Coronary revascularization is regulated by epicardial and endocardial cues and forms a scaffold for cardiomyocyte repopulation

SUMMARY

Defective coronary network function and insufficient blood supply are both cause and consequence of myocardial infarction. Efficient revascularization after infarction is essential to support tissue repair and function. Zebrafish hearts exhibit a remarkable ability to regenerate, and coronary revascularization initiates within hours of injury, but how this process is regulated remains unknown. Here we show that revascularization requires a coordinated multi-tissue response culminating with the formation of a complex vascular network available as a scaffold for cardiomyocyte repopulation. During a process we term “coronary-endocardial anchoring”, new coronaries respond by sprouting 1) superficially within the regenerating epicardium, and 2) intra-ventricularly towards the activated endocardium. Mechanistically, superficial revascularization is guided by epicardial Cxcl12-Cxcr4 signaling, and intra-ventricular sprouting by endocardial Vegfa signaling. Our findings indicate that the injury-activated epicardium and endocardium support cardiomyocyte replenishment initially through the guidance of coronary sprouting. Simulating this process in the injured mammalian heart should help its healing.

Graphical Abstract

eTOC Blurb

Marín-Juez et al. find that the epicardium and endocardium help to reestablish the coronary network after cardiac injury in adult zebrafish. Coronary revascularization is regulated by the coordinated action of Cxcl12/Cxcr4 and Vegfa signaling. Regenerating coronaries provide a scaffold available for cardiomyocyte repopulation. Perturbing Cxcl12/Cxcr4 or Vegfa signaling impairs coronary revascularization and cardiomyocyte repopulation.

INTRODUCTION

Timely revascularization after myocardial infarction (MI) improves cardiac function and is key to preventing post infarction pathophysiological remodeling. Upon MI in humans, coronary vessels occasionally extend collaterals to reperfuse the infarct, thereby decreasing the size and extent of the injury (Habib et al., 1991; Sabia et al., 1992). The patients requiring revascularization undergo percutaneous coronary intervention or bypass surgery. Despite their proven benefit, both of these approaches are invasive techniques that cannot be used in a substantial portion of patients due to their anatomical or clinical characteristics (McFalls et al., 2004). Motivated by the need for alternative treatments, therapeutic angiogenesis aims to promote reperfusion of the infarcted area by stimulating the growth of new vessels from pre-existing ones. A number of growth factors have attracted therapeutic interest due their ability to stimulate angiogenesis. Despite no shortage of candidates including VEGF (Ruixing et al., 2007), PlGF (Kolakowski et al., 2006), FGF (Baffour et al., 1992), and HGF (Morishita et al., 1999), clinical studies designed to boost therapeutic angiogenesis have thus far proven inefficient (Robich et al., 2011).

Endothelial cells (ECs) are the most numerous cardiac cell type in humans, mice (Pinto et al., 2016) and zebrafish (Patra et al., 2017). Using zebrafish as a model for cardiac regeneration, we recently found that after injury the adult heart exhibits the unique ability to mount a fast and strong revascularization response (Marin-Juez et al., 2016), supporting the regeneration of damaged tissue (Lepilina et al., 2006). Similarly, stabilization of new vessels after cardiac injury in non-regenerative model organisms has been shown to improve survival (Zangi et al., 2013) and reduce scarring (Lai et al., 2017). However, how regenerative revascularization is regulated and how it takes place so quickly after injury remains unknown.

Here, we show that after cryoinjury of the adult zebrafish heart, new coronaries regenerate superficially to engulf the injured area, as well as towards the cardiac lumen. By loss- and gain-of-function manipulations, we find that epicardial Cxcl12/Cxcr4 signaling and endocardial Vegfa signaling regulate these two modes of coronary revascularization. We also report that coronaries form a vascular scaffold available for cardiomyocytes (CMs) during regeneration as well as during development, and that perturbation of coronary revascularization affects CM replenishment. Our findings reveal how by orchestrating a multi-tissue response, the regenerating zebrafish heart grows a vascular scaffold that aids in CM repopulation of the injured area.

RESULTS

Regenerating coronaries sprout superficially and intra-ventricularly

To understand how revascularization occurs in a regenerative model, we used the zebrafish and started by examining coronary revascularization after cardiac cryoinjury using the coronary-specific Tg(−0.8flt1:RFP) line (Bussmann et al., 2010) (Fig. S1A). We previously reported revascularization on the surface of the ventricle as early as 15 hours post cryoinjury (hpci) (Marin-Juez et al., 2016). To better understand coronary endothelial cell (cEC) behavior during this process, we first examined their proliferation at different times after injury (Fig. 1A, Fig. S1B-F), and observed that cEC proliferation peaked at 96 hpci and returned to basal levels by 30 days post cryoinjury (dpci) (Fig. 1A, Fig. S1B-F). Notably, at 7 dpci we identified two types of coronary vessels: some covering the surface of the injury (hereafter termed superficial), and others sprouting towards the ventricular lumen (hereafter termed intra-ventricular) (Fig. 1B). To further examine this process, we performed light-sheet imaging of cleared ventricles at 7 dpci (Fig. 1C,D). 3D-reconstructed images revealed several intra-ventricular sprouts in the injury border zone at this stage (Fig. 1C, Fig. S1G, Video S1). Moreover, we also observed intra-ventricular sprouting in areas distant from the border zone, closer to the apex (Fig. 1D, Video S2). Importantly, we did not observe intra-ventricular sprouting of coronaries in uninjured hearts. Hence, we identified two different types of regenerating coronaries: superficial and intra-ventricular, the latter ones only observed during regeneration.

Figure 1. Regenerating coronaries sprout superficially and towards the cardiac lumen.

(A) cEC proliferation at different times after injury (n=4 ventricles per time point). (B) Ventricle at 7 dpci stained for coronaries (green), CMs (red) and DNA (blue); (i, ii) high-magnification images of coronary vessels sprouting into the ventricle (white arrowheads) and superficially (magenta arrowheads). (C, D) 3D image reconstructions of the injured area after tissue clearing and staining for coronaries (red) and CMs (green). (C) Luminal view of intra-ventricular coronary sprouting (arrowheads) in the border zone (white arrows). Sprouts (yellow arrowheads) shown in high magnification (n=3). (D) Intra-ventricular sprouting vessels in the apex (arrowheads). High magnification images of sprouts (yellow arrowheads) shown in i and ii (n=3). Asterisk marks the border zone. (E) Tg(−0.8flt1:RFP); TgBAC(apln:EGFP) ventricle at 96 hpci. Yellow dotted line delineates the area with apln:EGFP⁺ coronaries (n=6). (F) Ventricular section at 96 hpci stained for coronaries (red), apln:EGFP expression (green), PCNA (white), and DNA (blue). High-magnification images of coronary vessels sprouting intra-ventricularly (Fi, white arrowheads) and superficially (Fii, white arrowheads) (n=6). (G) qPCR analysis of ppargc1a, apln, ndufb5, and atp5j mRNA levels in sorted −0.8flt1:RFP⁺ cECs at 96 hours post sham (hps) (n=4) and 96 hpci (n=4). (H) qPCR analysis of ppargc1a, apln, ndufb5, and atp5j mRNA levels in sorted −0.8flt1:RFP⁺ cECs from 8 months post fertilization (mpf) untouched WT (n=4) and ppargc1a−/− (n=4) ventricles. (I) qPCR analysis of ppargc1a, apln, ndufb5, and atp5j mRNA levels in sorted −0.8flt1:RFP⁺ cECs from 8 months post fertilization (mpf) WT (n=4) and ppargc1a−/− (n=4) ventricles at 96 hpci. (J) cEC proliferation in 8 mpf WT (n=5), ppargc1a−/− (n=5), and apln−/− (n=5) ventricles. (K) Section of a TgBAC(apln:CreERT2); Tg(ubb:GSR) ventricle at 7 dpci stained for GFP (white), mCherry (green), CMs (red), and DNA (blue); right panel shows high magnification image of recombined coronaries (white arrows) (n=6). White dotted lines delineate injured area. Data in graphs expressed as mean ± SEM. ns, no significant difference, *P < 0.05, **P < 0.01, ***P < 0.001. Scale bars: 50 μm. See also Figure S1; Figure S2; Figure S3; Video S1; and Video S2.

Superficial revascularization is metabolically regulated

Recent studies in mouse have reported that after MI, sprouting coronaries upregulate Apelin (Apln), a GPCR ligand gene (Liu et al., 2015). To gain more insight into coronary revascularization in zebrafish, we developed a TgBAC(apln:EGFP) line. No GFP signal was observed in sham operated ventricles (Fig. S2A); however, new apln:EGFP⁺ coronary sprouts could be detected in the injured area as early as 15 hpci (Fig. S2B), supporting our previous findings (Marin-Juez et al., 2016). The sprouting coronary front exhibited strong apln:EGFP expression at 48 (Fig. S2C) and 96 (Fig. 1E) hpci. In addition, we confirmed by in situ hybridization on injured zebrafish ventricles that TgBAC(apln:EGFP)^bns157 expression recapitulates endogenous apln expression: apln was expressed by regenerating coronaries at 48 (Fig. S2D) and 96 (Fig. S2E) hpci, as was the TgBAC(apln:EGFP)^bns157 transgene (Fig. S2F,G). Some border zone CMs also appeared to express apln (Fig. S2E), as well as TgBAC(apln:EGFP) (Fig. S2F) and TgBAC(apln:CreERT2) (Fig. S2H). Notably, at 96 hpci, only the superficial, and not intra-ventricular, regenerating coronaries were clearly apln:EGFP+ (Fig. 1F, Fig. S2G). By 7 dpci, TgBAC(apln:EGFP) expression was mostly downregulated, and the few apln:EGFP⁺ vessels observed at this stage were superficially located (Fig. S2I). APELIN has been shown to modulate cellular metabolism (Bertrand et al., 2015), and promote glucose and fatty acid uptake in ECs (Pi et al., 2018). Thus, we reasoned that in order to rapidly colonize the injured area, the superficial coronaries might experience metabolic changes during this period of active sprouting (Fig. S2J) (Marin-Juez et al., 2016). To test this hypothesis, we performed qPCR analyses on sorted cECs at 96 hpci, when these cells exhibit their proliferative peak. We found that the mRNA levels of peroxisome proliferator-activated receptor gamma coactivator 1a (ppargc1a), which encodes a master regulator of oxidative metabolism, as well as of its downstream targets NADH:ubiquinone oxidoreductase subunit b5 (ndufb5) and ATP synthase peripheral stalk subunit f6 (atp5j) (Arany et al., 2005), were reduced, whereas those of apln were increased (Fig. 1G). These results suggest that after injury, cECs switch their metabolism to rely less on oxidation and become more glycolytic, consistent with previous data showing that ECs use glycolysis as their main source of energy during migration and proliferation (De Bock et al., 2013). To further investigate the metabolic switch that cECs experience during regeneration, we developed a ppargc1a mutant line by targeting the transcriptional activation domain (Fig. S3A,B). qPCR analysis of uninjured mutant cECs showed that the mRNA levels of ppargc1a as well as of its downstream targets ndufb5 and atp5j were significantly lower compared to wild type (WT), whereas those of apln were significantly higher (Fig. 1H). Moreover, we found that ppargc1a−/− ventricles exhibited significantly reduced oxidative capacity and increased reliance on glycolysis (Fig. S3C). We also examined apln mRNA levels in ppargc1a−/− cECs after cardiac cryoinjury and found a significant upregulation compared to WT at 96 hpci (Fig. 1I). Importantly, we found that, compared to WT, superficial cEC proliferation at 96 hpci was increased in ppargc1a−/− ventricles while it was reduced in apln−/− ventricles (Fig. 1J, Fig. S3D-F), and so was superficial coronary revascularization (Fig. S3G-I). Altogether these data indicate that, superficial revascularization is regulated at least in part by Apelin signaling and that metabolic reprograming of cECs is involved in this process.

Superficial and intra-ventricular coronaries have a common origin

We next investigated whether the superficial and intra-ventricular sprouting coronaries derive from different cEC populations or, instead, have a common origin. To address this question, we developed a TgBAC(apln:CreERT2) line, which in combination with the Tg(ubb:GSR) line, allowed us to trace the descendants of superficial coronaries. We induced recombination by administration of 4-OHT from 48 hpci to 5 dpci, when cEC proliferation and apln:EGFP expression peak, and analyzed the ventricles at 7 dpci when the injured area is colonized by cECs and both types of vessels are clearly present. At 7 dpci we observed recombined cECs in both the superficial and intra-ventricular coronaries (Fig. 1K). Thus, after recombination of the reporter line occurred in superficial apln:EGFP⁺ cECs, the labeled vessels sprouted towards the ventricular lumen indicating that both types of regenerating coronaries have a common origin.

Cxcl12b-Cxcr4a signaling regulates superficial revascularization

In view of these results, we hypothesized that these two modes of coronary revascularization are regulated by different mechanisms. To test this hypothesis, we first investigated mechanisms underlying superficial revascularization and focused on the adjacent tissue, the epicardium. To analyze whether the epicardium was required for superficial revascularization, we carried out ablation experiments using the TgBAC(tcf21:NTR-mCherry) line (Fig. S4A,B) in combination with the TgBAC(etv2:EGFP) line, which allows discrimination between coronary and endocardial endothelial cells (Fig. S4C, Table S1). While control fish displayed both superficial and intra-ventricular vessels (Fig. 2A), epicardial ablated hearts exhibited a strong reduction in the number of superficial cECs (Fig. 2B,C) without a significant change in the number of border zone intra-ventricular vessels (Fig. 2B,D).

Figure 2. Epicardial chemokine expression directs superficial coronary revascularization.

(A, B) Ventricles at 10 dpci stained for epicardium (white), endothelium (green), CMs (red), and DNA (blue). High magnification images of intra-ventricular sprouting coronaries (Ai, Bi; GFP⁺ yellow arrowheads), and superficial coronaries (Aii, Bii; GFP⁺, magenta arrowheads) in DMSO (A) and MTZ (B) treated animals. Number of superficial cECs (C) and intra-ventricular coronaries (D) in DMSO (n=6) and MTZ (n=6) treated animals at 10 dpci. (E) 7 dpci ventricle stained for cxcl12b:YFP expression (white), coronary ECs (green), CMs (red), and DNA (blue). High magnification images of regenerating epicardium upregulating cxcl12b:YFP expression. (F) qPCR analysis of aldh1a2 and cxcl12b mRNA levels in sorted tcf21:RFP⁺ cells from 7 days post sham (dps) and 7 dpci ventricles (n=4). (G) 7 dpci ventricle stained for cxcl12b:YFP expression (green), activated epicardium and endocardium (white), CMs (red), and DNA (blue). High magnification images of regenerating cxcl12b:YFP⁺ epicardium (yellow arrowheads), and activated endocardium negative for cxcl12b:YFP expression (magenta arrowheads). Wholemount images of Tg(fi1a:EGFP) WT (H) and cxcr4a−/− (I) ventricles at 7 dpci. (I) High magnification image shows an intra-ventricular sprouting coronary in a cxcr4a−/− heart (white arrowhead). Number of superficial cECs (J) and intra-ventricular coronaries (K) in WT (n=6) and cxcr4a−/− (n=6) animals at 7 dpci. Scale bars: 100 μm. See also Figure S4; Figure S5; and Table S1.

Since we previously found that the TgBAC(cxcr4a:YFP) line labels superficial coronaries during regeneration (Marin-Juez et al., 2016), we postulated that activated epicardial cells produce Cxcl12b, a Cxcr4 ligand, which in mouse and zebrafish is required for coronary development and regeneration (Itou et al., 2012; Cavallero et al., 2015; Harrison et al., 2015; Das et al., 2019). To test this hypothesis, we cryoinjured TgBAC(cxcl12b:YFP); Tg(−0.8flt1:RFP) ventricles. While sham operated hearts exhibited cxcl12b:YFP expression only in perivascular cells (Fig. S4D), previously shown to be epicardial-derived (Harrison et al., 2015), cryoinjured hearts exhibited cxcl12b:YFP expression in the epicardium lining the injured area at 7 (Fig. 2E) and 14 (Fig. S4E) dpci. This upregulation at 7 dpci was confirmed by qPCR analysis of tcf21:RFP⁺ sorted cells (Fig. 2F) and by in situ hybridization (Fig. S4F,G). Additionally, we co-stained injured TgBAC(cxcl12b:YFP) hearts for activated epicardium and endocardium and observed no endocardial cxcl12b:YFP expression (Fig. 2G).

To test the role of Cxcl12/Cxcr4 signaling in superficial coronary sprouting, we cryoinjured Tg(fli1a:EGFP);cxcr4a−/− hearts and observed that at 7 dpci, while the injured area in WT hearts was fully covered by new superficial coronaries (Fig. 2H, Fig. S4H), mutant hearts were largely devoid of superficial coronaries while border zone intra-ventricular vessels were still present (Fig. 2I, Fig. S4I). These observations were quantified (Fig. 2J,K). Moreover, chemical inhibition of Cxcr4 signaling also blocked superficial coronary revascularization (Fig. S4J,K). This defective coronary revascularization might be responsible for the impaired regenerative capacity of cxcr4a−/− hearts reported by Harrison et al. (2015), further supporting the importance of superficial revascularization in cardiac regeneration. Overall, these results indicate that the regenerating epicardium guides superficial coronary revascularization at least in part via Cxcl12-Cxcr4 signaling.

Hypoxia triggers epicardial cxcl12b expression via Hif1a

During heart regeneration, epicardial and epicardial-derived cells (EPDCs) give rise to perivascular cells via an epithelial-to-mesenchymal transition (EMT) (Kikuchi et al., 2011a; Gonzalez-Rosa et al., 2012). During development in avian models and mouse, epicardial EMT is regulated by Hif1a (Tao et al., 2013; Guimaraes-Camboa et al., 2015). Moreover, in mouse hearts, Cxcl12 expression is stimulated by hypoxic preconditioning (Hu et al., 2007). To test whether hypoxia triggers epicardial cxcl12b expression after cardiac injury, we analyzed mRNA levels of egln3, a well-known hypoxia responsive gene (Santhakumar et al., 2012), in sorted epicardial cells at 7 dpci, and observed higher levels than in controls (Fig. 3A). To further test whether the regenerating epicardium is hypoxic at this stage, we utilized a Hypoxyprobe which marks cells with low oxygen tension. Whereas in sham operated hearts a signal was hardly detectable (Fig. 3B), 7 dpci ventricles exhibited a strong Hypoxyprobe signal in regenerating epicardial cells (Fig. 3C). Next, we analyzed cxcl12b mRNA levels in WT and hif1aa−/−;hif1ab−/− (hereafter hif1a−/−) (Gerri et al., 2017) hearts after injury. At 7 dpci, cxcl12b mRNA levels in hif1a−/− ventricles were significantly decreased compared to WT, showing levels similar to those in sham operated samples (Fig. 3D). Accordingly, we found that superficial cEC proliferation in hif1a−/− was significantly reduced compared to WT (Fig. 3E-G). To test whether hypoxia stimulates superficial cEC proliferation, we treated injured animals with DMOG, which mimics hypoxia by stabilizing Hif1a (Jaakkola et al., 2001). Supporting our findings, DMOG injected animals exhibited increased superficial cEC proliferation at 96 hpci when compared to controls (Fig. 3H-J). Altogether, these data indicate that during regeneration, hypoxia triggers epicardial cxcl12b expression via Hif1a to promote superficial coronary revascularization.

Figure 3. Epicardial hypoxia stimulates epicardial cxcl12b expression and coronary EC proliferation via Hif1a.

(A) qPCR analysis of aldh1a2 and egln3 mRNA levels in sorted tcf21:RFP⁺ cells from 7 dps and 7 dpci ventricles (n=4). 7 dps (B) and 7 dpci (C) ventricles stained for epicardium (green), hypoxyprobe (red), and DNA (blue). Insets show high magnification images of sham (B) and regenerating (C) epicardium. (D) qPCR analysis of cxcl12b mRNA levels in WT ventricles at 7 dps (n=4) and 7 dpci (n=4), and in hif1a−/− ventricles at 7 dpci (n=4). WT (E) and hif1a−/− (F) ventricles stained for endothelial nuclei (red), PCNA (green), and DNA (blue). Insets show high-magnification images of proliferating cECs (white arrowheads). (G) cEC proliferation in WT (n=4), and hif1a−/− (n=4) ventricles at 96 hpci. Ventricles from PBS (H) and DMOG (I) treated animals stained for endothelial nuclei (red), PCNA (green), and DNA (blue). Insets show high-magnification images of proliferating cECs (white arrowheads). (J) cEC proliferation in PBS (n=4), and DMOG (n=4) treated ventricles at 96 hpci. White dotted lines delineate injured area. Data in graphs expressed as mean ± SEM. ns, no significant difference, *P < 0.05, **P < 0.01. Scale bars: 100 μm.

Vegfaa signaling regulates intra-ventricular sprouting

To investigate mechanisms of intra-ventricular sprouting, we analyzed Tg(fli1a:EGFP) ventricles at 7dpci and found that intra-ventricular regenerating vessels sprout towards the activated endocardium as assessed by Aldh1a2 expression (Kikuchi et al., 2011b) (Fig. 4A). In addition, we found that intra-ventricular coronaries are also positive for cxcr4a:YFP expression (Fig. S5A), likely due to their proximity to the regenerating epicardium and the stability of the EGFP protein. Moreover, we confirmed by colocalization with Tg(−0.8flt1:RFP) expression that TgBAC(cxcr4a:YFP) expression was present specifically in regenerating coronaries (Fig. S5B,C). Analyzing TgBAC(cxcr4a:YFP) ventricles at 7 dpci, we also observed that regenerating intra-ventricular coronaries had sprouted towards the activated endocardium (Fig. 4B). Previous studies in adult mice have suggested that the endocardium is a source of cECs (Miquerol et al., 2015; Dube et al., 2017) during cardiac regeneration but recent lines of evidence have challenged this model (Tang et al., 2018). To evaluate whether activated endocardial cells could be detected in the regenerating coronary plexus, we injured the endocardial specific ET(krt4:EGFP)^sqet33-1A line (Munch et al., 2017). Examining GFP and Aldh1a2 expression at 7 and 14 dpci, this later time point being when the new coronary network is well established (Marin-Juez et al., 2016), we failed to detect ET(krt4:EGFP)^sqet33-1A positive cells in the coronary plexus (Fig. S5D,E). Therefore, these data, together with previously published lineage tracing data (Zhao et al., 2014), suggest that preexisting coronaries might be the main source of regenerated cECs in zebrafish. However, endocardial specific lineage tracing will be needed to conclusively assess the extent of endocardial contribution to the regenerated coronary plexus.

Figure 4. Vegfaa signaling directs intra-ventricular coronary sprouting.

(A) 7 dpci ventricle stained for endothelium (green), Aldh1a2 (red), and DNA (blue). High magnification images show intra-ventricular sprouting coronaries (EGFP+,Aldh1a2−) (white arrowhead) in close proximity to activated endocardium (EGFP+,Aldh1a2+) (n=4). (B) 7 dpci ventricle stained for regenerating coronaries (green), Aldh1a2 (white), CMs (red), and DNA (blue). Regenerating superficial coronaries (yellow arrowheads). High magnification images show a cross section of regenerating intra-ventricular coronaries surrounded by activated endocardium (n=6) (red arrowheads). 15 hpci (C, D) (n=6) and 30 dpci (E) (n=5) ventricles stained for vegfaa:EGFP expression (green), endocardium (white), CMs (red), activated endocardium (D, white), and DNA (blue). High magnification images of vegfaa:EGFP⁺ border zone endocardium (Ci; yellow arrowheads), endocardium distal to the injury (Cii), and vegfaa:EGFP⁺ activated endocardium (D, E; white arrowheads). WT (F) and flt1−/− (G) ventricles stained for cardiac endothelium (green) and cECs (red). High magnification images show intra-ventricular sprouting coronaries (white arrowheads). Number (H) and length (I) of intra-ventricular coronaries in WT (n=6) and flt1−/− (n=4) animals at 96 hpci. White dotted lines delineate injured area. Data in graphs expressed as mean ± SEM. **P < 0.01, ***P < 0.001. Scale bars: 100 μm. See also Figure S5.

We previously reported that induced expression of a dominant negative form of the angiogenic factor Vegfaa blocked nearly all revascularization after cardiac injury, with only a few superficial coronaries left in the injured area, and no evidence of intra-ventricular coronaries (Marin-Juez et al., 2016). To further investigate the role of Vegfa signaling during intra-ventricular sprouting, we cryoinjured TgBAC(vegfaa:EGFP) (Karra et al., 2018) ventricles. This vegfaa:EGFP transgene has been shown to be to be expressed by epicardial cells in untouched and amputated zebrafish ventricles, and to be upregulated in the border zone endocardium at 3 days post amputation (Karra et al., 2018). Moreover, Vegfa has been shown to be expressed by the endocardium during mouse development (Miquerol et al. 1999). By colocalization with expression of the endothelial specific line Tg(kdrl:mCherry) (Fig. 4C) and Aldh1a2 immunostaining (Fig. 4D), we observed vegfaa:EGFP expression in the border zone endocardium as early as 15 hpci (Fig. 4C), coinciding with the initiation of coronary revascularization (Marin-Juez et al., 2016). Endocardial vegfaa:EGFP expression persisted during the time of active cEC proliferation (48 hpci and 7 dpci, Fig. S5F,G) and was still detectable at 30 dpci (Fig. 4E). Since blocking Vegfa signaling had a strong impact on intra-ventricular sprouting, we set out to test whether gain-of-function approaches would have the opposite effect. For this purpose, we utilized flt1−/− fish, which exhibit excessive Vegfaa signaling (Matsuoka et al., 2016), and found that they displayed a significant increase in the number and length of intra-coronary vessels at 96 hpci (Fig. 4F-I). Altogether, these data suggest that endocardial upregulation of vegfaa after injury plays an important role in intra-ventricular coronary sprouting.

Coronaries form a scaffold for cardiomyocytes during regeneration as well as during development

To better understand the functional significance of these two sets of regenerating vessels, we analyzed ventricles at 7 dpci, when coronaries have covered the injured area and CM proliferation peaks. Interestingly, we observed that many of the CMs repopulating the injured area were associated with regenerating vessels (Fig 1Bi,ii, Fig. 5A). Analyzing 30 dpci ventricles, we observed regenerating CMs closely associated with superficial vessels (Fig. 5B, Fig. 4B,E) as well as with intra-ventricular vessels (Fig. 5B). Upon injury, CMs that upregulate gata4:GFP expression are important contributors to the regenerated cardiac muscle (Kikuchi et al., 2010). gata4:GFP expression is upregulated by cortical CMs after injury (Kikuchi et al., 2010) (Fig. S6A,B), and we observed that most of these CMs were in close proximity to regenerating vessels at 7 (Fig. 5C) and 14 (Fig. S6C) dpci. Analysis of Tg(gata4:GFP) ventricles at 7 dpci revealed that in the injured area, gata4:GFP⁺ CMs in close proximity to regenerating coronaries were three times more proliferative than gata4:GFP⁻ CMs (Fig. 5D,E).

Figure 5. Coronaries constitute a scaffold available to cardiomyocytes during regeneration as well as during development.

(A) 7 dpci ventricle stained for coronaries (green), CMs (red), and DNA (blue) (n=8). CMs associated with regenerating coronaries in the injured area (white arrowhead). (B) 30 dpci ventricle stained for cxcr4a:YFP expression (green), activated endocardium and epicardium (white), CMs (red), and DNA (blue) (n=6). White arrowheads point to regenerating coronaries scaffolding the regenerating myocardium. High magnification images of regenerating intra-ventricular coronaries scaffolding the regenerating myocardium (red arrowheads). (C) 7 dpci ventricle stained for coronaries (green), gata4:GFP expression (white), DNA (blue), and CMs (red) (n=6). High magnification image shows regenerating gata4:GFP⁺ CMs associated with coronaries in the injured area (red arrowheads). (D) 7 dpci ventricle stained for gata4:GFP expression (white), CM nuclei (red), PCNA (green), and DNA (blue). High magnification image shows cortically located and proliferating gata4:GFP⁺ CMs. (E) Proliferation index of gata4:GFP⁺ vs gata4:GFP⁻ CMs at 7 dpci (n=6). (F-F”) Wholemount images of hearts from 20 mm long 7 wpf fish (n=6). High magnification images of developing coronaries (RFP⁺) and cortical CMs (GFP⁺). White arrowheads point to vessels at the sprouting front of the developing coronary network (not associated with developing cortical CMs); yellow arrowhead points to cortical CMs (associated with coronaries). (G) Ventricle from a 7 wpf fish stained for coronary ECs (white), gata4:GFP expression (red), CMs (green), and DNA (blue). Yellow arrowheads point to developing cortical CMs. Coronary vessel (white arrowhead) flanked by gata4:GFP⁺ CMs. White dotted lines delineate injured area. Data in graph expressed as mean ± SEM. ***P < 0.001. Scale bars: 100 μm (A-F); 50 μm (G). See also Figure S6.

gata4:GFP is also expressed by cortical CMs during development (Gupta et al., 2013), which happens at a similar developmental stage as coronary formation (Harrison et al., 2015). Therefore, we hypothesized that coronary vessels might also serve as a scaffold to developing cortical CMs. To analyze both processes simultaneously, we imaged Tg(gata4:GFP); Tg(−0.8flt1:RFP) ventricles at 7 weeks post fertilization (wpf), a stage when both coronaries and the cortical myocardium are forming. At this stage, while some developing coronaries could be observed without cortical CMs (Fig. 5F), most gata4:GFP⁺ CMs appeared to be associated with developing coronaries (Fig. 5F,G). Analysis of larger fish, which exhibit a more developed coronary network, suggests that gata4:GFP⁺ CMs might use the coronary network as a scaffold to migrate, as the areas lacking coronary vasculature were also devoid of cortical CMs (Fig. S6D). Collectively, these results suggest that coronary vessels serve as a scaffold available to CMs during both development and regeneration.

The formation of a coronary scaffold is required for cardiomyocyte repopulation

Next, we tested whether superficial and intra-ventricular coronaries regulate different aspects of CM regeneration. To this end, we first analyzed CM regeneration in cxcr4a−/− as they display reduced superficial revascularization (Fig. 2H-K). We measured the distance that cortically located CMs spanned into the injured area at 7 dpci and found a significant reduction in cxcr4a−/− ventricles (Fig. 6A,B). To block intra-ventricular vessels, we used the HOTcre system (Hesselson et al., 2009) and generated a Tg(hsp70l:LSL-dnvegfaa) line. Using this line together with the Tg(myl7:CreERT2) line gives one spatial and temporal control over the expression of dominant negative Vegfaa (dnvegfaa) (Fig. S6E-G), which was previously shown to block sprouting angiogenesis (Rossi et al., 2016). Similar to our observations on coronary revascularization, regenerating CMs first engulf the injured area (Gonzalez-Rosa et al., 2011). Therefore, we administered Tg(myl7:CreERT2); Tg(hsp70l:LSL-dnvegfaa) fish with EtOH (control) or with 4-OHT at 30 dpci and heatshocked them daily until 90 dpci. Using this protocol, we mostly affected intra-ventricular revascularization by overexpressing dn-Vegfaa after allowing for coronary and CM engulfment of the injured area. While in control hearts CMs had largely repopulated the injured area by 90 dpci (Fig. 6D), upon induction of dn-Vegfaa we observed large luminal areas devoid of CMs (Fig. 6E). To better characterize the effect that blocking intra-ventricular revascularization has during CM regeneration, we performed Acid Fuchsin Orange G (AFOG) staining at 90 dpci to determine the presence and size of a scar. In control ventricles at 90 dpci, we observed minor or no scars as well as several contact points between regenerated and spared muscle (Fig. 6F). Conversely, late induction of dn-Vegfaa expression in regenerating hearts impaired intra-ventricular CM repopulation leading to the retention of a fibrotic scar (Fig. 6G,H).

Figure 6. Perturbations in coronary revascularization impair cardiomyocyte repopulation.

WT (A) and cxcr4a−/− (B) ventricles at 7 dpci stained for CMs (red) and DNA (blue). Presence (A) and absence (B) of cortically located regenerating CMs is indicated by white arrowhead and asterisk, respectively. (C) Distance spanned into the injured area by cortically located CMs in WT (n=6) and cxcr4a−/− (n=6) at 7 dpci. Control (D, +EtOH) and recombined (E, +4-OHT) ventricles at 90 dpci stained for CMs (red) and DNA (blue). Asterisk marks an area devoid of CMs (E). Control (F, +EtOH) and recombined (G,+4-OHT) ventricles at 90 dpci stained with AFOG to identify muscle (orange), collagen (blue) and fibrin (red). Asterisks mark contact areas between spared and regenerated myocardium. Magenta arrowheads point to regenerating CMs inside the remaining scar (F). Green dashed lines outline the scar area. (H) Quantification of scar area in control (n=5) and recombined (n=4) ventricles at 90 dpci. Ventricles from control (I, PBS) and Poly(I:C) (J) injected medaka at 14 dpci stained for CMs (red), DNA (blue) and endothelium (green). Yellow arrowheads point to CMs expanding beyond the border zone endocardium into the injured area; white arrowheads point to ECs in close proximity to CMs in the injured area. (K) Quantification of EC to CM association index in control (n=4) and Poly(I:C) (n=4) injected medaka at 14 dpci. White (A,B,D,E,I,J) and black (F,G) dotted lines delineate injured area. Data in graphs expressed as mean ± SEM. *P < 0.05, ****P < 0.0001. Scale bars: 100 μm.

We next asked whether by stimulating revascularization we could also stimulate CM repopulation. Working with medaka, a fish unable to regenerate its heart (Ito et al., 2014), we recently reported that Poly (I:C) injection after cardiac injury appears to stabilize endocardial derived vessel-like structures, and leads to reduced scarring (Lai et al., 2017). Therefore, we decided to test whether stabilization of the vessel-like structures in medaka improved CM repopulation. Indeed, we found that while in PBS injected medaka, CMs were not present beyond the border zone endocardium (Fig. 6I), in Poly (I:C) injected animals a large number of CMs populated the injured area (Fig. 6J), and they were in close proximity to endothelial cells (Fig. 6K). Altogether, these findings indicate that superficial and intra-ventricular vessels are required for superficial and luminal CM regeneration respectively, and that stimulating revascularization might stimulate CM repopulation in non-regenerative species.

DISCUSSION

Overall, we describe how during cardiac regeneration in adult zebrafish, new coronaries sprout from preexisting vessels to cover the ventricular surface as well as towards the trabecular layer to reach the activated endocardium in a process we refer to as coronary-endocardial anchoring. After cardiac damage in non-regenerative models including the adult mouse (Miquerol et al., 2015; Dube et al., 2017; Tang et al., 2018) and medaka (Lai et al., 2017), the endocardium undergoes major morphological remodeling and upregulates coronary markers, possibly to compensate for the lack of an efficient coronary network (Carmeliet, 2005; Lai et al., 2017). Our data suggest that during cardiac regeneration in adult zebrafish, the endocardium provides guidance cues to the new coronary plexus at least in part via Vegfaa signaling. We also found that the epicardial hypoxic microenvironment induces superficial coronary sprouting at least in part via Cxcl12/Cxcr4 signaling. Recent data indicate that the Cxcl12/Cxcr4 signaling pathway facilitates cardiac regeneration in neonatal mice as well (Das et al., 2019). In this study, Das et al. (2019) showed that upon cardiac injury, capillaries upregulate Cxcl12-DsRed expression, and that upon endothelial deletion of Cxcr4, regeneration is impaired. Similar to our findings in zebrafish, studies in mouse reported Cxcl12 expression in embryonic EPDCs (Cavallero et al., 2015) and upregulation of Cxcl12 and CXCL12 in adult EPDCs upon MI (Zhou et al., 2011). To our knowledge, whether CXCL12 becomes upregulated by EPDCs in neonatal mice after cardiac injury remains to be determined.

Notably, we found that after cryoinjury, the superficial coronary vessels became glycolytic, a metabolism favored in low oxygen conditions (Robey et al., 2005). Moreover, we show that by manipulating Hif1a we were able to modulate cEC proliferation during regeneration. A number of epicardial factors (e.g Fgf, Pdgf, Nox/Duox) have been shown to regulate cardiac regeneration and revascularization (Kapuria et al., 2018). Which factors are regulated by epicardial hypoxia, and to what extent, remains to be determined.

The multi-tissue coordination of Vegfa, Apelin and Cxcl12-Cxcr4 signaling culminates with the formation of a vascular scaffold that spans the injured area. We previously reported that blocking revascularization early after injury reduces CM proliferation (Marin-Juez et al., 2016). Here, our data indicate that CMs can use the regenerated vascular scaffold to repopulate the injured area. Similarly, in vitro studies have reported that preformed endothelial networks improve the survival and spatial organization of CMs (Narmoneva et al., 2004). In addition, we found that after stimulating cardiac repair by Poly (I:C) administration in medaka, regenerating CMs were also in close proximity to endocardial derived vessel-like structures in the injured area. It will be important to further investigate the crosstalk between coronaries and CMs during regeneration as well as during development. Our results highlight the importance of blood vessels during cardiac regeneration beyond their role as a transport system. These findings have therapeutic significance and indicate that recapitulation of timely and injury spanning revascularization is a first step towards cardiac regeneration.

STAR METHODS

LEAD CONTACT AND MATERIALS AVAILABILITY

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Didier Y.R. Stainier (didier.stainier@mpi-bn.mpg.de). All unique/stable reagents generated in this study are available from the Lead Contact without restriction.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Animal studies

Zebrafish and medaka were reared and handled in compliance with institutional (MPG) and national animal welfare legislation and maintained according to standard protocols (http://zfin.org). Male and female zebrafish and medaka were used in this study. We used the previously published zebrafish lines Tg(fli1a:EGFP)^y1 (Lawson and Weinstein, 2002), Tg(tcf21:DsRed2)^pd37 (Kikuchi et al., 2011a), TgBAC(tcf21:mCherry-NTR)^pd108 (Wang et al., 2015), Tg(kdrl:mCherry)^is5 (Sacilotto et al., 2013), Tg(−0.8flt1:RFP)^hu5333 (Bussmann et al., 2010), TgBAC(etv2:EGFP)^ci1 (Proulx et al., 2010), TgBAC(cxcr4a:YFP)^mu104 (Xu et al., 2014), ET(krt4:EGFP)^sqet33-1A (Poon et al., 2010), TgBAC(cxcl12b:YFP)^mu105 (Bussmann and Schulte-Merker, 2011) Tg(−14.8gata4:GFP)^ae1 (Heicklen-Klein and Evans, 2004) (abbreviated as Tg(gata4:GFP)), TgBAC(vegfaa:EGFP)^pd260 (Karra et al., 2018), TgBAC(apln:EGFP)^bns157 (Helker et al. in preparation), Tg(−3.5ubb:loxP-EGFP-loxP-mCherry)^cz1701 (Mosimann et al., 2011) (abbreviated as Tg(ubb:GSR)) Tg(myl7:CreERT2,cryaa:DsRed)^pd10 (Kikuchi et al., 2010) (abbreviated as Tg(myl7:CreERT2)), hif-1aa^bns89 and hif-1ab^bns90 (Gerri et al., 2017), cxcr4a^um20 (Siekmann et al., 2009), flt1^bns29 (Matsuoka et al., 2016) and apln^mu267 (Helker et al., 2015). We used the Tg(fli1∷GFP)^tg1206 medaka line (Shinagawa-Kobayashi et al., 2018). For 4-hydroxytamoxifen (4-OHT) labeling using the TgBAC(apln:CreERT2); Tg(ubb:GSR) and Tg(myl7:CreERT2); Tg(hsp70l:LSL-mTomato-P2A-dnvegfaa) lines, adult fish were injected intraperitoneally daily from 2 to 5 dpci as previously described (Kikuchi et al., 2010). Tg(myl7:CreERT2); Tg(hsp70l:LSL-mTomato-P2A-dnvegfaa) larvae were treated with 5μM 4-OHT from 24 to 96 hours post fertilization (hpf) and heatshocked as previously described (Marin-Juez et al., 2016). For epicardial ablation experiments TgBAC(tcf21:mCherry-NTR) fish were bathed from 7 until 10 dpci in daily refreshed system water containing 1 mM MTZ as previously described (Curado et al., 2007; Pisharath et al., 2007; Curado et al. 2008; Wang et al., 2015). IT1t (Merck) was dissolved to a final concentration of 80 μM in PBS and 10 μl were injected daily intraperitoneally from 3 to 7 dpci. Cryoinjuries were performed as described (Gonzalez-Rosa et al., 2011). In the case of cxcr4a−/− fish, only animals with a developed coronary network were used for experiments. Procedures involving animals were approved by the veterinary department of the Regional Board of Darmstadt.

METHOD DETAILS

Generation of transgenic and mutant lines

The TgBAC(apln:CreERT2)^bns310 line was generated by modification of a Bacterial Artificial Chromosome (BAC) clone as previously described (Bussmann and Schulte-Merker, 2011; Helker et al., 2019). Briefly, the transcriptional start site codon of apln in the RP71-2G21 clone was replaced with the CreERT2 cassette. After Tol2 transposon LTR insertion into the vector backbone, the recombination was performed by amplifying a CreERT2 cassette using primers with 50 bp of homology flanking the start codon. Primer sequences are shown in Table S2. The BAC construct was purified with Nucleobond BAC kit (Clontech) and co-injected with tol2 transposase mRNA into zebrafish embryos at the one-cell stage. To establish the Tg(hsp70l:loxP-Stop-LoxP-mTomato-codOptP2A-T46Avegfaa; cryaa:Cerulean)^bns288 line (abbreviated as Tg(hsp70l:LSL-dnvegfaa)), one-cell staged zebrafish embryos were injected with the construct Tg(hsp70l:loxP-Stop-LoxP-mTomato-codOptP2A-T46Avegfaa; cryaa:Cerulean) and Isce-I enzyme. These injected embryos were grown to adulthood and screened for founders using Cerulean expression in the larval eyes. CRISPR/Cas9 was used to generate ppargc1a^bns176 mutants as previously described (Gerri et al., 2017). The following guide sequence: GGAAAATGAGGCCAACTTGC targeting exon 3 was injected (50 pg) together with Cas9 mRNA (100 pg) into zebrafish embryos at the one-cell stage. Mutant alleles were identified by high-resolution melt analysis. Primer sequences are shown in Table S2.

Hypoxyprobe and DMOG administration

Pimonidazole and DMOG were diluted in PBS to a final concentration of 5 mg/ml and 300 μM, respectively. In each case, 10 μl were injected intraperitoneally daily during the course of the experiment.

Poly (I:C) administration

PBS and Poly (I:C) injections were performed as previously described (Lai et al., 2017). Briefly, medaka were injected intraperitoneally with 10 μl of PBS or Poly I:C (1 μg/μl, R&D Systems) immediately after cryoinjury.

Histological Analysis and Imaging

Histological analyses were performed as previously described (Marin-Juez et al., 2016). Tissue clearing was performed using the X-CLARITY Tissue Clearing System (Biozym Scientific) following manufacturer’s indications. Primary antibodies used in this study include anti-MHC (Developmental Studies Hybridoma Bank, 1:50), anti-GFP (Aves, 1:500), anti-DsRed (Clontech, 1:500), anti-tRFP (Evrogen, 1:500), anti-PCNA (Santa Cruz Biotechnology, 1:500), anti-Fli1 (Abcam, 1:100), anti-Aldh1a2 (Genetex, 1:100), anti-Mef2 (Santa Cruz Biotechnology, 1:100) and DAPI (DNA, Sigma). Secondary antibodies used in this study include Alexa Fluor 647, Alexa Fluor 488, Alexa Fluor 568 (Thermo Fisher, 1:500), HP-Red549 (Hypoxyprobe, 1:100), and Alexa Fluor 568 phalloidin (Thermo Fisher, 1:50).

Analyses of cEC proliferation were performed by counting nuclei which were 0.8flt1:RFP⁺/PCNA⁺/DAPI⁺, etv2:EGFP⁺/Aldh1a2⁻/PCNA⁺/DAPI⁺,fli1a:EGFP⁺/Aldh1a2⁻/PCNA⁺/DAPI⁺, or Fli1⁺/PCNA⁺/DAPI⁺. Cardiomyocyte proliferation was analyzed as previously described (Marin-Juez et al., 2016) by counting Mef2⁺/gata4:GFP⁺/PCNA⁺/DAPI⁺ nuclei. Quantification of cEC proliferation was performed in the 100 μm area directly adjacent to the border zone and in the injured area of three non-superficial/nonconsecutive midsagittal sections from each heart. For analyses of ECs in close proximity to CMs in medaka, we counted fli1:EGFP⁺ cells and F-actin⁺ cells in the injured area. For all quantifications, we selected three non-superficial/non-consecutive midsagittal sections from each heart. Imaging of heart sections was performed using a Zeiss LSM 700 confocal microscope. Imaging of larval hearts was performed using a Zeiss LSM 700 or a Zeiss CSU-X1 Yokogawa confocal microscope. Whole-heart and larval imaging was performed on a Nikon SMZ25 stereomicroscope. Cleared ventricles were imaged using a Zeiss Lightsheet Z.1 microscope. Image analysis was carried out using Imaris (Bitplane), Zen Blue (Zeiss) and Zen Black (Zeiss).

Cell sorting

Adult zebrafish epicardial (tcf21:RFP⁺) and endothelial (−0.8flt1:RFP⁺) cells were isolated following manufacturer’s instructions (Pierce Primary Cardiomyocyte Isolation Kit, Thermo Scientific) with the following modifications: incubation was performed at 30°C with gentle shaking for 45 minutes followed by careful resuspension in 1xHBSS, 0.25% BSA, 10 mM HEPES buffer. Resuspended cells were immediately sorted using an FACSAria III (BD) sorter for RFP+/PI− (epicardial) and RFP+/PI− (endothelial) cells.

Quantitative PCR

For quantitative PCR (qPCR) expression analysis of whole ventricles, total RNA was isolated from sham and cryoinjured ventricles using an RNeasy Mini kit (Qiagen) following manufacturer’s instructions. Three ventricles were pooled per biological replicate, and 500 ng of total RNA was reverse transcribed with a Maxima First Strand cDNA Synthesis Kit (Thermo Scientific). For qPCR analysis of sorted cells, total RNA was isolated using an RNeasy Micro kit (Qiagen). Eight ventricles were pooled per condition, and at least 100 ng of total RNA was reverse transcribed with a Maxima First Strand cDNA Synthesis Kit. mRNA levels were normalized against the mRNA levels of rpl13a as a housekeeping gene. Primer sequences and Cq values are shown in Table S2.

Glycolytic flux and mitochondrial respiration

Zebrafish ventricles where dissociated following the procedure described above and 5×10⁴ cells per well were seeded on a Seahorse XF96 collagen pre-coated culture plate (Agilent, California, USA) in XF DMEM Base Medium without Phenol Red. For Glycolytic Flux measurements, the Base medium was supplemented with 1 mmol/L sodium pyruvate. For Oxidative Phosphorylation measurements, the Base medium was supplemented with 10 mmol/L glucose. Measurements were performed with the XF 96 extracellular flux analyzer (Agilent) at 28°C. For extracellular acidification rate (ECAR) measurements, glucose (10 mmol/L), Oligomycin (2.5 μmol/L) and 2-deoxy-D-glucose (2DG; 50 mmol/L) were used. For oxygen consumption rate (OCR) measurements, oligomycin (2.5 μmol/L), carbonyl cyanide 4-(trifluoromethoxy)phenylhydrazone (FCCP; 1.5 μmol/L), rotenone (0.5 μmol/L) and antimycin A (0.5 μmol/L) were used.

QUANTIFICATION AND STATISTICAL ANALYSIS

For analysis of cEC and CM proliferation, expression patterns, coronary length and number, mutant phenotypes, epicardial ablation, and compound administration, at least four biological replicates were examined. For qPCR analysis, at least four biological replicates, each of them consisting of three ventricles, were used. For cell sorting, at least eight ventricles were pooled per sample. For assessment of glycolytic flux and mitochondrial respiration, an average of three technical replicates was measured from four separate isolations, each isolation consisting of two zebrafish ventricles. Sample sizes are indicated in each figure legend. Statistical differences of qPCR expression data were analyzed by Mann–Whitney nonparametric test and were considered significant at P<0.05. cEC and CM proliferation as well as intra-ventricular coronary length and number were analyzed by the Student’s t-test (two tailed) and considered significant at P<0.05. Glycolytic flux and mitochondrial respiration were analyzed by two-way analysis of variance (ANOVA) with Bonferroni’s multiple comparison test. P values are indicated in the figure legends.

DATA AND CODE AVAILABILITY

This study did not generate any unique datasets or code.

Supplementary Material

1

Supplemental Video 1. Border zone intra-ventricular coronaries (Related to Figure 1)

3D image reconstruction of a 7 dpci border zone after tissue clearing and staining for coronaries (red) and CMs (green).

2

Supplemental Video 2. Intra-ventricular coronaries sprouting in the apex (Related to Figure 1)

3D image reconstruction of a 7 dpci injured area after tissue clearing and staining for coronaries (red) and CMs (green).

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Mouse monoclonal anti-MHC	DSHB	Cat#MF20; RRID:AB_2147781
Mouse monoclonal anti-PCNA	Santa Cruz	Cat#sc-56; RRID:AB_628110
Rabbit monoclonal anti-FLI1	Abcam	Cat#ab133485; RRID:AB_2722650
Rabbit polyclonal anti-Aldh1a2	Genetex	Cat#GTX124302; RRID:AB_11177627
Rabbit polyclonal anti-MEF2	Santa Cruz	Cat#sc-313; RRID:AB_631920
Chicken polyclonal anti-GFP	Aves	Cat#GFP-1010; RRID:AB_2307313
Rabbit polyclonal anti-DsRed	Takara	Cat# 632496; RRID:AB_10013483
Rabbit polyclonal anti-tRFP	Evrogen	Cat#AB233; RRID:AB_2571743
Alexa Fluor 568® phalloidin	Thermo Fisher Scientific	Cat# A12380; RRID:AB_2759224
Alexa Fluor® 488, Goat anti-Mouse IgG (H+L)	Thermo Fisher Scientific	Cat#A-11029; RRID:AB_138404
Alexa Fluor® 568, Goat anti-Mouse IgG (H+L)	Thermo Fisher Scientific	Cat#A-11004; RRID:AB_2534072
Alexa Fluor® 488, Goat anti-Rabbit IgG (H+L)	Thermo Fisher Scientific	Cat#A-11008; RRID:AB_143165
Alexa Fluor® 568, Goat anti-Rabbit IgG (H+L)	Thermo Fisher Scientific	Cat#A-11036; RRID:AB_143011
Alexa Fluor® 647, Goat anti-Rabbit IgG (H+L)	Thermo Fisher Scientific	Cat#A-21244; RRID:AB_141663
Alexa Fluor® 488, Goat anti-Chicken IgG (H+L)	Thermo Fisher Scientific	Cat#A-11039; RRID:AB_142924
Alexa Fluor® 647, Goat anti-Chicken IgG (H+L)	Thermo Fisher Scientific	Cat#A-21449; RRID:AB_2535866
Chemicals, Peptides, and Recombinant Proteins
Metronidazole	Sigma-Aldrich	Cat#M1547
DAPI	Sigma-Aldrich	Cat#D954
Alexa Fluor 568 phalloidin	Sigma-Aldrich	Cat#A12380
IT1t	Merk	Cat#1258011-83-4
DMOG	Sigma-Aldrich	Cat#D3695
Poly (I:C)	R&D Systems	Cat#4287
4-Hydroxytamoxifen	Sigma-Aldrich	Cat#H7904
Critical Commercial Assays
Hypoxyprobe™ Red549 Kit	Hypoxyprobe, Inc.	Cat#HP7-x
X-CLARITY Tissue Clearing System	Biozym Scientific	Cat# LGBED0001
Experimental Models: Organisms/Strains
Zebrafish: Tg(fli1a:EGFP)y1	Lawson et al. 2002	y1
Zebrafish: Tg(tcf21:DsRed2)pd37	Kikuchi et al. 2011	pd37
Zebrafish: TgBAC(tcf21:mCherry-NTR)pd108	Wang et al. 2015	pd108
Zebrafish: Tg(kdrl:mCherry)is5	Sacilotto et al. 2013	is5
Zebrafish: Tg(−0.8flt1:RFP)hu5333	Bussmann et al. 2010	hu5333
Zebrafish: TgBAC(etv2:EGFP)ci1	Proulx et al. 2010	ci1
Zebrafish: TgBAC(cxcr4a:YFP)mu104	Xu et al. 2014	mu104
Zebrafish: ET(krt4:EGFP)sqet33-1A	Poon et al. 2010	sqet33-1A
Zebrafish: TgBAC(cxcl12b:YFP)mu105	Bussmann et al. 2011	mu105
Zebrafish: Tg(−14.8gata4:GFP)ae1	Heicklen-Klein et al. 2004	ae1
Zebrafish: TgBAC(vegfaa:EGFP)pd260	Karra et al. 2018	pd260
Zebrafish: Tg(−3.5ubb:loxP-EGFP-loxP-mCherry)cz1701	Mosimann et al. 2011	cz1701
Zebrafish: TgBAC(apln:EGFP)bns157	Helker et al. in preparation	bns157
Zebrafish: TgBAC(apln:CreERT2)bns310	This paper	bns310
Zebrafish: Tg(hsp70l:loxp-Stop-loxp-mTom-codOptP2A-T46Avegfaa)bns288	This paper	bns288
Zebrafish: Tg(myl7:CreERT2,cryaa:DsRed)pd10	Kikuchi et al. 2010	pd10
Zebrafish: hif-1aabns89	Gerri et al. 2017	bns89
Zebrafish: hif-1abbns90	Gerri et al. 2017	bns90
Zebrafish: cxcr4aum20	Siekmann et al. 2009	um20
Zebrafish: flt1bns29	Matsuoka et al. 2016	bns29
Zebrafish: aplnmu267	Helker et al. 2015	mu267
Zebrafish: ppargc1abns176	This paper	bns176
Medaka: Tg(fli1a∷GFP)tg1206	Shinagawa-Kobayashi et al. 2018	tg1206
Oligonucleotides
qPCR	Table S2	N/A
Genotyping	Table S2	N/A
BAC generation	Table S2	N/A
Software and Algorithms
Zen 2012 (Blue Edition)	Carl Zeiss Microscopy	Version 1.1.2.0
Zen 2011 SP3 (Black Edition)	Carl Zeiss Microscopy	Version 8.1.0.484
Imarisx64	Bitplane	Version 9.3.0

Highlights.

• Regenerating coronaries respond by sprouting superficially and intra-ventricularly

• Superficial revascularization is regulated by epicardial Cxcl12/Cxcr4 signaling

• Intra-ventricular revascularization is regulated by endocardial Vegfa signaling

• Regenerating coronaries provide a scaffold available for cardiomyocyte repopulation