hapln1 defines an epicardial cell subpopulation required for cardiomyocyte expansion during heart morphogenesis and regeneration

Abstract

BACKGROUND:

Certain non-mammalian species like zebrafish have an elevated capacity for innate heart regeneration. Understanding how heart regeneration occurs in these contexts can help illuminate cellular and molecular events that can be targets for heart failure prevention or treatment. The epicardium, a mesothelial tissue layer that encompasses the heart, is a dynamic structure that is essential for cardiac regeneration in zebrafish. The extent to which different cell subpopulations or states facilitate heart regeneration requires research attention.

METHODS:

To dissect epicardial cell states and associated pro-regenerative functions, we performed single-cell RNA-sequencing and identified 7 epicardial cell clusters in adult zebrafish, 3 of which displayed enhanced cell numbers during regeneration. We identified paralogs of hyaluronan and proteoglycan link protein 1 (hapln1) as factors associated with the extracellular matrix (ECM) and largely expressed in cluster 1. We assessed HAPLN1 expression in published scRNA-seq datasets from different stages and injury states of murine and human hearts, and we performed molecular genetics to determine the requirements for hapln1-expressing cells as well as functions of each hapln1 paralog.

RESULTS:

A particular cluster of epicardial cells had the strongest association with regeneration and was marked by expression of hapln1a and hapln1b. hapln1 paralogs are expressed in epicardial cells that enclose dedifferentiated and proliferating cardiomyocytes during regeneration. Induced genetic depletion of hapln1-expressing cells or genetic inactivation of hapln1b altered deposition of the key ECM component hyaluronic acid (HA), disrupted cardiomyocyte proliferation, and inhibited heart regeneration. We also found that hapln1-expressing epicardial cells first emerge at the juvenile stage, when they associate with and are required for focused cardiomyocyte expansion events that direct maturation of the ventricular wall.

CONCLUSIONS:

Our findings identify a subset of epicardial cells that emerges in post-embryonic zebrafish and sponsors regions of active cardiomyogenesis during cardiac growth and regeneration. We provide evidence that, as the heart achieves its mature structure, these cells facilitate HA deposition to support formation of the compact muscle layer of the ventricle. They are also required, along with the function of hapln1b paralog, in production and organization of HA-containing matrix in cardiac injury sites, enabling normal cardiomyocyte proliferation and muscle regeneration.

INTRODUCTION

Heart failure is a severe, disabling, and costly human condition of insufficient cardiac output to meet the body’s demands for oxygen and blood. The primary risk factors contributing to heart failure are coronary artery disease, high blood pressure, and previous myocardial infarction events, all of which coincide with significant cardiac muscle loss and fibrosis¹. Although the damaged adult mammalian heart has minimal capacity to restore lost cardiac muscle, mammals display measurable regeneration at early neonatal stages, and human studies suggest a low turnover rate in cardiomyocytes (CMs) throughout an individual’s lifetime^{2, 3, 4, 5}. Thus, a prevailing view is that the adult mammalian heart has endogenous regeneration capabilities but cannot initiate or complete this process after injury.

Zebrafish have exceptionally high cardiac regenerative capacity, based on the ability of adult zebrafish CMs to dedifferentiate and proliferate after major injury^6–10. While many studies have explored the cell-intrinsic basis of the competency for CM division upon injury during zebrafish heart regeneration^{7, 8, 11, 12}, other reports have highlighted the essential role of non-myocardial cells, including epicardial, vascular, lymphatic, and neural cells^13–16. The epicardium is a mesothelial tissue that envelops the cardiac outer surface and is rapidly activated to proliferate, migrate, and cover injuries^17–23, and genetic depletion of the epicardium blocks zebrafish heart regeneration²⁴. Conversely, manipulations of the epicardium have been reported to enhance mammalian heart repair^24–28. As the epicardium is a heterogeneous structure, its cellular constituents require further exploration, especially as to how they might support customized roles during cardiac regeneration.

Here, we applied single-cell RNA-sequencing (scRNA-seq) analysis in zebrafish to examine epicardial subpopulations and how they might contribute to myocardial growth and regeneration. Our studies identify a requirement for a large, regeneration-associated cluster of cells expressing genes encoding the hyaluronic acid (HA)-organizing factors hapln1a and hapln1b. We find that these cells envelop proliferative CMs during key cardiogenic events during both heart morphogenesis and regeneration, when they are required for normal HA organization and cardiac muscularization.

METHODS

Data Availability

The authors declare that all data that support the findings of this study are available within the article and its Supplemental Material. The data, analytic methods, and study materials will be available to other researchers for purposes of reproducing the results or replicating the procedure. The scRNA-seq dataset has been submitted to Gene Expression Omnibus (GEO) and the accession number is GSE172511.

Zebrafish and heart injuries

Outbred EK or EK/AB zebrafish 4–10 months old were used for ventricular resection⁶ and cardiomyocyte (CM) ablation⁸. For depletion of hapln1a:mCherry-NTR-expressing cells, zebrafish were used at 5–12 weeks (juvenile) and 4–12 months (adult). Animal density was maintained at ~4 fish/L in all experiments. To deplete hapln1a+ cells, hapln1a:mCherry-NTR animals and control siblings were treated with vehicle or 10 mM metronidazole (Mtz) in a 1.5 L mating tank for 14 hours/day over 2 (juvenile) or 3 (adult) days. Transgenic strains described elsewhere were gata4:EGFP²⁹, Z-CAT and tcf21:nucEGFP⁸, and tcf21:mCherry-NTR²⁴. Newly constructed strains are described in Supplement Methods. All transgenic strains were analyzed as hemizygotes. All animal procedures were performed in accordance with Emory University and Duke University guidelines.

Histological methods

Histological analyses were performed on 10 μm cryosections of paraformaldehyde-fixed hearts. Primary antibodies used here include: anti-Mef2 (rabbit; Santa Cruz Biotechnology), anti-MF20 (mouse; Developmental Studies Hybridoma Bank), anti-Lcp1 (rabbit, GeneTex), and anti-PCNA (mouse; Sigma). Alexa Fluor secondary antibodies (Invitrogen) used here were: 488 (goat anti-rabbit), 594 (goat anti-rabbit and goat anti-mouse), and 633 (goat anti-mouse).

CM proliferation cycling were quantified as previously described²⁴ and MF20 staining of 30 dpa hearts for muscularization was performed as previously described¹¹. Quantification of gata4:EGFP+ expression in adult fish were performed as described⁷. To quantify gata4:EGFP⁺ cell growth, 3 fields of EGFP⁺ cells in each heart were observed and followed for two days. Images were captured using a stereomicroscope. The extent of EGFP⁺ growth was determined in pixels by ImageJ software.

To quantify hapln1b:EGFP and hapln1a:mCherry-NTR colocalization in uninjured and regenerating ventricles, whole-mounted specimens were selected from each group and imaged. Images of ventricles were captured using a 20x objective lens (1,024 × 1,024 pixels). The colocalization between EGFP⁺ and mCherry⁺ signals was determined in pixels using JACoP from ImageJ software.

To calculate HA signals, ventricular sections were stained with biotinylated-HABP³⁰ and fluorescently labeled streptavidin-Alexa-647 conjugate. For adults, images of the ventricle were captured using a 20x objective lens (1,024 × 1,024 pixels), while images of single optical slices of ventricles were acquired using a 63x objective lens (1,024 × 1,024 pixels) in the same position of each ventricle for juveniles. The HA signals were measured in pixels by ImageJ software for signals in the edge of the injury site (163 × 163 μm) of each ventricle. As the number of HA aggregates indicates HA instability, we calculated HA aggregates by the amount of HA dots per unit area of HA signals to determine the HA stability.

To quantify compact muscle thickness, 3 medial, longitudinal sections were selected from each heart and imaged. Single optical slices of the ventricle were acquired using a 63x objective lens (1,024 × 1,024 pixels) in the same position of each ventricle, from sections stained with MF20 antibody. Compact muscle thickness was measured in pixels by ImageJ software.

Single-cell RNA-sequencing

To prepare epicardial cells for single-cell RNA-sequencing analysis, tcf21:nucEGFP and tcf21:nucEGFP:Z-CAT (tcf21:nucEGFP;cmlc2:CreER;bactin2:loxp-mCherry-STOP-loxp-DTA) transgenic fish were raised to adulthood. Hearts were collected from tcf21:nucEGFP;Z-CAT and tcf21:nucEGFP transgenic fish at 7 days post 4-hydroxytamoxifen (4-HT) treatment, and digested with 0.26 U/ml Liberase™ Thermolysin Medium (TM) based on a previously published protocol³¹. Dissociated cells were spun down and live EGFP⁺ cells sorted by flow cytometry. Isolated cells were sent to Emory Integrated Genomics Core (EIGC) center for 10x single-cell RNA-sequencing. Approximately 6,666 cells were loaded per channel for an expected recovery of 4,000 cells. Single-cell RNA-seq libraries were prepared using the Chromium Single Cell 3’ Library & Cell Bead Kit v3.1 (Cat. No. 1000128, 1000127, 120262; 10x Genomics) according to manufacturer’s protocol. Libraries were sequenced with an Illumina NextSeq550 using mid output 150-cycle kits according to manufacturer specifications. The newly generated scRNA-seq data were demultiplexed, aligned, and quantified using Cell Ranger Single-Cell Software. Preliminary filtered data generated from Cell Ranger were used for downstream analysis by Seurat R package according to standard workflow.

STATISTICAL ANALYSIS

All data are presented as mean ± SEM. All statistical analyses were performed using Prism 7 software (GraphPad). The Mann-Whitney Rank Sum test was used for assessing statistical differences between 2 groups. The Kruskal-Wallis test was performed for 3 group comparisons. The Chi-squared test was used for categorical variables. The results with P values <0.05 were considered statistically significant.

RESULTS

Identification of epicardial cell states during heart regeneration

To identify epicardial subpopulations as defined by gene expression states, we performed single-cell RNA-sequencing (scRNA-seq) analysis using dissociated epicardial cells. Cells activating the regulatory sequences of the pan epicardial maker tcf21 were purified from hearts of uninjured tcf21:nEGFP animals⁸ or those regenerating after induced genetic ablation of ~50% of cardiomyocytes (CMs; Fig. 1A). Genetic ablation injuries produce areas of regeneration throughout the ventricle and assist profiling from whole ventricles, whereas partial resection or cryoinjury leaves only a small area of regenerating tissue^{8, 32, 33}. We applied stringent quality filtering and discarded a small number of GFP⁻/tcf21⁻ cells and other non-epicardial cell types like erythroid hematopoietic cells (gata1a)³⁴, myeloid or leukocyte hematopoietic cells (lcp1)³⁵, cardiomyocytes (myl7)³⁶ and endothelial cells (fli1a)³⁷. We obtained high-quality transcriptomes of 3,644 and 4,225 epicardial cells from uninjured and regenerating hearts, respectively (Figure S1). Unsupervised clustering identified 7 epicardial cell clusters, with each cluster possessing distinct gene expression patterns during regeneration (Fig. 1B and 1C, and Figure S2 and S3). Cluster 1 had the largest cell population, expressing an enrichment of genes associated with muscle development, heart regeneration and extracellular matrix (ECM) organization, for instance expressing high levels of nrg1, fstl1a, inhbaa, col8a1b, and has1^{25, 38–41}. Clusters 2, 3, and 6 are three groups of cells that express genes implicated in regulating immune responses, such as ccl25b, c4b, cfd, and tnfsf12^42–45. These clusters had a high degree of overlap, and it is unclear if they represent distinct cellular identities or different stages of epicardial cells. However, we observed a notable difference in cluster 3 with high levels of crabp1a and Frzb^{46, 47}, implicated in negative regulation of the retinoic acid and Wnt signaling pathways, respectively. Cluster 4 cells displayed notable expression of cell cycle genes such as mki67, fen1, mcm2, PCNA, and rpa2^48–52, indicative of a proliferative state. Cluster 5 expressed genes like dhx58, mxb, rsad2, and saa^53–56, indicative of a defense response. cxcl12b expression in cluster 7 suggested responses to chemokines in these cells⁵⁷. Among the identified clusters, clusters 1, 4, and 5 increased in representation during regeneration, with the others decreasing (Fig. 1D). These increased clusters acquire shared functional characteristics during regeneration; for example, they are each associated with the gene ontology (GO) terms tissue regeneration and ECM organization, and clusters 1 and 5 are linked to muscle structure development and angiogenesis (Fig. 1E).

Figure 1. Single-cell RNA-sequencing reveals distinct epicardial cell clusters in zebrafish during heart regeneration.

(A) Schematic representation of the scRNA-seq workflow. Z-CAT: zebrafish cardiomyocyte ablation transgenes (cmlc2:CreER;bactin2:loxp-mCherry-STOP-loxp-DTA). One hundred tcf21:nEGFP and 70 tcf21:nEGFP;Z-CAT were used.

(B) Uniform manifold approximation and projection (UMAP) clustering of single-cell samples.

(C) Heatmap of the top 10 markers for the epicardial cells from tcf21:nucEGFP; Z-CAT animals at 7 days post 4-HT treatment.

(D) Cell clusters as a percentage of epicardial cells in uninjured and regenerating hearts.

(E) Identification of cell clusters based on marker gene expression.

(F) Feature plot of hapln1a expression in epicardial clusters of uninjured and regenerating hearts. There are 787 hapln1a⁺ cells from uninjured hearts and 1702 hapln1a⁺ cells from regenerating hearts.

To better define relationships between the epicardial clusters, we performed RNA velocity analysis⁵⁸, which infers the direction and rate of cellular state changes based on the relative abundance of spliced and unspliced transcripts. RNA velocity analysis indicated a dynamic movement among all clusters of epicardial cells (Figure S4). We observed a predicted transition from cluster 2 to cluster 1 in both uninjured and regenerating hearts, suggesting that cluster 1 cells originate from cluster 2. Interestingly, there is a predicted transition from cluster 1 to cluster 4 in regenerating hearts, suggesting that proliferating epicardial cells originate from cluster 1 cells.

hapln1a labels epicardial cells that associate with regenerating cardiomyocytes

As cluster 1 was the most heavily represented of the 3 enhanced cell states and enriched with genes related to injury-induced heart regeneration (Fig. 1D and 1E), we focused attention here. Previous studies revealed extracellular matrix (ECM) regulation as a key function of epicardial cells during regeneration, through dynamic production of critical components like fibronectin and connective tissue growth factor^{59, 60}. Among the enriched genes with ECM function were fstl1a, col8a1b, has1, and hapln1a^{39, 40} (Figure S2 and S3). In particular, we noted 33% of tcf21+ cells express hapln1a in uninjured hearts, with an increase to 49.3% in regenerating hearts (Fig. 1F). hapln1a encodes an ECM component that belongs to the hyaluronic acid superfamily and stabilizes hyaluronan-proteoglycan complexes^{61, 62}. Previous reports indicated that hapln1a regulates heart development in zebrafish by instructing areas of ECM expansion, and that HAPLN1 is required for cardiac wall morphogenesis in mice^63,64. For additional context, we re-analyzed published scRNA-seq datasets from samples of human and mouse hearts^65–68. There exists a relative abundance of HAPLN1-expressing cells in embryonic human and mouse hearts (Fig. 2A and 2B), with expression observed in fibroblasts and endothelial cells in human and mouse hearts and in mouse CMs (Fig. 2C and 2D), by comparison with adult hearts from these species which have few cells with detectable HAPLN1 expression (Fig. 2E and 2F). In damaged human and mouse hearts, there is a small increase in the number of cells expressing HAPLN1, but this representation is much lower than in embryonic human and mouse hearts (Fig. 2G–2J). These findings indicate that the cellular representation of HAPLN1 (hapln1) expression differs after cardiac injury between species with disparate regenerative capacities.

Figure 2. HAPLN1 expression in mammalian fetal, adult and damaged hearts, assessed by re-analyzing published scRNA-seq datasets.

(A, B) Box plots comparing cardiac cells with HAPLN1 expression in fetal and adult human hearts (A), and fetal and adult mouse hearts (B).

(C, D) UMAP plot indicating the clusters with HAPLN⁺ cells in fetal human (C) and mouse (D) hearts based on the expression of HAPLN1. FB, fibroblast; EC, endothelial cells; CM, cardiomyocytes.

(E, F) UMAP plot indicating the clusters with HAPLN1⁺ cells in adult human (E) and mouse (F) hearts based on the expression of HAPLN1.

(G) Box plots comparing cardiac cells with HAPLN1 expression in adult human healthy and failure hearts.

(H) Quantification of HAPLN1⁺ cells in fetal, adult, and failure human hearts in (G).

(I) Box plots comparing cardiac cells with mice HAPLN1 expression in sham control, and myocardial-infarction (MI)- injured mouse hearts by 3 and 7 days post damage.

(J) Quantification of HAPLN1⁺ cells in fetal, adult, and damaged mouse hearts in (I).

To visualize hapln1a expression, we conducted in situ hybridization in adult zebrafish hearts, observing localized signals in the ventricular wall consistent with epicardial tissue (Fig. 3A). To better assess areas of expression, we also generated hapln1a:EGFP and hapln1a:mCherry-NTR BAC transgenic animals containing sequences 194 kb upstream of the hapln1a translation initiation codon and 13 kb downstream of the stop codon. We observed prominent fluorescence signals in the ventricular wall and bulbus arteriosus, with no detectable cardiac expression elsewhere (Fig. 3B, and Figure S5), mimicking endogenous hapln1a mRNA expression. hapln1a:EGFP⁺ cells (occasionally referred to for brevity as hapln1a⁺ or hapln1a-expressing cells) increased in number during regeneration after genetic myocardial ablation, consistent with our observations from scRNA-seq analysis of tcf21⁺ cells (Figure S5). As predicted by scRNA-seq, hapln1a-directed expression occupied a subset of cells marked by transgenic tcf21:mCherry fluorescence in uninjured hearts⁶⁹ (Fig. 3C), indicating that hapln1a⁺ cells are epicardial or epicardial-derived cells. hapln1a-directed fluorescence signals did not colocalize with reporter transgenes marking endothelial or myocardial cells^{7, 70} (Fig. 3D and 3E). hapln1a:EGFP was not present in the extreme surface layer of epicardium positive for Raldh2 in uninjured hearts⁷, whereas cells expressing both of these markers emerged upon injury (Figure S5).

Figure 3. hapln1a labels an epicardial subpopulation that is associated with proliferating cardiomyocytes.

(A) In situ of hapln1a mRNA in a section of an uninjured heart. White arrows indicate in situ signals. n = 8. Scale bar, 50 μm.

(B) hapln1a⁺ cells visualized in whole cardiac sections from adult hapln1a:EGFP animals. Signals are present in epicardium and smooth muscle of the outflow tract. n = 9. Scale bar, 50 μm.

(C) Confocal slices indicating hapln1a⁺ cells in uninjured hapln1a:EGFP;tcf21:mCherry ventricles. White arrows represent hapln1a⁺/tcf21⁺ cells, and arrowheads represent hapln1a⁻/tcf21⁺ cells. n = 8. Scale bar, 50 μm.

(D, E) Confocal slices indicating hapln1a⁺ cells in adult hapln1a:EGFP;cmlc2:mCherry (D) and hapln1a:EGFP;fli1a:mCherry (E) ventricles. n = 6–10 in each group. Scale bars, 50 μm.

(F) Visualization of hapln1a⁺ cells in hapln1a:EGFP;cmlc2:mCherry ventricles without injury and during regeneration at 7, 14 and 30 dpa. Dashed line indicates amputation plane. n = 9 to 12 animals for each timepoint. Scale bars, 50 μm.

(G) Section images of hapln1a:EGFP;tcf21:mCherry ventricles during regeneration at 7 dpa, assessed for hapln1a⁺ cells and tcf21⁺ cells in the injury site. These images display a colocalization of EGFP⁺ signals with a portion of mCherry⁺ signals in the injury site, with EGFP⁺/mCherry⁺ cells comprising 21.23% of the total mCherry⁺ cells in the injury site. Dashed line indicates amputation plane. n = 10. Scale bar, 50 μm.

(H) Whole-mount view of hapln1a⁺ cells and gata4:EGFP⁺ CMs in the injured area of hapln1a:mCherry-NTR;gata4:EGFP fish hearts at 7 dpa. Dashed line indicates amputation plane. n = 12. Scale bar, 50 μm.

(I, J) Section images of hapln1a:mCherry-NTR;gata4:EGFP ventricles at 7 dpa, assessed for hapln1a⁺ cells and gata4:EGFP⁺ CMs in the injury site. The box area in (I) is shown in higher magnification in (J). Arrowheads represent hapln1a⁺ cells lining and surrounding gata4:EGFP⁺ CMs. Dashed line indicates amputation plane. n = 15. Scale bar, 50 μm.

To assess the distribution of hapln1a⁺ cells during regeneration, we examined hapln1a:EGFP hearts at various timepoints after partial resection of the ventricular apex. hapln1a-expressing cells accumulated adjacent to injuries by 7 days post resection (dpa), and they remained in the wounds and regenerating muscle at 14 and 30 dpa (Fig. 3F). We noticed that 21.23% of tcf21⁺ cells express hapln1a in the injury site by 7 dpa, suggesting hapln1a⁺ cells represent a minority of tcf21⁺ cells in the regenerating area after resection surgery (Fig. 3G). Activation of regulatory sequences of the cardiogenic transcription factor gata4 expression is induced in regenerating CMs and considered a marker of dedifferentiation⁷. We observed hapln1a⁺ cells surrounding dedifferentiated and proliferating gata4:EGFP⁺ CMs during regeneration (Fig. 3H–3J, and Figure S5), suggesting that they may function in muscle regeneration.

Genetic depletion of hapln1⁺ cells inhibits cardiac muscle regeneration

To test the requirement for hapln1a-expressing cells during heart regeneration, we employed the bacterial nitroreductase (NTR) system for inducible tissue ablation⁷¹. We generated a new transgenic line and incubated otherwise uninjured hapln1a:mCherry-NTR adults with the pro-drug metronidazole (Mtz). This treatment depleted ~97% of ventricular hapln1a+ cells, without apparent effects on animal survival or CM proliferation (Figure S6). We then incubated adult hapln1a:mCherry-NTR fish and control siblings for 3 days with Mtz at 2 days after resection of the ventricular apex, and collected hearts at various timepoints post-injury.

At 7 dpa, a timepoint at which many CMs in the injury site proliferate, hearts with ablated hapln1a⁺ cells displayed a 39% reduction in the percentage of cycling CMs compared to control siblings (Fig. 4A and B). To assess effects on regenerating cardiomyocytes, we imaged and quantified gata4:EGFP signals in 14 dpa injury sites (Fig. 4C and D). Mtz-treated hapln1a:mCherry-NTR animals displayed a 57% reduction in gata4:EGFP tissue on average when compared to controls. By 30 dpa, most hapln1a:mCherry-NTR fish exhibited large gaps in the myocardial wall indicative of defective regeneration (Fig. 4E and F). These results indicate that cardiac muscle regeneration requires the presence of the hapln1a-expressing epicardial cell population.

Figure 4. hapln1a⁺ cells are required for myocardial regeneration.

(A) CM proliferation in ventricular sections from Mtz-treated wild-type siblings and hapln1a:mCherry-NTR animals at 7 dpa. Arrowheads represent proliferating Mef2⁺PCNA⁺ CMs. Brackets, injury site used for quantification. Dashed line indicates amputation plane. Scale bar, 50 μm.

(B) Quantification of CM proliferation in injury sites in experiments from (A). A total of 23 wild-type siblings and 20 hapln1a:mCherry-NTR fish hearts were assessed from two experiments. Mann-Whitney Rank Sum test.

(C) Section images of ventricles of vehicle- or Mtz-treated hapln1a:mCherry-NTR;gata4:EGFP at 14 dpa, assessed for gata4:EGFP⁺ CMs in the injury site. Brackets, injury site used for quantification. Dashed line indicates amputation plane. Scale bars, 50 μm.

(D) Quantification of EGFP⁺ pixels in cardiac wounds from experiments in (C). A total of 21 vehicle-treated and 25 Mtz-treated hapln1a:mCherry-NTR fish hearts were assessed from two experiments. Mann-Whitney Rank Sum test.

(E, F) Section images (E) of Mtz-treated wild-type siblings or hapln1a:mCherry-NTR ventricles at 30 dpa assessed for muscle recovery, and quantification of regeneration indices (F). A total of 20 hapln1a⁺ cell-depleted animals and 20 control siblings was examined from two experiments. Myocardial regeneration is categorized as: 1, complete regeneration of a new myocardial wall; 2, partial regeneration; and 3, a strong block in regeneration. Five of 20 control siblings and 18 of 20 hapln1a⁺ cell-depleted ventricles showed prominent myocardial gaps. Chi-squared test. Dashed line indicates amputation plane. Scale bar, 50 μm.

hapln1b, but not hapln1a, is required for heart regeneration

Analysis of our scRNA-seq dataset indicated that hapln1a-expressing epicardial clusters also express hapln1b (Fig. 5A, and Figure S7). DNA sequence homology between hapln1a and hapln1b coding sequences is 69%, and amino acids are 58% identical (Figure S7). Protein homology comparisons between Hapln1a and b revealed the same functional domains. We generated a new hapln1b:EGFP BAC reporter strain to visualize hapln1b-expressing cells, observing that hapln1b signals localized to the epicardium and showed 99% colocalization with hapln1a⁺ cells in uninjured ventricles, and 89% colocalization at sites of regeneration (Fig. 5B–5D, and Figure S7). Thus, we postulated that key Hapln1 functions in hapln1a-expressing cells could be carried out by Hapln1a, Hapln1b, or both factors.

Figure 5. Hapln1b is required for injury-induced cardiomyocyte proliferation and heart regeneration.

(A) Feature plot of hapln1b expression in epicardial clusters from uninjured heart samples.

(B) Visualization of hapln1a⁺ and hapln1b⁺ cells in ventricular sections from uninjured adult hapln1a:EGFP;hapn1b:mCherry animals. Arrowheads represent hapln1a⁺/hapln1b⁺ cells. n = 11. Scale bar, 50 μm.

(C) Visualization of hapln1a⁺ and hapln1b⁺ cells in injury sites in ventricular sections from adult hapln1a:mCherry-NTR;hapn1b:EGFP animals at 14 dpa. White dashed lines indicate the amputation plane. Arrowheads represent hapln1a+/hapln1b+ cells. n = 7. Scale bar, 50 μm.

(D) Quantification of hapln1a⁺ cells as a percentage of total hapln1b⁺ cells in experiments from (B) and (C), based on areas of fluorescence. Mann-Whitney Rank Sum test. Scale bar, 50 μm.

(E, F) Section images (E) and quantification (F) of hapln1a mutant and wild-type sibling ventricles, assessed for muscle recovery at 30 dpa. A total of 23 hapln1a mutant animals and 18 wild-type siblings were assessed from two experiments. One of 18 wild-type siblings and 5 of 23 hapln1a mutants showed partial myocardial regeneration. Chi-squared test. Dashed line indicates amputation plane. Scale bar, 50 μm.

(G, H) Section images (G) and quantified regeneration indices (H) of injured hapln1b mutant and wild-type sibling ventricles, assessed for muscle recovery at 30 dpa. A total of 22 hapln1b mutant animals and 21 wild-type siblings were assessed from two experiments. Zero of 21 wild-type siblings and 18 of 22 hapln1b mutants displayed prominent myocardial gaps. Chi-squared test. Dashed line indicates amputation plane. Scale bar, 50 μm.

(I) Cardiomyocyte (CM) proliferation in hapln1b mutant and wild-type sibling hearts at 7 dpa, stained with antibodies against Mef2 for CM nuclei and PCNA. Arrowheads represent proliferating CMs. Brackets, injury site used for quantification. Scale bar, 50 μm.

(J) Quantification of CM proliferation in injury sites in experiments from (I). A total of 16 hapln1b mutant animals and 21 wild-type siblings were assessed from two experiments. Mann-Whitney Rank Sum test.

To determine functions of Hapln1a and Hapln1b during regeneration, we generated large deletions to remove their coding sequences with CRISPR-Cas9 gene-editing. As expected, these mutations disrupted their respective gene products, as determined by Real-time PCR (qPCR) and Western blotting (Figure S7). Similar to alleles generated in a previous study, homozygous hapln1a mutant animals showed a transient and minor cardiac dysmorphology at the embryonic stage that did not affect viability⁶³ (Figure S7). We examined homozygous hapln1a and hapln1b mutant animals at juvenile and adult stages, and observed similar ventricular morphology of hapln1a mutant animals to their wild-type siblings with no obvious difference detected at juvenile and adult stages. Interestingly, hapln1b mutant animals show measurably thinner ventricular walls compared with their wild-type siblings at the juvenile stage. Adult hapln1b mutant animals and their siblings have similar ventricular wall morphology (Figure S7).

While hapln1a mutant animals trended toward a lower regeneration index score at 30 dpa, most were able to renew a contiguous wall of heart muscle (Fig. 5E and 5F, and Figure S7). These results indicate that hapln1a is largely dispensable for heart regeneration. However, in contrast with hapln1a mutants, we observed consistently disrupted restoration of the myocardial wall in hapln1b mutants at 30 dpa. This phenotype was still evident by 60 dpa (Fig. 5G, 5H, and Figure S7). We then assessed whether hapln1b mutations are essential for injury-induced proliferation of CMs. There was a 35% decrease in the 7 dpa CM proliferation index in hapln1b mutations compared with their wild-type clutchmates (Fig. 5I and 5J). Together, our results reveal that hapln1a and b expression mark a subset of epicardial cells that are preferentially affiliated with regenerating CMs, and that Hapln1b is essential for normal heart regeneration.

Depletion of hapln1a-expressing cells or inactivation of hapln1b result in defective HA deposition during cardiac muscle regeneration

To elucidate the mechanism by which hapln1-expressing epicardial cells and Hapln1b impact CM proliferation and regeneration, we assessed the levels and distribution of its substrate, hyaluronic acid (HA). HA has been explored as a biomaterial to assist repair after myocardial infarction^72–74. Moreover, Tsang and colleagues reported its requirement for heart regeneration^{41, 75}, finding that inhibition of Has enzymes with UDP glucuronic acid blocked epicardial epithelial-mesenchymal transition (EMT) and migration, reducing CM cycling and muscle regeneration. We observed increased HA presence in the injury site using a fluorescent dye as previously reported³⁰; additionally, has1 is present in cluster 1 in regenerating hearts (Fig. 6A). Notably, HA signals were closely associated with hapln1a:EGFP⁺ cells within the injury area (Fig. 6B), as well as near regions of proliferating CMs and gata4:EGFP⁺ cells, consistent with our Hapln1 analyses and with a role in muscle regeneration (Fig. 6C).

Fig. 6. Depletion of hapln1a⁺ cells or inactivation of hapln1b results in defective hyaluronic acid deposition in regenerating hearts.

(A) Violin-plot of has1 expression in tcf21⁺ epicardial cell clusters in uninjured and regenerating hearts.

(B) Section images of hapln1a:EGFP ventricles without injury and at 7 dpa, assessed for hapln1a⁺ cells and HA signals in the injury site. The boxed area was visualized with high magnification. Arrows represent hapln1a⁺ cells lining with HA signals. n = 8. Dashed line indicates amputation plane. Scale bar, 50 μm.

(C) Section images of gata4:EGFP ventricles at 7 dpa, assessed for gata4:EGFP⁺ cells and HA signals in the injury site. The boxed area was visualized with high magnification. Arrowheads represent gata4:EGFP⁺ cells lining and surrounding with HA signals. n = 7. Scale bar, 50 μm.

(D) Section images of ventricles of hapln1a:mCherry-NTR or control siblings with Mtz treatment at 7 dpa, assessed for HA signals in the injury site. Scale bar, 50 μm.

(E, F) Quantification of HA⁺ fluorescence pixels (E) and HA signals per area (F) in injury edges from experiments in (D). A total of 15 hapln1a:mCherry-NTR and 15 control sibling hearts were assessed from two experiments. Mann-Whitney Rank Sum test.

(G) Section images of ventricles of hapln1a −/− and wild-type siblings at 7dpa, assessed for HA signals in the injury site. Scale bar, 50 μm.

(H, I) Quantification of HA⁺ fluorescence pixels (H) and HA signal area per HA aggregate (I) in the injury edges from experiments in (G). Eleven hapln1a −/− and 13 wild-type siblings were analyzed. Mann-Whitney Rank Sum test.

(J) Section images of ventricles of hapln1b −/− and wild-type siblings at 7dpa, assessed for HA signals in the injury site. Scale bar, 50 μm.

(K, L) Quantification of HA⁺ fluorescence pixels (K) and HA signal area per HA aggregate (L) in the injury edges from experiments in (J). Total twelve hapln1b −/− and 16 wild-type siblings were analyzed from two experiments. Mann-Whitney Rank Sum test.

To test requirements for hapln1a⁺ epicardial cells in regulation of cardiac HA, we examined HA in hearts of hapln1a:mCherry-NTR animals after Mtz treatment and ventricular resection (Fig. 6D). Amounts of HA were assessed by quantifying fluorescence intensity, and the size of HA aggregates was estimated by quantifying the area of each focus of HA. We found that 7 dpa injury sites of hapln1a⁺ cell depleted animals displayed ~38% less HA (Fig. 6E), and ~47% smaller puncta, compared with wild-type siblings (Fig. 6F). To test whether one or both of the Hapln1 enzymes are required for HA localization, we examined HA distribution at 7 dpa in each hapln1 mutant line. HA features in regenerating hapln1a mutant ventricles were normal (Fig. 6G–6I). By contrast, hapln1b mutants displayed a similar quantified amount of HA in 7 dpa wounds as wild-types (Fig. 6J and 6K), however with 51% smaller aggregates (Fig. 6L). The different effects of the hapln1a and hapln1b mutants on HA deposition are likely to contribute to why only the hapln1b mutants display defects in heart regeneration. Our results indicate a mechanism for Hapln1 function during regeneration, in which HA, deposited and organized by epicardial cells, is important for CM proliferation, dedifferentiation, and regeneration during zebrafish heart regeneration.

Regionally localized hapln1-expressing cells support the morphogenesis of the ventricular wall in juvenile animals

Cardiac chambers develop in zebrafish by 24 hours post-fertilization (hpf), and their maturation continues throughout both juvenile and early adult life stages. In juvenile zebrafish ventricles, a small number of gata4:EGFP⁺ CMs initially emerge on the chamber surface at ~5 weeks of age, where they proliferate to form the initial clones of compact muscle. These initial clones gradually expand and converge with others to eventually encapsulate the ventricle and create a contiguous wall of compact muscle that is characteristic of adult hearts^{76, 77}.

To examine roles for hapln1a⁺ cells in ventricular morphogenesis, we assessed hearts of embryonic and juvenile hapln1a:EGFP animals. We initially observed hapln1a:EGFP⁺ signals in CMs by 2 dpf, with no detectable expression in endothelial or epicardial cells. EGFP⁺ cells were not visible in the embryonic heart at 5 dpf (Figure S8). Interestingly, while tcf21⁺ epicardial cells coat the juvenile ventricle, we found that they do not activate hapln1a regulatory sequences until 5 weeks post-fertilization (wpf). Then, a small cluster of hapln1a⁺ cells expressing tcf21, but not Raldh2 or CM markers, appears at the ventricular base and gradually expands and spreads over the entire ventricle by 12 wpf (Fig. 7A and Figure S8). To examine a potential association with compact wall formation, we analyzed hearts of hapln1a:mCherry-NTR;gata4:EGFP fish from 5 to 12 wpf. Imaging of the ventricular surface revealed enclosure of gata4:EGFP⁺ CMs by hapln1a⁺ cells at all timepoints (Fig. 7B and 7C), an observation confirmed by analyses of tissue sections (Fig. 7D). To identify dynamics of HA during heart morphogenesis, we assessed its localization on the ventricular surface in juvenile animals. We found strong concordance of HA signals with both hapln1a:EGFP⁺ epicardial cells and gata4:EGFP+ CMs (Fig. 7E and 7F), suggesting a similar role for Hapln1 factors in organizing HA as during regeneration.

Figure 7. hapln1a⁺ cells associate with emergent gata4:EGFP⁺ cardiomyocytes and HA in juvenile ventricles.

(A) Whole-mount views of hapln1a⁺ cells on the hapln1a:EGFP ventricular surface (opposite the atrioventricular junction) from 5 to 12 weeks post fertilization. Dashed lines indicate ventricular periphery. n = 10–20 animals for each timepoint. Scale bars, 100 μm.

(B) Whole-mount visualization of hapln1a⁺ cells and gata4:EGFP⁺ CMs on the hapln1a:mCherry;gata4:EGFP ventricular surface (opposite the atrioventricular junction) from 5 to 12 weeks post fertilization. A total of 50 juvenile animals were assessed. Scale bars, 100 μm.

(C) Magenta areas (B1, B2, and B3) in (B) were enlarged, highlighting close juxtaposition of gata4⁺ and hapln1a⁺ cells. Scale bars, 50 μm.

(D) Visualization of hapln1a⁺ cells and gata4:EGFP⁺ CMs in sections of juvenile hapln1a:mCherry;gata4:EGFP ventricles. Ten animals were assessed. Scale bars, 50 μm.

(E) hapln1a⁺ cells and hyaluronan acid (HA) fluorescence signals visualized in whole-mount view of the ventricles of juvenile hapln1a:EGFP animals at the age of 2 months post fertilization (mpf), assessed by HA staining. HA signals are present in hapln1a⁺ cell-distributed area. The boxed area is enlarged. n = 11. Scale bars, 100 μm.

(F) Section views of gata4:EGFP⁺ CMs, hapln1a⁺ cells and HA fluorescence signals from ventricles of juvenile gata4:EGFP;hapln1a:mCherry-NTR animals, assessed by HA fluorescence staining. Arrows indicategata4:EGFP⁺ CMs lining with hapln1a⁺ cells and HA⁺ fluorescence signals. n = 7. Scale bar, 50 μm.

To test requirements of hapln1a⁺ cells in genesis of the compact myocardium, we ablated hapln1a⁺ cells at 6–7 weeks post fertilization. Juvenile hapln1a:mCherry-NTR; gata4:EGFP animals and gata4:EGFP siblings were treated with Mtz, and their hearts were collected for histological analysis at 10 and 30 days post-treatment (dpt). HA amounts in the ventricular wall were reduced by ~53% compared with wild-type siblings at 3 dpt (Fig. 8A and 8B). We found that hapln1a+ cell-depleted hearts displayed 35% lower gata4:EGFP fluorescence density when compared with control siblings at 10 dpt (Fig. 8C and 8D). At 30 dpt, fish with a depleted hapln1a⁺ cell population showed a normal overall cardiac shape but disorganized compact muscle, with gaps in the ventricular wall and a 52% reduction in the thickness of the compact layer (Fig. 8E and 8F, and Figure S9). By 60 dpt, these animals still displayed a 32% reduction in compact layer thickness (Figure S9). We also assessed gata4:EGFP⁺ CMs after depleting hapln1a⁺ cells in juvenile hearts in an ex vivo setting that enables survival of beating hearts in tissue culture for several weeks²⁴. Whereas gata4:EGFP⁺ CMs expanded on the ventricular surface in the presence of the full complement of hapln1a-positive tissue, such growth rarely occurred in hapln1a:mCherry-NTR ventricles, with gata4:EGFP⁺ CM expansion reduced by ~90% compared with controls (Fig. 8G and 8H).

Figure 8. hapln1a⁺ cells are required for morphogenesis of the compact myocardium.

(A) Section images of the ventricles of Mtz-treated juvenile hapln1a:mCherry-NTR or wild-type siblings at 3 days post Mtz-treatment (dpt), assessed for HA antibody staining in the ventricular wall. Scale bar, 50 μm.

(B) Quantification of HA⁺ fluorescence pixels in the ventricular wall from experiments in (A). A total of 8 Mtz-treated hapln1a:mCherry-NTR fish and 8 control sibling hearts were assessed from two experiments. Mann-Whitney Rank Sum test.

(C) Visualization of gata4⁺ cardiomyocytes in whole-mounted juvenile gata4:EGFP and gata4:EGFP;hapln1a:mCherry-NTR hearts after Mtz treatment. Boxed area is enlarged with high magnification in the panels below. Arrowheads represent EGFP signals in hapln1a cell-depleted hearts. AVJ, Atrium-ventricle junction. n = 18–20, in each group. Scale bars, 50 μm.

(D) Quantification of the percentage of EGFP⁺ pixels on the ventricular surface from experiments in (C). Mann-Whitney Rank Sum test.

(E) Section view of the compact muscle layer in hapln1a:mCherry-NTR and wild-type ventricles 30 days after Mtz treatment. Arrowheads represent discontinuities in the cortical wall in hapln1a cell-depleted hearts. 16–18 animals were assessed for each group and repeated once. Dashed line represents the boundary separating ventricular wall and the inner ventricle. Scale bar, 50 μm.

(F) Quantification of ventricular wall thickness in (E). Mann-Whitney Rank Sum test.

(G) (Left) Schematic to visualize cardiomyocyte growth in control and hapln1a⁺ cell-depleted juvenile hearts ex vivo. (Right) Visualization of gata4:EGFP⁺ tissue growth in juvenile gata4:EGFP and gata4:EGFP;hapln1a:mCherry-NTR hearts over 2 days after Mtz treatment. gata4:EGFP⁺ tissue growth is clear in gata4:EGFP ventricles (top; n = 11 out of 11); growth of this extent is rarely observed in hapln1a:NTR;gata4:EGFP ventricles (bottom; n = 2 out of 13). Red boxes represent gata4:EGFP⁺ cardiac muscle expansion. Chi-square test. Scale bar, 300 μm.

(H) Quantification of distances of gata4:EGFP⁺ tissue expansion over two days, from experiments in (G). Mann-Whitney Rank Sum test.

In total, our observations reveal a hapln1a-expressing epicardial cell subpopulation that emerges at the juvenile stage and is necessary for key sites of localized cardiomyogenic activity. By controlling the ECM microenvironment, these cells contribute to compact muscle growth that completes, and later helps regenerate, the layering of the zebrafish ventricle.

DISCUSSION

Here we describe cellular heterogeneity in the epicardium of adult zebrafish, including the appearance of 3 cellular states associated with heart regeneration. Several recent studies using single-cell technologies have revealed the presence of cells or cell states that are preferential to a regeneration context, including skeletal progenitors in axolotl limbs, planarian stem cells, mammalian CMs from different life stages, murine hair follicle stem cells and hepatic cells^78–82. Although the epicardium is established by the fourth day of life in zebrafish, we report here a latent subpopulation of epicardial cells that emerges coincident with compact muscle in 5-week-old juvenile animals. Our results indicate a key function of hapln1a-expressing cells is to oversee principal cardiogenic activities of compact muscle. These include its initial emergence and maintenance as muscle clones on the ventricular surface, and during CM proliferation to replace wall areas lost after a massive injury.

Epicardial cells are known to secrete developmental factors, including many ECM components, during heart regeneration^{25, 83}, ostensibly as a process of building a nurturing scaffold. Hapln1 enzymes are responsible for the processing and organization of HA within the ECM, which has been recently implicated in heart regeneration. HA and its derivatives are biodegradable; biocompatible; promote faster healing of injured tissues; and support cells in relevant processes including survival, proliferation, and differentiation. Injectable HA-based therapies for cardiovascular disease are rapidly gaining attention because of the benefits obtained in preclinical models of myocardial infarction. HA-based hydrogels, especially as a vehicle for stem cells, have been associated with improvements to the process of cardiac repair by stimulating angiogenesis, reducing inflammation, and supporting local and grafted cells in their reparative functions⁷⁵. We find tight associations of HA deposition during regeneration both with hapln1 gene expression itself, as well as with CM proliferation and the gata4:EGFP indicator of dedifferentiation. Moreover, HA deposition is altered in certain manipulations of hapln1 genes or hapln1a-expressing cells. Thus, we infer that HA organization in cardiac injuries is the key mechanism by which the expression of hapln1 genes in epicardial cells influences cardiomyogenesis. While our initial investigations focused on the hapln1a gene, our genetic manipulations indicate that the hapln1a gene product is dispensable for HA organization and indicators of regeneration, whereas hapln1b is required for these events. There are many potential reasons for this, including genetic compensation, differential post-translational modifications, differential in vivo activity, and differential interactors. Understanding the developmental timing of the appearance of hapln1-expressing epicardial cells and how they control foci of cardiogenesis can inspire new ideas for targeted tissue regeneration.

CLINICAL PERSPECTIVE.

What’s New?

1. A subset of epicardial cells emerges in post-embryonic zebrafish, expresses hapln1 paralogs, and surrounds regions of active cardiomyogenesis in contexts of heart morphogenesis and injury-induced regeneration.

2. Induced genetic depletion of hapln1a-expressing cells or genetic inactivation of hapln1b disrupts cardiomyocyte proliferation and inhibits heart regeneration.

3. hapln1-expressing cells or activity is required to produce a matrix rich with organized HA in injury sites.

What Are the Clinical Implications?

1. Targeting HA regulation by manipulation of HAPLN1 in human epicardial cells could potentially modulate cardiac repair after myocardial infarction.

2. Matrix regulation by subsets of epicardial cells might also be key for localized cardiomyocyte proliferation during human heart development, which would be relevant to understanding the basis of congenital cardiac disease.

Non-standard Abbreviations and Acronyms

Hapln1

hyaluronan and proteoglycan link protein 1

HA

hyaluronic acid

Mtz

metronidazole

NTR

nitroreductase

Mef2

myocyte enhancer factor-2