replication stress

Abstract

In contrast to mammals, adult zebrafish achieve complete heart regeneration via proliferation of cardiomyocytes. Surprisingly, we found that regenerating cardiomyocytes experience DNA replication stress, which represents one reason for declining tissue regeneration during aging in mammals. Pharmacological inhibition of ATM and ATR kinases revealed that DNA damage response signaling is essential for zebrafish heart regeneration. Manipulation of Bone Morphogenetic Protein (BMP)-Smad signaling using transgenics and mutants showed that BMP signaling alleviates cardiomyocyte replication stress. BMP signaling also rescues neonatal mouse cardiomyocytes, human fibroblasts and human hematopoietic stem and progenitor cells (HSPCs) from replication stress. DNA fiber spreading assays indicate that BMP signaling facilitates re-start of replication forks after replication stress-induced stalling. Our results identify the ability to overcome replication stress as key factor for the elevated zebrafish heart regeneration capacity and reveal a conserved role for BMP signaling in promotion of stress-free DNA replication.

Fig. 1. Signatures of DNA damage are induced in regenerating zebrafish cardiomyocytes (CMs).

A Gene Set Enrichment Plot showing enrichment of transcripts associated with the GO term “DNA repair” in zebrafish hearts at 7 days post ventricular resection injury (dpi). Fold changes were calculated with DESeq2 using the Wald test. B qRT-PCR confirms upregulation of DNA damage response genes in ventricles at 7 days post cryoinjury (dpi), relative to sham-treated hearts. n_E (biological replicates) = 3, n_A (animals) = 4 per replicate. Numbers indicate p-value. HR, homologous recombination; NHEJ, non-homologous end-joining; BER, base excision repair. C Dot plot visualization of single cell RNASeq expression data of border zone and remote CM clusters at 7 dpi for the genes upregulated in (B). Source data is available in: GSE130940. D Western blotting of wound border zone tissue shows increased levels of γH2a.x at 7 dpi relative to uninjured hearts. n_E = 3, n_A = 10 per replicate. E Immunofluorescence on cryosections of myl7:GFP transgenic hearts reveals γH2a.x accumulation in GFP+ CMs located within 150 µm from the wound border at 7 dpi. White box, magnified region. n_E = 1, n_A = 5 for uninjured, 3 and 7 dpi groups; 7 for 14 dpi; 6 for 30 and 60 dpi groups, n_C (analyzed CMs) = 38200 across all time-points. Scale bar, 100 μm. F The fraction of γH2a.x+ wound border CMs correlates with the fraction of cycling CMs. Fraction of γH2a.x+ CMs presented in (E) is overlaid with data of PCNA+ cycling CMs as published in ref. . n_E = 1, n_A = 5 for uninjured, 3 and 7 dpi groups; 7 for 14 dpi; 6 for 30 and 60 dpi groups, n_C (analyzed CMs) = 38200 across all time-points. G γH2a.x+ CMs are found close to the wound border at 7 dpi. n_E = 1, n_A = 6, n_C = 16400 total. H Oxidative DNA damage is induced by hydrogen peroxide treatment of explanted hearts, but does not differ between uninjured and regenerating hearts as determined by 8‐OHdG ELISA at 7 dpi. n_E = 2, n_A = 7 per group. B–H Two tailed Student´s t-test; (E, F) Ordinary one-way ANOVA with Bonferroni correction; (G) Ordinary one-way ANOVA with Dunnet´s multiple comparison test; (A–H) Data are presented as mean values. Error bars indicate confidence interval (CI) 95%. Source data are provided in the Source Data file. Some elements were created in BioRender (Agreement number: AH27UPOG7H; Posadas, D. (2025) https://BioRender.com/w88l548).

Fig. 2. CM replication stress is a specific feature of regeneration and not of physiological growth or aging.

A Immunofluorescence for γH2a.x and GFP combined with EdU detection in myl7:GFP transgenic hearts at 7 dpi. Cycling CMs were labeled by daily EdU injections from 4 to 6 dpi. n_E = 1, n_A = 6, n_C = 7170. Scale bar, 10 μm. Arrows point to EdU+ γH2a.x+ CM nuclei. B In untreated uninjured juvenile fish at 30 days post fertilization (30 dpf), cycling CMs (labeled by immersion of fish in EdU for 8 h) are not γH2a.x +  while hydroxyurea (HU) treatment induces γH2a.x+ accumulation. Yellow box, magnified region. n_E = 1, n_A = 6 per treatment, n_C = 4560 untreated, 3660 HU. Scale bar, 50 μm. C Most of γH2a.x+ CMs at the wound border at 7 dpi present pan-nuclear staining (upper image) versus distinct puncta (lower image). n_E = 1, n_A = 6, n_C = 6450. Scale bar, 10 μm. D Western blotting shows increased levels of p-Rpa32 (S33) in the wound border at 7 dpi. n_E = 3, n_A = 10 per replicate. E Treatment of cryoinjured fish with the cdk4/6 inhibitor PD-0332991 for 1 d reduces the fraction of mitotic pH3+ wound border CMs at 7 dpi. Nocodazole treatment was used to block cytokinesis, increasing detectable pH3+ cells. n_E = 1, n_A = 5 untreated, 6 treated, n_C = 10114 untreated, 11826 treated. Scale bar, 10 μm. F PD-0332991 treatment decreases the fraction of γH2a.x+ wound border CMs. n_E = 1, n_A = 6 per treatment, n_C = 2352 untreated, 2688 treated. G Stimulated growth condition increases the fraction of PCNA+ CMs without an associated increase in γH2a.x+ CMs in uninjured juvenile fish at 30 dpf (16 mm SSL). n_E = 2 for PCNA and 1 for γH2a.x, n_A = 8 restricted growth (RG) PCNA, 8 stimulated growth (SG) PCNA, 8 RG γH2a.x, 7 SG γH2a.x. n_C = 83980 RG, 95600 SG. H The fraction of wound border γH2a.x+ CMs at 7 dpi does not increase with fish age. n_E = 1 (42 days), 2 for 6 months and 2 for 2 years, n_A = 6 (42 days uninjured), 8 (42 days, 7 dpi), 13 (6 months uninjured), 14 (6 months, 7 dpi), 11 (2 years, uninjured), 12 (2 years, 7 dpi), n_C = 31700 (total for all conditions). A–H Data are presented as mean values. Error bars, CI 95%. Two tailed Student´s t-test. Source data are provided in the Source Data file. Some elements were created in BioRender (Agreement number: AH27UPOG7H; Posadas, (2025) https://BioRender.com/w88l548).

Fig. 3. DNA damage repair is essential for zebrafish CM regeneration.

A Very few caspase-3+ apoptotic CMs can be detected at the wound border at 7 dpi, while γ-irradiation induces CM apoptosis. Arrow points to perinuclear caspase-3+ areas in a single apoptotic CM. n_E = 1, n_A = 6 non-irradiated, 5 γ-irradiation, n_C = 3930 non-irradiated, 5950 irradiated. B Treatment of cryoinjured fish with the ATR inhibitor VE821 for 24 h reduces the fraction of pH3+ wound border CMs in myl7:H2b-GFP transgenic hearts at 7 dpi. n_E = 1, n_A = 7 per condition, n_C = 18400 DMSO, 21100 ATRi. C Combined treatment with the ATR and ATM inhibitors VE821 and KU55933 for 24 h reduces wound border CM mitosis. n_E = 2, n_A = 9 per treatment, n_C = 16720 DMSO, 15610 ATRi + ATMi. D Combined intermittent treatment (2 days on, 1 day off) of cryoinjured fish with VE821 and KU55933 from 1 dpi to 21 dpi results in a reduction in wound resorption compared to DMSO treated fish. Wound size (dotted lines) is determined by area lacking GFP expression in myl7:GFP transgenics. n_E = 1, n_A = 5 for 3 dpi, 8 for 21 dpi DMSO and 7 for 21 dpi ATRi + ATMi. Scale bar, 250 μm. A–D Data are presented as mean values. Error bars, CI 95%. Two tailed Student´s t-test. Source data are provided in the Source Data file. Some elements were created in Biorender (Agreement number: AH27UPOG7H; Posadas, (2025) https://BioRender.com/w88l548).

Fig. 4. BMP signaling alleviates CM replication stress during zebrafish heart regeneration.

A Cartoon depicting BMP signaling. Heat-shock-induced overexpression of the BMP ligands 2, 4 or 7 from transgenes activates Smad-mediated and “noncanonical” signaling pathways (gain-of-function, GOF), while overexpression of the secreted inhibitor Noggin3 (Nog3) inhibits them (loss-of-function, LOF). B BMP-LOF caused by overexpression of noggin3 via daily heat-shock from 1 to 7 dpi increases the fraction of γH2a.x+ CMs at 7 dpi relative to heat-shocked wild-type siblings, while BMP-GOF via bmp2b overexpression strongly reduces it. Double transgenics with myl7:GFP were used to identify CMs, “wild-type” refers to absence of the BMP modifying transgene. n_E = 3, n_A = 24 wild-type (wt) siblings to hsp70l:nog3, 23 hsp70l:nog3, 11 wt siblings to hsp70l:bmp2b, 12 hsp70l:bmp2b, n_C = 41460 total across all groups. Scale bar, 50 μm. C BMP-GOF is sufficient to reduce the fraction of γH2a.x+ CMs within 6 h after a single heat-shock of hsp70l:bmp2b transgenics at 7 dpi. n_E = 2, n_A = 12 wt, 11 hsp70l:bmp2b, n_C = 4872 wt, 6470 hsp70l:bmp2b. D bmp7b overexpression reduces the fraction of γH2a.x+ CMs. n_E = 1, n_A = 7 per group, n_C = 4900 wt, 3500 hsp70l:bmp7b;myl7:GFP. E bmp7a –/– mutants display an increased fraction of γH2a.x+ CMs at 7 dpi compared to wild-type siblings. n_E = 2, n_A = 13 per group, n_C = 7800 wt, 7600 bmp7a –/–. F BMP-LOF using hsp70l:nog3 enhances exogenous, HU-induced replication stress in border zone CMs, while BMP-GOF using hsp70l:bmp2b transgenics rescues CMs from HU-induced γH2a.x accumulation. n_E = 1 for wt vs hsp70l:nog3 and 2 for wt vs hsp70l:bmp2b, n_A = 6 wt siblings to hsp70l:nog3, 5 hsp70l:nog3, 13 wt siblings to hsp70l:bmp2b, 14 hsp70l:bmp2b, n_C = 47860 total CMs across all groups. B–F Data are presented as mean values. Error bars, CI 95%. Two tailed Student´s t-test. Source data are provided in the Source Data file. Some elements were created in Biorender (Agreement number: AH27UPOG7H; Posadas, (2025) https://BioRender.com/w88l548).

Fig. 5. BMP-mediated alleviation of replication stress requires Smad signaling.

A Cartoon depicting specific inhibition of the canonical BMP-Smad pathway via overexpression of the inhibitory smad6b. B Immunofluorescence for nuclear Tomato (nT, white arrows) reveals mosaic expression of the hsp70l:nT-p2a-smad6b transgene 6 h after a single heatshock at 7 dpi. p-Smad1/5/9 immunostaining reveals inhibition of BMP-Smad signaling in nT+ CMs (identified by Troponin staining), while nT– CMs in transgenic hearts are equally likely to be p-Smad1/5/9+ than CMs in heat-shocked wild-type fish. n_E = 2, n_A = 11 wt, 13 hsp70l:nT-p2a-smad6b, n_C = 6400 wt, 4540 nT + , 2950 nT–. Scale bar, 100 μm. C BMP-Smad LOF using hsp70l:nT-p2a-smad6b increases γH2a.x accumulation in nT+ CMs compared with wild-type or nT– CMs after a single heat-shock 6 h before harvest at 7 dpi. n_E = 2, n_A = 11 wt, 13 hsp70l:nT-p2a-smad6b, n_C = 5800 wt, 2600 nT + , 2780 nT–. D Smad6 co-expression blocks the ability of BMP2b GOF to alleviate CM replication stress, since the fraction of γH2a.x+ CMs in hsp70l:nT-p2a-smad6b; hsp70l:bmp2b double transgenics is not different from that in heat-shocked wild-type fish 6 h after a single heat-shock at 7 dpi. n_E = 2, n_A = 9 wt, 6 hsp70l:bmp2b, 6 hsp70l:nT-p2a-smad6b, 10 hsp70l:nT-p2a-smad6b and hsp70l:bmp2b, n_C = 13250 total across all groups. The observed relative difference between wild-type and double transgenics is 2%, the calculated smallest significant difference 22%, which is smaller than the difference between hsp70l:bmp2b and double transgenics of 47%. Thus, this experiment had enough power to reveal biologically relevant effects. E AFOG staining reveals increased wound collagen content in hsp70l:nT-p2a-smad6b transgenics at 21 dpi following daily heat-shock. Myocardium, brown; collagen, blue; fibrin, red. Lower images show masks used to quantify the blue collagen+ areas. n_E = 2, n_A = 16 wt, 17 hsp70l:nT-p2a-smad6b. Scale bar, 50 μm. B–D Ordinary one-way ANOVA with Bonferroni correction; Two tailed Student´s t-test. B–E Data are presented as mean values. Error bars, CI 95%. Source data are provided in the Source Data file. Some elements were created in Biorender (Agreement number: AH27UPOG7H; Posadas). (2025) https://BioRender.com/w88l548).

Fig. 6. BMP signaling promotes stress-free replication and relieves CMs from cell cycle delay.

A BMP-Smad LOF using hsp70l:nT-p2a-smad6b reduces CM mitosis within 24 h after a single heat-shock. n_E = 2, n_A = 13 wt, 14 hsp70l:nT-p2a-smad6b, n_C = 19377 wt, 17328 hsp70l:nT-p2a-smad6b. B Cycling CMs were labeled by EdU incorporation from 4 to 6 dpi, followed by EdU and PCNA co-staining at 7 dpi. CMs that are EdU+ PCNA+ display an ongoing cell cycle. BMP-LOF using hsp70l:nog3 enhances the fraction of EdU+ PCNA+ CMs at 7 dpi over that observed in heat-shocked wild-type fish. n_E = 2, n_A = 19 wt, 19 hsp70l:nog3, n_C = 14830 wt, 14981 hsp70l:nog3. C Cycling CMs were labeled by EdU incorporation from 4–6 dpi, followed by EdU and γH2a.x co-staining at 7 dpi. Double-positive CMs are considered to experience replication stress, EdU+ γH2a.x– CMs are stress-free. hsp70l:nog3 enhances the fraction of EdU+ γH2a.x+ stressed CMs at 7 dpi. n_E = 1, n_A = 11 per group, n_C = 5440 wt, 5660 hsp70l:nog3. D bmp7a –/– mutants display an increase in the fraction of CMs that experience replication stress at 7 dpi. n_E = 2, n_A = 12 wt, 13 bmp7a −/−, n_C = 1500 wt, 1300 bmp7a –/–. E HU treatment increases the fraction of EdU+ PCNA+ CMs at 7 dpi, which can be reversed by BMP-GOF using hsp70l:bmp2b. n_E = 1, n_A = 6 per group, n_C = 26000 CMs across all groups. F BMP-GOF using hsp70l:bmp2b is sufficient to increase the fraction of CMs that experience stress-free replication at 7 dpi, even when additional stress is induced by HU treatment. n_E = 1, n_A = 6 wt (untreated), 5 hsp70l:bmp2b (untreated), 6 wt (HU), 6 hsp70l:bmp2b (HU). n_C = 18910 CMs across all groups. A–F Data are presented as mean values. Error bars, CI 95%. Two tailed Student´s t-test. Source data are provided in the Source Data file. Some elements were created in Biorender (Agreement number: AH27UPOG7H; Posadas, (2025) https://BioRender.com/w88l548).

Fig. 7. The ability of BMP signaling to alleviate replication stress is conserved in mouse and human cells.

A Immunofluorescence for γH2a.x and Troponin shows that BMP2 or BMP7 rescue cultured primary neonatal mouse CMs from HU-induced replication stress. Data points represent the fraction of γH2a.x+ CMs per well. n_E = 2, n_wells = 10 media, 10 BMP7, 9 BMP2, 12 media (HU), 9 BMP7 (HU), 10 BMP2 (HU). n_C = 2795 media, 2658 BMP7, 2106 BMP2, 3593 media (HU), 2704 BMP7 (HU), 2533 BMP2 (HU). Scale bar, 20 μm. B The p38 MAPK inhibitor SB202190 does not protect CMs from HU-induced replication stress. n_E = 2, n_wells = 6 media, 4 p38i, 8 media (HU), 4 p38i (HU), n_C = 1384 media, 765 p38i, 1937 media (HU), 876 p38i (HU). The observed relative difference between HU (media) and HU (p38i) groups is 11%, the calculated smallest significant difference 36%, which is smaller than the effect size between the HU (media) and HU (BMP7) groups of 44% (from Fig. 7A). We conclude that this experiment had enough power to detect effects of that magnitude. C Immunofluorescence for γH2a.x shows that pretreatment with BMP2 and BMP4 protects U2OS cells from HU-induced replication stress. Plots show integrated intensity of γH2a.x nuclear levels. n_E = 1, n_wells = 3 for each treatment. Scale bar, 20 μm. D Immunofluorescence for γH2a.x and Phalloidin in human primary neonatal foreskin dermal fibroblasts shows that pre-treatment with BMP2 and BMP4 alleviates HU-induced replication stress. Data points represent quantification of γH2a.x nuclear levels. n_E = 1, n_C = 14 BSA, 11 BMP2 + 4, 14 BSA (HU), 22 BMP2 + 4 (HU). Scale bar, 20 μm. E Pre-treatment with BMP2 rescues survival and proliferative capacity of primary human bone-marrow derived HSPCs treated with HU. Data represent the colony forming units observed in HSPCs isolated from 3 individual donors normalized to BSA-treated controls. n_E = 5, n_donors = 3, n_plates = 2 for donors 1 and 2; 1 for donor 3. A–C, E Ordinary one-way ANOVA with Bonferroni correction; D Kruskal-Wallis followed by Dunn´s correction. A–E Data are presented as mean values. Error bars, CI 95%. Source data are provided in the Source Data file. Some elements were created in Biorender (Agreement number: CH27UPPBB6; Posadas, (2025) https://BioRender.com/n30v908).

Fig. 8. BMP signaling resolves replication stress by enhancing replication fork speed and re-start.

A DNA fiber spreading assays show that treatment with either BMP2 or BMP4 ligands for 48 h increases the speed of replication fork progression measured by track length of either CldU (green) or IdU (red) in human cord blood HSPCs. Data points represent length of individual CldU or IdU tracks. n_E = 3, n_fibers = 670 per treatment. B Treatment with BMP4 increases replication fork progression in U2OS cells, while BMP2 does not have a significant effect. n_E = 2, n_fibers = 250 per treatment. C Pre-treatment with BMP2 or BMP4 ligands increases the fraction of replication forks that restart after HU-mediated stalling in U2OS cells. Fork arrest is indicated by tracks that are only labeled by CldU, restarted forks by CldU tracks that are followed by IdU. Data points represent fraction of restarted forks out of all analyzed forks. n_E = 2, n_fibers = 600 per treatment. A–C Data are presented as mean values. Error bars, CI 95%. Kruskal-Wallis followed by Dunn´s correction. Source data are provided in the Source Data file. Some elements were created in Biorender (Agreement number: CH27UPPBB6; Posadas, (2025) https://BioRender.com/n30v908).

Fig. 9. Model for the role of BMP signaling in alleviation of replication stress in zebrafish heart regeneration and mammalian cells.

Zebrafish hearts efficiently regenerate via proliferation of pre-existing cardiomyocytes, although cardiomyocytes experience regeneration-induced replication stress. BMP signaling, activated by BMP2, 4 and 7 ligands, and mediated via Smad-signaling, allows CMs to overcome the replication stress, which otherwise leads to S-phase arrest and failure or regeneration. The unique ability of BMP signaling to promote stress-free replication is conserved in mouse cardiomyocytes, human cell lines, primary fibroblasts and hematopoietic stem and progenitor cells. Some elements were created in BioRender (Agreement numbers: AH27UPOG7H; https://BioRender.com/w88l548; CH27UPPBB6; Posadas, D. (2025) https://BioRender.com/n30v908).