Decoding an organ regeneration switch by dissecting cardiac regeneration enhancers

Adult mammals permanently lose a substantial number of cardiomyocytes (CMs) after cardiac injury. This limited ability to regenerate lost CMs leads to heart failure and increased morbidity and mortality. By contrast, adult zebrafish possess a remarkable capacity to regenerate damaged hearts (Tzahor and Poss, 2017). In combination with available genetic tools, this ability makes zebrafish a powerful model system for deciphering the mechanisms underlying heart regeneration (Gonzalez-Rosa et al., 2017).

Heart regeneration is a complex process, in which all distinctive components of the cardiac tissue are appropriately coordinated. This elaborate and exquisite event is accomplished by orchestration of both CMs and non-cardiomyocytes. After cardiac injury, non-cardiomyocytes, including those in the epicardium and endocardium, sense injury cues and are rapidly activated to secrete paracrine factors that facilitate heart regeneration (Cao and Poss, 2018; Gonzalez-Rosa et al., 2017). For example, within a day of cardiac injury, the epicardium and endocardium robustly produce retinoic acid, which stimulates CM proliferation (Kikuchi et al., 2011). The endocardium of injured hearts induces vascular endothelial growth factor Aa (vegfaa), which mediates vasculogenesis, another crucial attribute for heart regeneration (Karra et al., 2018; Marin-Juez et al., 2019). Injury-responsive induction of such factors is required for heart regeneration, and it is crucial that their expression is transient and restricted to the site of injury to ensure proper regeneration. Spatiotemporal induction is important because continuous expression of these factors can have deleterious consequences, such as heart enlargement, cardiac failure and scarring (Gabisonia et al., 2019; Gemberling et al., 2015; Karra et al., 2018; Monroe et al., 2019). However, the mechanism by which spatiotemporal expression of injury-induced factors is achieved during heart regeneration remains a major gap in our knowledge.

Gene transcription in eukaryotes is a complex process that requires specific interactions between transcription factors (TFs) and cis-regulatory DNA elements. Among the cis-regulatory elements, enhancers are the key to direct spatiotemporal gene expression (Spitz and Furlong, 2012). The activities of enhancers are determined by the binding of TFs to their binding sites, sequence-specific motifs (Shlyueva et al., 2014). Recently, several groups have used regenerative systems, including cardiac tissue, to identify injury-responsive or regeneration-associated enhancers that direct gene expression in injured tissues (Goldman et al., 2017; Harris et al., 2016, 2020; Kang et al., 2016; Lee et al., 2020; Rodriguez and Kang, 2020; Soukup et al., 2019; Suzuki et al., 2019; Thompson et al., 2020; Vizcaya-Molina et al., 2018; Wang et al., 2020, 2019). Comparative analysis combined with transgenic assays in mice identified epicardial enhancers that direct developmental expression, as well as injury-responsive induction in the epicardium (Huang et al., 2012). Unbiased genome-wide analysis revealed multiple tissue regeneration enhancer elements (TREEs) that govern regeneration-dependent gene expression in adult zebrafish and killifish fins and hearts (Goldman et al., 2017; Kang et al., 2016; Thompson et al., 2020; Wang et al., 2020). Although the number of injury/regeneration enhancers, such as TREEs, has expanded, many questions remain regarding the control of their activities and whether multiple mechanisms or a predominant mechanism is responsible for conferring regeneration-induced transcription. Such knowledge is crucial to our understanding of how injury cues are transduced to establish regeneration programs.

Previously, we identified the zebrafish regeneration enhancer linked to leptin b (lepb) that directs regeneration-specific gene expression in the hearts (Kang et al., 2016). Here, we undertake a molecular dissection of lepb regeneration enhancer (LEN) to decipher how the regulatory elements in regeneration enhancers instruct regeneration-dependent gene expression in the heart. We discovered that a 22 bp DNA segment harbors a repressive sequence and that repression emerges during cardiac maturation. Thus, we propose a new model for regeneration-dependent transcription involving a two-component mechanism; one component mediates injury-responsive activation, and the other component participates in developmental stage-dependent repression. The integration of these components ensures spatiotemporal gene expression during heart regeneration.

We demonstrated previously that LEN is strongly activated during fin and heart regeneration (Kang et al., 2016). To determine the extent of regeneration-specific activation of LEN, LEN activation was compared in two different injury models: a whole fin amputation model, which causes massive loss of multiple tissues, such as bones, fibroblasts and epidermis, and a fin incision injury model that does not cause tissue loss. A small incision within an inter-ray region induced EGFP signal in LEN:EGFP, but the signal intensity was noticeably less robust than that observed after whole fin amputation (Fig. S1B). These results indicate that LEN is activated by injury and is a TREE.

Our previous study demonstrated that LEN contains two tissue-specific modules that function independently in the fin and heart (Kang et al., 2016). The fin LEN regeneration module (fLEN) is located in the central portion of LEN, whereas the cardiac LEN regeneration module (cLEN) is located proximally, in nucleotides 1034-1350 of LEN (Fig. 1A). Although injury-responsive activity of cLEN was examined using adult hearts in a previous study (Kang et al., 2016), the activity of the 317 bp cLEN fragment has not been studied during heart development and regeneration.

To determine the regeneration-dependent activity of cLEN, EGFP expression was determined in cLEN:EGFP reporter fish, which carry cLEN coupled to the lepb 2 kb minimal promoter (P2) and EGFP cassette, in uninjured and regenerating hearts during development. cLEN:EGFP fish were crossed with the cardiac myosin light chain 2 (cmlc2):mCherry-nitroreductase (NTR) strain, which enables genetic ablation of CMs (Chen et al., 2013; Curado et al., 2007). In this strain, the bacterial NTR gene is controlled by the cmlc2 regulatory element (Rottbauer et al., 2002), resulting in CM-specific NTR expression. The innocuous prodrug metronidazole (Mtz) is converted into a cytotoxic molecule by NTR in CMs of cmlc2:mCherry-NTR fish, thereby ablating a significant number of CMs. cLEN activity was examined in uninjured and injured hearts at 5 days post-fertilization (dpf), 7 weeks post-fertilization (wpf) and 4 months post-fertilization (mpf), representing larval, juvenile and adult stages, respectively. This assay demonstrated that EGFP is undetectable in uninjured hearts but is strongly induced upon ablation at all three stages (Fig. 1C-F), indicating that cLEN is not a developmental enhancer in the hearts and that injury signals are required for its activity. Cardiac ablation in larvae results in a pericardial edema phenotype, potentially inducing a systemic injury response owing to insufficient circulation and subsequently activating cLEN in the epidermis (Fig. S2A). By contrast, cardiac ablation in adults did not direct ectopic expression in the epidermis (Fig. S2A). Epidermal induction upon local injury was not observed in amputated adult caudal fins, suggesting that ectopic cLEN activation in the epidermis is repressed in adults (Fig. S1C). Mtz treatment did not induce EGFP in the absence of cmlc2:mCherry-NTR, confirming the specificity of cardiac injury-responsive activation of cLEN (Fig. S2B).

To explore cLEN activity further during adult heart regeneration, cLEN:EGFP expression was quantified throughout heart regeneration in adult fish. Ventricular apices of cLEN:EGFP fish were amputated, and hearts were collected at 3, 7, 14 and 42 days post-amputation (dpa), in addition to uninjured hearts (Fig. 1G). EGFP expression was examined by immunostaining cardiac sections with antibodies against EGFP, myosin heavy chain (MHC), a CM marker, and Raldh2, an endocardial and epicardial marker. EGFP was undetectable in uninjured hearts but was strongly induced at 3 dpa, mainly in endocardial cells, indicating that cardiac injury activates cLEN (Fig. 1H,I; Fig. S3A). EGFP intensity peaked at 7 dpa and remained high at 14 dpa, suggesting that cLEN activity was maintained during heart regeneration (Fig. 1H,I). However, EGFP expression was barely detectable at 42 dpa, when heart regeneration was completed (Fig. 1H,I). This dynamic cLEN activity during regeneration demonstrates that cLEN is a cardiac TREE.

Enhancer activity is determined by the binding of transcription-promoting TFs to their binding sites (Spitz and Furlong, 2012). Cardiac injury-responsive activation motifs are poorly characterized owing to the low number of regeneration enhancer studies (Rodriguez and Kang, 2020; Soukup et al., 2019); hence, we focused on motifs representing tissues or cells displaying enhancer activity. cLEN is active in the endocardium, the innermost layer of endothelial cells in the heart (Fig. S3A). Motif analysis of cLEN using JASPAR, a database of TF binding profiles (Khan et al., 2018), and searching for consensus binding sequence information revealed that cLEN contains predicted binding sites for TFs associated with endothelial/endocardial cells, such as nuclear factor of activated T-cells (NFAT), GATA, Forkhead (FOX) and ETS (De Val and Black, 2009; De Val et al., 2008; Meadows et al., 2011; Nemer and Nemer, 2002; Norrmén et al., 2009; Park et al., 2013; Zhou et al., 2005). There is one NFAT, five GATA, six FOX and two ETS sites within cLEN (Fig. 1B; Fig. S4).

In silico motif analysis predicts potential TF binding sites, but we probed the activity of potential cis-regulatory elements using transgenic enhancer assays (Kvon, 2015). An enhancer reporter system with various cLEN fragments was used to identify the essential cis-regulatory elements in cLEN that mediate injury-responsive gene activation. Initially, cLEN was fragmented into two parts: 144 bp of distal cLEN(1-144), named cLEN-dis, and 177 bp of proximal cLEN(141-317), named cLEN-pro. Transgenic reporter lines carrying cLEN-dis or cLEN-pro were unable to direct endocardial EGFP expression in 3 dpa hearts, indicating that these fragments are not sufficient for injury-responsive expression in the endocardium (Fig. 2A,B; Fig. S5).

Notably, cLEN(131-164) overlaps parts of cLEN-dis and cLEN-pro and contains binding motifs for FOX, NFAT and ETS within a single 34 bp region (Fig. S4), leading us to examine whether the FOX-NFAT-ETS (FNE) motifs are essential for endocardial expression. We generated a new transgenic line carrying a small (54 bp) fragment of cLEN(118-171), named cLEN-FNE, and examined their activity in hearts (Fig. 2A). Similar to cLEN-dis and cLEN-pro, cLEN-FNE was unable to direct endocardial EGFP expression in 3 dpa hearts, demonstrating that the FNE motifs in cLEN are not sufficient to drive regeneration-dependent endocardial expression (Fig. 2B; Fig. S5). Given that cLEN-dis, cLEN-pro and cLEN-FNE cover the full sequence of cLEN, these results suggest that at least two regulatory elements located in cLEN-dis and cLEN-pro are required for regeneration-dependent activation.

Activator protein 1 (AP-1) is a well-characterized injury-responsive factor in various contexts, including Schwann cells, fins and the heart (Alfonso-Jaume et al., 2006; Beisaw et al., 2020; Hung et al., 2015; Ishida et al., 2010; Shirali et al., 2018; Yates and Rayner, 2002), and cLEN contains two AP-1 sites (Fig. S4). A recent study also demonstrated that GATA and ETS motifs are crucial components of regeneration enhancers (Soukup et al., 2019). Thus, the involvement of AP-1, GATA and FNE motifs in regeneration enhancer activity was tested. The results indicated that cLEN(107-238), named cLEN-act, which contains AP-1, GATA and FNE motifs, was able to drive regeneration-dependent endocardial expression (Fig. 2B; Fig. S5). Developmental enhancers are known to be regulated by multiple regulatory elements (Spitz and Furlong, 2012); hence, these results indicate that cardiac regeneration enhancers are also regulated by multiple cis-regulatory elements.

To test the involvement of the AP-1 binding sites in injury-responsive activation, two AP-1 motifs in cLEN were mutated to generate cLEN^AP-1m, and their activity was examined in the F₀ mosaic injured hearts. Several constructs were generated, including P2:EGFP (minimal promoter, negative control), cLEN:EGFP (positive control) and cLEN^AP-1m:EGFP (Fig. 2C). These constructs were injected into the one-cell stage of cmlc2:mCherry-NTR embryos, and EGFP expression was examined before and after cardiac ablation (Fig. 2D). None of these constructs was able to direct EGFP expression in uninjured hearts. Upon cardiac injury, a significant number of larvae carrying the intact cLEN directed EGFP induction (Fig. 2E). However, the majority of larvae with P2 or cLEN^AP-1m were not able to direct injury-responsive expression (Fig. 2E), indicating that AP-1 binding sites are required for injury-responsive activity.

We next attempted to build a synthetic version of a cardiac regeneration enhancer by combining TF binding sites present in cLEN-act. Three copies of AP-1-FOX-NFAT-ETS-GATA sites were coupled with the P2 minimal promoter and EGFP sequences to produce AFNEGx3, and its activity was examined in the F₀ mosaic injured hearts. EGFP was undetectable in 3 dpf uninjured hearts. Noticeable EGFP induction was detected in some mosaic injured hearts carrying AFNEGx3 (Fig. S6D); however, the majority of larvae did not demonstrate injury-responsive EGFP induction (Fig. 2E). These data suggest that further dissection of cLEN is required to create a functional synthetic regeneration enhancer. In addition, enhancer assays with stable lines will be required to deduce the activity of AFNEGx3.

A recent study demonstrated that co-injection of mRNAs of the AP-1 complex components, junba and fosl1a, into one-cell-stage embryos is sufficient to activate the expression of AP-1 target genes (Beisaw et al., 2020). To determine whether AP-1 TFs can activate cLEN, mCherry, AP-1 (junba/fosl1a), nfatc1 or nfatc1+AP-1 mRNAs were injected into the one-cell stage of cLEN:EGFP heterozygote embryos, and EGFP expression was assayed at 1 day (Fig. S6A). Although mCherry or nfatc1 mRNA injection was unable to induce EGFP, we observed that a significant number of embryos injected with AP-1 or nfatc1+AP-1 demonstrated EGFP expression (Fig. S6C), suggesting that AP-1 contributes to the activation of cLEN. Overall, our data suggest that AP-1 is an important injury-activating factor for cLEN.

Although our data suggest that FNE motifs are not sufficient for regeneration-dependent activation, it is unclear whether FNE motifs are necessary. To examine this, we generated another transgenic line, which carries cLEN without a 47 bp sequence harboring the FNE motifs: cLEN(1-317Δ122-168), referred to as cLENΔ47 (Fig. 3A). Unexpectedly, although cLEN was inactive, cLENΔ47 exerted ectopic activity in the uninjured hearts (Fig. 3B; Fig. S7). Three independent lines of cLENΔ47 demonstrated endocardial EGFP expression in uninjured hearts. Although EGFP was undetectable in uninjured hearts of cLEN lines (0.001 and 0.016 µm²/100 µm² in line 3 and 7, respectively), EGFP was expressed in cLENΔ47 uninjured hearts (0.912 and 1.04 µm²/100 µm² in line 3 and 6, respectively) (Fig. 3C). This ectopic expression pattern suggests the presence of an inhibitory element in the deleted 47 bp sequence (cLEN-47).

Next, we explored how repression is lifted upon cardiac injury. To this end, we examined the EGFP expression patterns of cLEN and cLENΔ47 in regenerating hearts. In 3 dpa hearts, both cLEN and cLENΔ47 directed EGFP expression in the border zone (5.66, 10.95, 12.99 and 15.62 µm²/100 µm² in cLEN line 3 and 7 and cLENΔ47 line 3 and 6, respectively), demonstrating their activation abilities in response to injury (Fig. 3B,D). The EGFP⁺ cells were colocalized with Raldh2⁺ cells, indicating that cLEN and cLENΔ47 were active mainly in the endocardial cells at the wound (Fig. S3A). By contrast, cLEN and cLENΔ47 displayed distinct enhancer activities in the region distant from the wound area (the remote zone). Although the EGFP⁺ area was very limited in the remote zone of cLEN (0.35 and 0.55 µm²/100 µm² in line 3 and 7, respectively), regenerating cLENΔ47 hearts demonstrated EGFP expression in a significant endocardial area of the remote zone (4.94 and 5.43 µm²/100 µm² in line 3 and 6, respectively) (Fig. 3E; Fig S3A,B; Fig. S8A-C). These results indicate that the inhibitory element in cLEN-47 is deactivated in the border zone but remains functional in the remote zone. cLEN activity returned to a level similar to that in the uninjured hearts at 42 dpa (Fig. 1H,I; Fig. S8D); however, significant EGFP expression was detected in the cLENΔ47 hearts at 42 dpa (Fig. S8D,E), suggesting that the repressive elements are functional and restrict TREE activity after the completion of regeneration. Our results provide evidence that cardiac regeneration enhancers are actively repressed in uninjured and regenerating hearts to prevent aberrant gene expression, revealing intricate modulation of regeneration enhancers in the heart.

To exclude the possibility that the 47 bp deletion places two activation elements close to each other or creates a de novo activation element, thereby aberrantly stimulating regeneration enhancers in the absence of injury, we created a series of five new 11 bp deletion lines, named cLENΔ11-1 to cLENΔ11-5, which together covered the deleted cLEN-47 fragment (Fig. 4A). Notably, our analyses with uninjured hearts demonstrated that cLENΔ11-1 and -2, but not cLENΔ11-3 to -5, drive ectopic endocardial expression in uninjured hearts (Fig. 4B). The EGFP⁺ cardiac tissues were 3.28 and 1.04 µm²/100 µm² in uninjured hearts of cLENΔ11-1 and -2, respectively. By contrast, EGFP expression was limited in uninjured hearts of cLENΔ11-3, -4, and -5 (0.03, 0.08 and 0.01 µm²/100 µm², respectively; Fig. 4C). In 3 dpa regenerating adult hearts, all five constructs were capable of directing injury-responsive gene expression in the border zone (16.77, 15.86, 20.98, 5.12 and 6.96 µm²/100 µm² for cLENΔ11-1 to -5, respectively; Fig. 4B,D), indicating that injury-responsive activity is intact regardless of these 11 bp deletions. cLENΔ11-4 and -5 had smaller areas of EGFP-expressing cells, compared with those in other cLENΔ11 lines, suggesting that deletion of the potential activation motifs, such as ETS, may influence injury-responsive activity in the border zone. In the remote zone, EGFP expression was limited in cLENΔ11-3 (0.69 µm²/100 µm²) and was barely detectable in cLENΔ11-4 and -5, (0.10 and 0.29 µm²/100 µm² for cLENΔ11-4 and -5, respectively), highlighting regional activation of the cardiac regeneration enhancer in the border zone. Although the EGFP⁺ area was larger in cLENΔ11-1 versus cLENΔ11-2 (3.96 versus 1.24 µm²/100 µm², respectively), a significant number of endocardial cells in the remote zone expressed EGFP in cLENΔ11-1 and cLENΔ11-2 (Fig. 4E). Similar to cLENΔ47, cLENΔ11-1 and -2 exhibited significant GFP expression at 42 dpa (Fig. S8D). These results confirm that ectopic activation is caused by deletion of a repressive element rather than altered spacing of chromatin structure. This repressive element resides within the 22 bp region of cLEN(121-142) or cLEN-22.

To identify an essential repressive element in cLEN, we explored orthologous cLEN regions in other fish species. Similar to zebrafish, the genomes of 15 other teleost fish models contain multiple leptin genes. These species include the golden-line barbel (Sinocyclocheilus grahami), horned golden-line barbel (Sinocyclocheilus rhinocerous), blind barbel (Sinocyclocheilus anshuiensis), common carp (Cyprinus carpio), goldfish (Carassius auratus), Mexican tetra (Astyanax mexicanus), red-bellied piranha (Pygocentrus nattereri), electric eel (Electrophorus electricus), Atlantic herring (Clupea harengus), denticle herring (Denticeps clupeoides) and Japanese medaka (Oryzias latipes). Zebrafish, Mexican tetra, red-bellied piranha, electric eel, Atlantic herring, denticle herring, and Japanese medaka possess two leptin paralogs: lepa and lepb (Gorissen et al., 2009). The lineage containing the barbel, carp and goldfish species underwent an additional round of genome duplication, resulting in four leptin genes: lepa1, lepa2, lepb1 and lepb2 (Gorissen and Flik, 2014; Huising et al., 2006). Using reference genomes from the Ensembl database, lepb upstream sequences from these fish species were compared with the zebrafish sequence via limited area global alignment of nucleotides (LAGAN) analysis (Brudno et al., 2007). This sequence comparison detected limited conservation or lack of conservation of cLEN upstream of lepb across these species (Fig. S9A,B). However, manual comparison revealed that cLEN-act is moderately conserved in zebrafish, barbel, carp and goldfish, all of which belong to the Cyprinidae family. An AP-1 binding site within cLEN-act appears to be conserved across the Cyprinidae species, whereas cLEN-22 containing a repressive element displays sequence diversity (Fig. S9C).

Owing to limited or undetectable conservation of cLEN in other teleost species, cLEN conservation was assessed in closely related species. Orthologous cLEN was evaluated within the Danio genus, to which Danio rerio (zebrafish) belongs. To compare the sequence, we extracted genomic DNA and isolated the cLEN equivalent fragments from Danio aesculapii, Danio kyathit and Danio albolineatus. Sequencing of cLEN identified that the AP-1 motif is present in the cLEN-act region (Fig. S9C). RT-qPCR or RT-PCR analysis in D. aesculapii and D. kyathit demonstrated that lepb is induced in response to cardiac injury, suggesting that the injury-responsive activity of cLEN is conserved in other Danio species (Fig. 5D,E). We next aimed to identify the repressive element in cLEN by comparison of cLEN-22. Intriguingly, counterparts of cLEN-22 displayed 6 or 9 bp DNA insertions (Fig. 5A). This observation raises a question of whether the functionality of the repressive element might be disrupted by these insertions. To assess the activity of cLEN-22 variants from other Danio species, a 6 bp insertion was introduced into the zebrafish cLEN to generate cLEN^ins6bp:EGFP transgenic fish. Two independent lines of cLEN^ins6bp:EGFP were able to induce EGFP expression in response to cardiac injury, indicating that a 6 bp insertion does not affect injury-responsive cLEN function (Fig. 5B). Unlike cLENΔ47, cLENΔ11-1 and cLENΔ11-2, cLEN^ins6bp:EGFP transgenic fish were unable to direct ectopic endocardial expression in uninjured hearts (Fig. 5B), and lepb transcript levels were not significantly increased in uninjured hearts of other Danio species (Fig. 5F). These results demonstrate that this 6 bp insertion does not interfere with the function of the repressive element.

Given that deletion of either cLEN11-1 or cLEN11-2 impedes repression, we focused on a sequence spanning these two fragments. The juxtaposition of cLEN11-1 and cLEN11-2 has a GT/CA sequence that is conserved in other Danio species and some barbel, carp and goldfish lepb paralogs (Fig. 5A; Fig. S9C). Our motif analysis of cLEN-22 using the JASPAR database predicted binding sites for two repressors: growth factor independent 1B transcription repressor (gfi1b) and PR domain containing 1a, with ZNF domain (prdm1a). gfi1b is the key repressor controlling hematopoiesis (Dahl et al., 2007; Li et al., 2010; Saleque et al., 2007). prdm1a acts as a repressor in the intestine and immune cells (Harper et al., 2011; Hohenauer and Moore, 2012; Kallies et al., 2006; Muncan et al., 2011). To test whether their binding sites are responsible for the repression, the JASPAR database was used to design a mutation that disrupts the Prdm1a and Gfi1b binding sites and is not predicted to create new binding sites or disturb spacing between other candidate sites. Based on this analysis, the GTCA sequence was mutated to CATT. To test whether the GTCA sequence comprises the crucial repressive motif, new transgenic fish with the GTCA to CATT mutation were generated, named cLENm^gtca. In 3 dpa hearts, two independent lines of cLENm^gtca:EGFP were able to drive injury-responsive expression in the border zone (Fig. 5C). By contrast, endocardial EGFP expression was limited in the remote zone of 3 dpa hearts (Fig. 5C). We next examined EGFP expression in uninjured hearts. Interestingly, cLENm^gtca directed ectopic EGFP expression in some endocardial cells in uninjured hearts (Fig. 5C). Although the GTCA mutation does not completely recapitulate the activity of cLEN deletion constructs in uninjured hearts, these results indicate that GTCA is a part of an essential repressive element in cLEN.

During development, the heart matures to form highly organized sarcomere structures (Fukuda et al., 2019; Huang et al., 2009). To determine when repression begins to function, EGFP expression in the cLEN and cLENΔ47 reporter lines was evaluated. First, we confirmed that CM ablation injuries drive EGFP induction in the cLEN, cLENΔ47 and cLENΔ11-1 to -5 reporter lines at the 5 dpf larval stage, indicating that activation elements respond to injury from early developmental stages (Fig. S10).

To examine when the repression mechanism begins to be operational during development, cardiac section samples were collected from cLEN:EGFP;cmlc2:mCherry and cLENΔ47:EGFP;cmlc2:mCherry at 5, 10, 14 and 35 dpf, and EGFP signals were amplified by immunostaining. EGFP was undetectable in 5 dpf hearts of cLEN and cLENΔ47 (n=17 and 13 for cLEN and cLENΔ47, respectively). Although EGFP expression was not observed in the cLEN hearts at later time points (n=17 and 20 at 10 and 14 dpf, respectively), EGFP in the cLENΔ47 hearts was visible at 10 dpf in 12 of 12 examined hearts and at 14 dpf in 14 of 14 examined hearts. At 35 dpf, uninjured hearts of all cLENΔ47 animals showed EGFP signals (n=6) (Fig. 6). A recent study demonstrated that trabecular CMs undergo significant morphological changes, including an increase in myofilament thickness, by 5 dpf, representing cardiac maturation (Fukuda et al., 2019). Given that cLENΔ47 becomes active in uninjured hearts after 5 dpf, our data indicate that repression is not operational in uninjured hearts during embryogenesis but emerges during cardiac maturation.

Tissue regeneration is accomplished by dynamic changes in gene transcription. Recently, multiple studies have demonstrated that regeneration-dependent or injury-responsive gene expression relies on the activities of enhancers, such as TREEs (Goldman et al., 2017; Harris et al., 2020; Huang et al., 2012; Kang et al., 2016; Lee et al., 2020; Rodriguez and Kang, 2020; Thompson et al., 2020; Vizcaya-Molina et al., 2018; Wang et al., 2020); yet underlying enhancer mechanisms have remained elusive. In the present study, we carried out a molecular dissection of the cardiac regeneration enhancer cLEN in zebrafish to elucidate mechanisms controlling regeneration enhancer activity. Regeneration-dependent activity of cLEN is detected mainly in the endocardium, which is the first cardiac tissue that responds to injury. An interesting feature of regenerative events in the endocardium is rapid organ-wide regenerative factor induction. This organ-wide gene activation is promptly restricted to the wound area within a day, and this restriction might be necessary to prevent ectopic regeneration outcomes in the remote zone, a region distant from the wound area where regeneration programs are inactive (Fang et al., 2013; Gamba et al., 2014; Kikuchi et al., 2011; Lepilina et al., 2006). Our results indicate that the repressive code embedded in regeneration enhancers, adjacent to the activation motifs, might contribute to this rapid deactivation. Moreover, we demonstrate that repression ensures a lack of regenerative factor induction in uninjured hearts. Thus, we provide evidence that regenerative tissues require repression to prevent aberrant gene expression. Our studies provide evidence that coordination of activation and repressive motifs confers regeneration enhancer function to govern spatiotemporal gene expression, which supports heart regeneration (Fig. 7).

Our results suggest that regulatory elements associated with heart regeneration are subject to repression during cardiac maturation. Similar observations were also reported in the mouse heart. The levels of insulin-like growth factor 2 mRNA binding protein 3 (Igf2bp3), a positive regulator of CM proliferation, are lower in postnatal day 8 matured hearts versus postnatal day 1 immature hearts (Wang et al., 2019). Interestingly, a repressive histone mark and DNA methylation are progressively enriched in the Igf2bp3 genomic locus during postnatal development, suggesting an emergence of a repressive component during cardiac maturation (Sim et al., 2015; Wang et al., 2019). One approach to promote cardiac regeneration is to revert cardiac maturity from mature to immature states, thereby increasing proliferative ability. Our findings reveal two strategies to achieve this goal, including attenuation of repression and promotion of activation. Triggering activation has been widely attempted, and this approach improves heart regeneration (Cui et al., 2020; Monroe et al., 2019; Tao et al., 2016). However, the impact of mitigating repression on heart regeneration has been little explored. Polycomb repressive complex 2 (PRC2) is the major repressive complex that catalyzes H3K27 trimethylation (H3K27me3) (Cao et al., 2002; Margueron et al., 2009). In mouse hearts, loss of Ezh1, a catalytic subunit of PRC2, results in reduced cardiac regeneration, suggesting a positive role of general repressive function in heart regeneration (Ai et al., 2017). In zebrafish hearts, global inhibition of the repressive histone mark H3K27me3 compromises the initiation of heart regeneration owing to the failure of structural gene silencing in the border zone (Ben-Yair et al., 2019). These studies suggest that attenuation of repression might be unable to promote heart regeneration (Ai et al., 2017; Ben-Yair et al., 2019). However, this perturbation might dysregulate vast numbers of genes, which complicates mechanistic interpretations. Further studies are needed to characterize specific repressors regulating regeneration enhancers and to explore whether attenuating repression helps in priming regeneration enhancers and, subsequently, improves heart regeneration.

The impact of repression in the uninjured tissues is a fundamental problem. Previous studies demonstrated that repression plays certain roles in supporting terminal differentiation and safeguarding proper tissue homeostasis (Ma et al., 2015; Sun et al., 2016). Schwann cells and liver constitute regenerative tissues in mammals. Loss of PRC2 in Schwann cells results in upregulation of injury-responsive genes, implying that repression is required for inhibition of aberrant gene expression in the absence of injury. Importantly, Schwann cells lacking PRC2 function display abnormalities of myelin, including morphological changes and progressive hypermyelination (Ma et al., 2015), emphasizing the impact of repression on the function of uninjured tissues. In the mammalian liver, Arid1a, a key component of the ATP-dependent chromatin-remodeling complex, plays repressive roles, which contribute to differentiation of the liver after birth (Sun et al., 2016). Although the mutation of lepb, a target gene of LEN, is unlikely to influence regeneration ability in zebrafish (Kang et al., 2016), it will be interesting to determine whether disruption of the endogenous repressive motif in cLEN yields ectopic lepb expression in uninjured hearts and, subsequently, influences endocardial function.

Our in-depth investigation of the cardiac regeneration enhancer at the molecular level provides evidence for a new model of regeneration enhancer function. In addition, studies in Drosophila imaginal disc demonstrated that the dual-component mechanism in regeneration enhancers influences regenerative capacity (Harris et al., 2016, 2020). It would be interesting to determine whether this dual-component mechanism can be extrapolated to regeneration enhancers operating in diverse cells and tissues as a general phenomenon. The results of this study can be adapted to other contexts to understand the mechanism of regeneration. A comprehensive understanding of transcriptional regulation at the level of regeneration enhancers will allow the construction of the gene regulatory network responsible for tissue regeneration.

Wild-type or transgenic male and female zebrafish of the outbred Ekkwill (EK) strain ranging up to 12 months of age were used for all zebrafish experiments. Wild types of Danio aesculapii, Danio kyathit and Danio albolineatus from 6 to 18 months of age were used for the experiments. The water temperature was maintained at 28°C for animals unless otherwise indicated. Partial ventricular resection surgery was performed as described previously (Poss et al., 2002), in which ∼20% of the cardiac ventricle was removed at the apex. For expression patterns to determine enhancer activity, at least four hearts, unless indicated, were examined per line. To ablate CMs, cmlc2:mCherry-NTR fish were used (Chen et al., 2013). To ablate CMs of larvae, juveniles and adults, 3 dpf, 7 wpf and 4 mpf cmlc2:mCherry-NTR zebrafish were incubated in 10, 10 and 5 mM metronidazole for 18, 24 and 24 h, respectively. Work with zebrafish and other Danio fish species was performed in accordance with University of Wisconsin-Madison guidelines.

To define the activity of various cLEN fragments, 26 additional new transgenic lines (listed in Table S2) were established in this study. I-SceI meganuclease transgenesis was used to generate transgenic animals. DNA sequences were amplified by PCR with the indicated primers (Table S1) and subcloned into the pCS2-P2:EGFP-I-SceI vector, in which I-SceI restriction sites were flanked by a multiple cloning site. The 2 kb upstream sequences of lepb (P2) were used as the minimal promoter. The In-Fusion Cloning technique was used for deletion and insertion of cLEN with the corresponding primers (Table S1). The rationale behind the deletion of 11 bp is to remove approximately a single helical turn of the DNA (Bond et al., 2010), potentially allowing for maintenance of the integrity of the chromatin after the deletion. The Q5 site-directed mutagenesis kit (NEB) was used to mutate GTCA to CATT. Both the In-Fusion Cloning technique and the Q5 site-directed mutagenesis kit were used to mutate the AP-1 binding sites. The AFNEGx3 synthetic enhancer fragment was generated by annealing two complementary single stranded oligonucleotides. These constructs were injected into one-cell-stage wild-type embryos using standard meganuclease transgenesis techniques. To isolate the stable lines, larvae were examined for transgene expression near the injury site in response to fin fold amputation. P2 induced transgene expression after fin fold amputation at larval stages, but there was no induction after caudal fin amputation in adults (Fig. S1A; Kang et al., 2016).

To generate mRNAs, the junba, fosl1a, nfatc1 and mCherry open reading frames were cloned downstream of the SP6 promoter in the pCS2-I-SceI vector. mRNA was synthesized using an SP6 mMESSAGE mMachine kit (Ambion) according to the manufacturer's instructions. A total of 60 pg of mRNA (30 pg each of junba and fosl1a for AP-1 and 20 pg each of junba, fosl1a and nfatc1 in AP-1+nfatc1 co-injection experiments) was injected into wild-type one-cell-stage cLEN:EGFP heterozygote embryos. Embryos were collected 1 day post-injection, and gene expression analysis was performed by imaging EGFP and mCherry signals and by RT-qPCR for EGFP. Four or five embryos per sample were used to extract the total RNA. Genomic DNA (gDNA) from D. aesculapii, D. kyathit and D. albolineatus was obtained by collecting fins and purifying the genomic DNA using a Gentra Puregene tissue kit (QIAGEN, 158667). cLEN orthologous regions from each species were amplified by PCR with gDNA using multiple combinations of forward and reverse primers targeting zebrafish cLEN. PCR was performed using Thermotaq DNA Polymerase (NEB, M0267) or Phusion high-fidelity DNA polymerase (NEB, M0530). Forward primers B1, B7, L233 and L246 and reverse primers B8, L193, L231 and L239 successfully amplified cLEN in all species (Table S1). Using primer pairs B1-B8, B7-B8, L233-L231 and L246-L239, all or part of cLEN in each species was amplified via PCR, and the amplicons were subcloned into pJET1.2/blunt cloning vectors using a CloneJET PCR cloning kit (Thermo Scientific, K1231). Plasmids were sequenced via Sanger sequencing using the primer targeting the T7 promoter (5′-TAATACGACTCACTATAGGG-3′) at Functional Biosciences, Inc. (Madison, WI, USA). Sequencing results were analyzed and aligned to zebrafish cLEN using SnapGene (GSL Biotech LLC).

Hearts were fixed with 4% paraformaldehyde overnight at 4°C or for 1 h at room temperature. Cryosectioning and immunohistochemistry were performed as described previously, with modifications (Kang et al., 2016). Hearts were cryosectioned at 10 µm thickness. Heart sections were equally distributed onto four or five serial slides such that each slide contained sections representing all areas of the ventricle. A solution comprising 5% goat serum, 1% bovine serum albumin, 1% dimethyl sulfoxide and 0.1% Tween-20 was used for blocking and antibody staining. The primary and secondary antibodies used in this study were as follows: anti-myosin heavy chain (mouse; F59; Developmental Studies Hybridoma Bank; 1:50), anti-EGFP (rabbit; A11122; Life Technologies; 1:200), anti-EGFP (chicken; GFP-1020; Aves Labs; 1:2000), anti-Ds-Red (rabbit; 632496; Clontech; 1:500), anti-Raldh2 (rabbit; GTX124302; Genetex; 1:200), Alexa Fluor 488 (mouse, rabbit and chicken; A11029, A11034 and A11039; Life Technologies; 1:500) and Alexa Fluor 594 (mouse and rabbit; A11032 and A11037; Life Technologies; 1:500). Whole-mount larval images were acquired using an AxioZoom stereofluorescence microscope (Zeiss) at the time points indicated. Images of cardiac tissue sections were acquired using an LSM 700 confocal microscope (Zeiss), Eclipse Ti-U inverted compound microscope (Nikon) or BZ-X810 fluorescence microscope (Keyence). Images were processed using ZEN (Zeiss), BZ-X800 analyzer (Keyence) or FIJI/ImageJ software. Image stitching was automatically processed using a ZEN or BZ-X800 analyzer. Further image processing was carried out manually using Photoshop or FIJI/ImageJ software.

Quantification of the EGFP intensity and EGFP⁺ area was performed using the FIJI/ImageJ software. One to five sections were used to determine the values in one cardiac sample. For quantification of the EGFP intensity (Fig. 1H; Fig. S8E), images of sections of wounded or uninjured apical regions were converted to 8 bit and adjusted with the threshold, which was determined by the autofluorescence level of CMs. Red channel (MHC) images were used to guide manual selection of heart outlines, which were saved to the region of interest (ROI) manager, after which values of integrated density in ROIs were measured. The EGFP intensity was calculated by normalizing to the intensity in wild-type uninjured heart samples. To quantify the EGFP intensity in Raldh2⁺ cells (Fig. S3C), the red channel (Raldh2) images were used to define ROIs, and the values of the integrated density in ROIs were calculated. The averaged EGFP intensity level of the border zone was divided by that of the remote zone in the same heart. To measure the EGFP⁺ area of uninjured hearts, ROIs were determined by manual selection of heart outlines, with guidance provided by the red channel (MHC) images. A remote zone was selected as a 155 µm×450 µm rectangular section at least 250 µm away from the border zone. To quantify the EGFP⁺ area, selected ROIs were converted to 8 bit, adjusted with the threshold, which was determined by the autofluorescence level of CMs, and the EGFP⁺ areas in ROIs were evaluated.

Zebrafish genomic sequences from dre-mir-129-1 through lepb were aligned via the VISTA alignment tool (http://genome.lbl.gov/vista/index.shtml) against equivalent sequences in golden-line barbel (Sinocyclocheilus grahami), horned golden-line barbel (Sinocyclocheilus rhinocerous), blind barbel (Sinocyclocheilus anshuiensis), German mirror carp (Cyprinus carpio carpio), Songpu mirror carp (Cyprinus carpio carpio), Hebao red carp (Cyprinus carpio wuyuanensis), Yellow River carp (Cyprinus carpio haematopterus), goldfish (Carassius auratus), Mexican tetra (Astyanax mexicanus), Pachon cavefish (Astyanax mexicanus), red-bellied piranha (Pygocentrus nattereri), electric eel (Electrophorus electricus), Atlantic herring (Clupea harengus), denticle herring (Denticeps clupeoides) and Japanese medaka (Oryzias latipes). The genomic location of each sequence is listed in Table S4. VISTA alignment results for dre-mir-129-1 though lepb, cLEN, cLEN-act and cLEN-22 were visualized using the VISTA-Point visualization tool.

RNA was isolated from uninjured and partly resected hearts using Tri-Reagent (ThermoFisher). Complementary DNA (cDNA) was synthesized from 300 ng to 1 μg of total RNA using a NEB ProtoScript II first strand cDNA synthesis kit (NEB, E6560). To design lepb qPCR primers for other Danio species, we initially ran a PCR on gDNA of each species using multiple combinations of primers targeting zebrafish lepb. L71 (forward primer, targets upstream of lepb exon 1) and L6 (reverse primer, targets lepb intron 1) successfully amplified the first exon of lepb in each species (Table S1). The amplicons were purified, subcloned into the pJET1.2/blunt cloning vectors, sequenced and analyzed as described above. Based on conserved sequences across all species, we designed the forward primer B27, which targets exon 1 (Table S1). To design a reverse qPCR primer targeting lepb exon 2, we first compared lepb exon 2 sequences of zebrafish, Pachon cavefish (Astyanax mexicanus), red-bellied piranha (Pygocentrus nattereri) and Japanese medaka (Oryzias latipes). Based on these analyses, we designed the forward (B23 and B24) and reverse primers (B25 and B6) with degenerated nucleotides at locations that differed across species (Table S1). Each of these degenerated primers amplified lepb fragments from gDNA of other Danio species. We then amplified lepb exon 2 from gDNA in each species (Table S1), and the amplicons were purified, subcloned into pJET1.2/blunt cloning vectors, sequenced and analyzed as described above. Based on conserved sequences across Danio species, the reverse primer B28 was designed to target exon 2. We confirmed that the B27 and B28 primers successfully amplified the lepb cDNA fragment. Quantitative PCR was performed using the qPCRBIO SyGreen Blue Mix Separate-ROX (Genesee Scientific, 17-507) and a Bio-Rad CFX Connect system. All samples were analyzed in biological triplicate with two technical repeats. The sequences of the primers used are listed in Table S1. lepb transcript levels were normalized to actb2 levels in all experiments.

We thank the University of Wisconsin-Madison School of Medicine and Public Health (SMPH) BRMS (Biomedical Research Models Services) staff, Lindsey Tushman and Jonah Mudge, for zebrafish care; Jonah Mudge and Daniel F. McGarry for assistance with sample preparation; and Jingli Cao, Chen-Hui Chen, Deneen M. Wellik and Emery H. Bresnick for comments on the manuscript. We also thank Kenneth Poss for sharing fish lines and for his generous support.

The peer review history is available online at https://dev.biologists.org/lookup/doi/10.1242/dev.194019.reviewer-comments.pdf