Hypoxia-mediated epicardial signaling coordinates coronary angiogenesis and myocardial expansion during zebrafish ventricle maturation

Abstract

During cardiac development, the myocardium expands in response to physiological demands to achieve proper cardiac morphology and functional contractility, while simultaneously integrating with the developing coronary vasculature. However, the mechanisms governing this ordered expansion remain poorly understood. Here, we found that regional hypoxia drives local tissue thickening, which in turn exacerbates a hypoxic microenvironment. We demonstrate that epicardial hypoxia serves as a central regulatory mechanism, coordinating both coronary angiogenesis and myocardial expansion during juvenile zebrafish heart development. This mechanism activates discrete spatial patterns of epicardial gene expression, including vegfaa, loxl2a, and col12a1b. Through live and fixed imaging, we find that cardiomyocytes and endothelial cells exhibit coordinated expansion patterns through third-party epicardial signals that are required for both coronary development and myocardial expansion. Using cxcr4a^um20 mutants lacking functional coronary vessels we show that coronary vessels provide negative feedback on epicardial hypoxia, while positively responding to the same hypoxic cues that drive myocardial expansion. Disruption of this negative feedback leads to increased myocardial stiffness through dysregulated extracellular matrix crosslinking as observed in pathological conditions such as cardiomyopathies. These findings establish the role of regional epicardial hypoxia within a fundamental regulatory network that drives appropriate regional tissue growth with integrated vascular supply during cardiac morphogenesis.

INTRODUCTION

Cardiac morphogenesis requires the intricate coordination of cardiac cell types to establish an optimally functional heart architecture. Critical to this process is coordinated development of the coronary vasculature necessary to support the metabolic demands of the growing myocardium. In zebrafish, the coronary vasculature is responsive to myocardial growth both in development and during regeneration in response to injury (1–4). Across species, coronary development coincides with thickening of the myocardium (1, 2, 5–8) and is essential to support myocardium regeneration (2, 4, 9), and cardiomyocyte (CM) proliferation (3, 4). However, the fundamental mechanisms that determine how the coronary vasculature influences chamber morphogenesis and how the myocardium drives optimal coronary growth remain unknown. Disruptions to this developmental program result in congenital malformations that lead to myocardial dysfunction, including hypertrophy, heart failure, and anomalous tissue mechanics characterized by increased stiffness and extracellular matrix remodeling (10–15).

In zebrafish, coronary vessels emerge post-embryonically around 21 days post-fertilization (dpf)/0.9 mm standard length (SL; snout to hypural) from endothelial cells at the atrio-ventricular canal. Their formation over the ventricle correlates most strongly with ventricle size (2) and is guided by Cxcr4a-Cxcl12b (2), Wnt/β-catenin (16), hyaluronic acid (17), and Vegfaa (9) signaling. Concurrent with this vascularization, the zebrafish ventricle undergoes expansion as internal CMs breach the ventricular wall to form the cortical layer that expresses gata4 and covers the surface to fortify the ventricular wall against biomechanical stress (18). This outer myocardial wall is covered by the epicardium, which supplies vascular smooth muscle cells and pericytes and also acts as a signaling hub to promote myocardial growth and regeneration (19–21). One such signal is Vegfaa that, in addition to regulating blood vessel development, also induces cardiac growth programs and hyperplastic myocardial thickening (3).

To explore how vascular supply integrates with and shapes tissue growth, we examined juvenile zebrafish during coronary vascularization and cortical layer formation. We demonstrate that hypoxia serves as a central regulatory mechanism coordinating coronary angiogenesis and myocardial expansion. Our observations suggest that local tissue stress and thickening create hypoxic microenvironments that regionally activate epicardial expression of factors, including pro-angiogenic and mitogenic protein Vegfaa, and regulators of extracellular (ECM) stiffness, Loxl2a and Col12a1b, required for coronary development and myocardial expansion. We conclude that epicardial hypoxia-responsive gene expression is reciprocally negatively regulated by vascular supply, establishing a feedback mechanism that balances myocardial and coronary vessel expansion to optimally shape the maturing ventricle. Disruption of this coordination leads to increased myocardial stiffness, offering mechanistic insights into tissue regeneration and the pathological progression to heart failure caused by microvascular dysfunction, hypertrophy and myocardial infarction.

RESULTS

Coronary vessel positioning and orientation are aligned with that of the myocardium.

The positioning of coronary vessels over the ventricle appears somewhat variable in zebrafish, but not completely random. We observed the structure of coronary vessels within the developed myocardium by imaging whole adult cardiac tissue expressing fli1:EGFP, which labels coronary vessels, and after staining for acetylated-tubulin (AcTub), which labels both axons and striated myocardial fibers on the outer surface of the ventricle (Figure 1A and B). Coronary vessels almost completely align along and within ribbons of compact myocardium to form cohesive cardiac tissue (mean EC orientation is ± 5.7° of nearest muscle fiber orientation, p < 0.0001, one sample t test, Figure 1C), suggesting that the two structures have coordinated development.

Figure 1. Alignment and coordinated expansion of coronary vessels and cardiomyocytes.

(A) Confocal projection of immunolabeled whole-mount adult zebrafish heart with blood vessels (fli1:GFP, red) and myocardium (acetylated tubulin, green). (B) IMARIS volume render of (A) showing vessel (green) integration within the myocardium (red). (C) Quantification of adult cortical CM and coronary EC alignment, relative angle of EC cells to nearest fiber binned in 5° increments (n = 220 cells/11 hearts, p < 0.0001, one-sample t and Wilcoxon test). (D) Confocal projection of fli1:GFP (green), gata4:DsRed (red), and acetylated tubulin (AcTub, blue) in 20 mm standard length (SL; 77 dpf) zebrafish. Acetylated tubulin is enriched on the ventricular surface and overlaps with gata4:DsRed+ cardiomyocytes; nascent coronary vessels align along these cardiomyocytes. (E) Confocal projection of fli1a:GFP (green; endothelial cells, EC), gata4:DsRed (red, gata4+ cardiomyocytes, CM), and cxcr4a:mCitrine (blue) at 25 mm SL (98 dpf, right aspect). Developing coronary vessels are enriched in gata4:DsRed+ regions. (F) Enlargement of boxed region in (E) showing vascularization/muscularization of an avascular region lacking cortical muscle. (G) Coronary vessel quantified as percentage of vessels within gata4:DsRed+ or – regions normalized to ventricle coverage, showing regional coordination despite asynchronous expansion (n = 40, P < 0.0001, paired t-test). (H) Confocal projection showing localized cortical and vessel formation with scored migrating cortical ECs and CMs (fli1:GFP and gata4:DsRed cells extending projections) in a representative 20 mm (91 dpf) heart. (I) Scoring of synchronous versus asynchronous sprouting (81 hearts, 12–27 mm/40–119 dpf). Blue dashed box: asynchronous sprouting (cell extending without simultaneous extension of the other) and red box: synchronous formation (CM and EC migrating at the same time). Scale bars, 50 μm.

To investigate this, we performed similar imaging analysis during juvenile development including the reporter gata4:DsRed, which labels actively expanding cortical CMs (18). AcTub is significantly enriched on the surface of the ventricle and overlaps with gata4:DsRed+ CMs (Figure 1D, S1A–D). Nascent coronary vessels align along these CMs and sprouting endothelial tip cells extend in the same orientation (Figure 1D, S1A–D). While there are regions containing only gata4+ CMs or coronary vasculature, developing coronary vessels are found disproportionally in gata4+ regions (3.0x average enrichment vs. 0.7x in gata4− regions), suggesting coordination or common responsiveness to shared developmental signals (Figure 1D–G, S1A–G). Consistent with regional neighborhoods of coordinated ventricle expansion, Gata4:DsRed+ regions also show significantly elevated rates of cell proliferation (Figure S1H and I).

Despite this broad regionalized coordination, most active vascular and myocardial expansion occurs asynchronously at the cellular level (Figure 1H and I, S2A–G). The majority (79%) of gata4:DsRed+ CMs and half of the coronary tip cells appear to expand without direct contact with each other (Figure S2A–G), suggesting that direct interactions between CMs and endothelial sprouts are limited or transient.

Developing and regenerating CMs directionally extend to interact transiently with coronary endothelial cells.

To assess interactions during earlier development, we employed microfluidic assisted ex vivo live imaging to observe CM projections as they expand over the ventricle. We imaged fli:GFP and gata4:DsRed hearts at 15 mm/56 dpf and 21 mm/79 dpf, when cortical layer expansion rapidly covers most of the ventricle. Live imaging revealed complex, multidirectional interactions between coronary ECs and gata4:DsRed+ CMs, including ECs migrating along CMs, CMs extending along nascent vessels, and heterotypic convergence into avascular regions (Figure S3A–D; Supplemental Movies 1–4). However, observations were limited by the slow rate of cell motility and a 40-hour imaging window during juvenile timepoints.

The zebrafish ventricle also regenerates following injury, with the wound site repopulated by CMs and coronary vasculature (9, 22–24). This regenerative process offers a spatially well-defined model with a synchronized start at the injury border and completion at the apex over 3 days (7–10 days post amputation), which can be entirely imaged ex vivo using a microfluidic device (23, 25). EC sprouts migrate from all sides of the wound and converge to form nascent coronary vessels spanning the regeneration site (Figure S3E, pink arrowheads; Supplemental Movie 5), consistent with fixed imaging despite being imaged ex vivo which can introduce artifacts (9, 22–24). CMs similarly extend filopodial projections and populate the wound site, albeit more slowly than ECs (Figure S3F; Supplemental Movie 6). Direct CM-EC interactions occur during regeneration, and the longer imaging window demonstrated that these interactions are transient, consistent with the minority of direct interactions observed in fixed and live imaging during development. The full spectrum of behaviors was observed, including CMs extending along vessels, vessels sprouting alongside CMs, mutual extension, and independent displacement (Figure S3G; Supplemental Movie 7). Together, these results led us to hypothesize that myovascular alignment is unlikely to depend solely on direct bidirectional EC-CM signaling, and that CMs and ECs instead respond to shared signals that coordinate their regional expansion and consequent alignment.

Coronary angiogenesis is associated with myocardial proliferation, and formed coronary vessels restrict myocardial expansion.

Given broad overlap of coronary vessel and myocardial expansion but transient EC-CM interactions, we investigated whether coronary vessel formation affects myocardial expansion. Proliferating CMs (EdU+) are found significantly closer to coronary vessel ECs in the maturing ventricle (average shortest distance 12.81 μm vs. 23.99 μm for non-proliferative CMs; Figure 2A–C). A similar relationship has been described previously in mammals(4); in zebrafish, these proliferating CMs are 6.3x more enriched near nascent non-lumenized and sprouting coronary vessels compared to mature lumenized vessels (normalized for region size, Figures 2A–B, D, S4A–C).

Figure 2. Coronary vasculature constrains myocardial expansion and suppresses hypoxia-responsive epicardial gene expression.

(A-B) Confocal projections of hearts labeled with EdU, Mef2c, and GFP (fli1:GFP; green; EC). Only EdU/Mef2c double-positive pixels (EdU+ CM, magenta) are shown. Proliferative CMs are observed near both mature lumenized vessels and active angiogenic sprouts. (C) Distance between coronary endothelial cells (coEC) and proliferative or non-proliferative CM nuclei (Mann-Whitney test, n = 142 EdU+, 294 EdU−). (D) Proliferative CM enrichment near vascular sprouts relative to mature vessels at 57 dpf/18 mm SL. Normalized proximity reflects nearest coEC subtype to proliferative CMs, corrected for subtype abundance (t-test, n = 4). (E-G) Mef2c/EdU labeling in cxcr4a^um20 mutants at 70 dpf/20 mm SL showing increased CM proliferation (Welch’s t test, n = 8 WT, 6 cxcr4a^um20). (H-I) fli1:GFP and gata4:DsRed hearts from cxcr4a wild-type or mutants showing loss of coronary vessels and disorganized gata4:DsRed+ CMs. (J) gata4:DsRed ventricular coverage in cxcr4a^um20 mutant versus wild-type (unpaired t test, n = 17 WT, 34 cxcr4a^um20). (K-L) Later-stage hearts (26 mm SL/84 dpf) showing disordered and overlapping ventricular expansion of gata4:DsRed in cxcr4a^um20 mutants versus wild-type. (M) qPCR of wild-type and cxcr4a^um20 ventricles at 70 dpf showing elevated hypoxia-responsive gene expression (hk2, phd3, slc38a5a) in mutants. (unpaired t test, n = 12). (N-O) Hif2a protein (gray) and tcf21:NLS-GFP (red) immunostaining showing increased Hif2a in the outer ventricular wall, particularly in tcf21:NLS-GFP+ epicardial cells. (P) Schematic of epicardial cell isolation for bulk RNAseq from cxcr4a^um20 or wild-type ventricles at 70 dpf/21 mm. (Q) Gene ontology (GO) analysis showing enrichment of transcripts associated with ECM organization (blue), metabolism (orange), and hypoxia (chartreuse) amongst others (black) in cxcr4a^um20 epicardium (enrichment >2.4 and p < 0.001). (R) Fold change of selected differentially expressed genes identified through GO analysis, including collagens, lysyl oxidase-like (Loxl) enzymes, glycolytic regulators, and hypoxia response factors, color-coded by GO class: ECM (blue), metabolism (orange), or hypoxia (chartreuse). (S-Y) Experimental schematic and whole-mount confocal projections of col12a1b:GFP, loxl2a:GFP and vegfaa:GFP (gray) hearts from PBS control and 2 days post-PHZ treatment. (Z) Epicardial surface coverage of col12a1b:GFP, loxl2a:GFP and vegfaa:GFP with and without PHZ (unpaired t tests, n = 5 PBS col12a1b, 4 PHZ col12a1b, 17 PBS loxl2a, 19 PHZ loxl2a, 16 PBS, vegfaa, 13 PHZ vegfaa). Scale bars, 50 μm

To directly test the impact of reduced coronary vasculature on myocardial expansion, we used cxcr4a^um20 mutant zebrafish that lack functionally developed coronary vessels and are viable to adulthood (Figure S4H) (2, 26, 27). As cxcr4a is expressed specifically in coronary ECs, the mutation’s effects are anticipated to be specific to coronary development, affording a unique opportunity to examine coronary vessel loss without directly affecting CM signaling (2). In cxcr4a^um20 mutants, CM proliferation was increased 4.2-fold compared to wild-type siblings, suggesting coronary vessels normally inhibit myocardial expansion (Figure 2E–G). Broadly, cxcr4a^um20 mutants have over twice as many proliferative cells per mm² than wild-type siblings (Figure S4D–F), and their ventricles are significantly enlarged relative to body length (141 ± 64% larger on average; Figure S4G). Furthermore, cxcr4a^um20 mutants show more cells expressing gata4:DsRed over the ventricle at 56 dpf (20 mm; Figure 2H–J). At later stages (84 dpf), cxcr4a^um20 mutant ventricles show higher-order cortical structural defects, with gata4:DsRed+ CMs appearing disorganized and overlapping rather than forming cortical ribbons (Figures 2K and L, S4I and J). Together, this suggests that despite proliferative CMs being enriched near coronary ECs in WT hearts, vessels restrict CM proliferation to allow ordered myocardial expansion.

Coronary vessels repress epicardial hypoxia and responsive signaling.

Coronary vasculature provides oxygenated blood to the thickening juvenile myocardium (2) and hypoxia impacts cell metabolism and proliferation including during myocardial development and regeneration (28–32). To investigate the role of hypoxia in myocardial expansion, we analyzed the effect of coronary vessel loss on ventricular hypoxia. hk2, phd3, and slc38a5a (33–35) transcript levels were all significantly elevated in cxcr4a^um20 ventricles compared to WT controls (Figure 2M). Hif2a was used to identify hypoxic regions, as it is more responsive to chronic hypoxia than Hif1a (36–38). Staining was strongest on the outside of the ventricle and in tcf21:nGFP+ epicardial cells, which are also more numerous in cxcr4a^um20 mutants (Figure 2O–P; Figure S4K–M). CMs beneath the epicardium displayed limited Hif2a staining, suggesting hypoxia is sensed by the epicardium, which then signals to the CMs.

The epicardium has a well-established role providing signals and progenitor cells during heart development (19–21). The juvenile epicardium appears particularly hypoxic and the loss of coronary vessels increases this hypoxia (Figure 2N–O). The tcf21:nGFP line allowed us to sort this population and characterize transcriptional changes by bulk RNA-seq in epicardial cells of cxcr4a^um20 mutant juvenile hearts (Figure 2P). The most enriched class of differentially upregulated genes have roles in the regulation of extracellular matrix (ECM), glycolysis, and hypoxia response (Figures 2Q and S5). Upregulated genes include fibril collagen genes (Figures 2R, S6; e.g., col5a3a, col11a1a, col11a1b, col27a1b) and Lysyl oxidase-like (Loxl) enzymes that crosslink fibril collagen. Like collagen crosslinking, fibril-associated collagens including Col12a1b regulate collagen fibril spacing and assembly during fibril genesis to set the mechanical properties of ECM (39). Together, these gene products regulate ECM stiffness (Figure 2R and S6A–D), which is known to be directly affected by hypoxia (40–43). Loxl genes are known to be Hif-targets, as are many collagen, glycolysis and angiogenesis genes (44). Prominent among the upregulated genes is Vegfa (Figure 2R) (44–47), with known roles in both angiogenesis and myocardial expansion (3, 48). Zebrafish were treated with Phenylhydrazine (PHZ) that induces hemolytic anemia causing systemic hypoxia and increased myocardial stress due to the resulting compromised cardiac output (49). Consistent with increased expression in hypoxic epicardial cells, PHZ treatment of juvenile zebrafish at 42 or 70 dpf rapidly increased col12a1b:GFP, loxl2a:GFP and vegfaa:GFP coverage of the ventricle within 2 days (Figure 2S–Z). We collectively designate genes upregulated in tcf21:nGFP+ cells of hearts lacking coronary vessels as hypoxia-responsive epicardial genes (HREGs).

HREGs are restricted to areas undergoing myocardial and coronary vessel expansion and repressed by coronary vessels.

Expression of vegfaa:GFP, loxl2a:GFP, and col12a1b:GFP was not uniform over the ventricle but restricted to specific epicardial regions (Figure 2S–Z). Prior to the onset of coronary vessel sprouting, vegfaa:GFP is expressed in distinct contiguous epicardial regions (Figure 3A, S7A). These regions become vascularized by flt1:tdTomato-expressing coronary vessels as ventricle maturation proceeds (Figure 3B, demarcated in yellow) and a second, strongly expressing perivascular population is observed next to the formed vessels (Figure 3B, arrowhead). Vegfaa:GFP becomes downregulated in contiguous epicardial patches but remains a discreet perivascular population along formed vessels from around 70 dpf (25 mm SL; Figure S7I–K).

Figure 3. Coronary vessel loss alters epicardial gene programs, promoting ECM remodeling and hypoxia-driven myovascular expansion.

(A) Whole-mount confocal projection showing vegfaa:GFP (green) expression in discrete epicardial domains (demarked yellow) enriched near the atrioventricular canal (AVC) prior to coronary vessel formation (flt1enh:tdTomato, magenta). (B) Confocal projection showing progressive restriction of vegfaa:GFP to perivascular mural cells (yellow arrowhead) as vascularization proceeds, with residual expression in epicardial patches undergoing vascularization (yellow). (C) loxl2a:EGFP (green) expression in avascular epicardial regions near the AVC at 9 mm/28 dpf. (D) At 19 mm/47 dpf, epicardial loxl2a:EGFP is expressed near vascular sprouts and perivascularly regions (yellow arrowhead). (E) col12a1b:EGFP (green) at 29 dpf/9 mm showing restricted epicardial expression near the AVC, which (F) later associates with angiogenic sprouts. (G) vegfaa:GFP (green) expression adjacent to gata4:DsRed+ CMs (magenta) during ventricular expansion. (H) In cxcr4a^um20 mutants, epicardial vegfaa:GFP is expanded across the ventricle. (I) Quantification of vegfaa:GFP ventricular coverage in wild-type or cxcr4a^um20 (unpaired t-test, n = 44 WT, 32 cxcr4a^um20). (J) loxl2a:GFP (green) association with gata4:DsRed+ CMs in wild-type at 59 dpf/21 mm. (K) loxl2a:GFP in cxcr4a^um20 ventricle. (L) loxl2a:GFP ventricular coverage in wild-type or cxcr4a^um20 (unpaired t-test, n = 46 WT, 41 cxcr4a^um20). (M) col12a1b:GFP (green) at 70 dpf/22 mm associated with expanding gata4:DsRed+ (magenta) myocardium. (N) col12a1b:GFP in cxcr4a^um20 hearts. (O) col12a1b:GFP coverage in wild-type or cxcr4a^um20 (unpaired t-test, n = 24 WT, 28 cxcr4a^um20). (P) Pressure–stretch curves of wild-type and cxcr4a^um20 ventricular myocardium; mutant ventricles (green) show significantly increased stiffness versus wild-type (gray) (n ≥ 6). (Q–S) gata4:DsRed-labeled compact myocardium (gray) following PBS (Q), DMOG (R), or PX-478 (S) injection. (T–V) fli1:GFP-labeled coronary vessels (gray) at 70 dpf following PBS (T), DMOG (U), or PX-478 (V) injection from 56 dpf. (W) Compact myocardium coverage quantification. (X) Coronary vessel coverage quantification. Data represent mean ± SEM. P-values by one-way ANOVA with Holm–Šídák test, n = gata4: 49 PBS, 49 DMOG, 46 PX-478; fli1a: 43 PBS, 48 DMOG, 45 PX-478. Scale bars, 50 μm.

The loxl2a:GFP expression pattern at 28 dpf/9 mm is similar to vegfaa:GFP, found in patches of the epicardium not yet vascularized or containing expanded myocardium (Figure 3C). Later during sprouting angiogenesis, loxl2a:GFP is expressed around the coronary sprout and long nascent vessels (Figure 3D). col12a1b:GFP expression is more restricted than vegfaa:GFP and loxl2a:GFP but still occurs in discrete small patches of epicardium associated with angiogenesis (Figure 3E–F, S7E–H).

Epicardial cells in direct contact with gata4:DsRed+ CMs express high levels of vegfaa:GFP and loxl2a:EGFP (Figure 3G and J, S7A–D), while col12a1b:EGFP is expressed next to a subset of gata4:DsRed CMs (Figure 3M, S7E–H). In cxcr4a^um20 mutant ventricles, vegfaa:GFP perivascular expression is no longer discernible but gata4:DsRed CM associated expression is more expansive (Figure 3H and I, S7L–N), with similar changes observed for loxl2a:EGFP and col12a1b:EGFP (Figure 3J–O, S7O–T). Expansion of HREG expression in cxcr4a^um20 mutant hearts precedes the emergence of gata4:DsRed myocardium (Figure S7L–T) suggesting that these factors, together with other upregulated HREGs (Figure 2S), are driving regional myovascular expansion and are reciprocally repressed in the epicardium by the coronary vasculature.

To investigate whether increased vessel density reduces HREG expression, we stably expressed cxcl12b in a region of the myocardium to increase coronary vessel density (2) and examined vegfaa:GFP, finding it was reduced in those regions suggesting it is repressed by coronary vasculature (Figure S7U–W).

Col12a1b controls the spacing and assembly of collagen fibrils which are then stabilized by Loxl2a-mediated crosslinking to elastin (50, 51). To determine the effect of increased collagen and collagen stabilizers on tissue tensile properties, we used pipette aspiration to compare focal deformation in adult WT and cxcr4a^um20 mutant ventricles. cxcr4a^um20 mutant ventricles were significantly stiffer, consistent with high levels of ECM crosslinking enzymes (Figure 3P; Figure S7X).

We next asked whether increasing hypoxia is sufficient to enhance myocardial expansion. We targeted Hif-signaling by treating juvenile zebrafish with Dimethyloxalylglycine (DMOG), a PHD inhibitor that increases Hif stabilization and downstream signaling (52), or PX-478, which blocks Hif stabilization (53, 54). Juvenile zebrafish (56 and 70 dpf) were injected with DMOG or PX-478 daily for two weeks, and coverage of coronary vessels and cortical CMs was observed. Two weeks of daily DMOG injection significantly increased coronary vessel coverage by 60.3% ± 17.1% (p = 0.001) and cortical CM coverage by 44.0% ± 16.1% (p = 0.022). In contrast, daily PX-478 injection decreased coverage of vessels by 45.2% ± 17.4% (p = 0.011) and cortical CM coverage by 55.7% ± 16.4% (p = 0.003) (Figure 3Q–X). These results indicate that Hif-signaling is a major driver of myocardial expansion and coronary vessel formation with coronary vessel formation itself repressing hypoxia and its downstream signals in a negative feedback loop that is lost in cxcr4a^um20 mutants.

Epicardial subpopulations are required for both coronary development and cxcr4a^um20 mutant myocardial hyper-expansion.

To investigate if Hif-driven signals promoting myocardial expansion are mediated through HREGs, we ablated the epicardial population during coronary and myocardial expansion. At 44 dpf/12 mm, the tcf21:mCherry-Nitroreductase (NTR) line showed more restricted expression than tcf21:nGFP and tcf21:BFP (Figure S8A–E), largely encompassing domains of vegfaa:GFP, loxl2a:GFP and col12a1b:GFP expression (Figure 4A–C). PHZ treatment resulted in expansion of tcf21:mCherry-NTR expression within 48 hours suggesting that the transgene is also hypoxia-responsive similar to vegfaa:GFP, loxl2a:GFP and col12a1b:GFP (Figure S8F–H). Nascent coronary vessels and sprouts were later found within the tcf21:mCherry-NTR+ domain and excluded from regions not covered by these epicardial cells (Figure 4D–E).

Figure 4. Regional epicardial signaling drives coronary vessel development and myocardial expansion.

(A–C) Confocal projections showing regional overlap of vegfaa:GFP (green, A), loxl2a:GFP (green, B), or col12a1b:GFP (green, C) with tcf21:mCherry-NTR (magenta) at 42 or 48 dpf (13 or 17 mm). (D) fli1:GFP (green) and tcf21:mCherry-NTR (magenta) hearts showing nascent coronary vessels within tcf21:mCherry-NTR+ regions (magenta). (E) Vessel coverage (%) in tcf21:mCherry-NTR positive versus negative regions (unpaired t-test, n = 9). (F–H) vegfaa:GFP (green, F), loxl2a:GFP (green, G), and col12a1b:GFP (green, H) with tcf21:mCherry-NTR (magenta) after MTZ treatment at 40 dpf. (I) fli1:GFP (green) and tcf21:mCherry-NTR (magenta) at 56 dpf/20 mm after 10 μM MTZ from 42 dpf. (J) Vessel coverage (% ventricle) in MTZ-treated versus control tcf21:mCherry-NTR hearts (unpaired t-test, n = 24 control, 27 MTZ). (K–N) gata4:GFP (green) with tcf21:mCherry-NTR (magenta) in wild-type (K, L) and cxcr4a^um20 (M, N) hearts without (K, M) or after (L, N) MTZ from 56 dpf. (O) gata4:GFP ventricular coverage in wild-type, untreated cxcr4a^um20, or MTZ-treated cxcr4a^um20 (one-way ANOVA with Tukey’s post hoc, n = 15 WT non-MTZ, 13 WT MTZ, 30 cxcr4a^um20 non-MTZ, 27 cxcr4a^um20 MTZ). (P–R) Control (P) and dnvegfaa-induced (heatshock 56–70 or 70–84 dpf, Q) hearts showing coronary vessels (green) and cortical myocardium (magenta). (R) Coronary vessel and cortical myocardium coverage quantification (unpaired t-test, n = fli1a: 11 WT, 14 dnvegfaa; gata4: 11 WT, 12 dnvegfaa). Data represent mean ± SEM. Scale bars, 50 μm.

NTR converts the pro-drug metronidazole (MTZ) to cytotoxic metabolites, resulting in the ablation of NTR-expressing cells (55). With MTZ addition, we observed near-complete loss of vegfaa:GFP, loxl2a:GFP, and col12a1b:GFP expressing cells, but not all epicardial cells (Figure 4F–H, S8I–K). Loss of this epicardial subpopulation resulted in significantly reduced coronary vessel expansion and perturbed myocardial expansion (Figure 4I–L, O). To test whether this requirement was independent of formed coronary vessels, we analyzed epicardial cell loss in cxcr4a^um20 mutant hearts (2). MTZ treatment reduced gata4:DsRed coverage to WT levels, rescuing the hyper-expansion observed in the cxcr4a^um20 mutant ventricle (Figure 4M–O).

Given the characterized effects of Vegfaa in cardiac development and pathogenesis (3, 4, 9, 56–58), we tested whether vegfaa induction contributes to hypoxia-mediated myovascular expansion. Induced expression of dominant negative Vegfaa (dnvegfaa) from 56 to 70 dpf or 70 to 84 dpf was sufficient to partially recapitulate the PX-478 phenotype by significantly reducing coronary vessel coverage and juvenile myocardial expansion (Figure 4P–R). Together, these results demonstrate that epicardial derived factors can drive both coronary vessel development and cardiac muscle expansion and that these programs are regionally restricted by coronary vessels.

Myocardial expansion drives regional epicardial activation and coronary angiogenesis.

We next investigated if myocardial expansion also impacts HREG expression to form a positive feedback loop. The gata4:DsRed+ cortical layer is adjacent to highly hypoxic epicardium surrounding gata4:DsRed+ CMs (Figure 5A and B, S9A). Interfering with Gata4 function is known to reduce cortical expansion (18). Treatment of conditional dominant-negative Gata4 transgenic zebrafish (actb2:LoxP-TagBFP-LoxP-mCherry,dngata4; myl7:CreER) with 4-hydroxytamoxifen (4OHT) at 28 or 35 dpf for 6 hours resulted in regional dnGata4 expression in the expanding myocardium (Figure S9B–E). The gata4:GFP reporter is expressed at high levels in dnGata4− regions, but is lost in dnGata4+ regions (red; Figure 5C and D). The gata4:GFP+ regions that are competent to form cortical myocardium start to envelop the dnGata4+ regions, but this is incomplete resulting in uneven thickening or thinning of the compact myocardium across the ventricle (Figures 5F, S9F and G). Hif2a signal intensity in epicardium adjoining unexpanded myocardium (Figure 5H–J, red arrow; S9H–K) is lower than in epicardium near expanding myocardium, indicating that unexpanded myocardium is less hypoxic (Figure 5I, blue arrow). Consistent with expression induced by localized epicardial hypoxia driven by myocardial expansion, vegfaa:GFP, loxl2a:GFP and col12a1b:GFP are strongly induced in dngata4− regions, and excluded from dngata4+ regions (red, Figures 5K–Q; S9L–N). Reciprocally dense coronary vasculature is observed in dngata4− regions but excluded from dngata4+ regions (Figure 5R–T). The dngata4+ regions also appear to balloon out in bulges between the dense vasculature, suggesting these regions lack the stiff ECM and structural rigidity provided by the high levels of loxl2a and col12a1b in dngata4− regions (Figure 5S, arrowhead). Earlier expression of dngata4 (with 24-hour 4OHT induction at 21 dpf) results in larger dngata4− regions, hypoplastic coronary arteries, and low survival beyond juvenile stages), with a reduction of proliferative ECs also observed (Figure S9O–S). This suggests regional myocardial expansion drives hypoxia and HREG expression, which in turn drives coronary vessel formation in a regionally appropriate manner incorporating a balance of positive and negative feedback from each, respectively (Figure 6).

Figure 5: Myocardial expansion reciprocally drives epicardial hypoxia and signaling.

(A) Heart section showing gata4:GFP+ cortical myocardium (magenta) adjacent to hypoxic epicardium (yellow). (B) Hif2a intensity within 20 μm of tissue edge in gata4:GFP+ versus gata4:GFP− domains (paired t-test, n = 15). (C–D) Schematic of clonal expansion and confocal projections showing gata4:GFP expression adjacent to and over clonally expanded dnGata4+ regions after 4OHT induction at 28 dpf (C) versus uninduced (D). (E) Gata4:GFP intensity in dnGata4, mCherry+ versus dnGata4, mCherry− domains (paired t-test, n = 5). (F) AFOG-stained section from dnGata4 (4OHT at 28 dpf) hearts showing regionally variable outer wall thickness (thick, blue arrow; thinner, red arrow). (G) Myocardial thickness in adjacent dnGata4, mCherry+ versus mCherry− domains (paired t-test, n = 36). (H–I) Hif2a levels in epicardium adjoining unexpanded (H; red arrow, I) versus expanded myocardium in dnGata4 hearts; thickened myocardial regions distal to the endocardium (Endo) show higher Hif2a and coronary vessels (CV). (J) Hif2a intensity within 20 μm of tissue edge in dnGata4, mCherry+ versus mCherry− domains (paired t-test, n = 9). (K–P) vegfaa:GFP (green, K, L), loxl2a:GFP (green, M, N), and col12a1b:GFP (green, O, P) are enriched in dnGata4− and reduced in dnGata4+ (red) regions. (Q) vegfaa, col12a1b, and loxl2a GFP intensity in dnGata4, mCherry+ versus mCherry− domains (paired t-tests, n = 9 vegfaa, 6 col12a1b, 6 loxl2a). (R–S) Coronary vasculature (fli1:GFP, green) in dnGata4− (R) and dnGata4+ (red, S) regions; dnGata4+ regions show bulging morphology (arrowhead). (T) Coronary vessel density in mCherry-labeled dnGata4-expressing versus non-expressing regions (paired t-test, n = 5). Data represent mean ± SEM. Scale bars, 50 μm.

Figure 6.

Model of epicardial coupled myocardial growth and coronary vessel development.

DISCUSSION

Our study reveals a regulatory network governing the coordinated development of the coronary vasculature and myocardial tissue in the zebrafish heart. Our findings demonstrate that coronary vessel positioning and myocardial expansion are interconnected developmental programs spatially and temporally coordinated through epicardial signaling. This coordination is important for balanced ventricle formation as the zebrafish grows, ensuring that the expanding myocardium maintains proper architecture and receives adequate vascular supply to meet increasing metabolic demands during juvenile development.

We define a role for epicardial hypoxia as a regulatory mechanism controlling coordinated coronary angiogenesis and myocardial expansion. The epicardium appears to be a highly hypoxia-sensitive tissue that orchestrates cardiac development through regulated expression of factors including Vegfaa, Loxl2a, and Col12a1b. This hypoxia-responsive gene expression program is spatially restricted to regions undergoing active myocardial and vascular expansion, suggesting a feed-forward mechanism in which tissue growth creates local hypoxic conditions that further drive developmental processes. Consistent with this, we observed increased expression of vegfaa, loxl2a, and col12a1b in hyperexpanded regions of partial dnGata4 hearts. Because these hearts display variable myocardial thickness, other factors, including epicardial stretch, systemic flow dynamics, and cardiac output, may also contribute to Hif-signaling and transcriptional changes (59–61). In addition to the epicardium, vegfaa:GFP and loxl2a:GFP expression is also observed in epicardial-derived mural cells (2, 19, 62–64) after vessel development, which may additionally contribute to gata4:DsRed+ CM sprouting along formed vessels (25.9% of expanding CMs; Figure 2), as previously observed after cryoinjury (65).

Under epicardial hypoxia, the most significantly impacted genes are involved in ECM organization, particularly collagen genes and Loxl enzymes. col12a1b:GFP exhibits spatial restriction to areas of active heart muscle expansion and given its role in collagen fibril spacing and assembly, this pattern suggests that ECM remodeling is tightly coordinated with tissue growth. These findings complement recent work identifying hapln1-expressing epicardial cell subpopulations essential for CM expansion compact muscle layer formation (17, 66), suggesting that multiple ECM components and regulatory mechanisms collaborate to maintain myocardial architecture.

The regional expression of epicardial factors in discrete patches overlaying areas of myocardial expansion may reflect local myocardial stress that creates hypoxic microenvironments, activating epicardial cells to promote angiogenesis and continued myocardial growth ensuring vascular supply keeps pace with myocardial demand. A shared program of ECM modification, hyperplastic myocardial thickening, and angiogenic cues including Vegfa (3, 4, 17, 62, 67) mediates the coordination of myovascular development. Consistent with this, our live imaging studies reveal that CMs and ECs exhibit coordinated extension patterns but largely transient direct interactions, suggesting that coordination between these cell types is primarily mediated by third-party signals rather than direct cell-to-cell communication. The HREG signaling program identified here likely represents this third-party coordinator. Collective reduction of HREG expression results in reduced CM and EC expansion, though how this gene network conjointly mediates its effect remains to be fully elucidated. We found that reducing Vegf-signaling was sufficient to decrease myocardial expansion back to WT levels in cxcr4a^um20 mutants lacking coronary vessels. Previous data established that inducing vegfaa expression increases CM proliferation, potentially through endocardial notch signaling (3), and VEGFA-VEGFR2 has been implicated in coordinated CM and coronary EC proliferation during mouse heart development (4). Interestingly, EC-specific deletion of VEGFR2 reduces CM proliferation, indicating that some myogenic effects are mediated via coronary and endocardial ECs (4). We establish Vegf-signaling as a conserved regulator of myocardial expansion during development that is inhibited by coronary vasculature formation. VEGF-signaling is also considered a major regulator of hypertrophy in response to pressure overload (68), with implications for understanding heart failure pathophysiology. Disruption in angiogenesis and its coordination with muscle growth can promote progression from adaptive cardiac hypertrophy to heart failure (68–70). We find that myocardial walls in zebrafish lacking coronary vessels have increased stiffness compared to WT, similar to that observed in heart failure with preserved ejection fraction for which microvascular dysfunction is a major contributing factor (71–74).

The indirect role of coronary vessels in repressing myocardial expansion appears at odds with the established role of blood vasculature in regenerative contexts, where coronary vessels promote myocardial expansion during regeneration (2, 9, 63). dnVegfaa-expressing zebrafish fail to revascularize after injury and do not regenerate, while cxcr4a and pdgfrb mutants, which lack normal coronary vasculature, show impaired regeneration (2, 9, 63). However, chronic vegfaa overexpression also inhibits zebrafish heart regeneration (3), suggesting that tight temporal regulation of Vegf signaling is critical: transient hypoxia-induced expression promotes early regenerative responses, whereas persistent elevation may disrupt later regenerative processes. Consistent with this, cxcr4a mutants with persistent vegfaa expression also fail to regenerate. Alternatively, the hyperproliferative immature myocardium in cxcr4a^um20 mutants may exhaust its proliferative potential during development or experience acute metabolic stress, leaving it with reduced capacity to mount proliferative responses following injury. Additionally, Vegf-inhibition may affect CM proliferation through mechanisms independent of angiogenesis, including direct effects on non-vascular cell types (57, 75–78). Independent of Vegf signaling, differences in the hypoxic environment between development and regeneration may also contribute to the differing impact of coronary vessels. During development, hypoxia provides growth-promoting signals, whereas post-injury hypoxia also triggers inflammatory responses that may actively inhibit myocardial expansion (79–82).

Despite these differences, the transient CM-EC interactions observed during regeneration mirror those seen in development, and the previously observed upregulation of epicardial vegfaa and col12a1b suggests that the signaling mechanisms identified here may also operate during cardiac repair (9, 62). These findings in zebrafish align with recent work in mice showing that physiological hypoxia can reactivate epicardial progenitor programs, promoting cell cycle re-entry and developmental gene expression (83, 84). Developmentally, hypoxia regulates epicardium-derived cell differentiation and migration in mammals while also inducing CM proliferation and hyperplasia (85–89). As observed in zebrafish, the mouse epicardium appears to be a hypoxic niche whose hypoxia increases after MI to drive responsive signaling (90, 91). This cross-species conservation suggests that hypoxia-mediated epicardial activation represents a fundamental mechanism for cardiac adaptation and repair, with epicardial hypoxia serving as a central integrator of myocardial expansion and coronary angiogenesis that provides insights for slowing cardiac disease progression and improving tissue regeneration.

METHODS AND MATERIALS

Zebrafish husbandry and transgenics

Zebrafish (Danio rerio) were maintained at 28°C on a 14:10 h light:dark cycle under IACUC-approved protocols (Weill Cornell Medical College). Transgenic and mutant lines are described in the SI Appendix. Cre-mediated recombination was induced by overnight 4-OHT treatment (10 μM), NTR-mediated ablation by metronidazole (10 mM MTZ) (55), and anemia by phenylhydrazine exposure (3 μg/mL, 30 min) (92). Heat shock was performed at 39μC for 1 hour three times weekly (9). For pharmacological experiments, zebrafish received intraperitoneal injections of DMOG (50 mM) or PX-478 (5 mM) three times weekly for two weeks (93).

Histology and imaging

Hearts were fixed in 4% PFA, cryoprotected in 30% sucrose, embedded in OCT, and cryosectioned at 10 μm. Fluorescence immunostaining used primary antibodies against GFP (chicken, 1:400), acetylated tubulin (mouse, 1:200), and HIF2α (mouse, 1:200), with appropriate fluorescent secondaries. Whole hearts were mounted in 1% low-melting-point agarose and imaged on a Zeiss LSM 980 confocal microscope (2, 25). Time-lapse imaging of isolated beating hearts was performed using a previously described microfluidic device on a Leica DMi8 inverted microscope (2, 25).

Transcriptomics and gene expression

Epicardial cells were isolated from Tg(tcf21:nEGFP) ventricles at 70 dpf by FACS and total RNA extracted for bulk RNA-seq. Libraries were prepared using SMART-Seq v4 and sequenced on an Illumina NovaSeq X Plus (paired-end 2×100 cycles); data are deposited under GEO accession GSE313651 (94). Gene expression was quantified by qPCR using the 2^−ΔΔCt method with ef1α as reference gene (95).

Tissue mechanics

Cardiac mechanical properties were assessed by micropipette aspiration of isolated hearts. Incremental suction was applied and tissue deformation analyzed to generate stretch ratio versus pressure curves, fitted with axial Cauchy stress models assuming incompressible material with exponential behavior; strain energy density was calculated for stretch ratios of 1–1.5.

Image analysis and statistics

Image analysis was performed in Fiji and ZEN software (96). Coronary vessel coverage was quantified by thresholding fli1a:GFP signal within gata4:DsRed positive and negative ventricular regions normalized to region area. EdU⁺ cardiomyocyte proximity to coronary endothelial cells was measured by co-labeling image arithmetic and distance calculation. Quantitative data are expressed as mean ± SEM. Two-group comparisons used Student’s t-test; three or more groups used one-way ANOVA with Tukey’s or Holm-Sidak post hoc tests. Where normality was not met (D’Agostino-Pearson test), the Mann-Whitney test was applied. All analyses were performed in GraphPad Prism 7.

SIGNIFICANCE STATEMENT.

The heart must grow new blood vessels (coronary vasculature) to meet the demands of expanding heart muscle (myocardium), yet how these two processes are coordinated remains poorly understood. We discover that coronary vessel growth occurs in regional waves that synchronize with rapid local expansion of the ventricular wall, and that this vascularization in turn restrains further muscle growth in those same regions. We identify the epicardium, the outer layer of cardiac cells, as a local regulator of this reciprocal relationship. This coordinated feedback shapes the ventricle’s architecture for optimal pumping function. Understanding this cell coupling is clinically important as its disruption underlies congenital heart defects and adult conditions such as heart failure, while insights into this process could advance cardiac regenerative therapies.