Hypoxia-activated scleraxis a mediates epicardial progenitor differentiation into a unique cardiac perivascular cell type

Abstract

The epicardium provides progenitor cells and paracrine signals essential for heart development and regeneration, yet the mechanisms regulating epicardial cell fate remain poorly understood. Here, we identify the transcription factor Scleraxis a (scxa) as a key regulator of epicardial progenitor differentiation in zebrafish. Single-cell transcriptomics, genetic lineage tracing, and cardiac injury models reveal transient scxa expression in activated epicardial progenitor cells (aEPCs) during developmental coronary angiogenesis and heart regeneration. scxa⁺ epicardial cells predominantly differentiate into a previously uncharacterized col18a1a⁺ perivascular population, termed epicardial-derived perivascular mesenchymal cells (Epi-PMCs), which is distinct from pericytes, vascular smooth muscle cells, and mammalian adventitial fibroblasts. Epi-PMCs closely associate with coronary vessels and may contribute to vascular stabilization and remodeling, potentially through collagen XVIII. Loss of scxa increases coronary vessel density. Hypoxia and Hif signaling induce scxa expression, identifying a hypoxia-responsive mechanism that promotes epicardial differentiation toward a vascular-supportive fate during heart development and regeneration.

Here they identify a hypoxia-responsive pathway that directs epicardial progenitor cells into a new vessel-supporting cell type, uncovering a key mechanism by which the epicardium orchestrates heart development and regeneration.

Introduction

The epicardium, a thin layer of mesothelial cells surrounding the heart, plays a pivotal role in heart development and regeneration. During embryonic heart development, the epicardium gives rise to various cell types, including fibroblasts, vascular smooth muscle cells (vSMCs), and pericytes, making it a critical source of progenitors essential for coronary vessel formation and myocardial development^1,2. Several signaling pathways and factors have been implicated in regulating epicardial cell differentiation. For instance, TGF-β promotes differentiation into fibroblasts and vSMCs, while Wnt or Notch signaling drives the vSMC fate^3–6. PDGF signaling was reported to guide epicardial-mesenchymal transition (EMT) and differentiation into vSMCs and pericytes^7,8. Transcription factors like Tcf21 and WT1 are crucial for epicardial progenitor specification, EMT, and fibroblast lineage commitment^9–12.

In response to cardiac injury, the epicardium is reactivated, reinitiating a developmental program that supports tissue repair and regeneration by contributing cells and releasing paracrine factors^2,13,14. During regeneration, several subpopulations of epicardial cells emerge, including an activated epicardial progenitor cell type (aEPC) that we recently identified¹⁵. In addition to renewing the epithelial epicardium, these aEPCs migrate to enclose the wound, undergo EMT, and differentiate into pdgfrb⁺ mural cells and hapln1a⁺pdgfra⁺ mesenchymal cells¹⁵. aEPCs upregulate a specific transcriptional program upon heart injury, including extracellular matrix (ECM) component gene col12a1b and pentraxin-related protein 3 (ptx3a)¹⁵. Despite these insights, it remains elusive what molecular and environmental cues determine whether epicardial progenitors differentiate into fibroblasts, vSMCs, pericytes, or other cell types and how epicardial cells coordinate with endothelial cells to form coronary vessels. In particular, the full spectrum of epicardial-derived mesenchymal cell fates remains incompletely defined, and it is unclear whether specialized epicardial-derived populations exist to support distinct stages of coronary vessel formation and maturation. Addressing these questions is critical for clarifying the regulatory mechanisms of epicardial differentiation and could significantly enhance strategies for tissue engineering and regenerative medicine.

Scleraxis (Scx) is a basic helix-loop-helix (bHLH) transcription factor best known for its pivotal role in embryonic mesodermal development, particularly in promoting tendon formation and tendon cell differentiation^16–20. Beyond its established role as a marker of tendon cell fate, Scx has broader functions in regulating cellular differentiation and tissue remodeling. In the heart, Scx was identified in the developing cardiac valves in mice, where it plays a key role in cell differentiation and extracellular matrix remodeling during valvulogenesis^21,22. Scx expression has also been observed in the pro-epicardial organ and embryonic epicardial cells and is downregulated after birth^23–25. Scx knockout mice exhibit ~50% fewer cardiac fibroblasts, suggesting a role in fibroblast lineage commitment²⁶, yet another study proposed an endothelial fate for Scx⁺ pro-epicardial cells²³, underscoring the need to clarify its diverse functions in epicardial and cardiac development. In addition, loss of Scx in mice results in hypocellular hearts, while prolonged post-injury Scx activation drives fibrotic myofibroblast differentiation²⁶. These observations suggest that Scx may regulate distinct epicardial-derived cell fates across different developmental and injury contexts, but the specific cell populations it controls and the mechanisms by which it influences coronary vessel formation remain poorly understood.

In the current study, we investigated the role of zebrafish scleraxis a (scxa) in aEPC cell fate and its contribution to heart development and regeneration. Using single-cell transcriptomics, we identified scxa expression in aEPCs and their mesenchymal progeny, with transient expression during coronary angiogenesis and strong induction under hypoxic conditions. Lineage tracing demonstrated that ventricular scxa⁺ cells originate from epicardial progenitors and mainly differentiate into a previously uncharacterized col18a1a⁺ perivascular cell population distinct from pericytes and vSMCs that we refer to as epicardial-derived perivascular mesenchymal cells (Epi-PMCs). The identification of Epi-PMCs expands the known repertoire of epicardial-derived cell fates and reveals a previously unrecognized epicardial contribution to the perivascular microenvironment during coronary vessel development and regeneration. Our findings reveal that hypoxia-driven scxa expression drives epicardial-to-perivascular differentiation, supporting coronary vasculature formation. We thus uncover a new transcriptional regulatory program for epicardial cell differentiation.

Results

scxa is expressed in differentiating epicardial progenitors in heart regeneration

In our search for regulators of epicardial cell differentiation, we looked for transcription factors enriched in aEPCs using our previously published single-cell transcriptomic dataset¹⁵. In this dataset, tcf21⁺ epicardial cells and their derivatives were isolated by FACS from tcf21:nucEGFP adult zebrafish hearts collected from uninjured controls and from hearts at 3 and 7 days post-ventricular amputation (dpa) injury. Following single-cell RNA sequencing and clustering analysis, 12 cell populations were identified, including aEPCs (cluster 5), epithelial epicardial cells (cluster 2), mural cells (cluster 6), and mesenchymal cells (clusters 0, 1, 3, 4, 7, and 9) (Fig. 1a). We found that the transcription factor scxa is expressed in a portion of the ptx3a⁺col12a1b⁺ aEPC cluster, as well as some epicardial progeny that take on a mesenchymal cell fate (hapln1a⁺) after heart injury (Fig. 1a, b, and Supplementary Fig. 1). scxa expression partially overlaps with col12a1b (Fig. 1c). To confirm scxa expression, we crossed a scxa:mCherry BAC (bacteria artificial chromosome) reporter²⁷ with the tcf21:nucEGFP line, which labels both the quiescent and activated epicardial cells and their progenies¹⁵. Without an injury, no or minimal scxa:mCherry signal was detected on the ventricular surface of adult hearts. By contrast, we observed prominent scxa expression at 3 dpa in the boundary zone of the wound, colocalizing with tcf21:nucEGFP (Fig. 1d). Co-expression of scxa and tcf21 is further confirmed in cryosections of hearts at 3 dpa. Notably, many of these scxa⁺ tcf21⁺ cells reside in the compacted myocardium, indicating a mesenchymal, rather than epithelial, identity (Fig. 1e).

Fig. 1. scxa is expressed in differentiating epicardial progenitors during heart regeneration.

a Cell trajectories suggested by pseudotime analysis with Monocle 3 are shown on a UMAP. b Normalized expression (log1p-transformed expression) of selected genes on UMAPs. c A Venn diagram showing numbers of cells expressing col12a1b and/or scxa at 3 dpa. d Whole-mount images of the ventricular surface showing expression of tcf21:nucEGFP (green) and scxa:mCherry (magenta) at 3 dpa. The framed region is enlarged to show details on the right. Arrows indicate representative mCherry⁺ EGFP⁺ cells. e A section image of a 3 dpa heart showing tcf21:nucEGFP (green) and scxa:mCherry (magenta) expression. DAPI staining is shown in blue. The framed region is enlarged to show details on the right. Arrows denote representative mCherry⁺ EGFP⁺ cells. f A section image of a 7 dpa heart showing col12a1b:EGFP (green) and scxa:mCherry (magenta) expression. HCR staining of hapln1a is shown in yellow. Single-channel images are shown in grayscale. Arrows denote representative mCherry⁺ hapln1a⁺ cells. g, h Whole-mount images of the ventricular surface showing expression of fli1a:EGFP (green) and scxa:mCherry (magenta) at 7 dpa (g) and 14 dpa (h). The framed regions are enlarged to show details on the right with single-channel images shown in grayscale. Arrows indicate representative mCherry⁺ cells adjacent to vessels.

We previously identified col12a1b and ptx3a as markers for aEPCs in zebrafish heart regeneration (Supplementary Fig. 1)¹⁵. To confirm scxa expression in aEPCs, we crossed the scxa:mCherry line with the aEPC reporter col12a1b:EGFP. As shown in Fig. 1f, scxa expression partially overlapped with that of col12a1b at the boundary zone of the wound at 7 dpa, while scxa^- col12a1b⁺ aEPCs occupied the wound. Similar colocalization patterns for ptx3a and scxa were observed (Supplementary Fig. 1b). We wondered whether some scxa⁺ cells gain mesenchymal identity by expressing the mesenchymal epicardial cell marker hapln1a¹⁵. We thus performed hybridization chain reaction (HCR) staining targeting hapln1a on cryosection samples from 7 dpa. As expected, some of the scxa⁺ cells in the inner layer of the ventricular wall turned off col12a1b expression but gained hapln1a expression (Fig. 1f). These expression patterns perfectly recapitulate our scRNA-seq result, confirming that scxa transcripts are found both in the col12a1b⁺ aEPCs as well as their mesenchymal progeny. This suggests that acquisition of scxa expression in aEPCs accompanies their transition to a mesenchymal cell fate.

The hapln1a⁺ mesenchymal epicardial cells are reported to support compact myocardium and coronary vessel growth and regeneration^15,28,29. To characterize the scxa⁺ mesenchymal epicardial cells, we analyzed the scxa expression in hearts carrying the pan-endothelial reporter fli1a:EGFP to visualize revascularization. We observed that many scxa⁺ cells are associated with regrowing blood vessel sprouts in the wound site at 7 dpa (Fig. 1g). By 14 dpa, heart wounds were still enriched in scxa⁺ cells where they form a network surrounding the regenerating vasculature (Fig. 1h). scxa⁺ cells are seen frequently associating with coronary blood vessels (insets in Fig. 1h), indicating a perivascular identity. Taken together, these results suggest that scxa is associated with epicardial cell differentiation to perivascular cells during heart regeneration.

Transient scxa⁺ epicardial cells are associated with coronary angiogenesis in heart development

The previously unperceived vascular association of scxa⁺ epicardial lineage is intriguing. Because scxa expression was first identified in aEPCs during heart regeneration, we next sought to determine the developmental origin, lineage relationships, and physiological function of these cells. Coronary vessel formation during juvenile heart development begins at around 5-6 weeks post fertilization (wpf)³⁰, providing a tractable model to investigate epicardial differentiation and angiogenesis in the absence of injury-induced inflammation and tissue remodeling. We therefore first characterized the identity, lineage, and function of scxa⁺ epicardial-derived cells during coronary vessel development and subsequently examined whether these developmental findings extend to heart regeneration. To determine whether scxa⁺ cells exhibit characteristics of aEPCs during juvenile heart development, we analyzed their co-expression with the aEPC marker col12a1b. Before 4 wpf, ventricular scxa expression and col12a1b expression were undetectable. At approximately 4 wpf, col12a1b:EGFP expression emerged in epicardial cells prior to the onset of scxa:mCherry expression, predominantly near the atrioventricular canal (AVC), where coronary vessel sprouts are known to initiate (Fig. 2a)³⁰. However, the coronary vessels had not developed at this stage. At around 5 wpf, ventricular scxa expression was observed in col12a1b⁺ cells near the AVC (Fig. 2b). This suggests that col12a1b activation in aEPCs precedes scxa expression, mirroring the pseudotime trajectory observed in heart regeneration.

Fig. 2. Transient scxa⁺ epicardial cells are associated with coronary angiogenesis.

a, b Whole-mount images showing scxa:mCherry (magenta) and col12a1b:EGFP (green) expression at 4 wpf (a) and 5 wpf (b). Arrows indicate representative cells. AVC, atrioventricular canal. Dashed lines indicate approximate AVC regions. c Whole-mount images showing scxa:mCherry (magenta) and fli1a:EGFP (green) expression at 5-6 wpf, with framed regions enlarged on the right. Single-channel images are shown in grayscale. The dashed line indicates approximate AVC. Arrows denote representative scxa⁺ cells. Single-channel images are shown in grayscale. d Whole-mount image demonstrating direct contact of scxa:mCherry⁺ (magenta) and fli1a:EGFP⁺ (green) cells with framed regions enlarged at the bottom. Single-channel images are shown in grayscale. Arrows denote scxa:mCherry⁺ cells. e Whole-mount images showing scxa:mCherry (magenta) and fli1a:EGFP (green) expression at around 7 wpf, with framed regions enlarged on the right. Single-channel images are shown in grayscale. Dashed lines indicate approximate AVC regions. Arrows mark scxa⁺ cells, while the arrowhead indicates an endothelial tip cell. f Whole-mount images showing scxa:mCherry (magenta) and fli1a:EGFP (green) expression at around 9 wpf, with framed regions enlarged on the right. The single-channel image is shown in grayscale. The dashed line indicates approximate AVC. Arrows denote representative scxa⁺ cells. g Whole-mount images showing scxa:mCherry (magenta) and fli1a:EGFP (green) expression in the adult heart. The dashed line indicates approximate AVC.

At 5-6 wpf, all hearts exhibiting scxa expression contained a subset of scxa⁺ cells associated with the vasculature (Fig. 2c). Vessel-associated scxa⁺ cells were observed either in contact with the vessel or as part of a surrounding cellular ensemble (insets, Fig. 2c). High-resolution imaging demonstrated direct cell contact between scxa⁺ cells and fli1a⁺ endothelial cells (Fig. 2d). Interestingly, in ventricular regions not yet vascularized, clusters of scxa⁺ cells were organized in parallel orientations. Endothelial tips from nearby vasculature appeared to orient toward these scxa⁺ clusters, suggesting directed growth of vessel sprouts toward these cells (Fig. 2c, frame 2 right side). This expression pattern continued to 7 wpf (Fig. 2e, frames) when scxa expression peaked. Decreased scxa expression was observed along with increased coronary vessel coverage from 7 to 9 wpf (Fig. 2e, f). In adult hearts with an established coronary vasculature, ventricular scxa expression was minimal (Fig. 2g). These observations suggest that transient scxa⁺ epicardial cells, present before the emergence of endothelial sprouts and during vessel growth, may facilitate coronary angiogenesis.

scxa⁺ epicardial progenitors predominantly differentiate into col18a1a⁺ mesenchymal cells

Noting the correlation of scxa⁺ cells with angiogenesis and their transient presence, we sought to determine their identity and progenies. We first explored the origin of ventricular scxa⁺ cells, and trajectory analysis indicated that these cells arise from the epithelial layer of the epicardium (Fig. 3a). To test this, we performed lineage tracing using animals carrying the tcf21:CreER^t2, ubb:loxP-TagBFP-stop-loxP-EGFP, and scxa:mCherry transgenes^31,32 (Fig. 3b). Following 4-hydroxytamoxifen (4HT) treatment during embryonic stages (2-4 dpf), analysis at 7 wpf revealed that all ventricular surface scxa⁺ cells expressed Cre-induced EGFP⁺, indicating their derivation from epicardial or pro-epicardial cells.

Fig. 3. scxa⁺ epicardial cells give rise to col18a1a⁺ cells.

a Cell trajectories suggested by pseudotime analysis with Monocle 3 are shown on a UMAP. Normalized expression (log1p transformed expression) of selected genes (col18a1a, vegfaa, and pdgfrb) is on UMAPs, with the mural cell cluster circled. b Whole-mount images showing scxa:mCherry (magenta) and epicardial progenies (EGFP⁺, green) at 7 wpf. c Whole-mount images showing scxa:mCherry (magenta) and pdgfrb:EGFP (green) expression at 5 wpf. The framed regions are enlarged to show details on the right with single-channel images shown in grayscale. Arrows indicate representative mCherry⁺ cells. AVC, atrioventricular canal. d Experimental design for (e–g). e, f Fate mapping results showing HCR staining of col18a1a (e, magenta), pdgfrb (e, blue), and vegfaa (f, magenta) in lineage-traced EGFP⁺ cells (green). Framed regions are enlarged to show representative colocalizations (arrows). g Quantification of cell fates as shown in (e). Data are presented as mean values +/- SD. Source data are provided as a Source Data file. h Experimental design for (i, j). i Fate-mapping analysis (maximum projections) showing HCR staining for col18a1a (magenta) and pdgfrb (blue) in lineage-traced EGFP⁺ cells (green) at 14 dpa (top) and 38 dpa (bottom). Framed regions are enlarged and shown as single optical sections to highlight representative colocalization events (arrows and arrowheads). Corresponding single-channel images are shown in grayscale. j Quantification of EGFP⁺ cells as shown in (i). Data are presented as mean values +/- SD. Source data are provided as a Source Data file.

We previously demonstrated that aEPCs give rise to pdgfrb⁺ pericytes or vSMCs during heart regeneration (Fig. 3a)¹⁵. Given the pericyte-like morphology of epicardial cell-derived scxa⁺ perivascular cells, we crossed scxa:mCherry fish with the pdgfrb:EGFP line that labels pericytes and vSMCs³³. Unlike pericytes that express high levels of pdgfrb and feature a characteristic elongated morphology, scxa⁺ perivascular cells mostly had a much lower-level expression of pdgfrb and exhibited a relatively round morphology with cell protrusions (Fig. 3c, arrows). These differences suggest that scxa is expressed in a perivascular cell population distinct from pericytes or vSMCs.

To define their identity, we searched for new markers in our scRNA-seq data. We found that col18a1a is highly enriched in the mesenchymal progenies of aEPCs but not in pericytes and vSMCs (Fig. 3a and Supplementary Fig. 1). COL18A1 is primarily localized in the basement membrane of blood vessels and epithelial tissues to maintain the structural integrity^34,35. COL18A1 can be cleaved into endostatin peptide, which was reported to inhibit angiogenesis and stabilize newly formed endothelial tubes^36–38. However, the cellular source and function of COL18A1 in coronary vasculature development remain unclear. To confirm whether scxa⁺ cells give rise to col18a1a⁺ cells in the heart ventricle, we performed lineage-tracing employing an inducible scxa:CreER^t2 line³⁹ crossed with the ubb:loxP-TagBFP-stop-loxP-EGFP reporter and applied labeling at juvenile stages with 3 overnight 4HT treatments at 6, 7, and 8 wpf (Fig. 3d). HCR staining of adult hearts confirmed that most ( ~ 72%) of the lineage-traced progenies express col18a1a with low or undetectable pdgfrb expression, which matched our prediction (Fig. 3a, e, g). We observed ~14% of EGFP cells expressing a high level of pdgrb^high indicating a mural cell identity (Fig. 3e, frame 2 and Fig. 3g). col18a1a expression was absent in pdgfrb^high mural cells (Fig. 3e, frame 1), further indicating that col18a1a⁺ cells are distinct from mural cells. In addition, we also observed ~14% of EGFP⁺col18a1a^-pdgfrb^- progenies (Fig. 3g), likely other mesenchymal progeny or scxa⁺ cells yet to differentiate. These observations indicate that col18a1a⁺ cells represent the dominant fate of scxa⁺ aEPCs, with mural cells as a minor fate. Moreover, scxa⁺ and col18a1a⁺ cells also express vegfaa (Fig. 3a, f), suggesting a potential pro-angiogenic function of these cells.

We next examined the fate of scxa⁺ cells following heart injury. Adult fish carrying the scxa:CreER^t2; ubb:loxP-TagBFP-stop-loxP-EGFP transgenes were subjected to ventricular amputation, followed by 4HT treatment at 6-8 dpa, when scxa expression is elevated in the wound region. Hearts were collected at 14 and 38 dpa (Fig. 3h). At 14 dpa, approximately 57% of lineage-traced EGFP⁺ cells expressed col18a1a with low or undetectable pdgfrb expression, consistent with col18a1a⁺ mesenchymal cell identity (Fig. 3i, arrows; Fig. 3j). In contrast, only ~1.7% of EGFP⁺ cells corresponded to col18a1a^- pdgrb^high mural cells, whereas the remaining ~41% were col18a1a^-pdgfrb^- mesenchymal cells (Fig. 3i, j). By 38 dpa, the proportion of col18a1a⁺ pdgfrb^low/- cells decreased to ~28%, while the proportion of col18a1a^- pdgrb^high mural cells and col18a1a^-pdgfrb^- mesenchymal cells increased to ~5% and 67%, respectively. These findings indicate that col18a1a⁺ mesenchymal cells persist as a stable scxa-lineage population during regeneration, although their relative abundance decreases over time as other scxa-derived mesenchymal populations expand. Notably, scxa⁺ progenitors consistently give rise to substantially more col18a1a⁺ mesenchymal cells than classical mural cells, consistent with our observations during heart development (Fig. 3g).

col18a1a⁺ mesenchymal cells are epicardial-derived perivascular cells distinct from mural cells

To better understand the col18a1a⁺ progeny and their relationship with the coronary vasculature, we performed HCR staining of endogenous col18a1a in the adult heart. As shown in Fig. 4a, col18a1a⁺ cells are adjacent to coronary vessels, supporting a perivascular cell identity. To further label these cells, we generated a col18a1a:EGFP BAC reporter and characterized its expression. In the adult heart, col18a1a:EGFP expression was observed across the entire ventricle in tcf21:H2A/Z-mCherry (or tcf21:H2R)⁺ cells, suggesting an epicardial origin (Fig. 4b). We crossed the line with the arterial reporter flt1:TdTomato and observed that col18a1a⁺ cells wrap around arteries (Fig. 4c, arrows). col18a1a⁺ cells were also found contacting vessels not visualized by flt1:TdTomato, likely the coronary veins or capillaries (Fig. 4c, arrowheads). These observations validate that the EGFP reporter faithfully recapitulates endogenous expression.

Fig. 4. col18a1a⁺ cells are perivascular cells distinct from pericytes and vascular smooth muscle cells.

a A whole-mount image of the ventricular surface showing HCR staining of col18a1a (magenta) and fli1a:EGFP (green) expression in an adult heart. The framed region is enlarged to show details at the bottom. b A whole-mount image of the ventricular surface showing expressions of tcf21:H2R (green) and col18a1a:EGFP (magenta) in an adult heart. The framed region is enlarged to show details at the bottom. Arrows indicate representative H2R⁺ EGFP⁺ cells. c A whole-mount image of the ventricular surface showing expressions of col18a1a:EGFP (green) and flt1:tdTomato (magenta) in an adult heart. The framed region is enlarged to show at the bottom. Arrows indicate representative EGFP⁺ cells adjacent to coronary vessels. d A whole-mount image of the ventricular surface showing HCR staining signals of pdgfrb in magenta. Expressions of col18a1a:EGFP and flt1:tdTomato are shown in green and yellow, respectively. DAPI staining is shown in blue. A single-channel image of pdgfrb signals is shown in grayscale on the right. Arrows indicate EGFP⁺pdgfrb^low cells. e Whole-mount image of the ventricular surface showing col18a1a (green) and scxa (magenta) expression in a 6 wpf heart. The framed region is enlarged to show cellular details, with corresponding single-channel images shown in grayscale. Arrows indicate representative scxa⁺col18a1a⁺ cells. AVC, atrioventricular canal. f–h Whole-mount images of the ventricular surface showing col18a1a (green) and flt1 (magenta) expression in 6 (f), 7 (g), and 10 wpf (h) hearts. Framed regions are enlarged on the right to highlight cellular details, with corresponding single-channel images shown in grayscale. AVC atrioventricular canal.

To further confirm the identity of col18a1a⁺ cells, we performed HCR staining of pdgfrb, the mural cell marker. As shown in Fig. 4d, col18a1a⁺ cells exhibit minimal pdgfrb expression compared with pdgfrb^high mural cells. We next assessed the temporal dynamics of col18a1a expression during coronary angiogenesis in juvenile hearts. We found that col18a1a:EGFP expression first emerged in a subset of scxa⁺ cells near the AVC at around 6 wpf (Fig. 4e), after coronary vessels had already expanded over the ventricular surface (Fig. 4f). At 7 wpf, when scxa expression reaches its peak, col18a1a expression remains low and is largely restricted to the ventricular surface adjacent to the AVC region (Fig. 4g). By 10 wpf, when the coronary vasculature is mostly established, col18a1a expression had expanded across the ventricular surface (Fig. 4h). Given that the COL18A1-derived endostatin peptide has been reported to stabilize newly formed endothelial tubes and inhibit angiogenesis^36–38, the delayed onset and progressive expansion of col18a1a expression are consistent with a role in vascular maturation, paralleling the spatiotemporal wave of coronary vessel development that originates near the AVC^30,40,41.

To further distinguish these col18a1a⁺ cells from mural cells at the molecular level, we reanalyzed our single-cell transcriptomic data (cell clusters shown in Fig. 3a). We found that col18a1a⁺ cells (clusters 0, 1, and 4) express little or no detectable levels of several established mural cell (cluster 6) markers, including myosin heavy chain 11a (myh11a), tropomyosin 1 (tpm1), regulator of G protein signaling 5a (rgs5a), and notch3^42,43 (Fig. 5a and Supplementary Fig. 1). Consistent with these transcriptional differences, pdgfrb⁺Acta2⁺ vSMCs in zebrafish predominantly surround larger coronary vessels, whereas col18a1a⁺ cells are primarily associated with smaller coronary vessels lacking Acta2 expression (Fig. 5b). To further compare these populations, we performed differential expression analysis between col18a1a⁺ cells (from clusters 0, 1, and 4) and mural cells (cluster 6). As shown in Fig. 5c, d and Supplementary Data 1, the two populations exhibit distinct transcriptional signatures. Genes enriched in mural cells were associated with Gene Ontology (GO) biological processes, including angiogenesis, smooth muscle cell development, and blood vessel morphogenesis, whereas genes enriched in col18a1a⁺ cells were predominantly associated with extracellular matrix organization and skeletal system development (Fig. 5e). These findings further support the molecular distinction between col18a1a⁺ mesenchymal cells and classical mural cells.

Fig. 5. Molecular property of col18a1a⁺ cells.

a Violin plots showing the expression of selected marker genes across single-cell clusters. Cluster 11 is a small heterogeneous cluster (53 cells from all timepoints) containing cells with mixed transcriptional identities, including endocardial, endothelial, and likely a small number of mural cells, recovered during tcf21⁺ cell isolation. Expression of mural cell markers in this cluster reflects its mixed cellular composition rather than a distinct mural cell identity. b Whole-mount image of the ventricular surface showing col18a1a:EGFP (green), flt1:tdTomato (magenta) expression together with Acta2 immunostaining (blue) in an adult heart. Single-channel images are shown in grayscale. c Heatmap showing top 25 differentially expressed genes in either the col18a1a⁺ or mural cell population in uninjured adult hearts. d Volcano plot showing differentially expressed genes between col18a1a⁺ and mural cells. Top genes are highlighted. e Gene Ontology (GO) biological process enrichment analysis of genes enriched in col18a1a⁺ or mural cell population. Selected GO terms discussed in the text are indicated by red arrows. Enrichment was tested with compareCluster(fun = “enrichGO”) in clusterProfiler (v4.10.1) using a one-sided hypergeometric test. P values were Benjamini-Hochberg adjusted within each gene set, and terms considered enriched at adjusted p < 0.05 and q < 0.2. f Dot plot showing expression of selected adventitial fibroblast markers with homologs of SCXA, COL18A1, ACTA2, and PDGFRB in human (top) or mouse adventitial fibroblasts. The plot was generated using the CZI Cellxgene platform.

Besides mural cells, adventitial fibroblasts are another class of perivascular mesenchymal cells that reside within the collagen-rich adventitial layer surrounding large arteries and veins in mammals^44,45. Unlike mammalian coronary arteries, which possess a well-defined adventitial layer containing fibroblasts, extracellular matrix, nerves, and immune cells⁴⁶, zebrafish coronary vessels have not been reported to contain a morphologically distinct adventitial compartment. Instead, perivascular cells are closely associated with the endothelial cells of the vessel³⁰. We compared col18a1a⁺ mesenchymal cells with published human and mouse adventitial fibroblast datasets^47–50 available through the CZI Cellxgene platform (https://cellxgene.cziscience.com). We found that expression of SCX was minimal in human exocrine gland, lung, and musculature adventitial fibroblasts, as well as in mouse lung adventitial fibroblasts (Fig. 5f). Likewise, although COL18A1 was detected at modest levels in human lung adventitial fibroblasts, its expression was minimal in the remaining adventitial fibroblast populations examined. Thus, while we cannot exclude the possibility that zebrafish col18a1a⁺ mesenchymal cells share some features with fibroblasts, their molecular profile is distinct from that of currently characterized mammalian adventitial fibroblasts.

Together with our developmental and lineage-tracing analyses, these findings support the conclusion that col18a1a marks a distinct epicardial-derived perivascular mesenchymal cell (Epi-PMC) population that is molecularly and anatomically distinct from classical mural cells and currently described mammalian adventitial fibroblasts.

Loss of scxa affects coronary angiogenesis but not cardiomyocyte proliferation

Because the scxa paralog scxb is also expressed in the zebrafish heart (Supplementary Fig. 2a), to determine the role of scleraxis genes in coronary angiogenesis, we first analyzed an allelic series of scxa ^kg170 and scxb ^kg107 mutants⁵¹ using the fli1a coronary vessel reporter. Adult scxa^-/- hearts exhibited a trend toward increased coronary vessel density compared with wild-type siblings, whereas scxb^-/- hearts showed no apparent vascular phenotype, indicating that scxa is the predominant scleraxis paralog regulating coronary vessel patterning (Fig. 6a, b). Consistent with the previous report, scxa^-/-scxb^-/- double mutants exhibited severe musculoskeletal defects and high mortality, typically dying before 3 wpf because of mandibular cartilage abnormalities that impair feeding⁵¹. By isolating mutant fish and providing supplemental feeding, we were able to raise fewer than 5% of double mutants to adulthood, although they remained substantially smaller than their wild-type siblings (Fig. 6c). Notably, the surviving double mutants lacked coronary vessels (Fig. 6d). While this phenotype indicates that scleraxis genes are collectively required for coronary vessel formation, the severe developmental defects and low survival rate of the double mutants precluded further mechanistic analyses. Given these findings, we focused subsequent analyses on scxa^-/- single mutants.

Fig. 6. Loss of scxa affects coronary angiogenesis but not cardiomyocyte proliferation.

a Representative whole-mount images showing ventricular coverage by fli1a:EGFP coronary vessels in hearts of the indicated genotypes. The framed regions are enlarged to show details on the bottom left. b Quantification of ventricular vessel coverage shown in (a). Data are presented as mean values +/- SD. Two-sided Student’s t-test. Source data are provided as a Source Data file. c Representative images showing developmental defects in scxa^-/-scxb^-/- fish. d Whole-mount images showing ventricular coverage of fli1a:EGFP vessels in scxa^+/+scxb^-/- (left) and scxa^-/-scxb^-/- (right) hearts. The framed regions are enlarged to show details. The residual GFP signals visible in the mutant hearts originate from endocardial cells within the heart rather than coronary endothelial cells on the ventricular surface. e, f Cryosection images showing antibody staining for PCNA (green) and Mef2 (magenta). The framed regions are enlarged and displayed at the top right. Dashed lines denote injury sites. The arrowheads indicate representative PCNA⁺Mef2⁺ cells. g Quantification of the CM proliferation index. Data are presented as mean values +/- SD. Two-sided Student’s t-test. Source data are provided as a Source Data file. h Left: schematic illustrating the imaging orientation of the ventricular apex at 14 dpa. Right: Whole mount images of the ventricular apex showing coronary revascularization visualized by fli1a:EGFP expression. i Quantification of ventricular vessel coverage shown in (h). Data are presented as mean values +/- SD. Two-sided Student’s t-test. Source data are provided as a Source Data file.

We next assessed col18a1a expression in the scxa single mutant. However, the number of col18a1a⁺ cells seems comparable in the scxa single mutant and wild-type siblings (Supplementary Fig. 2b, c), indicating that loss of scxa alone does not measurably alter the abundance of Epi-PMCs. This may reflect functional compensation by the closely related paralog scxb, although additional studies will be required to distinguish this possibility from more subtle effects on Epi-PMC differentiation or function. To determine whether the increased vessel density influences cardiac regeneration, we subjected scxa^-/- mutants (with or without a scxb^+/- allele) and wild-type siblings to ventricular amputation. CM proliferation at 7 dpa was indistinguishable between the four groups (Fig. 6e–g). We further quantified coronary revascularization within the injury site at 14 dpa using fli1a:EGFP reporter and observed a trend toward increased vessel density in scxa^-/- hearts, consistent with the developmental phenotype (Fig. 6h, i). Together, these findings suggest that loss of scxa alters coronary vessel patterning without detectable effects on CM proliferation during regeneration.

To gain insight into the molecular programs associated with scxa expression and to facilitate interpretation of the vascular phenotype observed in scxa mutants, we reanalyzed our single-cell transcriptomic dataset by comparing gene expression between scxa⁺ and scxa^- epicardial-derived cells at 3 dpa (Supplementary Fig. 3a). Differential expression analysis identified a cohort of genes co-enriched with scxa, including ECM-associated genes (col12a1a, col12a1b, col5a, and fn1a), the regenerative factor mdka (also fn1a and col12a1b)^15,52,53, the angiogenesis-associated gene serpine1²⁸, and additional mesenchymal-associated genes (Supplementary Fig. 3b, c; Supplementary Data 2). GO analysis further revealed enrichment of biological processes related to extracellular matrix organization, tissue regeneration, wound healing, regulation of proteolysis, and embryonic development (Supplementary Fig. 3d). Together, these findings indicate that scxa expression is associated with a regenerative mesenchymal transcriptional program characterized by extracellular matrix remodeling and tissue repair.

Hypoxia induces scxa expression

To identify factors regulating scxa expression, we performed CUT&Tag analysis of histone H3K4me1 modifications in aEPCs. We isolated aEPCs at 3 dpa using tcf21:nucEGFP and ptx3a:mScarlet-NTR reporter for EGFP⁺mScarlet⁺ cells (Fig. 7a). EGFP⁺ cells from uninjured hearts were used as a control for quiescent epicardial cells. We combined the new CUT&Tag data with our published RNA-seq, ATAC-seq, and Chip-seq results⁵⁴. As shown in Fig. 7b, we observed 4 putative promoter and/or enhancer regions upstream of the scxa gene that gained accessibility and enhancer markers (H3K4me1 and H3K27Ac) after heart injury. Motif analysis using JASPAR⁵⁵ predicted multiple putative HIF1A (Hypoxia-Inducible Factor 1 Alpha)-binding sites within these regions (Fig. 7c; profile score > 4), implying hypoxia and Hif signaling as regulators of scxa expression.

Fig. 7. Hypoxia induces scxa expression in adults.

a Schematic for experiment design. Heart resection injuries were performed in tcf21:nucEGFP; ptx3a:mScarlet animals. Ventricles were collected at 3 dpa together with those of uninjured siblings (Ctrl). Ventricles were dissociated, and EGFP⁺ or EGFP⁺ mScarlet⁺ epicardial cells were isolated by FACS for the CUT&Tag assay. b A browser track of the genomic region comprising gene scxa showing the transcripts, chromatin accessibility profiles, and CUT&Tag marks of H3K4me1 in the epicardium across replicates from Ctrl, 3 dpa, and 7 dpa samples. The whole-ventricle H3K27Ac profile of the uninjured (Ctrl) and regenerating (Reg) hearts is shown. Gray boxes indicate putative enhancer regions that show increased accessibility and enhancer marker activity during regeneration. c Putative HIF1A binding sites in the 4 enhancer (Enh) regions shown in (b). The motif profile is downloaded from JASPAR under the Creative Commons Attribution 4.0 International License. d Whole-mount images showing scxa:mCherry expressions without (Ctrl, left) or with PHZ treatment (middle). The framed region is enlarged to show detail in the top right, with the coronary vessel marker fli1a:EGFP shown in magenta in the overlaid image (bottom right). The control heart is outlined. e Quantification of ventricular coverage of scxa:mCherry signals shown in (d). Data are presented as mean values +/- SD. Two-sided Student’s t-test. Source data are provided as a Source Data file. f Whole-mount images showing scxa:mCherry (magenta) and col18a1a:EGFP (green) expressions without (Ctrl, top) or with PHZ treatment (bottom). Single-channel images are shown in grayscale. The control heart is outlined. The framed regions are enlarged to show details on the bottom right. g Quantification of ventricular coverage of col18a1a:EGFP signals shown in (f). Data are presented as mean values +/- SD. Two-sided Student’s t-test. Source data are provided as a Source Data file.

To test the impact of excessive hypoxia on scxa expression, we used phenylhydrazine (PHZ) to induce acute hemolysis and hypoxia^56–58. After treatment with PHZ (1.5 μg/ml for 20 min, every other day for seven days), we observed a striking re-expression of scxa:mCherry in the entire ventricular surface (Fig. 7d, e). Importantly, the ectopic scxa expression is enriched adjacent to coronary vessels (Fig. 7d inset). We next examined whether col18a1a expression is also regulated by systemic hypoxia. Unlike the marked increase in scxa expression following PHZ treatment, col18a1a expression remained unchanged (Fig. 7f, g). These results indicate that, unlike scxa, col18a1a is not acutely responsive to hypoxic stimulation.

To evaluate whether the hypoxia effect on scxa expression is reproducible in juvenile heart development, we applied PHZ treatment to juveniles for one week from 7 wpf and assessed scxa expression at 8 wpf. As expected, PHZ treatment significantly increased scxa expression in juveniles (Fig. 8a, b). To further test the role of HIF-mediated hypoxic signaling, we treated juveniles with Dimethyloxalylglycine (DMOG) from 6 to 8 wpf, which stabilizes Hif factors and induces hypoxic responses^59,60. Compared with vehicle-treated controls, DMOG treatment resulted in increased scxa expression (Fig. 8c, d).

Fig. 8. Hypoxia and Hif signaling regulate scxa expression.

a Whole-mount images showing scxa:mCherry expressions in control (Ctrl, left) and PHZ-treated (right) hearts. The framed region is enlarged to show detail at the bottom. b Quantification of ventricular coverage of scxa:mCherry signals shown in (a). Data are presented as mean values +/- SD. Two-sided Student’s t-test. Source data are provided as a Source Data file. c Whole-mount images showing scxa:mCherry expressions with vehicle (Ctrl, left) or DMOG (right) treatment at 8 wpf after a 2-week treatment. The framed regions are enlarged at the bottom for details. d Quantification of ventricular coverage of scxa:mCherry signals at 8 wpf. Data are presented as mean values +/- SD. Two-sided Student’s t-test. Source data are provided as a Source Data file. e Cryosection images showing scxa expression (magenta) together with Hif2a immunostaining (green) in uninjured (Ctrl, top), 1 dpa (middle), and 4 dpa (bottom) hearts. Framed regions are enlarged to highlight representative scxa⁺Hif2a⁺ cells (arrowheads). Corresponding Hif2a single-channel images are shown in grayscale. f Quantification of Hif2a staining intensity in the wound. Data are presented as mean values +/- SD. Two-sided Student’s t-test. Source data are provided as a Source Data file. g Working model. In response to hypoxia induced by cardiac injury or developmental cues, epicardial cells activate Hif signaling and upregulate scxa expression. A subset of scxa⁺pdgfrb^low cells becomes associated with coronary vessels and subsequently differentiates into col18a1a⁺pdgfrb^low epicardial-derived perivascular mesenchymal cells (Epi-PMCs), distinct from classical pdgfrb^high mural cells. Differentiation into mural cells represents a minor fate of scxa⁺pdgfrb^low cells. In addition, scxa⁺ epicardial progenitors give rise to other mesenchymal cell populations.

To determine whether endogenous Hif signaling is activated following cardiac injury, we next examined Hif protein expression by immunostaining. While the Hif1a antibodies we tested did not produce reliable staining in zebrafish heart tissues under our experimental conditions, the Hif2a antibody yielded robust and reproducible staining. Compared with uninjured hearts, Hif2a staining was markedly increased at both 1 and 4 dpa in endocardial cells as well as in scxa⁺ epicardial cells (Fig. 8e, f), consistent with activation of hypoxia-responsive signaling during heart regeneration. Together, these findings indicate that hypoxia and Hif signaling promote scxa expression during both heart development and regeneration, which further supports the differentiation of epicardial progenitors toward the Epi-PMC fate.

Discussion

The epicardium is recognized as a dynamic and multipotent contributor to heart development, repair, and regeneration. In this study, we identify Scxa as a critical transitional regulator within the epicardial lineage that governs differentiation toward a previously unexplored col18a1a⁺ perivascular population that we term epicardial-derived perivascular mesenchymal cells (Epi-PMCs) (Fig. 8g). These cells are molecularly distinct from canonical mural cell types and mammalian adventitial fibroblasts and are characterized by a unique combination of ECM components and angiogenic regulators. Our findings expand the spectrum of epicardial-derived fates and provide new insights into how hypoxia-responsive transcriptional programs shape the coronary perivascular microenvironment during heart development and regeneration.

Although Scx is historically associated with tendon development, emerging evidence links it to cardiac processes. In mammals, Scx has been linked to pro-epicardial organ development, embryonic epicardial cells, cardiac fibroblast formation, and valve morphogenesis^21–26. Its prolonged activation post-injury promotes a fibrotic myofibroblast phenotype through upregulation of fibrillar collagen (Col1a2)²⁶. Interestingly, loss of Scx in mice results in hypocellular, underdeveloped hearts, indicating a broader role in promoting mesenchymal cell differentiation²⁶. These seemingly opposing outcomes underscore a key principle: Scx functions are highly context- and time-dependent, and their cellular effects are shaped by local signaling environments and lineage states. In the zebrafish heart, we find that scxa expression is transient and tightly linked to a hypoxic microenvironment. This limited temporal window may allow epicardial cells to flexibly adopt supportive roles in angiogenesis. In contrast, the adult mammalian epicardium likely lacks this flexibility and typically responds to injury by generating fibrotic tissue rather than functional vasculature. Understanding how scxa is temporally gated and environmentally regulated could provide valuable insights into the failure of regenerative programs in the adult human heart.

Our findings further underscore the growing appreciation that perivascular cell populations extend beyond the classical pericyte and vSMC lineages. A previous study using Scx lineage tracing suggested an endothelial cell fate for Scx⁺ proepicardial cells²³. Although we have yet to define the fate of the remaining col18a1a^-pdgfrb^-/low cells of the scxa-lineage, our study suggests that scxa⁺ epicardial progenitors predominantly adopt a perivascular rather than an endothelial fate. Whether these col18a1a^-pdgfrb^-/low cells represent a quiescent progenitor population that can be further activated upon heart injury remains an important question for future investigation. Previous studies in zebrafish have described perivascular fibroblast populations in the trunk vasculature that exhibit low pdgfrb expression and contribute to vessel stability⁶¹. These cells were shown to support vessel integrity and, in some cases, differentiate into pericytes. The col18a1a⁺ Epi-PMCs identified here resemble these perivascular fibroblasts in low pdgfrb expression and morphology but likely carry distinct functions suited to the complexity of the coronary vasculature.

We found that zebrafish Epi-PMCs are molecularly distinct from both classical mural cells and currently characterized mammalian adventitial fibroblasts. Nevertheless, whether scxa⁺ epicardial progenitors can generate adventitial fibroblasts in mammals, or whether subsets of col18a1a⁺ Epi-PMCs possess the potential to differentiate into mural cells or fibroblast-like populations, remains unknown. It is therefore possible that zebrafish Epi-PMCs represent an evolutionarily related, yet distinct, perivascular mesenchymal population. Additional heterogeneity is also likely to exist within the Epi-PMC compartment, and future single-cell lineage tracing and functional studies will be required to define potential subtypes and their specific roles. Also, although our lineage-tracing experiments demonstrate that scxa⁺ epicardial progenitors contribute substantially to the col18a1a⁺ Epi-PMC population, we cannot conclude that they represent the exclusive cellular source. Because tamoxifen-inducible Cre-mediated recombination is inherently incomplete, not all scxa-expressing cells are expected to be lineage labeled. In addition, other epicardial or non-epicardial progenitor populations may also contribute to a subset of col18a1a⁺ cells. Future studies using complementary lineage-tracing approaches will be required to fully resolve the developmental origins of Epi-PMCs. Furthermore, although col18a1a serves as a useful marker for Epi-PMCs, whether the paralog col18a1b or alternative col18a1a isoforms are expressed in other cardiac cell types and contribute to vascular regulation remains to be determined.

The increased coronary vessel coverage observed in scxa single mutants suggests that Scxa may act, at least in part, as a negative regulator of angiogenesis in our experimental setting. This phenotype is consistent with previous studies showing that the COL18A1-derived endostatin peptide possesses anti-angiogenic and vessel-stabilizing activities^36–38. In support of this, Col18a1 knockout mice show increased vascular density in the retina⁶², implicating this matrix protein in vascular remodeling. Although this hypothesis requires direct functional testing, our observed late onset of col18a1a expression during coronary angiogenesis in zebrafish is consistent with a role in the maturation phase rather than the initiation of vessel growth. Simultaneously, scxa⁺ aEPC and scxa progenies also express vegfaa (Fig. 3a, f), a potent pro-angiogenic factor. This duality raises the possibility that Scxa, responding to hypoxia, coordinates the temporal transition from a pro-angiogenic epicardial progenitor state toward a more mature perivascular mesenchymal state.

Of note, the abundance of col18a1a⁺ Epi-PMCs was unchanged in scxa mutants, indicating that loss of scxa does not measurably affect Epi-PMC abundance or their formation. One likely explanation is functional compensation by the closely related paralog scxb, consistent with our allelic analysis showing that combined loss of scxa and scxb results in a much more severe coronary vascular phenotype than loss of scxa alone. Alternatively, the transcriptional changes may be too subtle to detect by HCR. Although the profound developmental defects in double mutants preclude definitive mechanistic interpretation, these findings indicate both functional redundancy and an essential requirement for scleraxis genes during coronary vessel formation. The contrasting outcomes of the scxa single mutant and the scxa/scxb double mutant may highlight the context-specific functions of Scx transcription factors, ranging from fine-tuning vessel growth to enabling vascular initiation.

Our transcriptomic analyses further provide clues to the molecular programs associated with Scxa activity. Previous studies in tendon, cardiac valve, and connective tissue lineages have demonstrated that Scx regulates extracellular matrix genes, including collagens and tenomodulin, and cooperates with TGF-β/Smad signaling to promote mesenchymal differentiation^21,63–66. Consistent with these observations, scxa⁺ epicardial-derived cells in our dataset were enriched for ECM genes (col12a1a, col12a1b, col5a, and fn1a), the regenerative factor mdka, and the angiogenesis-associated gene serpine1, together with Gene Ontology terms related to ECM organization, tissue regeneration, and wound healing. These findings suggest that Scxa promotes a regenerative mesenchymal program associated with ECM remodeling during coronary vessel development and repair. Defining the direct transcriptional targets of Scxa in the heart will require genome-wide chromatin occupancy studies combined with cell-type-specific transcriptomic analyses.

One intriguing aspect of this study is the connection between hypoxia and scxa induction. Our epigenomic analyses identified multiple HIF1A motifs in injury-responsive putative regulatory regions near scxa, and systemic induction of hypoxia dramatically elevated scxa expression in epicardial cells. Furthermore, Hif2a protein accumulated in scxa⁺ epicardial cells following cardiac injury, supporting activation of hypoxia-responsive signaling during heart regeneration. These findings align with previous reports showing that hypoxia induces epicardial cell differentiation and vascular precursor cell invasion into the myocardium^67–69. It was recently found that stabilization of hypoxic signaling in the epicardium promotes regenerative outcomes in mammalian models of myocardial infarction⁷⁰. Together, our data provide a conceptual bridge between hypoxia-responsive signaling, epicardial cell fate specification, and coronary vessel development.

In conclusion, this study identifies scxa as a hypoxia-responsive regulator of epicardial progenitor differentiation during coronary vessel development and heart regeneration. These findings expand the known repertoire of epicardial-derived cell fates and provide new insight into how hypoxia-responsive transcriptional programs contribute to the establishment of the coronary perivascular microenvironment. More broadly, our work highlights the remarkable plasticity of the regenerating zebrafish epicardium and provides a framework for investigating whether analogous hypoxia-responsive epicardial programs can be harnessed to promote vascular repair and regeneration in the injured mammalian heart.

Methods

Animal maintenance and procedures

All animal procedures (Protocol #2018-0044) were approved by the Institutional Animal Care and Use Committee (IACUC) at Weill Cornell Medical College. Zebrafish (aged 4 weeks −12 months) of the Ekkwill (EK) and EK/AB strains were maintained as previously described^71,72. Animals of both sexes were used for experiments. Water temperature was maintained at 28 °C, and fish were kept on a 14/10 light/dark cycle at a density of 5-10 fish per liter. Heart amputation injury was performed as previously described^72,73. For tamoxifen treatments, juvenile and adult fish were placed in a mating tank of aquarium water containing 5 μM 4-Hydroxytamoxifen (4HT, Sigma-Aldrich). Embryos were treated with 5 μM 4HT in dishes. Fish were maintained for 16 h, rinsed with fresh aquarium water, and returned to the recirculating aquatic system. For phenylhydrazine (PHZ, Sigma-Aldrich) treatment, adult fish were placed in a mating tank of aquarium water containing 1.5 μg/ml PHZ for 20 min daily. Fish were rinsed with fresh aquarium water and returned to the recirculating aquatic system. For Dimethyloxalylglycine (DMOG) treatment, juvenile fish were injected (intraperitoneal) with 1-2 μl (based on the fish size) of 50 mM DMOG every other day from 6 to 8 wpf. The following published lines were used: Tg(tcf21:nucEGFP)^pd41 (³²), Tg(tcf21:H2A-mCherry)^pd252 (⁷⁴), Tg(tcf21:CreER^t2)^pd42 (³²), Tg(pdgfrb:EGFP)^ncv22 (³³), Tg(col12a1b:EGFP)^wcm108 (¹⁵), Tg(ptx3a:mScarlet-P2A-NTR)^wcm107 (¹⁵), TgBAC(ubb:loxP-TagBFP-STOP-loxP-EGFP)^vcc18 (³¹), Tg(scxa:mCherry)^fb301 (²⁷), Tg(scxa:CreER^t2)^fb316 (³⁹), scxa^kg170 (⁵¹), and scxb^kg107 (⁵¹). Newly generated lines are described below. The generation of new lines is described below. All reporters were analyzed as hemizygotes.

Generation of Tg(col18a1a:EGFP) zebrafish

To make a BAC construct, the translational start codon of col18a1a in the BAC clone DKEY-160c21 was replaced with the EGFP-polyA cassette by Red/ET recombineering technology (Gene Bridges)⁷⁵. The 5’ and 3’ homologous arms for recombination were a 50-base pair (bp) fragment upstream and downstream of the start codon and were included in PCR primers to flank the EGFP-polyA cassette. The primer sequences are: Forward, 5’- CAACAACAATCCCATGTAGCTATCGGGACGGAGCAGTTTCACGAACTGAAATGGTGAGCAAGGGCGAGGAGCTG-3’; Reverse, 5’- CTTAAAATATATACTCCAAAACTCATGTATCCTCATGTATACCTTCTTGCGCCCTTGAAGTTCCTATTCTCTAG-3’. The final BAC was purified with the Nucleobond BAC 100 kit (Clontech) and co-injected with PI-Sce I into one-cell-stage zebrafish embryos. Stable transgenic lines with bright fluorescence were selected. The full name of this line is Tg(col18a1a:EGFP)^wcm113.

CUT&Tag assay and data analysis

CUT&Tag assay targeting H3K4me1 was performed using the EpiCypher CUTANA™ CUT&Tag Kit following the manual and a published protocol⁷⁶. Briefly, heart amputation surgery was performed on adult fish (6 months of age) carrying tcf21:nucEGFP and ptx3a:mScarlet-P2A-NTR reporters. Hearts were isolated at 3 dpa, together with the uninjured control (2 biological replicates each), on ice, and washed several times to remove blood cells. Ventricles were digested in an Eppendorf tube with 0.5 ml HBSS plus 0.13 U/ml Liberase DH (Roche) at 37 °C, while stirring gently with a Spinbar® magnetic stirring bar (Bel-Art Products). Supernatants were collected every 5 min and neutralized with sheep serum. Dissociated cells were spun down and re-suspended in DMEM plus 10% fetal bovine serum (FBS) medium and sorted using a Sony MA900 sorter for EGFP⁺ cells (uninjured) and EGFP⁺mScarlet⁺ (3 dpa). Nuclei preparation and CUT&Tag library preparation using rabbit anti-H3K4me1 antibody (Abcam, ab8895) were performed using the EpiCypher CUTANA™ CUT&Tag Kit following the manual. Two replicates with 5000–12,000 nuclei for each sample were used. Sequencing was performed using an Illumina NextSeq2000 with 11 − 44 million 50-nt pair-end reads obtained for each library. CUT&Tag data were processed and analyzed following a published pipeline^77,78. Briefly, reads were trimmed for adaptors using Trimmomatic (v0.39) before aligning to the DanRer10 genome using Bowtie2 (v2.5.3) with parameters for 10-700 bp length. Peaks were called using SEACR (v1.3) with p-value < 0.01.

Single-cell RNAseq analysis

Our previously published single-cell RNA-seq dataset (GEO accession: GSE202836)¹⁵ was loaded in R (V4.3.2). Published clustering annotations were used to generate UMAP visualizations of cell populations and gene expression patterns. To compare mural cells and col18a1a⁺ cells, cluster 6 (mural cells) and col18a1a⁺ cells from mesenchymal clusters 0, 1, and 4 (logCounts > 0) were extracted for differential expression analysis. Marker genes enriched in each population were identified using the multiMarkerStats function from scran (v1.30), combining a two-sided Welch’s t-test and a two-sided Wilcoxon rank-sum test, with p values Benjamini-Hochberg adjusted across genes within each test. For each gene, the maximum adjusted p-value (FDR) from the two tests was used as the combined significance score, providing a conservative and robust estimate of differential expression. Genes were considered significant only when supported by both statistical tests (Supplementary Data 1). Volcano plots were generated using t.summary.logFC and t.FDR values. To identify genes associated with scxa expression during regeneration, a similar differential expression analysis was performed comparing scxa⁺ and scxa^- epicardial-derived cells at 3 dpa (Supplementary Data 2). Gene Ontology (GO) biological process over-representation analysis was performed on differentially expressed genes (combined FDR < 0.05, |log₂FC | > 0.5), split by the sign of the log₂ fold change into two gene sets per comparison. Enrichment was tested with compareCluster (fun = “enrichGO”) in clusterProfiler (v4.10.1) using a one-sided hypergeometric test. P values were Benjamini-Hochberg adjusted within each gene set, and terms considered enriched at adjusted p < 0.05 and q < 0.2.

Histology and microscopy

Freshly isolated hearts were fixed overnight at 4 °C in 4% paraformaldehyde (PFA). Fixed hearts were either mounted with Fluoromount G (Southern Biotechnology, cat#0100-01) between two coverslips to visualize both ventricular surfaces or processed for cryosectioning at a thickness of 10 μm. Hybridization Chain Reaction (HCR) staining was performed on whole-mount hearts and cryosections following established protocols⁷⁹. HCR probes targeting col18a1a and pdgfrb were obtained from Molecular Instruments Inc. For immunostaining of whole-mount hearts and cryosections^74,80, samples were blocked with 2% bovine serum albumin (BSA, VWR, cat#97061), 1% DMSO, 0.5% goat serum (Thermo Fisher, cat#16210), and 0.5% Triton X-100 in PBS for 1 h at room temperature (RT). Primary antibodies were diluted in the blocking buffer and incubated with hearts overnight at 4 °C. Hearts were then washed with PBS plus 0.1% Tween 20 and incubated with the secondary antibody diluted in the blocking buffer for 1.5 h at room temperature. Hearts were stained with DAPI (Thermo Fisher, D3571) to visualize nuclei. The primary antibodies used in this study included mouse anti-GFP monoclonal antibody (3E6) (Thermo Fisher, A11120), rabbit anti-dsRed/mCherry (1:200, Takara, 632496), chick anti-GFP (Aves Labs, GFP-1020), rabbit anti-Hif2a (Thermo Fisher, PA1-16510), mouse anti-PCNA (Sigma, p8825), and rabbit anti-Mef2 (Boster Bio, dz01398-1). Secondary antibodies (Thermo Fisher) included Alexa Fluor 488 goat anti-rabbit, goat anti-mouse, goat anti-chick, Alexa Fluor 546 goat anti-rabbit, goat anti-mouse, and Alexa Fluor 633 goat anti-mouse. Fluorescent images of whole-mount and cryosectioned heart samples were acquired using a Zeiss LSM 800 confocal microscope.

RT-PCR

Five hearts per sample were dissected, washed in sterile PBS, collected in 500 μl QIAzol (Qiagen), and mixed with 100 μl of 0.5 mm RNase-free stainless-steel beads. Samples were homogenized using a Next Advance Bullet Blender for 5 minutes and subsequently stored at −80 °C. Total RNA was extracted using the RNeasy Mini Kit (Qiagen) according to the manufacturer's protocol. cDNA synthesis was performed with the SuperScript III First-Strand Synthesis Kit (Invitrogen) following the manual. PCR amplification of target genes was conducted using the DreamTaq PCR system under the following conditions: 30 cycles with an annealing temperature of 59 °C. Primer sequences were as follows:

scxa-forward: 5′-GTCAACGCAATGCTGCCAAT-3′;

scxa-reverse: 5′-TCAGCAGGCTCTGTAGGGAT-3′;

scxb-forward: 5′-GCGAGGGTTCTGGATCTGAG-3′;

scxb-reverse: 5′-ACTGTTCGTTCGTTCTCGCT-3′;

gapdh-forward: 5′-GATGGTCATGCAATCACAGTCTA-3′;

gapdh-reverse: 5′-ATCATACTTGGCAGGTTTCTCAA-3′;

rpl13a-forward: 5′-TCTGGAGGACTGTAAGAGGTATGC-3′;

rpl13a-reverse: 5′-AGACGCACAATCTTGAGAGCA-3′.

Statistics and reproducibility

For each treatment, clutchmates were randomly assigned to different experimental groups. Sample sizes were determined based on prior studies and the nature of the experiments, with details provided in each figure legend. Statistical results are presented as the mean ± standard deviation (S.D.). Information on sample sizes, statistical tests, and P-values can be found in the figures or their corresponding legends. Two-tailed Student’s t-tests were used for comparisons when assumptions of normality and equal variance were satisfied. All experiments were performed successfully, yielding similar results across at least 3 biological replicates. For all fluorescent images showing representative results, at least 6 zebrafish were analyzed with consistent results.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Conceptualization: J.C. Methodology: B.P., Y.X., P.Z., D.B. and J.C. Investigation: B.P., Y.X., J.Y., I.V.W., M.Q., A.G.C.Y., M.N., P.Z. and A.Y. Resources: S.L.T., K.K., J.L.G., M.R.M.H. and J.C. Writing and editing: B.P., T.E., M.R.M.H. and J.C. Funding Acquisition: M.R.M.H. and J.C.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Funding

This work was supported by the New York State Stem Cell Science program (NYSTEM) predoctoral training grant position to B.P., the Louis and Rachel Rudin Foundation fellowship to Y.X., and a Research Fellowship award from the Belgian American Educational Foundation to I.V.W. American Heart Association (AHA) Career Development Award (AHA941434) and National Institutes of Health (NIH) grant R01NS126209 to M.R.H., AHA Transformational Project Award (23TPA1077772) and NIH grants (R01HL155607, R01HL166518) to J.C.

Data availability

The CUT&Tag datasets generated in this study have been deposited at NCBI’s Gene Expression Omnibus (GEO) under accession numbers GSE299367. The previously published RNA-seq, ATAC-seq, and single-cell RNAseq datasets of epicardial cells are available under access numbers GSE89444 and GSE202836. Ventricle Chip-seq (H3K27Ac) dataset was downloaded from GEO under accession number GSE75894. Source data for quantifications are provided as a Source Data file. Source data are provided with this paper.

Code availability

No custom code was generated for this study. The single-cell RNA-seq dataset and associated analysis pipeline, including clustering annotations, were previously published¹⁵. Data visualization and downstream analyses were performed in R (v4.3.2) using standard functions and packages as described above.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-77008-y.