Attenuation of epicardial activation and myofibroblast abundance via the Fbln2–Nupr1b axis stimulates cardiac regeneration in zebrafish
Gülsüm Kayman Kürekçi, Gursimran Kaur Bajwa, Shaoqiu Zhang, Séverine Leclerc, Emilie de Chantal, Darrell Belke, Gregor Andelfinger, Justin F. Deniset, Rubén Marín-Juez
TL;DR
The study shows how fibulin-2 and nuclear protein 1b control heart regeneration in zebrafish by managing fibrosis after injury.
Contribution
The novel finding is the identification of the Fbln2–Nupr1b axis as a regulator of epicardial cell activation and regeneration.
Findings
Fibulin-2 regulates activated epicardial cells to balance fibrosis and regeneration.
Nupr1b controls myofibroblast abundance and rescues fbln2 mutant phenotypes.
Modulating the Fbln2–Nupr1b axis enhances cardiac regeneration in zebrafish.
Abstract
After injury, the adult human heart fails to regenerate and forms a persistent fibrotic scar. By contrast, fibrosis is transient in the injured zebrafish heart, facilitating cell recruitment and providing regenerative cues. The mechanisms that restrain excessive fibrosis while enabling regeneration remain poorly understood. Here we show that fibulin-2 (Fbln2) regulates specific populations of activated epicardial cells to balance the response to cardiac injury. Using genetic tools for Fbln2 dosage, we find that attenuation of epicardial activation stimulates regenerative programs. Mechanistically, we identify epicardial nuclear protein 1b (Nupr1b) as an Fbln2 effector. Using gain- and loss-of-function approaches, we show that Nupr1b controls epicardial myofibroblast abundance. Notably, epicardial-specific overexpression of nupr1b rescued fbln2 mutant phenotypes. These findings shed…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
Figure 10
Figure 11
Figure 12
Figure 13
Figure 14
Figure 15
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
Figure 7
Figure 8
Figure 9- —https://doi.org/10.13039/501100000024Gouvernement du Canada | Canadian Institutes of Health Research (Instituts de Recherche en Santé du Canada)
- —https://doi.org/10.13039/501100000156Fonds de Recherche du Québec - Santé (Fonds de la recherche en sante du Quebec)
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsCongenital heart defects research · Cardiac Fibrosis and Remodeling · Developmental Biology and Gene Regulation
Main
Myocardial infarction (MI) is the leading cause of mortality and morbidity worldwide^1,2^. To prevent cardiac rupture after MI, fibroblasts activate, differentiate into myofibroblasts and deposit extracellular matrix (ECM) proteins, forming a fibrotic scar. Fibrosis provides immediate structural support but ultimately undergoes remodeling and impairs long-term cardiac function. In contrast to adult mammalian hearts, adult zebrafish hearts can regenerate after different types of insults^3–7^. Like mammals, zebrafish fibroblasts activate and produce ECM following cardiac injury^8–10^. However, fibrosis is transient, and ECM deposits are resorbed as the heart regenerates^6,8^. The fibrotic tissue not only provides structural support but also facilitates cell recruitment and provides regenerative signals^11^.
An important player in cardiac regeneration is the epicardium, a mesothelial envelope engaging in molecular and cellular events following cardiac injury in a process referred to as epicardial activation. The activated epicardium constitutes an important source of cells and signaling^12,13^. After activation, epicardial cells undergo epithelial-to-mesenchymal transition (EMT) to give rise to epicardium-derived cells (EPDCs)^12,14^, which differentiate into mural cells and fibroblasts^9,12,15^. EPDCs secrete paracrine factors that modulate the response to injury^16–18^. Growth factors, including transforming growth factor β (TGFβ)^19^, regulate epicardial activation. In zebrafish, TGFβ signaling modulates key processes during regeneration, such as proliferation and fibrosis, in a cell- and time-dependent manner^20^. TGFβ contributes to the generation of myofibroblasts and the subsequent fibrosis, both in regenerative and non-regenerative organisms^21^. Yet, the mechanisms preventing excessive fibrosis while enabling sufficient regeneration remain poorly understood.
Fibulins are ECM glycoproteins that regulate different cell functions and signaling pathways, particularly in the context of tissue remodeling^22–24^. Specifically, FIBULIN-2 (FBLN2) interacts with various ECM components and ligands^25–29^. FBLN2 is deposited at EMT sites during development and in response to skin, atherosclerotic and neuronal lesions^30–33^. FBLN2 positively regulates TGFβ signaling across diverse biological contexts^34–37^. In tumor development, FBLN2 can show both pro- and anti-tumorigenic effects^38,39^, highlighting its pivotal role in mediating the interplay between seemingly opposite mechanisms.
In our attempt to identify regulators of transient fibrosis, we found that fbln2 is dynamically expressed during cardiac regeneration. We found that two distinct loss-of-function fbln2 alleles show reductions in epicardial TGFβ signaling and myofibroblasts following cardiac cryoinjury. Early CM (cardiomyocyte) regeneration was similarly declined in both mutants. However, late-stage fibrosis differs between both alleles. While downregulation of fbln2 attenuates transient fibrosis, its complete loss results in scarring. We identified nuclear protein 1b (Nupr1b) as an Fbln2 effector. nupr1b mutants show reduced numbers of epicardial myofibroblasts and decreased CM proliferation, recapitulating fbln2 mutant phenotypes, and its epicardial overexpression rescues these phenotypes in fbln2 mutants.
Overall, our data identify Fbln2 as a key regulator of the balance between cardiac fibrosis and regeneration through Nupr1b-dependent epicardial myofibroblast regulation.
Results
fbln2 is upregulated in EPDCs and endothelial cells after cardiac injury in zebrafish
During mouse cardiogenesis, Fbln2 is expressed in the endocardial cushion, coronary vasculature, epicardium and cardiac valves^40,41^. To assess the expression pattern of fbln2 in adult zebrafish hearts, we used in situ hybridization chain reaction (HCR) combined with immunostaining. While fbln2 was not detected in uninjured ventricles (Fig. 1a), at 4 days post-cryoinjury (dpci), fbln2 expression mainly co-localized with tcf21:mCherry expression, marking EPDCs (Fig. 1b). Co-staining with the endothelial marker cdh5 revealed low fbln2 expression in the cardiac endothelium (Fig. 1c). To confirm these observations, we analyzed fbln2 expression in Et(krt4:EGFP) and Tg(**-0.8flt1:RFP) ventricles, marking endocardial cells and coronary endothelial cells (cECs), respectively (Fig. 1d,e). fbln2 was weakly expressed in some border-zone krt4:EGFP^+^ cells at 4 dpci (Fig. 1d) and remained detectable in cECs at 7 dpci (Fig. 1e). At 7 dpci, fbln2 expression was prominent in tcf21:mCherry^+^ EPDCs (Fig. 1f), and by 14 dpci, its expression was largely reduced and restricted to EPDCs at the injury border zone (Fig. 1g). To further assess the expression dynamics of fbln2, we engineered a TgBAC(fbln2:EGFP) transgene. In uninjured ventricles, fbln2:EGFP expression was detected in cardiac valves and epicardial cells (Fig. 1h). At 4 dpci, the transgene was upregulated in EPDCs and border-zone endocardium (Fig. 1i), recapitulating the HCR data. Altogether, these data indicate that fbln2 is strongly upregulated in EPDCs after injury and its expression is dynamically regulated during heart regeneration in adult zebrafish.Fig. 1fbln2 is upregulated in EPDCs and endothelial cells after cardiac cryoinjury in zebrafish.a,b, In situ HCR for fbln2 (green) and immunostaining for mCherry (magenta), and DAPI (DNA, blue), on sections of uninjured (a) and 4 dpci (b) TgBAC(tcf21:mCherry) ventricles and DAPI (DNA, blue). c, In situ HCR for fbln2 (green) and cdh5 (magenta) on sections of wild-type ventricles at 4 dpci. High-magnification images show border-zone EPDCs. d, In situ HCR for fbln2 (green) and immunostaining for mCherry (magenta) and EGFP (white), and DAPI (DNA, blue), on sections of 4 dpci TgBAC(tcf21:mCherry); Et(krt4:EGFP) ventricles. White arrowheads point to fbln2^+^ EPDCs and yellow arrowheads point to fbln2^+^ endocardial cells. e, In situ HCR for fbln2 (green) and immunostaining for RFP (magenta) on sections of 7 dpci Tg(-0.8flt1:RFP) ventricles. High-magnification images show border-zone coronary endothelial cells expressing fbln2. f, In situ HCR for fbln2 (green) and immunostaining for mCherry (magenta), and DAPI (DNA, blue), on 7 dpci TgBAC(tcf21:mCherry) ventricles. g, In situ HCR for fbln2 (green) and DAPI (DNA, blue) on wild-type ventricles at 14 dpci. High-magnification images of border-zone epicardium. Arrowheads point to fbln2^+^ EPDCs. h, Immunostaining of uninjured ventricle sections of fish injected with TgBAC(fbln2:EGFP) construct and stained for EGFP (green) and DAPI (DNA, blue). High-magnification images of cardiac valves ((i), asterisks) and epicardial cells (ii). i, Immunostaining of sections of cryoinjured ventricles of fish injected with TgBAC(fbln2:EGFP) construct and stained for EGFP (green) and DAPI (DNA, blue) at 4 dpci. High-magnification images of the border-zone EPDCs (i) and endocardium (ii). In all panels, white arrowheads point to fbln2^+^ cells. White dashed lines delineate the injured area. Scale bars, 20 µm (a–f) and 100 µm (g–i).
Fbln2 regulates epicardial TGFβ signaling and myofibroblast number after cardiac injury
To investigate the role of Fbln2 during heart regeneration, we set out to generate fbln2 mutants. FBLN1 is upregulated in Fbln2 mutant mice^42,43^. To prevent transcriptional adaptation and potential compensation, we generated a fbln2 promoter-less allele (fbln2^PL^)^44,45^ (Fig. 2a). Although antibodies for zebrafish Fbln2 are not available, mRNA expression analyses in fbln2^PL^^−^^/^^−^ showed a 65% reduction in fbln2 levels, with no upregulation of fbln1 and efemp2a (fbln4), the zebrafish orthologs of Fbln2 paralogs expressed in murine cardiac tissues^41^^,46^ (Fig. 2b). fbln2^PL^^−^^/^^−^ survive to adulthood, show no overt morphological defects and are fertile.Fig. 2fbln2^PL^^−^^/^^−^ shows reduced TGFβ signaling and decreased numbers of epicardial myofibroblasts.a, Schematic representation of the fbln2 locus showing gRNAs (pink lines) targeting 5′ UTR and exon 1 to generate a promoter-less (fbln2^PL^) allele. Protein-coding exons are shown as gray boxes and UTRs as white boxes. The arrow indicates the transcription start site. The gray dashed line represents the deleted sequence. b, RT-qPCR analysis of fbln2, fbln1 and efemp2a mRNA levels of fbln2^+/+^ and fbln2^PL^^−^^/^^−^ siblings at 5 dpf; n = 5 (pools of 5). Data are shown as means ± s.d. and bars represent medians. Statistical test: two-tailed Student’s t-test. c, Immunostaining of sections of cryoinjured ventricles of fbln2^+/+^ and fbln2^PL^^−^^/^^−^ siblings at 7 dpci stained for pSmad3 (green), Cav1 (magenta) and DAPI (DNA, blue). Arrowheads point to epicardial pSmad3^+^ cells. d, Number of total, intraventricular (intrav.) and epicardial pSmad3^+^ cells per mm^2^ injured area in fbln2^+/+^ and fbln2^PL^^−^^/^^−^ ventricles at 7 dpci; n = 7. Box plots show the median and interquartile range (IQR; 25th–75th percentiles), with whiskers spanning minimum (min) to maximum (max) values. Statistical test: two-tailed Student’s t-test. e, Immunostaining of sections of cryoinjured ventricles of fbln2^+/+^ and fbln2^PL^^−^^/^^−^ siblings at 7 dpci stained for αSMA (green), myosin heavy chain (MyHC, magenta) and DAPI (DNA, blue). Arrowheads point to epicardial αSMA^+^ cells. f, Number of total, endocardial and epicardial αSMA^+^ cells per mm^2^ injured area in fbln2^+/+^ and fbln2^PL^^−^^/^^−^ ventricles at 7 dpci; n = 7. Box plots show the median and IQR (25th–75th percentiles), with whiskers spanning min–max values. Statistical test: two-tailed Student’s t-test. Scale bars, 100 µm.Source data
Fbln2 has been shown to positively regulate TGFβ signaling^35,37,47,48^. TGFβ signaling regulates various processes in response to cardiac damage, both at early and late stages after injury^49^. Therefore, we hypothesized that Fbln2 regulates TGFβ signaling during cardiac regeneration. It has been shown that FBLN2 induces TGFβ signaling via the Smad pathway^35,37^. To assess TGFβ signaling activation levels, we performed immunostaining for pSmad3 expression on ventricular sections from fbln2^PL^^−^^/^^−^ and wild-type siblings at 7 dpci (Fig. 2c). We found a significant reduction in the total number of pSmad3^+^ cells in fbln2^PL^^−^^/^^−^ ventricles compared with wild-type siblings, with a more prominent decrease in pSmad3^+^ cells localized at the epicardial region marked by Cav1 expression (Fig. 2c,d). These data indicate that Fbln2 is required for TGFβ signaling activation after injury.
TGFβ is a primary inducer of fibroblast activation and differentiation into myofibroblasts^49,50^. During cardiac regeneration in zebrafish, EPDCs are a potential source of α-smooth muscle actin^+^ (αSMA) myofibroblasts^9^. Therefore, a reduction in TGFβ signaling in EPDCs could lead to a reduction in αSMA^+^ myofibroblasts. To test this possibility, we performed immunostaining for αSMA expression on ventricular sections from fbln2^PL^^−^^/^^−^ and wild-type siblings at 7 dpci, the onset of collagen deposition in the injury (Fig. 2e). As myofibroblasts can potentially differentiate from EPDCs and endocardial cells^8,9,21^, we separately quantified αSMA^+^ cells based on their spatial distribution within the injury as localized at the subepicardial region or the injury border zone^51–53^. At 7 dpci in wild-type ventricles, αSMA^+^ cells were predominantly observed at the periphery of the injury, underlining the thickened epicardium (Fig. 2e), as previously reported^51–53^. While the total number of αSMA^+^ cells did not differ between wild-type and fbln2^PL^^−^^/^^−^ ventricles, we identified distinct phenotypes in the number of αSMA^+^ cells localized at the subepicardial region and the border zone. Specifically, we observed a significant reduction in epicardial αSMA^+^ myofibroblasts, while no difference was found in the number of αSMA^+^ cells at the border zone (Fig. 2f).
Collectively, these data show that Fbln2 regulates epicardial TGFβ signaling and myofibroblast number during heart regeneration, suggesting that epicardial activation is Fbln2 dependent.
fbln2PL−/− shows reduced CM and endothelial proliferation and attenuated fibrosis
After cardiac injury in zebrafish, regenerating CMs undergo dedifferentiation and proliferation, supported by epicardial and endothelial cues^18,54^. Chemical inhibition of TGFβ signaling impairs CM proliferation after cardiac injury in zebrafish, indicating a role in early regenerative processes^20,55,56^. Therefore, we reasoned that CM regeneration could be affected in fbln2^PL^^−^^/^^−^. To test this possibility, we analyzed CM proliferation and dedifferentiation at 7 dpci and found that both were significantly reduced in fbln2^PL^^−^^/^^−^ ventricles compared with wild-type siblings (Fig. 3a–d). Next, we examined endothelial proliferation as fbln2 is transiently expressed in border-zone endocardial cells. We found that fbln2^PL^^−^^/^^−^ ventricles have reduced endothelial proliferation compared with wild-type siblings at 4 dpci (Fig. 3e,f).Fig. 3fbln2^PL^^−^^/^^−^ ventricles present reduced cardiomyocyte and endothelial proliferation and retain smaller collagen deposits.a, Immunostaining of sections of cryoinjured ventricles of fbln2^+/+^ and fbln2^PL^^−^^/^^−^ siblings at 7 dpci stained for Mef2 (myocyte enhancer factor 2; CMs, magenta), PCNA (proliferating cell nuclear antigen; proliferation marker, green) and DAPI (DNA, blue). Arrowheads point to PCNA^+^ CMs. b, Percentage of PCNA^+^ CMs in the injury border zone of fbln2^+/+^ (n = 6) and fbln2^PL^^−^^/^^−^ (n = 8) siblings at 7 dpci. Statistical test: two-tailed Student’s t-test. c, Immunostaining of sections of cryoinjured ventricles of fbln2^+/+^ and fbln2^PL^^−^^/^^−^ siblings at 7 dpci stained for Mef2 (CMs, magenta), N2.261 (embryonic MyHC, green) and DAPI (DNA, blue). Arrowheads point to N2.261^+^ CMs. d, Percentage of N2.261^+^ CMs in the injury border zone of fbln2^+/+^ (n = 5) and fbln2^PL^^−^^/^^−^ siblings (n = 5) at 7 dpci. Statistical test: two-tailed Student’s t-test. e, Immunostaining of sections of cryoinjured ventricles of fbln2^+/+^ and fbln2^PL^^−^^/^^−^ siblings at 4 dpci stained for Fli1 (endothelial cells, magenta), PCNA (proliferation marker, green) and DAPI (DNA, blue). Arrowheads point to PCNA^+^ endothelial cells. f, Percentage of PCNA^+^ endocardial cells in the injury of fbln2^+/+^ (n = 4) and fbln2^PL^^−^^/^^−^ siblings (n = 4) at 4 dpci. Statistical test: two-tailed Student’s t-test. g, Representative images of AFOG staining of sections of fbln2^+/+^ and fbln2^PL^^−^^/^^−^ ventricles at 7 dpci. Orange, muscle; red, fibrin; blue, collagen. h, Percentage of collagen area within the injury at 7 dpci of cryoinjured fbln2^+/+^ (n = 6) and fbln2^PL^^−^^/^^−^ ventricles (n = 9). Statistical test: two-tailed Student’s t-test. i, Representative images of sections of cryoinjured of fbln2^+/+^ and fbln2^PL^^−^^/^^−^ siblings at 7 dpci stained for CHP (green) and DAPI (DNA, blue). j, Quantification of CHP intensity (a.u.) per μm^2^ injured area of fbln2^+/+^ (n = 6) and fbln2^PL^^−^^/^^−^ (n = 5) siblings at 7 dpci. Statistical test: two-tailed Student’s t-test. k, Representative images of AFOG staining of sections of fbln2^+/+^ and fbln2^PL^^−^^/^^−^ ventricles at 90 dpci. Orange, muscle; red, fibrin; and blue, collagen. Number of animals analyzed and with the phenotype observed has been indicated. l, Percentage representation of groups with different collagen area sizes at 90 dpci for cryoinjured fbln2^+/+^ (n = 10) and fbln2^PL^^−^^/^^−^ ventricles (n = 10). White dashed lines delineate the border of the injured tissue, black dashed lines delineate the area with collagen deposits and red dashed lines delineate the regenerated myocardium. Scale bars, 100 µm. In all panels, box plots show the median and IQR (25th–75th percentiles), with whiskers spanning min–max values.Source data
Reduced endothelial proliferation, along with impaired CM proliferation and dedifferentiation, typically results in defective myocardial replenishment and is associated with compromised regeneration and persistent fibrotic scarring^18,51^. In view of the CM and endothelial regeneration phenotypes observed, we hypothesized that fibrosis resolution might be compromised in injured fbln2^PL^^−^^/^^−^ ventricles. To assess the extent of collagen deposition and fibrosis, we quantified collagen areas in injured ventricles using acid fuchsin orange G (AFOG) staining. At 7 dpci, collagen deposition was significantly reduced in fbln2^PL^^−^^/^^−^ ventricles when compared with wild-type siblings (Fig. 3g,h). Next, to better understand fibrosis dynamics, we analyzed collagen remodeling using collagen hybridizing peptide (CHP) staining at 7 dpci. CHP binds to single-stranded collagen, thereby marking immature or partially degraded collagen^57,58^. CHP intensity at 7 dpci was significantly lower in fbln2^PL^^−^^/^^−^ ventricles than in wild-type siblings, suggesting altered early collagen remodeling (Fig. 3i,j). Interestingly, at 90 dpci, collagen deposits were significantly smaller in fbln2^PL^^−^^/^^−^ ventricles (Fig. 3k,l). Like in wild types, the remaining collagen was luminally localized with the trabecular and cortical myocardium surrounding it, suggesting no regeneration defects in fbln2^PL^^−^^/^^−^ ventricles (Fig. 3k). Altogether, these data show that despite a significant reduction in early CM and endothelial proliferation, fibrosis is reduced in fbln2^PL^^−^^/^^−^ ventricles.
fbln2FLD−/− shows a further reduction in TGFβ signaling and impaired fibrosis resolution
Our findings indicate that Fbln2 regulates TGFβ signaling activation and the initiation of early proliferative responses following cardiac injury. However, the downregulation of fbln2 does not impede long-term regeneration but rather seems to enhance fibrosis resolution. Therefore, we reasoned that the observation that fbln2^PL^^−^^/^^−^ ventricles present smaller collagen deposits at 90 dpci may be linked to the hypomorphic nature of the allele (Fig. 2b). To test this possibility and avoid triggering transcriptional adaptation, we generated a fbln2 full-locus deletion mutant (fbln2^FLD^) (Fig. 4a). We confirmed successful gene body ablation by genotyping (Extended Data Fig. 1b) and verified the absence of fbln2 expression in fbln2^FLD^^−^^/^^−^ larvae (Fig. 4b). First, we analyzed CM proliferation at 7 dpci and found a significant reduction in CM proliferation in fbln2^FLD^^−^^/^^−^ ventricles compared with wild-type siblings (Fig. 4c,d). Next, we assessed TGFβ signaling by quantifying pSmad3 levels in fbln2^FLD^^−^^/^^−^ ventricles at 7 dpci (Fig. 4e). The number of pSmad3^+^ cells was significantly decreased in fbln2^FLD^^−^^/^^−^ ventricles compared with both wild-type and fbln2^PL^^−^^/^^−^ ventricles (Fig. 4f). This reduction was observed across the injury site, affecting both intraventricular and epicardial regions. Notably, the extent of pSmad3^+^ cell loss correlated with fbln2 levels, with the strongest reduction observed in fbln2^FLD^^−^^/^^−^, where epicardial pSmad3^+^ cells were hardly detectable (Fig. 4e,f). These results suggest that Fbln2 levels determine the extent of TGFβ signaling activation during cardiac regeneration. Then, we quantified the number of αSMA^+^ myofibroblasts in fbln2^FLD^^−^^/^^−^ ventricles at 7 dpci (Fig. 4g). We found a significant reduction in epicardial αSMA^+^ myofibroblasts in fbln2^FLD^^−^^/^^−^ ventricles compared with wild types (Fig. 4h). The total number of αSMA^+^ cells in fbln2^FLD^^−^^/^^−^ ventricles was similar to those observed in fbln2^PL^^−^^/^^−^.Fig. 4TGFβ signaling is further decreased and fibrosis resolution altered in cryoinjured ventricles of fbln2 full-locus fbln2 deletion mutants.a, Schematic representation of the fbln2 locus showing gRNAs (pink lines) targeting the 5′ UTR and 3′ UTR to create a full-locus deletion (fbln2^FLD^) allele. Protein-coding exons are shown as gray boxes and UTRs as white boxes. The arrow indicates the transcription start site. The gray dashed line represents the deleted sequence. b, RT-qPCR analysis of fbln2 mRNA levels of fbln2^+/+^ (n = 4) and fbln2^FLD^^−^^/^^−^ (n = 5) siblings at 5 dpf. Data are shown as means ± s.d. and bars represent medians. Statistical test: two-tailed Student’s t-test c, Immunostaining of sections of cryoinjured ventricles of fbln2^+/+^ and fbln2^FLD^^−^^/^^−^ siblings at 7 dpci stained for Mef2 (CMs, magenta), PCNA (proliferation marker, green) and DAPI (DNA, blue). Arrowheads point to PCNA^+^ CMs. d, Percentage of PCNA^+^ CMs in the injury border zone of fbln2^+/+^ (n = 6) and fbln2^FLD^^−^^/^^−^ (n = 7) siblings at 7 dpci. Box plots show the median and IQR (25th–75th percentiles), with whiskers spanning min–max values. Statistical test: two-tailed Student’s t-test. e, Immunostaining of sections of cryoinjured ventricles of fbln2^+/+^ and fbln2^FLD^^−^^/^^−^ siblings at 7 dpci stained for pSmad3 (green), Cav1 (magenta) and DAPI (DNA, blue). Arrowheads point to epicardial pSmad3^+^ cells. f, Number of total, intraventricular and epicardial pSmad3^+^ cells per mm^2^ injured area in fbln2^PL^^−^^/^^−^ (n = 8), fbln2^FLD^^−^^/^^−^ (n = 8) and their respective fbln2^+/+^ siblings (n = 10) at 7 dpci. Box plots show the median and IQR (25th–75th percentiles), with whiskers spanning min–max values. Statistical test: one-way ANOVA followed by Tukey’s post hoc test. g, Immunostaining of sections of cryoinjured ventricles of fbln2^+/+^ and fbln2^FLD^^−^^/^^−^ siblings at 7 dpci stained for αSMA (green), MyHC (magenta) and DAPI (DNA, blue). Arrowheads point to epicardial αSMA^+^ cells. h, Number of total, border-zone and epicardial αSMA^+^ cells per mm^2^ injured area in fbln2^PL^^−^^/^^−^ (n = 7), fbln2^FLD^^−^^/^^−^ (n = 8) and their respective fbln2^+/+^ siblings (n = 12) at 7 dpci. Box plots show the median and IQR (25th–75th percentiles), with whiskers spanning min–max values. Statistical test: one-way ANOVA followed by Tukey’s post hoc test. i, Representative images of AFOG staining of sections of fbln2^+/+^ and fbln2^FLD^^−^^/^^−^ ventricles at 7 dpci. Orange, muscle; red, fibrin; and blue, collagen. j, Percentage of collagen area within the injury at 7 dpci of cryoinjured fbln2^+/+^ (n = 6) and fbln2^FLD^^−^^/^^−^ ventricles (n = 8). Box plots show the median and IQR (25th–75th percentiles), with whiskers spanning min–max values. Statistical test: two-tailed Student’s t-test. k, Representative images of AFOG staining of sections of fbln2^+/+^, fbln2^FLD^^−^^/^^−^ and fbln2^FLD+/^^−^ ventricles at 90 dpci. Orange, muscle; red, fibrin; and blue, collagen. Number of animals analyzed and with the phenotype observed has been indicated. l, Percentage representation of groups according to collagen area sizes at 90 dpci for cryoinjured fbln2^+/+^ (n = 11), fbln2^FLD^^−^^/^^−^ (n = 7) and fbln2^FLD+/^^−^ (n = 7) ventricles. m, Representative images of sections of cryoinjured fbln2^+/+^ and fbln2^FLD^^−^^/^^−^ ventricles at 7 and 20 dpci stained for CHP (green) and DAPI (DNA, blue). n, Quantification of CHP intensity per μm^2^ injured area of fbln2^+/+^ (n = 5) and fbln2^FLD^^−^^/^^−^ (n = 5) ventricles at 7 and 20 dpci. Box plots show the median and IQR (25th–75th percentiles), with whiskers spanning min–max values. Statistical test: one-way ANOVA followed by Tukey’s post hoc test. o, RT-qPCR analysis of mmp9, mmp13a and mmp14b mRNA levels in fbln2^+/+^, fbln2^PL^^−^^/^^−^ and fbln2^FLD^^−^^/^^−^ cryoinjured ventricles at 4 dpci (n = 4 or 5, pools of 2 ventricles each). Data are shown as means ± s.d. and bars represent medians. Statistical test: one-way ANOVA followed by Tukey’s post hoc test. White dashed lines delineate the border of the injured tissue, black dashed lines delineate the area with collagen deposits and red dashed lines delineate the regenerated myocardium. Scale bars, 100 μm.Source data
Finally, we assessed collagen deposition and fibrosis resolution in fbln2^FLD^^−^^/^^−^. We found that early collagen deposition was reduced in fbln2^FLD^^−^^/^^−^ ventricles compared with wild types at 7 dpci (Fig. 4i,j). Interestingly, fbln2^FLD^^−^^/^^−^ ventricles failed to resolve fibrosis and retained high levels of collagen at 90 dpci indicating failure to regenerate (Fig. 4k,l). These data show that fbln2 expression levels determine the extent of fibrosis retention at late stages after injury. To further test this possibility, we analyzed collagen areas in fbln2^FLD+/^^−^ ventricles at 90 dpci. Similar to fbln2^PL^^−^^/^^−^, fbln2^FLD+/^^−^ show a 70% reduction in fbln2 expression (Extended Data Fig. 1c). Notably, fibrosis was reduced in fbln2^FLD+/^^−^ ventricles at 90 dpci, recapitulating the fbln2^PL^^−^^/^^−^ phenotype (Fig. 4k,l). Next, to test whether collagen remodeling was affected in fbln2^FLD^^−^^/^^−^, we analyzed CHP intensity at 7 dpci and did not observe any difference between mutants and wild types (Fig. 4m,n). We reasoned that the fibrotic phenotypes observed in fbln2^FLD^^−^^/^^−^ could be caused by defects in fibrosis resolution at later time points after injury. Therefore, we performed CHP staining at 20 dpci and found a significant increase in CHP intensity in fbln2^FLD^^−^^/^^−^ ventricles (Fig. 4m,n). These data suggest the retention of partially degraded collagen that cannot be efficiently resolved in fbln2^FLD^^−^^/^^−^ ventricles, indicating disrupted collagen remodeling. As alterations in collagen remodeling may indicate defects in collagen degradation, we analyzed the expression levels of matrix metalloproteinase-encoding genes mmp9, mmp13a and mmp14b, all shown to be upregulated early after injury^58,59^. Interestingly, we found that all three genes were significantly upregulated in fbln2^PL^^−^^/^^−^ ventricles while their expression remained unchanged in fbln2^FLD^^−^^/^^−^ ventricles compared with wild types at 4 dpci (Fig. 4o). These data, together with the observed reduction in CHP intensity at 7 dpci, suggest increased collagen degradation in fbln2^PL^^−^^/^^−^ ventricles at early time points after injury. Overall, these results indicate that, while the reduction of fbln2 levels facilitates fibrosis resolution, its abrogation leads to scarring.
fbln2 reduction attenuates fibrosis via modulation of epicardial activation
Our data show that transient fibrosis is reduced in fbln2^PL^ mutants, suggesting an improved response to injury. To further investigate the molecular mechanisms driving this response, we performed single-cell RNA sequencing on fbln2^+/+^, fbln2^PL^^−^^/^^−^ and fbln2^FLD^^−^^/^^−^ ventricles at 4 dpci, when fbln2 expression peaks and early regeneration phenotypes are observed. All major cardiac cell types were identified and annotated based on the expression of established marker genes (Fig. 5a,b).Fig. 5. Reduced activation of EPDCs in cryoinjured fbln2^PL^^−^^/^^−^ ventricles.a, UMAP representation of all samples combined and individual groups with inferred cellular identity after scRNA-seq analysis. Lympho, lymphocytes; neutro, neutrophils; mes, mesenchymal cells; macro, macrophages. b, Dot plot depicting average expression (avg. exp) and abundance (Pct. exp) of marker genes across all cell types. c, Proportion of cells per cluster within each sample. d, UMAP representation of EPDC subclusters. Purple dashed lines indicate epithelial EPDC clusters (podxl^+^), blue dashed lines indicate mesenchymal EPDC clusters (hapln1a^+^) and orange dashed lines indicate activated EPDC clusters. e, Dot plot depicting average expression and abundance (Pct. exp) of marker genes across all EPDC subclusters. f, Normalized expression of selected marker genes of Ep0 (ptx3a and stra6), Ep4 (manf and hyou1) and Ep10 (top2a) clusters in UMAPs (all samples combined). g, UMAP representation of pseudotime along the inferred trajectory of EPDCs. White circles represent leaf nodes, and black circles represent branch points. h, Comparison of pseudotime distribution of EPDCs per group at the single-cell level (each group is one pool of four fish). Box plots show the median and IQR (25th–75th percentiles), with whiskers spanning min–max values. Statistical test: Kruskal–Wallis. The statistical test reflects within-group cellular variability only. i, UMAP representation of the trajectory partitioned into early (0–0.33), intermediate (0.33–0.67) and late (0.67–1) pseudotime stages. j, Proportion of cells in the early, intermediate and late stages by sample. Bars represent median values. Statistical test: two-sided chi-square test. The statistical test reflects within-group cellular variability only. k, Violin plots showing expression of ptx3a, twist2, scxa, tgfb1a, col12a1a, col12a1b and fn1b in EPDCs across all samples. Uninj, uninjured. l, Heatmap showing aggregated expression levels across samples and cell clusters of top downregulated genes in fbln2^PL^^−^^/^^−^ EPDCs at 4 dpci compared with wild types. m, Normalized nupr1b expression levels along the inferred trajectory of EPDCs. n, Normalized nupr1b expression across all samples.
First, we analyzed cell proportions per cluster within each sample at 4 dpci and found a gradual decrease in the proportion of EPDCs in fbln2^PL^^−^^/^^−^ and fbln2^FLD^^−^^/^^−^ ventricles compared with fbln2^+/+^ (Fig. 5c), suggesting the impairment of the epicardial response after injury. In view of the defects in EPDCs, we conducted a subclustering analysis of EPDCs across all samples and identified 13 subclusters (Fig. 5d,e). On the basis of previously characterized markers, we defined two main populations in uninjured ventricles: podxl^+^ epithelial EPDCs^15^ (clusters Ep2 and Ep7) and hapln1a^+^ mesenchymal EPDCs^15^ (clusters Ep1, Ep3, Ep5 and Ep8) (Extended Data Fig. 3a). Clusters Ep0, Ep4 and Ep10 that were largely absent in uninjured ventricles and emerged at 4 dpci were designated as activated EPDCs (Fig. 5d). The Ep0 cluster was characterized by the expression of the retinoic acid signaling response gene stra6 and ptx3a, the latter recently reported to mark transiently activated epicardial progenitor cells^15^ (Fig. 5f). By contrast, the Ep4 cluster showed elevated expression levels of cellular stress genes, such as manf (mesencephalic astrocyte-derived neurotrophic factor), as well as hypoxia-associated genes, including hyou1 (hypoxia up-regulated 1) (Fig. 5f). These data suggest that cells in the Ep4 cluster probably represent a population of EPDCs adapting to the hypoxic environment, a well-known feature of epicardial activation^18,60,61^. The Ep10 cluster was marked by expression of top2a, indicating that it is a cluster of proliferating EPDCs (Fig. 5f). Cells in the Ep6 cluster co-expressed podxl and hapln1a, along with endothelial markers such as spock3 and cdh5 (Extended Data Fig. 3a). Finally, Ep9, Ep11 and Ep12 contained few cells expressing macrophage and cardiomyocyte markers suggesting that these cells may probably represent doublets (Extended Data Fig. 3a).
To assess whether epicardial cell state transitions are disrupted in fbln2 mutants, we conducted trajectory analysis. We first identified root cells in the fbln2^+/+^ sample to establish a consistent reference across all conditions, then calculated pseudotime for each condition (Fig. 5g). The inferred trajectory shows EPDC activation and differentiation originating from epithelial and mesenchymal clusters (leaf nodes 1 and 2), converging in one main branch at cluster Ep0 (node 1), eventually leading to cluster Ep4 (Fig. 5g). Notably, cluster Ep0 (ptx3a^+^) appears as a transitional population of early activated cells as previously reported^15^, progressing toward the more mature Ep4 cluster (manf^+^). Comparison of cell distribution along pseudotime showed significantly divergent pseudotime distributions between the wild type and mutants (Fig. 5h). To better dissect this divergence, we divided the pseudotime into three intervals corresponding to ‘early’-, ‘intermediate’- and ‘late’-stage EDPCs (Fig. 5i). Both fbln2 mutants showed a gradual shift toward earlier pseudotime states compared with the pseudotime of wild-type EPDCs, indicating impaired progression along the activation trajectory (Fig. 5j). Notably, late-stage EPDCs were strongly reduced in fbln2^FLD^^−^^/^^−^, implying impaired progression to maturation. Conversely, fbln2^PL^^−^^/^^−^ mutants showed a significant reduction in intermediate EPDCs, suggesting an attenuation at an earlier stage of activation that does not impair the subsequent maturation phase (Fig. 5j).
To further investigate the molecular mechanisms underlying the distinct fbln2 mutant phenotypes, we analyzed the expression of key genes involved in epicardial activation. We found that ptx3a, twist2, scxa and tgfb1a were downregulated in fbln2^PL^^−^^/^^−^ EPDCs, and their expression was further reduced in fbln2^FLD^^−^^/^^−^ at 4 dpci compared with fbln2^+/+^ (Fig. 5k). Reduced ptx3a expression was also detectable at 7 dpci in fbln2 mutants by HCR (Extended Data Fig. 3b). Twist2 is a canonical downstream target of TGFβ signaling and a well-established EMT marker^62^. SCXA, a TGFβ target, regulates cardiac fibroblast activation in response to pro-fibrotic signals, and its knockout reduces fibrosis in pressure-overload-induced heart failure^63–65^. The expression levels of the TGFβ target genes twist2, scxa and gsc were strongly downregulated in fbln2^FLD^^−^^/^^−^ (Fig. 5k) and rescued upon fbln2 mRNA administration (Extended Data Fig. 4). These findings are consistent with the reduced epicardial TGFβ signaling observed in fbln2 mutant ventricles at 7 dpci. Conversely, col12a1a, col12a1b and fn1b, genes known to be upregulated by activated cardiac fibroblasts^8,10^, were downregulated in fbln2^PL^^−^^/^^−^ but upregulated in fbln2^FLD^^−^^/^^−^ (Fig. 5k).
To gain deeper insight into the epicardial defects observed in fbln2 mutants, we next analyzed the genes most differentially expressed in EPDCs at 4 dpci. We found that genes highly expressed in the wild-type Ep0 cluster were the most downregulated in fbln2^PL^^−^^/^^−^ EPDCs (Fig. 5l). As such, kng1 (kininogen-1), manf and nupr1b, all coding for stress-response factors, were among the top downregulated genes in fbln2^PL^^−^^/^^−^ EPDCs at 4 dpci (Fig. 5l). KNG1 is a pro-oxidant factor, with its overexpression leading to increased oxidative stress in the heart^66^. MANF, an Ep4 cluster marker, is a cardiomyokine (a secreted factor produced by cardiomyocytes) induced upon cardiac damage due to ischemia and ER stress^67^. As an endoplasmic-reticulum-resident chaperone, MANF exerts its cardioprotective role by enhancing protein folding and myocyte viability during reductive ER stress^68^. NUPR1 functions as a transcriptional regulator that is transiently induced in response to cellular stress conditions, including hypoxia^69^. Notably, nupr1b was the most downregulated gene in fbln2^PL^^−^^/^^−^ ventricles at 4 dpci across all cell types (Extended Data Fig. 5a). We confirmed the downregulation of nupr1b in fbln2 mutants by qPCR and HCR (Extended Data Fig. 5b,c). In view of these data, we examined nupr1b expression along the pseudotime trajectory and found that its expression was strongly associated with activated EPDC clusters (Ep0 and Ep4) (Fig. 5m). Interestingly, nupr1b expression was downregulated in both Ep0 and Ep4 clusters in fbln2^PL^ mutants, whereas in fbln2^FLD^ mutants, it was primarily lost in the Ep0 cluster, which is also markedly underrepresented in fbln2^FLD^ hearts (Fig. 5n). These patterns suggest that a stronger reduction in fbln2 levels may hamper early nupr1b-dependent epicardial activation.
Altogether, these data reveal a graded alteration of the epicardial response to injury that correlates with fbln2 levels, observing an attenuation in fbln2^PL^^−^^/^^−^ and a more severe impairment in fbln2^FLD^^−^^/^^−^. The downregulation of epicardial activation genes and cellular stress genes in EPDCs might underlie the reduction in myofibroblasts observed in fbln2^PL^^−^^/^^−^ during regeneration.
Nupr1b drives epicardial myofibroblast abundance and rescues defects in fbln2PL−/− ventricles
Our transcriptomic analyses identified nupr1b as the most downregulated gene in fbln2^PL^^−^^/^^−^ ventricles after injury. NUPR1 has been implicated in fibrotic disorders^70–72^. In renal fibrosis, Nupr1 promotes EMT and activates fibroblasts^73^. In the heart, Nupr1 is required for the induction of matrix metalloproteases (MMPs) via TNF in cardiac fibroblasts^74,75^. Importantly, NUPR1 has been shown to be regulated by TGFβ signaling^76,77^. On the basis of these observations, we hypothesized that nupr1b functions downstream of Fbln2–TGFβ signaling and may mediate the epicardial and fibrotic phenotypes observed in fbln2^PL^^−^^/^^−^.
We first analyzed the expression levels of nupr1b in ventricles at different time points after injury by qPCR. nupr1b expression was strongly upregulated up to 4 dpci and returned to basal levels by 21 dpci (Fig. 6a). Our single-cell RNA sequencing (scRNA-seq) data identified a robust induction of nupr1b expression in injury-activated EPDCs at 4 dpci (Fig. 5l). To gain spatial resolution, we analyzed nupr1b expression in wild-type ventricles after cardiac injury by in situ HCR. While nupr1b was nearly undetectable in uninjured ventricles, it was strongly induced in EPDCs at 4 dpci, as shown by colocalization with tcf21:mCherry (Fig. 6b). No expression of nupr1b was detected in border-zone endocardium at 4 dpci (Extended Data Fig. 6a).Fig. 6. Epicardial overexpression of nupr1b rescues the number of epicardial myofibroblasts in fbln2^PL^^−^^/^^−^ ventricles after cryoinjury.a, RT-qPCR analysis of nupr1b mRNA levels in uninjured ventricles at 2, 4 and 21 dpci (n = 4 or 5, pools of 2 ventricles). Data are shown as means ± s.d. and bars represent medians. Statistical test: one-way ANOVA followed by Tukey’s post hoc test. b, In situ HCR for nupr1b (green) and immunostaining for mCherry (magenta) on sections of uninjured TgBAC(tcf21:mCherry) ventricles and at 4 dpci. Arrowheads point to nupr1b^+^ EPDCs. c, Immunostaining of sections of cryoinjured ventricles of nupr1b^+/+^ and nupr1b^FLD^^−^^/^^−^ siblings at 7 dpci stained for Mef2 (CMs, magenta), PCNA (proliferation marker, green) and DAPI (DNA, blue). Arrowheads point to PCNA^+^ CMs. d, Percentage of PCNA^+^ CMs in the injury border zone of nupr1b^+/+^ and nupr1b^FLD^^−^^/^^−^ siblings at 7 dpci (n = 8). Box plots show the median and IQR (25th–75th percentiles), with whiskers spanning min–max values. Statistical test: two-tailed Student’s t-test. e, Immunostaining of sections of cryoinjured ventricles of nupr1b^+/+^ and nupr1b^FLD^^−^^/^^−^ siblings at 7 dpci stained for αSMA (green), MyHC (magenta) and DAPI (DNA, blue). Arrowheads point to αSMA^+^ cells in the injury periphery. f, Number of total, border-zone (BZ) and epicardial αSMA^+^ cells per mm^2^ injured area in nupr1b^+/+^ (n = 4) and nupr1b^FLD^^−^^/^^−^ (n = 7) ventricles at 7 dpci. Box plots show the median and IQR (25th–75th percentiles), with whiskers spanning min–max values. Statistical test: one-way ANOVA followed by Tukey’s post hoc test. g, Representative images of AFOG staining of sections of nupr1b^+/+^ and nupr1b^FLD^^−^^/^^−^ ventricles at 90 dpci. Orange, muscle; red, fibrin; and blue, collagen. h, Percentage representation of groups according to collagen area sizes at 90 dpci for cryoinjured nupr1b^+/+^ (n = 10) and nupr1b^FLD^^−^^/^^−^ (n = 9) ventricles. i, Schematic representation of 4-hydroxytamoxifen (4-OHT) or ethanol (EtOH) treatments followed by cardiac cryoinjury (CI) and heat-shock treatments (arrows). Immunostaining of sections of 7 dpci ventricles of fbln2^PL^^−^^/^^−^, TgBAC(tcf21:CreERT2); Tg(hsp70l:LBL-nupr1b-t2A-mCherry) zebrafish treated with 4-OHT or EtOH; sections stained for αSMA (green), MyHC (magenta) and DAPI (DNA, blue). Arrowheads point to epicardial αSMA^+^ cells. Number of total, border-zone and epicardial αSMA^+^ cells per mm^2^ injured area in fbln2^PL^^−^^/^^−^, TgBAC(tcf21:CreERT2); Tg(hsp70l:LBL-nupr1b-t2A-mCherry) ventricles treated with EtOH or 4-OHT at 7 dpci; n = 6. Box plots show the median and IQR (25th–75th percentiles), with whiskers spanning min–max values. Statistical test: one-way ANOVA followed by Tukey’s post hoc test. j, Immunostaining of sections of 7 dpci ventricles of fbln2^PL^^−^^/^^−^, TgBAC(tcf21:CreERT2); Tg(hsp70l:LBL-nupr1b-t2A-mCherry) zebrafish treated with EtOH or 4-OHT; sections stained for Mef2 (cardiomyocytes, magenta), PCNA (proliferation marker, magenta) and DAPI (DNA, blue). Arrowheads point to PCNA^+^ CMs. Percentage of PCNA^+^ CMs in the injury border zone of fbln2^PL^^−^^/^^−^, TgBAC(tcf21:CreERT2); Tg(hsp70l:LBL-nupr1b-t2A-mCherry) ventricles treated with EtOH (n = 6) or 4-OHT (n = 7) at 7 dpci. Box plots show the median and IQR (25th–75th percentiles), with whiskers spanning min–max values. Statistical test: two-tailed Student’s t-test. k, Representative images of AFOG staining and percentage of representation of groups according to collagen area sizes at 90 dpci of EtOH- (n = 5) and 4-OHT-treated (n = 7) fbln2^PL^^−^^/^^−^, TgBAC(tcf21:CreERT2); Tg(hsp70l:LBL-nupr1b-t2A-mCherry) ventricles. Orange, muscle; red, fibrin; and blue, collagen. White dashed lines delineate the border of the injured tissue, black dashed lines delineate the area with collagen deposits and red dashed lines delineate the regenerated myocardium. Scale bars, 100 µm.Source data
Next, to investigate the role of nupr1b during cardiac regeneration, we generated a nupr1b full-locus deletion mutant (nupr1b^FLD^) (Extended Data Fig. 6b). nupr1b expression was undetectable in nupr1b^FLD^^−^^/^^−^ ventricles, and the expression levels of its paralog nupr1a remained unchanged in mutants (Extended Data Fig. 6c). We then examined whether loss of nupr1b affects the regenerative response and found that CM proliferation was significantly reduced in nupr1b^FLD^^−^^/^^−^ compared with wild-type siblings at 7 dpci (Fig. 6c,d).
NUPR1 has been implicated in fibroblast activation, and our data show that it is expressed in EPDCs in response to injury. Thus, we hypothesized that nupr1b downregulation might be a key driver of the alterations in epicardial αSMA^+^ myofibroblast observed in fbln2^PL^^−^^/^^−^. To test this possibility, we quantified αSMA^+^ cells in nupr1b^FLD^^−^^/^^−^ ventricles at 7 dpci and found that the number of epicardial αSMA^+^ cells was significantly reduced in mutant ventricles (Fig. 6e,f). Moreover, nupr1b^FLD^^−^^/^^−^ showed high levels of collagen at 90 dpci compared with nupr1b^+/+^ siblings, indicating fibrosis resolution failure and scarring (Fig. 6g,h). These findings further support a role for Nupr1b in regulating the number of epicardial αSMA^+^ myofibroblasts and show that its complete ablation impairs regeneration.
As nupr1b deletion phenocopied the myofibroblast reduction observed in fbln2^PL^^−^^/^^−^ ventricles, we sought to determine whether nupr1b overexpression could rescue this phenotype. To recapitulate the tissue-specific and transient nature of nupr1b expression after injury, we generated a heat-shock-inducible Tg(hsp70l:LBL-nupr1b-t2A-mCherry) overexpression line using the HOTCre system^78^. In combination with the TgBAC(tcf21:CreERT2) line^79^, this strategy allows spatial and temporal control of nupr1b expression in EPDCs (Fig. 6i). We found that overexpression of nupr1b in EPDCs of fbln2^PL^^−^^/^^−^ ventricles increased the number of epicardial αSMA^+^ cells when compared with uninduced fbln2^PL^^−^^/^^−^ ventricles at 7 dpci (Fig. 6i). Moreover, CM proliferation at 7 dpci was also increased in fbln2^PL^^−^^/^^−^ ventricles after epicardial overexpression of nupr1b compared with EtOH-treated fbln2^PL^^−^^/^^−^ siblings (Fig. 6j). Finally, we reasoned that nupr1b overexpression might prevent the improvement in fibrosis resolution observed in fbln2^PL^^−^^/^^−^. Indeed, we found that epicardial overexpression of nupr1b in fbln2^PL^^−^^/^^−^ leads to increased late-stage fibrosis at 90 dpci compared with EtOH-treated fbln2^PL^^−^^/^^−^ siblings (Fig. 6k).
Rescue of epicardial myofibroblast numbers through nupr1b overexpression in fbln2^PL^ mutants suggests the existence of an epicardial Fbln2–Nupr1b axis, potentially acting through TGFβ signaling. To test this possibility, we used pharmacological inhibition of ALK5/7/4 receptors with SB431542 in wild-type fish (Extended Data Fig. 7). SB431542 treatment in wild types reduced nupr1b expression at 4 dpci, indicating that nupr1b is downstream of TGFβ signaling. Similar nupr1b downregulation was observed in fbln2^PL^ mutants. Interestingly, upon SB431542 treatment, nupr1b expression levels remained unchanged, consistent with already impaired TGFβ signaling in these mutants. These data further support the Fbln2–ALK5/7/4-mediated TGFβ signaling axis as an upstream regulator of nupr1b. Moreover, the number of epicardial pSmad3^+^ cells was increased upon nupr1b induction in fbln2^PL^^−^^/^^−^ ventricles (Extended Data Fig. 8a,b).
Overall, we show that Nupr1b regulates the number of epicardial αSMA^+^ myofibroblasts after injury. Our data show that epicardial overexpression of nupr1b rescues regeneration phenotypes in fbln2^PL^ mutants and identify Nupr1b as a Fbln2 effector required for TGFβ-mediated epicardial responses during regeneration.
Discussion
Fibrosis develops after cardiac injury. While the initial fibrotic response is critical for preserving structural integrity and providing cellular support, failure to resolve it as a consequence of maladaptive fibrosis results in scarring. Zebrafish are capable of efficiently resolving injury-induced cardiac fibrosis^6–8^. The mechanisms that maintain the balance between fibrosis and regeneration are crucial yet remain poorly understood. Here we identify Fbln2 as a regulator of this balance and show how its manipulation can promote either transient or maladaptive fibrosis. Mutants with reduced fbln2 expression showed attenuated transient fibrosis, whereas complete abrogation led to scarring. These phenotypes correlated with the degree of epicardial TGFβ inhibition: mild attenuation improved fibrosis resolution, while strong suppression impaired regeneration.
Beyond their fibrogenic role, TGFβ signaling and cardiac fibroblasts also support regeneration. Early ablation of col1a2-expressing cardiac fibroblasts or inhibition of TGFβ signaling results in reduced CM proliferation^8,20^. In line with these data, we observed a reduction in early CM regeneration programs and EC proliferation in fbln2^PL^^−^^/^^−^ ventricles, reinforcing the notion that Fbln2 is essential for efficient regeneration, in addition to its fibrogenic function. Manipulations leading to reduced CM and EC proliferation are typically associated with loss of regenerative capacity in zebrafish^18,20,51,52^. It has been shown that ablation of collagen-expressing cells reduced early CM proliferation without impacting long-term regenerative outcomes^8^. Similarly, regeneration was not affected in fbln2^PL^ mutant ventricles, which showed reduced fibrosis at stages of advanced regeneration. It is possible that this initial reduction in cell proliferation does not directly translate into a net loss of cells at the end of the regeneration process. In view of these data, it is tempting to hypothesize that the response to cardiac injury in zebrafish may be initially excessive and its attenuation, to some extent, does not impair ultimate regeneration.
Our scRNA-seq analyses revealed that epicardial activation was attenuated in fbln2^PL^ and further suppressed in fbln2^FLD^. Trajectory analyses revealed that, in contrast to fbln2^FLD^ mutants*, fbln2*^PL^ mutants show a decrease in intermediate EPDCs, pointing to an earlier alteration in EPDC activation that does not hinder later maturation (Extended Data Fig. 9). Along with the previously characterized markers of activated epicardial progenitor cells such as ptx3a and col12a1b (ref. ^15^), we identified a subgroup of activated EPDCs enriched in stress-responsive genes such as manf and hyou1. The upregulation of these genes in activated EPDCs suggests a mechanism for cellular adaptation to injury, which favors EMT and prevents cellular damage. The downregulation of these genes in fbln2^PL^^−^^/^^−^ ventricles may indicate a disruption in this response, thereby potentially resulting in attenuation of epicardial myofibroblasts. Several downregulated EPDC genes in fbln2^PL^ mutants were downstream targets of TGFβ signaling, such as twist2, tgfb1a, scxa and nupr1b, supporting a Fbln2–TGFβ signaling axis regulating epicardial activation.
Our data implicate Nupr1b as a Fbln2–TGFβ signaling effector regulating epicardial activation and myofibroblast abundance after injury. In mice, Nupr1 is necessary for both cardiac collagen deposition and induction of collagen-degrading MMP expression in fibroblasts^74,75^. Nupr1-deficient mice show a complex cardiac phenotype characterized by elevated basal levels of fibrotic markers and collagens, yet present attenuated fibrosis following transverse aortic constriction-induced cardiac hypertrophy^75^. This paradoxical phenotype may result from the compensatory upregulation of alternative MMPs^75^. Here we show that nupr1b is strongly upregulated in EPDCs early after cardiac injury and that nupr1b^FLD^ mutants show defects in epicardial myofibroblasts. Rescue of the number of myofibroblasts and pSmad3^+^ cells through nupr1b overexpression suggests that Nupr1b functionally contributes to these processes in fbln2 mutants. Rescue of CM proliferation defects further supports the restoration of a wild-type (WT)-like regenerative milieu. Pharmacological epistasis results suggest a model in which Fbln2–ALK5/7/4-mediated TGFβ signaling acts upstream of nupr1b. Yet, the precise mechanism by which Nupr1b regulates myofibroblast specification remains to be fully elucidated. These findings support the existence of an epicardial Fbln2–TGFβ–Nupr1b axis that is essential for cardiac regeneration.
Overall, we propose that Fbln2 is an EPDC-secreted factor capable of modulating the extent of TGFβ signaling activation, which in turn influences the critical balance between regeneration and transient fibrosis. We show that manipulating Fbln2 can alter this balance and shape the long-term cardiac regenerative outcome. Our findings underscore the complex cross talk between regenerative and fibrogenic processes, emphasizing the pleiotropic roles of signaling pathways that govern them. Notably, our study identifies a Fbln2–Nupr1b pathway that selectively regulates the number of epicardial myofibroblasts. Dissecting the mechanisms involved in transient fibrosis will help better identify and target cellular processes toward improved regeneration and restrained fibrosis.
Methods
Zebrafish lines
All experimental protocols involving zebrafish (Danio rerio) were approved by the Comité Institutionnel des Bonnes Pratiques Animales en Recherche (protocol number 2021-3308 #833) of the Azrieli Research Center of Centre Hospitalier Universitaire Sainte-Justine (Université de Montréal). Both male and female zebrafish aged between 3 and 12 months old were used in this study. We used TgBAC(tcf21:mCherry-NTR)^pd108^ (ref. ^80^) (hereafter TgBAC(tcf21:mCherry)), TgBAC(cryaa:EGFP, tcf21:CreERT2)^pd42^ (ref. ^79^) (hereafter TgBAC(tcf21:CreERT2)), Tg(-0.8flt1:RFP)^hu5333^ (ref. ^81^) and Et(krt4:EGFP)^sqet331AEt^ (ref. ^82^). Adult zebrafish were anesthetized in 0.016% tricaine and cryoinjuries were performed as previously described^4,6,7^. For TGFβ inhibitor treatment, adult fish were treated with 10 μM of SB431542 (BOC Sciences, catalog number B0084-211014) or DMSO, starting from 1 day before cryoinjury until 4 dpci with the solution refreshed every 2 days.
Generation of mutant and transgenic lines
CRISPR–Cas9 was used to generate fbln2^sjr3^ promoter-less (hereafter fbln2^PL^), fbln2^sjr7^ full-locus deletion (hereafter fbln2^FLD^) and nupr1b^sjr14^ full-locus deletion (hereafter nupr1b^FLD^) alleles. Guide RNAs (gRNAs) targeting fbln2 (GRCz11, NC_007122.7) and nupr1b (GRCz11, NC_007114.7) were synthesized using the cloning-free method previously described^83^. Oligos used to synthesize gRNAs are listed in Supplementary Table 1. One-cell-stage embryos were injected with 80–120 pg of gRNAs and 2.5 μM of Cas9 protein (Integrated DNA Technologies, catalog number 1081058). Fish were genotyped by PCR and Sanger sequencing (Extended Data Figs. 1a,b and 4b, and Supplementary Table 1). The fbln2^PL^ allele consists of a 1,594-bp deletion spanning the transcription start site, fbln^FLD^ consists of a 109,378-bp deletion spanning 5′ UTR (untranslated region) and 3′ UTR, and nupr1b^FLD^ consists of an 884-bp deletion from 5′ to 3′ UTR. To create the Tg(hsp70l:loxP-BFP-loxP-nupr1b-t2A-mCherry)^sjr18^ line (hereafter Tg(hsp70l:LBL-nupr1b-t2A-mCherry)), the nupr1b coding sequence (NM_001281922.1) was amplified by PCR and cloned under the hsp70l promoter into a Tol2 plasmid using T4 DNA ligase (NEB, catalog number M0202). One-cell-stage embryos were injected with 20 pg of hsp70l:LBL-nupr1b-t2A-mCherry plasmid along with 30 pg of Tol2 mRNA. Injected embryos that were positive for BFP after heat-shock treatment were raised and founders identified by screening the offspring for BFP expression following three heat-shock treatments at 39 °C for 1 h each (Extended Data Fig. 2a). Two independent founders were identified and one generation raised. The precise number of insertion sites per line has not been investigated. For the HOTCre approach^78^, fbln2^PL^^−^^/^^−^; TgBAC(tcf21:CreERT2); Tg(hsp70l:LBL-nupr1b-t2A-mCherry) fish were recombined with incubation in the dark in system water supplemented with 5 µM 4-hydroxytamoxifen (Sigma, catalog number H7904) or vehicle (ethanol) for 12 h d^−1^, for 3 consecutive days before cryoinjury. For heat-shock treatments following cryoinjury, all fish were incubated in preheated system water at 39 °C for 1 h every 12 h until 4 dpci. Control animals carried the HOTCre transgenes and received the heat-shock regimen but were given vehicle (ethanol) instead of tamoxifen. The confirmation of recombination has been done using qPCR to show induction of nupr1b in recombined fish and mCherry expression to confirm recombination in the tcf21^+^ cells (Extended Data Fig. 2a–c). For fbln2 mRNA injections, the pT7-fbln2-Amp plasmid containing fbln2 CDS was ordered from Twist Bioscience and used as a template for in vitro transcription using a HiScribe T7 Quick High Yield RNA Synthesis Kit (New England Biolabs, catalog number E2050S). One-cell-stage fbln2^FLD^^−^^/^^−^ embryos were injected with 100 pg of fbln2 mRNA and collected at 3 days post-fertilization (dpf) for subsequent RNA isolation.
Histological analyses, imaging and quantification
Adult zebrafish hearts were fixed in 4% paraformaldehyde overnight at 4 °C, transferred in 30% sucrose and embedded in optimal cutting temperature compound. Immunostaining was performed on 8-μm cryosections as previously described, with an initial additional step of antigen retrieval using citrate-based antigen unmasking solution (MJS BioLynx, catalog number VECTH3300) at 96 °C for 10 min (ref. ^54^). Antibodies and working concentrations are listed in the Reporting Summary. DAPI was used as a nuclear counterstain. For CHP staining (3Helix, catalog number RED60), 8-μm cryosections were stained overnight at 4 °C using a 10 μM working solution prepared according to the manufacturer’s instructions^84^. For AFOG staining, 8-μm heart cryosections were hydrated in distilled water for 5 min and incubated in Bouin’s solution for 2 h at 60 °C followed by an additional hour at room temperature (RT). Slides were rinsed in running tap water for 30 min, then incubated in 1% phosphomolybdic acid for 90 s. Slides were rinsed in distilled water for 5 min and incubated in AFOG staining solution (0.5% aniline blue, 1% orange G, 1% acid fuchsin, pH 1.09) for 3 min. After rinsing in distilled water for 2 min, slides were dehydrated twice in 95% EtOH, followed by two washes in absolute EtOH for 5 min each. Slides were washed twice in xylene for 2 min each and mounted in Entellan (Sigma, catalog number 1079610500). Sections from AFOG staining were imaged using a Nikon SMZ18 stereomicroscope. For HCR RNA-FISH (Molecular Instruments), the manufacturer’s protocol for fixed fresh frozen tissue sections was followed by omitting the proteinase K treatment, as previously described^85^. Imaging was performed on a Leica TCS SP8 laser scanning confocal microscope.
Imaging and quantifications were done averaging two or three non-superficial and non-consecutive sections for each biological replicate. Quantification of CM proliferation and dedifferentiation was performed by counting cells within the 100-μm area from the border zone of the injury. For the quantification of pSmad3^+^ cells, cells co-localizing with Cav1 were designated as ‘epicardial’ and the remaining positive cells located within the injury were designated as ‘intraventricular’. For αSMA^+^ cells, cells located at the superficial periphery of the injury and underlining the epicardium were designated as ‘epicardial’ and cells located at the injury border zone as ‘border zone’ αSMA^+^ cells. αSMA^+^ vascular smooth muscle cells and cardiomyocytes were excluded from quantification. In both cases, the surface density of cells was calculated as the number of positive cells relative to the area of the injury. CHP fluorescence intensity within the injured area was quantified using Fiji (ImageJ). For each section, the integrated fluorescence intensity of the CHP channel was measured within the injured area. The total integrated density was then normalized to the corresponding injury area (µm^2^) to obtain the CHP intensity per µm^2^ of injury. Background fluorescence measured in a non-injured area of the ventricle was subtracted from all values before normalization. For collagen area analyses, the collagen area was determined by color deconvolution using the Masson Trichrome setting in Fiji^86^, and was calculated as a ratio from the total ventricular surface. All non-superficial sections were stained, and the sections with the largest collagen areas were chosen for AFOG quantification at 90 dpci.
Quantitative real-time PCR
Total RNA was isolated from pools of five mechanically homogenized larvae at 5 dpf or from pools of two ventricles per biological replicate using TRIzol (Invitrogen, catalog number 15596026) following the manufacturer’s instructions. For cDNA synthesis, the RevertAid H Minus First Strand cDNA Kit (Thermo Scientific, catalog number K1632) was used following the manufacturer’s protocol. RT-qPCRs were performed using iTaq Universal SYBR Green Supermix (Bio-Rad, catalog number 1725122) on the CFX Opus 96 Real-Time PCR System (Bio-Rad). Expression levels of genes of interest were normalized to the expression of rpl13a, and expression fold changes were calculated using the ΔΔCt method^87^. The primer sequences used are provided in Supplementary Table 1.
Single-cell dissociation
Pools of four uninjured ventricles and cryoinjured ventricles at 4 dpci were extracted from WT, fbln2^PL^^−^^/^^−^ and fbln2^FLD^^−^^/^^−^ zebrafish. Anesthetized fish were dissected and ventricles extracted in calcium- and magnesium-free HBSS (Sigma, catalog number H9394) supplemented with 20 U ml^−1^ of heparin (Sigma, catalog number H3393). Ventricles were washed in fresh HBSS and digested into 1 ml of 1 mg ml^−1^ of collagenase (Gibco, catalog number 17101015) in HBSS for 30 min on a thermal mixer at 32 °C and 700 rpm gentle agitation. Ventricles were mechanically disrupted by pipetting every 10 min. Samples were centrifuged at 500 × g for 5 min at RT and supernatants discarded. Cells were digested with 1 ml of 1× TrypLE Express Enzyme (Gibco, catalog number 12605010) for 15 min on a thermal mixer at 32 °C and 700 rpm gentle agitation and pipetted every 5 min. Enzymatic digestion was stopped by adding 1 ml of 20% fetal bovine serum in HBSS to extracts. Tissue extracts were strained through a 70-μm cell strainer and centrifuged at 500 × g for 5 min at RT. The supernatants were discarded, and pellets were washed with 1 ml of 20% fetal bovine serum in HBSS. After centrifugation at 500 × g for 5 min at RT, cells were resuspended in 80 μl of 20% fetal bovine serum in HBSS and strained through a 40-μm cell strainer. Cells were counted and viability (>85%) was assessed using 0.4% trypan blue. Cells were kept on ice and immediately processed for scRNA-seq.
scRNA-seq and analysis
For the scRNA-seq experiment, 14,000 cells were used for each sample and the Chromium Next GEM Single Cell 3′ Reagent Kit v3.1 (10x Genomics, catalog number PN-1000269) was used following the manufacturer’s protocol. Each sample was indexed individually using the Dual Index Kit TT Set A (10x Genomics, catalog number PN-1000215). After quality control, libraries were run on a NovaSeq 6000 system (Illumina) at a depth of approximately 300 M reads per sample. FASTQ files were processed individually in 10x Genomics Cloud Analysis using the Cell Ranger Count v6.1.2 pipeline, and reads were aligned to the GRCz11 v4.3.2 reference genome^88^. Filtered gene expression matrices were imported and further processed in Seurat v4 (ref. ^89^). Samples were integrated to create a single aggregated Seurat object, and low-quality cells containing less than 200 unique features were removed. Batch effect correction was performed using the data integration and ScTransform functions in Seurat. To increase the retrieval of cardiomyocytes characterized with high mitochondrial RNA content, an initial filtering was applied to remove cells with more than 40% mitochondrial RNA^90^. After initial clustering and annotation of CMs, non-CM cells with greater than 10% mitochondrial RNA were excluded from further analysis. The remaining cells were then re-clustered and annotated based on the expression of marker genes. Following normalization, data visualization was done in Seurat, incorporating previously published R code with modifications^90^. Monocle3 was used for trajectory analyses on the integrated Seurat object using the uniform manifold approximation and projection (UMAP) embedding of EPDCs^91^. First, we constructed a reference trajectory on WT cells. To define the root, we first selected leaves by identifying graph leaf nodes, computing their centroids and selecting as root the leaf closest to the manually selected predefined cell. The centroid of these WT root cells was then used to anchor a joint trajectory for WT, fbln2^PL^^−^^/^^−^ and fbln2^FLD^^−^^/^^−^ samples, and the leaf whose centroid was closest to the WT root centroid was selected as root, after which pseudotime was computed for all cells. To better dissect pseudotime distribution across samples, we partitioned the pseudotime into three intervals: early (0–0.33), intermediate (0.33–0.67) and late (0.67–1). We compared pseudotime distributions and proportions of cells in each stage across genotypes using nonparametric Kruskal–Wallis and chi-square tests.
Statistics and reproducibility
Data were statistically analyzed and graphics created on GraphPad Prism v.10. For each dataset, normality was assessed using the Shapiro–Wilks normality test (alpha value of 0.05). When two groups were compared, comparative statistics were performed using two-tailed Student’s t-test for parametric data or Mann–Whitney U test for nonparametric data, as stated in the figure legends. One-way ANOVA followed by Tukey’s post hoc tests was used when more than two groups were compared. Exact P values are indicated in the figures. In all bar graphs, error bars represent mean ± s.d. and bars represent median values. All box plots show the median (center line), interquartile range (25th–75th percentiles; box) and whiskers indicating minimum and maximum values. Imaging and quantifications were done averaging two or three non-superficial and non-consecutive sections for each biological replicate. For in situ HCR, stainings were repeated in at least three independent experiments.
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
Supplementary information
Reporting Summary Supplementary TablesSupplementary Tables 1–3.
Source data
Source Data Fig. 2. Statistical source data. Source Data Fig. 3. Statistical source data. Source Data Fig. 4. Statistical source data. Source Data Fig. 6. Statistical source data. Source Data Extended Data Fig. 1. Statistical source data. Source Data Extended Data Fig. 1. Unprocessed gel. Source Data Extended Data Fig. 2. Statistical source data. Source Data Extended Data Fig. 4. Statistical source data. Source Data Extended Data Fig. 5. Statistical source data. Source Data Extended Data Fig. 6. Statistical source data. Source Data Extended Data Fig. 6. Unprocessed gel. Source Data Extended Data Fig. 7. Statistical source data. Source Data Extended Data Fig. 8. Statistical source data.
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1Global Burden of Disease 2021: Findings from the GBD 2021 Study (Institute for Health Metrics and Evaluation (IHME), 2024).