Distinct features of the regenerating heart uncovered through comparative single-cell profiling

Myocardial infarction (MI), commonly known as a heart attack, contributes significantly to human morbidity and mortality (Roth et al., 2020). During an MI, a blockage in a coronary artery cuts off blood flow to the heart muscle causing cell death and the eventual formation of a non-contractile scar. In adult mammals, including humans, this scar is permanent and impairs cardiac function (Lam and Sadek, 2018). In contrast, many types of fish and amphibians possess the remarkable ability to clear cardiac scar tissue and regrow damaged muscle as adults (Cutie and Huang, 2021; Vivien et al., 2016). These observations have sparked intensive studies of regenerating species in hopes of discovering evolutionarily conserved mechanisms to enable regeneration in humans (Godwin, 2014). Such comparative studies are confounded, however, by the large evolutionary divergence between mammals and regenerating vertebrates. This distant evolutionary relationship results in often unclear gene orthology to mammals and manifests in the distinct simplified heart anatomy of fish and amphibians. Thus, despite many advances, the precise molecular, cellular, and genetic factors that enable some animals to regenerate as adults remain incompletely defined.

Zebrafish have emerged as a powerful model for studying adult heart regeneration (González-Rosa et al., 2017; Poss et al., 2002). Experimentally induced ventricular cryoinjury is frequently used in zebrafish to mimic infarction events seen in humans (González-Rosa et al., 2011). Following injury with a liquid nitrogen-cooled probe, a lesion of necrotic tissue forms, triggering an acute inflammatory response that recruits various immune cell types to the wound (Bevan et al., 2020). The activities of these immune cells play a crucial role in the subsequent remodeling and regeneration processes in zebrafish. Macrophages and regulatory T cells, in particular, are indispensable for successful regeneration (Hui et al., 2017; Sanz-Morejón et al., 2019). Additionally, fibroblast cells derived from both the endocardium and epicardium become activated and deposit the collagenous matrix that makes up the scar and stabilizes the injured ventricle (Hu et al., 2022; Sánchez-Iranzo et al., 2018b). In zebrafish, activated fibroblasts also provide critical signals that foster a regenerative niche by promoting neovascularization of the wound area and dedifferentiation and proliferation of existing cardiomyocytes in the wound border zone. This signaling is mediated in part by molecules such as nrg1, secreted from epicardial-derived cells (Gemberling et al., 2015), and retinoic acid, chiefly produced by the endocardial compartment (Kikuchi et al., 2011). The synergistic effects of these signals promote the regrowth of coronary vessels and replacement of scar tissue with healthy myocardium. Although these cellular behaviors are well-established in zebrafish, it remains unclear whether non-regenerating species might share some or all of these characteristics. Therefore, comparative studies are still needed to determine which behaviors of immune cells and components of the signaling environment are truly unique to the regenerating heart.

Recent surveys of cardiac regeneration capabilities among different teleost fish species have yielded surprisingly contrasting results, demonstrating that zebrafish cardiac injury responses are not representative of all teleosts. While ventricular regeneration is found in some fish species (Grivas et al., 2014; Wang et al., 2020), several others, including the grass carp (Long et al., 2022), Mexican cavefish (Stockdale et al., 2018), and Japanese medaka (Ito et al., 2014), exhibit permanent scarring similar to adult mammals. The cellular and molecular behaviors that distinguish these non-regenerating species from zebrafish, however, have only begun to be characterized at the cellular level. Japanese medaka, Oryzias latipes, have recently emerged as a non-regenerating counterpart to the zebrafish model system (Ito et al., 2014). Bulk RNA sequencing analyses uncovered critical differences in the immune pathways activated in the post-injury ventricle in medaka. Notably, stimulation of an immune response through the injection of double-stranded RNA, ostensibly through activation of interferon signaling, promotes revascularization and cardiomyocyte proliferation in medaka in a macrophage-dependent manner (Lai et al., 2017). These findings implicate the immune system as a critical source of phenotypic diversity in cardiac injury responses. The existence of non-regenerating teleosts offers a unique opportunity to compare and contrast the differing regeneration phenotypes across relatively short evolutionary distances to determine which cellular features are unique to regenerating species. Given that heart regeneration was likely an ancestral trait of teleosts (Cutie and Huang, 2021; Kikuchi et al., 2011), understanding the evolutionary path that led to the loss of this ability in some species may offer parallel insights into why mammals lose the ability to regenerate as adults.

In this study, we used comparative single-cell transcriptomics to create detailed time-course maps of the cardiac injury response in zebrafish and the non-regenerating Japanese medaka. These fish share similar body and heart anatomy and shared a common teleost ancestor ∼140 million years ago (Naruse et al., 2004). Our approach revealed key differences in both pre- and post-injury hearts that may be responsible for the contrasting regeneration outcomes. We found differences in immune cell recruitment and behavior, epicardial and endothelial cell signaling, and alterations in the structure and makeup of the myocardium. Overall, our findings shed new light on the factors that coordinate heart regeneration and generate new hypotheses for the mechanisms that underlie the loss of this ability in certain species.

There have been few direct cross-species comparative studies of cardiac injury thus far (Lai et al., 2017; Stockdale et al., 2018). Single-cell RNA sequencing (scRNA-seq) is a powerful approach that enables simultaneous comparison of cellular composition and gene expression across all major cell types in the heart. Therefore, we sought to create a comparative cell atlas of cardiac injury response in zebrafish (Danio rerio) and medaka (O. latipes) ventricles. We used cardiac cryoinjury to induce comparable injury in the ventricles of both zebrafish and medaka (see Materials and Methods). We collected and analyzed uninjured hearts, as well as hearts at 3 and 14 days post-injury (d.p.i.), from both species with several biological replicates at each time point (Fig. 1A). Dissected ventricles were dissociated into a single-cell suspension and prepared for scRNA-seq. Consistent with prior studies (González-Rosa et al., 2011), cryoinjury-induced lesions in the myocardium and formation of a fibrous clot by 3 d.p.i., with deposition of a collagen-rich scar evident by 14 d.p.i. (Fig. 1B).

A total of 47,741 non-erythroid cells that passed quality control and filtering parameters were obtained across all samples. To improve downstream data integration, gene names were standardized to unify names where one-to-one orthology was supported (Fig. S1A). The resulting cells were log normalized and integrated using Seurat v4 (Stuart et al., 2019). Cell integration anchors were calculated using canonical correlation analysis, a technique that benchmarks well for joint-embedding of shared cell types while maintaining species-specific cell types (Song et al., 2023). After integration, dimensionality reduction and clustering yielded 22 cell clusters (Fig. 1C). Analysis of marker gene expression revealed five main cell types including cardiomyocytes (marked by myl7 and actc1a expression), endothelial cells (fli1a/kdrl; Habeck et al., 2002; Lawson and Weinstein, 2002), epicardial cells (tcf21/tbx20; Peralta et al., 2014), vascular mural cells (pdgfrb/rgs5a; Shih et al., 2021), and leukocytes (coro1a/ptprc; Rougeot et al., 2019) (Fig. 1D). Cross-species data integration was effective as both zebrafish and medaka cells were represented in each major cluster (Fig. S1B,C).

To facilitate dissemination of the data to the research community, we created a freely available web-based application for exploration of the single-cell dataset (Ouyang et al., 2021). This tool allows users to examine and compare gene expression and cell type compositions across both species at all time points. Access and instructions for usage can be found at https://github.com/clay-carey/medaka_zebrafish_regeneration.

We first compared injury-induced immune responses to determine how medaka might respond differently than zebrafish to immunostimulatory signals released after cryoinjury. Interferon signaling can modulate gene expression in various cell types and may promote regeneration in tissues such as the intestine and skeletal muscle (Cheng et al., 2008; Leibowitz et al., 2021). Previous studies have shown that injection of poly(I:C), a synthetic double-stranded RNA species and a potent activator of interferon signaling via TLR interaction, can enhance revascularization and cardiomyocyte proliferation in injured medaka hearts (Lai et al., 2017; Marín-Juez et al., 2019). To explore whether endogenous interferon signaling may be activated in zebrafish after cryoinjury, we examined the expression of 12 genes previously identified as zebrafish interferon-stimulated genes (ISGs) across different tissue types in our dataset (Levraud et al., 2019). We observed a surge in ISG expression at 3 d.p.i. in the zebrafish endothelial, epicardial, and leukocyte compartments, but not in the myocardium. In contrast, we did not detect ISG upregulation at 3 d.p.i. or 14 d.p.i. in medaka (Fig. 2A). Using RNA in situ hybridization of injured zebrafish hearts, we observed expression of isg15, a highly conserved ISG across vertebrates, in the wound area co-mingled with kdrl⁺ endothelial cells at 3 d.p.i., but not in 14 d.p.i hearts. Consistent with the single-cell data, we detected little isg15 expression in injured or uninjured medaka hearts (Fig. 2B). Examination of isg15 expression by cell cluster revealed upregulation in most zebrafish endothelial, epicardial, and mural cell clusters at 3 d.p.i., consistent with a systemic interferon response to injury in zebrafish (Fig. S2A).

To investigate the source of post-injury interferon signals in zebrafish, we examined the expression of each interferon isoform annotated in the zebrafish genome. Our analysis revealed that only the ifnphi1 isoform was expressed in a significant number of zebrafish cells, and it was primarily restricted to endothelial cells (Fig. 2C; Fig. S2B). After cell quantification, we found that there was an approximately threefold increase in the number of ifnphi1⁺ endothelial cells following injury, which returned to baseline levels by 14 d.p.i. (Fig. 2D). These results indicate that endothelial cells are a major source of interferon signaling after injury in zebrafish, possibly in response to damage-associated molecules, which may include immunostimulatory nucleic acids released from necrotic cells (Jounai et al., 2013). The medaka genome lacks annotation of interferon genes, making it difficult to assess whether interferons are produced in response to injury. We identified five medaka genes with putative homology to zebrafish interferon genes based on Ensembl orthology assessments (Fig. S2C). Only one of these genes were expressed in any medaka cell in our dataset, a putative ortholog of ifng, which was expressed in some medaka leukocytes, but did not increase in expression after injury (Fig. S2D,E). Therefore, while production of interferon signals cannot be fully ruled out, medaka display a conspicuous lack of ISG response after injury compared to zebrafish.

Immune cells play a critical role in facilitating immediate injury responses and in modulating tissue regeneration and may represent a key point of phenotypic variation across species (Aurora and Olson, 2014; Godwin et al., 2017). In prior studies, analysis of bulk transcription patterns highlighted the differential expression of immune-related genes in zebrafish and medaka hearts following injury (Lai et al., 2017). We further investigated the immune response in these species by examining immune cell populations through the subsetting and re-clustering of all cell clusters identified as leukocytes (Fig. 3A). By using marker gene expression, we identified five distinct cell clusters, corresponding to macrophages (cxcr3.2/lgmn), proliferating macrophages (mki67/pcna), T lymphocytes (lck/zap70), B lymphocytes (cd79a/pax5), and granulocytes (mpx/lyz) (Fig. 3B).

Macrophages are essential for regeneration in zebrafish, but medaka display altered macrophage recruitment (Lai et al., 2017). To further investigate how medaka macrophages respond to injury, we quantified macrophage dynamics in the intact and injured areas of the zebrafish and medaka ventricle with RNA in situ hybridization and imaging of cells expressing macrophage marker mpeg1.1. By imaging healthy and injured ventricles, we found that macrophages are highly enriched in the wound area in zebrafish compared to medaka, localizing near cells expressing isg15 (Fig. 3C). Quantification of mpeg1.1⁺ cells in zebrafish demonstrated that macrophage density in the wound is approximately 5-fold higher than in uninjured myocardium at 3 and 14 d.p.i., but macrophage density in the unaffected tissue remains constant (Fig. 3D). In contrast, the wound area in medaka ventricles had a similar density of mpeg1.1⁺ cells to the intact uninjured ventricle, but the macrophages appeared to be depleted from intact tissue after injury in medaka (Fig. 3E). These results show that medaka have few signs of post-injury macrophage recruitment or proliferation, while zebrafish macrophage populations are highly expanded in the wound area.

We next compared gene expression in zebrafish and medaka macrophages to determine whether they may have different behaviors in response to injury. Strikingly, medaka macrophages highly upregulated tnfa expression at 3 d.p.i. compared to zebrafish, a marker of inflammatory M1-like macrophages (Fig. 3F). Conversely, zebrafish macrophages express higher levels of cd9b after injury, a marker of anti-inflammatory M2-like macrophages (Brosseau et al., 2018) (Fig. 3G). Cell quantification from single-cell data showed a slight increase in tnfa⁺ macrophages in 3 and 14 d.p.i. zebrafish, but uninjured medaka ventricles display a substantially higher population, further increasing after injury (Fig. 3H). Our findings show that although medaka have lower macrophage recruitment, they exhibit a more pro-inflammatory response to injury, highlighting a crucial difference in the immune response to injury between the two species.

Formation of a fibrotic scar following cardiac injury is an evolutionarily conserved trait among vertebrates (Vivien et al., 2016). In zebrafish, the scar is chiefly deposited by activated fibroblasts derived from pre-existing epicardial and endothelial cells, but it remains unclear how this process might differ in species that exhibit permanent scarring (González-Rosa et al., 2012; Hu et al., 2022). We investigated whether medaka might differ in their fibrotic response to injury compared to zebrafish. To identify individual cell types contributing to fibrosis, we first conducted a re-clustering of all cells expressing the endothelial markers kdrl and fli1a, along with perivascular cells expressing mural cell markers rgs5a and pdgfrb. This process led to the identification of 11 cell clusters, categorized using marker gene expression as endocardial endothelium cells (eECs), fibroblast-like endothelial cells (fECs), coronary endothelium cells (cECs), and lymphatic endothelium cells (lECs) (Fig. S4A,B). Cells in the fEC cluster expressed high levels of collagen isoforms col1a2 and col5a1, as well as postna and twist1b, markers of a transition to a mesenchymal activated fibroblast state but maintained expression of endothelial cell markers fli1a/kdrl but not epicardial markers tcf21/tbx18 (Fig. S3A,C).

To investigate the contributions of epicardial-derived cells to the fibrotic response, we re-clustered all cells expressing epicardial-specific markers tcf21 and tbx18, resulting in cell clusters consisting of three epicardial cell types: canonical epicardial cells (cEPs), fibroblast-like epicardial cells (fEPs), and a zebrafish-specific cluster of epicardial cells (zEPs) (Fig. 4C). The fEP cells are characterized by the expression of epithelial to mesenchymal transition marker twist1b, along with fibrotic-response genes periostin-b (postnb) and collagen isoforms such as col12a1b. In contrast, the zEP cells express cell-adhesion proteins, including cldnc and jam2b, distinct from other epicardial cell types. Cells in the cEP instead maintain tcf21/tbx18 expression without expressing these functional markers (Fig. 4D; Fig. S3B).

Having identified activated fibroblast-like cells from both endothelial and epicardial compartments, we next investigated how these cell populations respond to injury in both species. We first calculated the proportion of endothelial cells in the fEC clusters in each sample. Both zebrafish and medaka display a 2-3-fold increase in fEC cell proportion at 3 d.p.i., and a return to baseline levels by 14 d.p.i. (Fig. 4E). We next investigated the factors that define the individual fEC clusters. Among these cells, clusters fEC2 and fEC3 were most strongly marked by expression of cell cycle progression markers (Fig. S3A). Furthermore, cell cycle scoring using Seurat v4 indicated that cells in clusters fEC2 and fEC3 are nearly uniformly in S or G2M phase, respectively (Fig. S3D). These proliferating fibroblast-like cells also increased in proportion after injury in both species and returned to normal levels by 14 d.p.i. (Fig. 4F). Among epicardial cells, we observed a similar increase in the proportion of collagen-producing fEP cells in both species (Fig. 4G). Thus, both zebrafish and medaka employ a strategy of fibroblast activation from both endothelial and epicardial compartments after injury, with a shared proliferative response in endothelium-derived cells that contribute to collagen production and scar formation.

We know relatively little about the makeup and dynamics of cardiac scar tissue between regenerating and non-regenerating species. To investigate whether zebrafish might have a unique scar composition, we compared expression of all detectable collagen isoforms in all epicardial and endothelial cell types in both species. It has been observed previously that zebrafish epicardial-derived fibroblasts upregulate collagen XII isoforms after injury, and that these populations promote regeneration (Hu et al., 2022; Marro et al., 2016). We observe a strong upregulation in expression and proportion of epicardial cells expressing collagens col12a1a and col12a1b in zebrafish, but a substantially more muted upregulation in medaka (Fig. S3E). In addition, while most collagens we examined were upregulated in both species after injury, zebrafish epicardial and endothelial cells upregulated collagen V and VI isoforms not seen in medaka (Fig. S3E). Matrix metalloproteinases (mmps) degrade collagen matrices and have been proposed to play important roles during regeneration and scar clearance (Gamba et al., 2017; Xu et al., 2018). We compared expression levels of all expressed mmp genes and found shared upregulation of several mmps in endothelial cells of both species, but also species-specific upregulation of mmp15/16 in zebrafish and mmp19 in medaka (Fig. S3F). Thus, while aspects of the fibrotic response program are shared between species, zebrafish fibroblasts upregulate a subset of collagens and matrix remodeling factors distinct from medaka. It remains to be determined what role, if any, these collagen and mmp isoforms play during regeneration.

Endothelial and epicardial cells are a major source of signals that promote revascularization and cardiomyocyte proliferation after cardiac injury in zebrafish (Fernandez et al., 2018; Masters and Riley, 2014). To determine which of these signals are unique to the regenerating heart, we examined expression patterns for four well-characterized pro-regenerative signals: neuregulin-1 (nrg1), retinoic acid synthesizing enzyme aldh1a2, ciliary neurotrophic factor (cntf), and the chemokine cxcl12a. Examination of their expression patterns revealed that each of the four genes were most predominantly expressed in epicardial and endothelial cells (Fig. 5A).

The secreted peptide nrg1 influences cellular survival and proliferation through interactions with epidermal growth factor receptors (D'Uva et al., 2015). After cardiac injury, nrg1 has been shown to be upregulated in epicardial-derived cells of the zebrafish heart and is sufficient to promote cardiomyocyte proliferation and cardiac hypertrophy, indicating a critical role in regeneration (Gemberling et al., 2015). We examined nrg1 expression patterns and found that it is mostly restricted to epicardial cells and a small number of endothelial cells (Fig. S4A). We quantified the number of nrg1+ epicardial cells and found a ∼2-fold increase at 3 d.p.i. in zebrafish, returning to base levels by 14 d.p.i. In contrast to zebrafish, we detected few nrg1+ medaka epicardial cells in any condition (Fig. 5B). The lack of nrg1 expression from the medaka epicardium may partially explain a lack of cardiomyocyte proliferation and regeneration.

Signaling by retinoic acid (RA), synthesized in zebrafish by the enzyme aldh1a2, is critical for development and regeneration across many types of organs and tissues (Ghyselinck and Duester, 2019). In zebrafish, injury-induced RA synthesis in the epicardium and endocardium is indispensable for cardiac regeneration (Kikuchi et al., 2011). We observed aldh1a2 expression in both epicardial and endothelial cells (Fig. 5A; Fig. S4B). Among endothelial cells, aldh1a2 was strongly induced among the zebrafish fibroblast-like fEC at 3 d.p.i., and to a lesser extent in medaka fEC. Zebrafish eEC cells also upregulated aldh1a2 after injury, but this response was not observed in medaka (Fig. 5C), consistent with prior reports that failed to detect endocardial RA synthesis response in medaka (Ito et al., 2014). Signaling by cntf has a well-established role in neural regeneration and has recently been shown to enhance ventricular regeneration when exogenously supplied to zebrafish (Bise et al., 2019; Sahenk et al., 1994). We find evidence for endogenous cntf expression predominantly in endothelial cells (Fig. S4C). Similar to aldh1a2 expression patterns, we observe a strong induction of cntf in fEC cells in both zebrafish and medaka, but expression of cntf among eEC cells was only observed in zebrafish (Fig. 5D). Together, these results suggest that the endocardium of medaka, the most abundant cell type in the heart, fails to mount a systemic pro-regenerative response to injury as seen in zebrafish.

The zebrafish ventricle rapidly revascularizes after injury, but medaka have limited re-growth of blood vessels after injury (Ito et al., 2014). This process is critical for regeneration, and evidence suggests that nascent blood vessels provide a scaffold for regrowing cardiomyocytes during regeneration (Marín-Juez et al., 2019). We investigated whether there was evidence of revascularization through analysis of apelin and cxcr4a expression, which are both upregulated in growing coronary vessels (Marín-Juez et al., 2019). Consistent with our understanding of previous literature, both apln and cxcr4a were elevated in zebrafish cEC cells after injury but remained low at all time points in medaka (Fig. 5E). Cxcl12a peptide acts as a ligand for cxcr4a, promoting angiogenesis and regeneration (Liekens et al., 2010). We observed that expression of cxcl12a was mostly restricted to the epicardial compartment (Fig. 5A; Fig. S4D). Following injury, cxcl12a was highly upregulated in fEP and cEP cells in zebrafish by 3 d.p.i., but only apparent by 14 d.p.i. In medaka (Fig. 5F). These results indicate that the zebrafish epicardium rapidly releases pro-angiogenic signals that promote rapid revascularization unique to this regenerating species.

When examining epicardial cells, we identified a small number of zebrafish-specific cells in the zEP cluster. In contrast to the fEP, the zEP cluster is only composed of zebrafish cells and does not change in abundance after injury (Fig. S4E). These cells express epicardial marker tcf21 and appear to have a unique injury-response profile. At 3 d.p.i., we observe an increased level of expression of pro-angiogenic signaling in zEP, including factors angpt1, vegfd, and cxcl8a, each known to play a role in promoting angiogenesis (El-Sammak et al., 2022; Morisada et al., 2006; Song et al., 2007) (Fig. 5G). Thus, while more evidence is needed to determine whether the zEP cells are truly unique to zebrafish, they represent a potential hub of pro-angiogenic signaling that may promote revascularization after injury in zebrafish.

Although medaka and zebrafish hearts are similar in overall morphology, they may not share similar populations of cardiomyocytes. It has been proposed that some cardiomyocyte populations might have outsized contributions to the pool of regenerating cells (González-Rosa et al., 2017). The zebrafish myocardium contains three main types of cardiomyocytes with distinct spatial arrangements: the inner mass of trabecular cardiomyocytes (tCMs), the outer layer of cortical cardiomyocytes (cCMs), and a single-cell-thick layer of primordial cardiomyocytes that sits underneath the cortical layer (Gupta and Poss, 2012). When comparing heart morphology, we noticed a distinctly thin outer cardiomyocyte layer in the medaka ventricle compared to the zebrafish ventricle, including in older fish (Fig. 6A). The cortical myocardium has been proposed to play an important role in regeneration (Kikuchi et al., 2010), yet some types of fish do not have this layer (Santer, 1985). We further compared the composition of the myocardium by reclustering cells expressing cardiomyocyte markers myl7 and actc1a, identifying four cell clusters (Fig. 6B). We identified one cluster of cCMs, which express canonical CM markers but lack expression of transcription factor tbx5a (Sánchez-Iranzo et al., 2018a). The remaining cell clusters expressed transcription factors tbx5a, as well as gata6 and mef2d, and are assigned as tCMs (Fig. 6C). Comparing the cellular compositions of each cardiomyocyte type revealed a skewing toward trabecular cardiomyocytes in medaka, and fewer cells clustering with cortical CMs, consistent with our imaging observations (Fig. 6D).

Primordial cardiomyocytes are of interest as a pro-regenerative cell population due to their expression of genes typically associated with embryonic cardiomyocytes (Pfefferli and Jaźwińska, 2017; Tsedeke et al., 2021). These cardiomyocytes make distinct cell junctions and act as a transitional layer between the cortical and trabecular layer, but do not act as a direct cellular source of regenerating cardiomyocytes (Lafontant et al., 2013). Initially, we did not observe a cell cluster expressing markers for the primordial myocardium (acta2/hey2/actn1) in the integrated single-cell object (Fig. 6B). However, species-specific clustering of uninjured cardiomyocytes revealed a small cluster of zebrafish cells expressing primordial CM makers, including acta2, hspa12b, clcn2a, and smoc1 (Fig. 6E,F). In contrast, clustering of uninjured medaka CMs failed to identify any group of cells expressing analogous cell markers (Fig. 6G,H). We examined the spatial patterns of top primordial CM markers acta2 and hey2 expression in the zebrafish ventricle and found co-localization in a thin sub-cortical cell layer of cardiomyocytes expressing both genes (Fig. 6I). We searched for a similar arrangement of CMs in medaka ventricles by probing for acta2 expression (hey2 is not annotated in the medaka genome). Imaging revealed no evidence of either acta2+ cardiomyocytes or an analogous primordial layer in medaka (Fig. 6J). Together, our transcriptomic and imaging data suggest that adult medaka lack primordial cardiomyocytes and have a diminished or absent cortical myocardium, factors that may influence the ability to regenerate.

Neonatal mammals are capable of heart regeneration but retain permanent scars if injured as adults. To investigate whether the neonatal mammal heart might share similarities with zebrafish, we compared injury responses in our dataset to a recently published single-nucleus RNA-seq dataset that assessed cardiomyocyte responses to injury in regenerating postnatal day 4 (p4) mice and non-regenerating postnatal day 11 (p11) mice (Cui et al., 2020). We used differential gene expression testing to identify all upregulated genes at 3 d.p.i. in cardiomyocyte cells in the p4 and p11 mice as well as in zebrafish and medaka cardiomyocytes. Using orthology assignments from Ensembl, we determined which upregulated genes with one-to-one orthology were shared or unique in each species. Indeed, zebrafish cardiomyocytes shared three times as many injury-upregulated genes with P4 mouse (33 genes) compared to the non-regenerating P11 mouse (11 genes) (Fig. 7A). In contrast, Medaka cardiomyocytes shared more genes in common with the p11 mouse (12 genes) (Fig. 7B). Remarkably, a large proportion of the shared injury-response genes between zebrafish and p4 mouse cardiomyocytes have well-described cardioprotective roles, promoting cardiomyocyte survival and guarding against metabolic stress (Table S1). Examination of gene expression of these cardioprotective genes revealed a conspicuous lack of induction in both medaka and p11 cardiomyocytes (Fig. 7C). These results indicate that regenerating vertebrates deploy an evolutionarily conserved cardioprotective program immediately after myocardial injury.

In this study, we used single-cell RNA sequencing to compare the injury responses of zebrafish and medaka heart ventricles. Our analysis revealed several previously undescribed differences between the two species, including variations in immune cell populations, endothelial and epicardial cell signaling, and the composition of the myocardium itself. In addition, our results corroborated and expanded upon previous observations in medaka, including altered macrophage recruitment and dampened retinoic acid signaling (Ito et al., 2014; Lai et al., 2017). These findings highlight the power of comparative scRNAseq to identify species-specific differences in biological processes and have generated hypotheses for future investigations into the mechanisms of cardiac regeneration in these species.

Our comparative analysis of immune cell populations and gene expression in response to injury revealed major differences between zebrafish and medaka. Previous comparisons showed that the hearts of these two species express a different repertoire of immune genes after injury (Lai et al., 2017). Remarkably, injection of poly(I:C), a potent activator of interferon signaling, can promote cardiac revascularization and regeneration in medaka. Interferon signaling seems to have pleiotropic effects depending on the tissue type, where it is deleterious in some cases but promotes regeneration in others (Cheng et al., 2008; Leibowitz et al., 2021). Our findings indicate a major difference between zebrafish and medaka in response to injury is the presence of endogenous interferon signaling specific to zebrafish. Interferon signals may play a role in modulating immune cell behaviors, perhaps explaining why zebrafish appear to recruit macrophages to the ventricle injury site while medaka do not. Additionally, while pro-regenerative genes such as aldh1a2 or nrg1 are not canonical ISGs, it remains to be determined whether indirect effects of interferon might promote their expression. Indeed, we find that the endocardium and epicardium, which are the main source of pro-regenerative signaling, exhibit the most robust ISG expression. Our study utilized timepoints at 3 and 14 d.p.i., representing an early and well-developed injury response. Although ISG expression was robust at 3 d.p.i. in zebrafish, it is possible that medaka may also mount an interferon response on an accelerated or delayed timescale we did not capture. Further experimental characterizations of the role of endogenous interferon signaling during heart regeneration represents a compelling direction for future inquiry.

Our examination of endothelial and epicardial cell populations uncovered several injury-response behaviors that are shared between zebrafish and medaka. Overall, both species shared a similar pattern of fibroblast activation from both the endothelial and epicardial compartments. Our comparison of the set of collagen and matrix remodeling factors made by each species showed a mostly overlapping pattern of induction after injury. This included collagen XII isoforms in epicardial-derived fibroblasts, which were upregulated in both species after injury, but to a higher degree in zebrafish. These collagen XII-producing fibroblasts are of special interest as they are required for proper regeneration in zebrafish (Hu et al., 2022). We also found evidence of some species-specific matrix-remodeling proteins and collagen isoforms, warranting further investigation into whether the makeup of the scar itself may promote regeneration in zebrafish.

While we observed similar strategies of fibroblast activation from the endothelial and epicardial compartments in both species, medaka failed to produce critical signals required for regeneration from these cells. In particular, medaka fibroblasts expressing epicardial markers had a conspicuous lack of nrg1 expression, which itself is sufficient to activate regenerative programs in the heart (Chen et al., 2016; Gemberling et al., 2015). Interestingly, we observed a post-injury upregulation of some pro-regenerative signals in medaka endothelial-like fibroblasts, including aldh1a2 and cntf. These same signals were highly upregulated in zebrafish endocardial clusters, but notably absent from medaka endocardium. In zebrafish, the endocardium undergoes systemic morphological changes in response to injury, coincident with strong upregulation of aldh1a2 and other signaling factors even in cells distal from the wound area (Kikuchi et al., 2011). Further experiments are needed to define the specific stimuli that trigger this systemic response and to determine why medaka endocardial cells appear to be less responsive to injury.

Finally, we investigated the cellular makeup of the myocardium in zebrafish and medaka. Our results reveal marked differences in the cardiomyocyte populations of these two species, which are discernible even in non-injured samples. Specifically, we found that medaka have a highly diminished or absent cortical myocardium, as evidenced by our single-cell sequencing and imaging analyses. This observation is notable, given that previous studies have indicated that the cortical myocardium is activated after injury in zebrafish and may serve as a source of pro-regenerative cardiomyocytes (Kikuchi et al., 2010). Additionally, we were unable to identify primordial layer cardiomyocytes in medaka using either single-cell gene expression analysis or by imaging of key markers. As the primordial layer acts as a transitional layer between cortical and trabecular myocardium with specialized cell junctions (Lafontant et al., 2013), it may be interesting to test whether the primordial layer is absent in all teleosts lacking cortical myocardium. In zebrafish, we observed that the primordial myocardium expresses acta2 in conjunction with hey2, an embryonic cardiomyocyte marker, indicating a distinct transcriptional state (Ihara et al., 2018). These findings are especially interesting in light of recent studies demonstrating that acta2+ embryonic-like cardiomyocytes are present in neonatal mammalian hearts but are lost as the animal ages, coincident with the decline of regenerative capacity (Cui et al., 2020, 2021). Thus, the presence of smooth muscle actin-expressing cardiomyocytes in uninjured hearts is notably correlated with regenerative capability. However, the mechanisms by which these specialized primordial cardiomyocytes might contribute to regeneration is unclear. Indeed, lineage tracing of primordial cardiomyocytes has demonstrated that these cells are not primary contributors to peri-injury zone regenerated myocardium, and the layer itself is deficient or delayed in re-forming around the injury area (Pfefferli and Jaźwińska, 2017). Further investigation is warranted, however, to explore any cell non-autonomous roles for primordial layer cardiomyocytes in heart regeneration.

Medaka and zebrafish have been used extensively as laboratory model vertebrates and have similar care requirements and body plans. While it was known that medaka are incapable of heart regeneration, it could have been assumed that medaka as fellow teleost fish would only have minor differences in cardiac structure. Our study found a surprising number of distinguishing characteristics of the medaka heart compared to zebrafish. These differences included not only changes in cellular behaviors but also substantial changes in the structure of the myocardium itself. Given the number of notable differences we observed between zebrafish and medaka, a wider phylogenetic survey of cardiac injury responses will be particularly useful to identify features that correlate with heart regeneration or non-regeneration across the teleost phylogeny. These observations highlight how biodiversity within shorter evolutionary distances can enable comparative studies that reveal fundamental insights about the gain or loss of complex traits.

Wild-type Tübingen zebrafish and CAB medaka, aged 6-18 months unless otherwise stated, were used for all experiments. All zebrafish and medaka work was performed at University of Utah's CBRZ zebrafish facility. This study was conducted under the approval of the Office of Institutional Animal Care and Use Committee (IACUC no. 20-07013) of the University of Utah's animal care and use program.

Cryoinjuries were performed on the ventricle apex of both medaka and zebrafish as described previously (González-Rosa et al., 2011). Briefly, 0.02% and 0.04% Tricaine (MS-222) was used to anesthetize zebrafish and medaka, respectively. Fish were mounted on a moist sponge. The ventricle apex was exposed by making a small thoracic incision using forceps and dissecting scissors. A cryoprobe was constructed as described previously using a 0.5 mm diameter copper wire (González-Rosa and Mercader, 2012). After submersion in liquid nitrogen for at least 2 min, the probe was placed in contact with the ventricle apex for exactly 23 s. After injury, fish were placed back into freshwater tanks to recover, then transferred back into the fish facility for monitoring.

Hearts were fixed in 4% paraformaldehyde overnight, then sucrose treated and embedded in Optimal Cutting Temperature (O.C.T.) embedding medium. 6 µm sections of the heart were cut using a Leica Cryostat. Bouin's solution was used to fix heart tissue before staining with PT/PM (phosphotungstic acid/phosphomolybdic acid) and AFOG (Methyl Blue, Orange G, and Acid Fuchsin) stain. After staining, heart tissue was washed with 0.5% acetic acid, 100% ethanol, and xylenes. Fibrinogen and collagen deposition from heart sections with representative injuries for both medaka and zebrafish were chosen to demonstrate comparative injury phenotypes at 3 and 14 d.p.i. Measurements were made using Fiji, ImageJ, software.

To prepare cells for scRNA-seq, adult fish were euthanized by immersion in ice-cold water. After dissection and removal of hearts, the remaining atrium and bulbus arteriosus tissue were cut away with scissors to isolate the ventricle. Ventricles were placed in PBS solution and gently squeezed with forceps to remove residual blood prior to digestion. Whole ventricles were placed in 200 µl dissociation solution containing 1 mg/ml liberase DH (Millipore Sigma, cat# 5401089001) in 1X HBSS. Ventricles were digested at 37°C for 30 min on a benchtop shaker at 250RPM with additional pipetting every 5 min to break apart the tissue. After complete digestion, the reaction was quenched with ice-cold PBS/50% fetal bovine serum and passed through a 40 µM strainer. The cells were then centrifuged at 250 RCF for 10 min at 4°C and resuspended in ice-cold PBS with 0.04% bovine serum albumin. Cells were then assessed for viability and dissociation quality using Acridine Orange and Propidium Iodide staining with a Nexcelom Cellometer automated cell imager. Each sample viability was assessed to be >85%. Cells were loaded onto a 10X Genomics Chromium controller and processed according to the manufacturer's specifications. cDNA libraries were prepared for sequencing on the Illumina Novaseq platform at a depth of approximately 400 M reads/sample.

For cross species comparisons, gene orthology was established using the Ensembl database (Harrison et al., 2024). Unless otherwise stated, genes compared across species sharing the same name were established as one-to-one orthologs using the Ensembl comparative genomics resources (Herrero et al., 2016). Genes with cardioprotective functions were identified from literature searches (Chen et al., 2022; Brundel et al., 2008; Denzel et al., 2010; Lizano et al., 2017; Morikawa et al., 2015; Nader 2019; Poch et al., 2022; Ponnalagu et al., 2022, Spelat et al., 2022).

Single-cell RNA seq reads were processed with Cellranger from 10X Genomics with the default settings, using GRCz11 v4.3.2 reference (Lawson et al., 2020) for zebrafish read alignments and ASM223467v1 for Medaka. Gene expression matrices were imported and processed in Seurat v4 for subsequent analysis. A preliminary filtering step was used to remove cells with less than 200 unique RNA features or cells with more than 40% mitochondrial RNA. To cluster cells across species and correct for batch effects, we used the data integration function in Seurat v4, which uses gene anchors with shared names to apply correction vectors to correct for batch and species effects and can allow for cross-species integration (Stuart et al., 2019). To maximize the number of usable anchors we renamed 596 medaka genes that were confidently assigned as one-to-one orthologs as determined by Ensembl orthology assessment. An initial round of clustering was performed on log normalized gene expression, and erythroid cells (hbba1/hbaa1 positive) and platelets (thbs1b/itga2b positive) were filtered out. An additional filtering criterion was then applied to remove non-myocyte cells with greater than 15% mitochondrial RNA. The filtered cells were then re-clustered and annotated by marker gene expression.

Prior to performing RNAscope (Advanced Cell Diagnostics, Hayward, CA, USA), fish hearts were fixed overnight in 4% PFA in 1x PBS, sucrose treated, embedded in OCT, and cryosectioned into 12 µm coronal sections. RNAscope^® Multiplex Fluorescent Detection Kit v2 protocol was completed according to the manufacturer's protocol for standard fixed-frozen tissue samples. See key resources table for probe details.

Brightfield images of AFOG stains were acquired with the Zeiss Axio Scan Z1 instrument. Fluorescence images were captured on a Zeiss 880 Airy Scan confocal microscope. Tiling and stitching were completed by ZEN Black software. Images displayed are maximum intensity projections. Adjustments of contrast and brightness were made using ImageJ. At each step of image collection, similar settings for both species were used. When images were used for quantification, all images were collected using the same laser settings (gain, laser power, ext.).

Images of RNAscope sections that marked cardiomyocytes (myl7), macrophages (mpeg1.1), and nuclei (DAPI) were prepared and processed with ImageJ. To distinguish between the intact ventricle and wound area, ImageJ thresholding of the myl7 channel was completed to make two regions of interest (ROI) – a region that marked just the intact ventricle and a region that marked the intact ventricle and the wounded area. These ImageJ ROIs were used to calculate the areas of each region. Then the ROIs were applied to the mpeg1.1 and DAPI channels to mask either the wound area or the intact ventricle. Masked images were uploaded to CellProfiler software (version 4.2.1) using a modified CellProfiler ‘Speckle Counting’ pipeline (Stirling et al., 2021). Nucleated cells were identified using the DAPI channel and the IdentifyPrimaryObjects CellProfiler module. mpeg1.1 speckles were identified using the mpeg1.1 channel and the IdentifyPrimaryObjects CellProfiler module. Objects were related using the RelateObjects CellProfiler module. Cells were identified as macrophages when a mpeg1.1 speckle fell within a mask created by the DAPI channel.

We acknowledge Cell Imaging Core at the University of Utah for the use of equipment (Zeiss 880 Airy Scan, Axio Scan.Z1, and Leica Cryostat) and thank Xiang Wang for assisting our imaging work. We thank the University of Utah high throughput genomics core and Opal Allen for assistance with single-cell sequencing. We thank Mike Shapiro and Chelsea Herdman for their helpful comments and review of the manuscript. We describe the experimental support provided by our core facilities here. These facilities did not provide financial support for this project.