The Hippo pathway effector Wwtr1 regulates cardiac wall maturation in zebrafish

Cardiac trabeculation is an important step to cardiac wall maturation and is key to the growth and function of the heart (Ozcelik et al., 2002; García-Rivello et al., 2005; Liu et al., 2010; Staudt and Stainier, 2012). During cardiac development, both the endocardial and myocardial layers participate in a series of signaling pathways that lead to the formation of a trabeculated ventricle. Cardiac trabeculation requires the neuregulin signaling pathway, as mutants (mouse, Nrg1 and Erbb2/4; zebrafish, nrg2a and erbb2) affected in this pathway lack trabecular structures (Gassmann et al., 1995; Lee et al., 1995; Meyer and Birchmeier, 1995; Liu et al., 2010; Rasouli and Stainier, 2017). Neuregulin ligands, which are expressed by endocardial cells (Meyer and Birchmeier, 1995; Grego-Bessa et al., 2007; Rasouli and Stainier, 2017), stimulate cardiomyocytes through ERBB2/4 receptors (Gassmann et al., 1995; Lee et al., 1995; Liu et al., 2010) to delaminate from the compact layer (Staudt et al., 2014). Nrg/Erbb2 signaling also drives cardiomyocyte proliferation (Bersell et al., 2009; Xin et al., 2013a; D'Uva et al., 2015). In mouse, the expression of neuregulin is modulated by Notch activity, as conditional deletion of Notch1, or its co-factor RBPJk, in the endothelium results in downregulation of Nrg1 expression (Grego-Bessa et al., 2007). In zebrafish, mutants/morphants with defective cardiac contractility exhibit hearts with reduced nrg2a expression (Rasouli and Stainier, 2017), which explains why tnnt2a (silent heart/sih) mutant hearts lack trabeculae (Chi et al., 2008; Samsa et al., 2015). However, many questions remain regarding how neuregulin signaling affects cardiomyocyte behavior during delamination and cardiac trabeculation.

YAP1 and WWTR1 (also known as TAZ) are nuclear effectors of the Hippo signaling pathway, and their nuclear localization and association with TEADs promote cell proliferation and survival (Johnson and Halder, 2014; Yu et al., 2015). The Hippo kinase cassette is a negative regulator of YAP1/WWTR1 activity (Wu et al., 2003; Huang et al., 2005; Zhao et al., 2010), and constitutive activity of YAP1/WWTR1 can lead to uncontrolled growth and/or tumorigenesis (Lu et al., 2010; Cox et al., 2016). The Hippo signaling pathway has been a topic of intense investigation in recent years, unveiling new discoveries for YAP1/WWTR1 function, such as modulating cell behavior (Kim et al., 2017; Kimelman et al., 2017) and metabolism (Cox et al., 2016; Kim et al., 2017), during development. The Hippo signaling pathway is also known to interact with other signaling pathways that participate in cardiac trabeculation. For example, YAP1 has been reported to interact with the intracellular domain of ERBB4, and the complex is shuttled into the nucleus by neuregulin stimulation (Komuro et al., 2003; Haskins et al., 2014). YAP1 also interacts with the Notch intracellular domain (NICD) (Manderfield et al., 2015), and recent studies have further shown that YAP1/WWTR1 can drive the expression of Notch ligands (Totaro et al., 2017), including JAG2 in chick skeletal muscle (Esteves de Lima et al., 2016). Additionally, the nuclear localization of YAP1 is regulated by mechanical forces, including skeletal muscle contraction (Esteves de Lima et al., 2016) and blood flow (Nakajima et al., 2017). However, previous studies investigating the Hippo signaling pathway and YAP1 in heart development have mostly focused on myocardial growth (Heallen et al., 2011; Xin et al., 2011; von Gise et al., 2012) and regeneration (Xin et al., 2013b; Heallen et al., 2013). Furthermore, the function of WWTR1, a vertebrate paralog of YAP1 (Hilman and Gat, 2011), in heart development has not been studied, leaving many questions about the Hippo signaling pathway during heart development unanswered. Thus, we used the zebrafish model in this study to investigate the role of Wwtr1 in cardiac development.

In the developing zebrafish heart, we find that Wwtr1 is predominantly localized to the nuclei of compact wall cardiomyocytes and, to a lesser extent to those of trabecular layer cardiomyocytes. In agreement with these observations, wwtr1^−/− cardiomyocytes in mosaic wild-type hearts prefer to populate the trabecular layer. However, we also find that wwtr1^−/− hearts exhibit reduced trabeculation. Investigating this paradox led to novel insights into the function of the Hippo signaling pathway, and specifically Wwtr1, during vertebrate heart development, especially in cardiac wall maturation.

We first investigated the spatiotemporal expression pattern of Wwtr1 in the developing zebrafish heart by immunostaining. An antibody against human WWTR1 (which cross-reacts with human YAP1) has been shown, using zebrafish wwtr1 mutant embryos as negative controls, to recognize zebrafish Wwtr1 in a number of tissues, including the periocular mesenchyme (Miesfeld et al., 2015), ectoderm and notochord (Kimelman et al., 2017) and, in this study, the heart (Fig. S1A). During embryonic and early larval stages, i.e. from 30 to 119 hours post fertilization (hpf), we observed that Wwtr1 immunostaining was predominantly localized to the nuclei of ventricular cardiomyocytes (Fig. 1). During the onset of trabeculation from 60 to 72 hpf, nuclei of delaminating or trabecular cardiomyocytes exhibited weaker Wwtr1 immunostaining than those of adjacent compact layer cardiomyocytes (cross-sections of Fig. 1B, and Fig. S1B).

To study the function of Wwtr1 during zebrafish heart development, we generated a wwtr1 mutant using the CRISPR/CAS9 technology and identified an allele with a 29 bp insertion in exon 2 (wwtr1^bns35). This out-of-frame insertion is predicted to cause the production of a truncated protein (p.Pro145Glnfs*15) (Fig. 2A). The mutant transcript was also less abundant than the wild-type transcript, as assessed by real-time PCR of whole embryos (data not shown) and RNA sequencing (RNAseq) of embryonic hearts (Table S1), suggesting nonsense-mediated decay. At 96 hpf, the luminal surface of wwtr1^−/− hearts was mostly smooth, appearing to lack the trabecular structures that line the outer-curvature of the wild-type ventricle (Fig. 2B). These trabecular structures are observed as ‘finger-like’ multicellular protrusions in a mid-sagittal confocal plane of wild-type animals (Fig. 2C). In contrast, the wwtr1^−/− ventricular wall, although composed of multiple layers of cardiomyocytes, did not exhibit ‘finger-like’ protrusions (Fig. 2C). To quantify ventricular cardiomyocytes, we first determined their total number and then separated them into two groups: the outermost layer cardiomyocytes (designated as compact wall cardiomyocytes); and the innermost layer cardiomyocytes (designated as trabecular layer cardiomyocytes). The number of cardiomyocytes in the trabecular layer of wwtr1^−/− hearts was significantly reduced at all stages examined (Fig. 2D), probably as a result of reduced delamination as well as reduced proliferation in the trabecular layer, as assessed by EdU incorporation (Fig. S2).

We have previously shown that inhibiting the neuregulin signaling pathway results in fewer cardiomyocytes undergoing apical constriction at 82-100 hpf (Jiménez-Amilburu et al., 2016), consistent with the lack of trabeculae in nrg2a/erbb2 mutant hearts. In contrast, we did not observe a significant change in the number of cardiomyocytes undergoing apical constriction in wwtr1^−/− hearts at 79 hpf when compared with wild type (Fig. S3A,B). Once cardiomyocytes undergo apical constriction and enter the trabecular layer, they are usually depolarized, as assessed by the localization of an EGFP-Podxl transgene; some cardiomyocytes may repolarize, but will eventually depolarize again (Jiménez-Amilburu et al., 2016). Although the majority of trabecular cardiomyocytes in both wwtr1^−/− and wild-type sibling hearts were depolarized, we found a twofold increase in the percentage of polarized cardiomyocytes in the trabecular layer of wwtr1^−/− hearts (Fig. S3C,D). In order to mosaically label cardiomyocytes in 96 hpf wwtr1 mutant and wild-type sibling hearts, we injected a myl7:mKate plasmid into one-cell stage embryos. We selected trabecular layer cardiomyocytes and visualized the expression pattern of myl7:LIFEACT-GFP. Cardiomyocytes in the trabecular layer of 96 hpf wwtr1^−/− hearts appeared flat and did not form any appreciable striated sarcomeric actin (Fig. S3E), in contrast to trabecular cardiomyocytes in wild-type hearts. In summary, wwtr1^−/− hearts exhibited reduced trabeculation and abnormal trabecular morphology.

Notch signaling has been associated with multiple steps of cardiac trabeculation (Grego-Bessa et al., 2007; Samsa et al., 2015; D'Amato et al., 2016; Han et al., 2016; Jiménez-Amilburu et al., 2016). Therefore, we investigated Notch signaling in wwtr1^−/− hearts by using a Notch reporter that expresses destabilized Venus under the control of the Notch-responsive tp1 element (Ninov et al., 2012). At 48 hpf, tp1 Notch reporter expression was present throughout the ventricular endocardium of wwtr1^+/+ and wwtr1^−/− hearts (data not shown) as reported before in wild-type hearts (Samsa et al., 2015). However, at 72 and 96 hpf, whereas a subset of compact layer cardiomyocytes in wwtr1^+/+ and wwtr1^+/− hearts expressed the tp1 Notch reporter, cardiomyocytes in most (7/8) wwtr1^−/− hearts did not (Fig. 3A,B). A zebrafish jag2b mutant was similarly reported to lack expression of the tp1 Notch reporter in cardiomyocytes (Han et al., 2016). In contrast to the reported hypertrabeculation phenotype (Han et al., 2016), we found that jag2b mutant hearts exhibited reduced trabeculation (Fig. S4), a phenotype similar to that of wwtr1 mutant hearts. In addition, we found that the majority of tp1⁺ cardiomyocytes exhibited nuclear localization of Wwtr1 (Fig. S5A), further suggesting that Wwtr1 plays a role in cardiomyocyte expression of the tp1 Notch reporter. We performed cell transplantation experiments to investigate whether Wwtr1 drives cardiomyocyte expression of the tp1 Notch reporter in a cell-autonomous manner. When wild-type cells were transplanted into wwtr1^+/− or wwtr1^−/− hosts, wild-type and wwtr1^+/−, but not wwtr1^−/−, cardiomyocytes could express the tp1 Notch reporter (Fig. 3C,C′). Conversely, when wwtr1^+/− or wwtr1^−/− cells were transplanted into wild-type hosts, again both wild-type and wwtr1^+/−, but not wwtr1^−/−, cardiomyocytes could express the tp1 Notch reporter (Fig. 3D,D′). These data indicate that Wwtr1 modulates cardiomyocyte expression of the tp1 Notch reporter in a cell-autonomous manner. We then mosaically expressed in wwtr1^−/− cardiomyocytes, wild-type Wwtr1, as well as Wwtr1 variants that affect the WW domain (WW*: W142A, P145A) (Linn et al., 1997) or Tead-binding domain (TBD*: S48A; ΔTBD: P46_D65del) (Vassilev et al., 2001; Zhao et al., 2008; Miesfeld et al., 2015) (Fig. S5B). Whereas some wwtr1^−/− cardiomyocytes that expressed the WT or WW* constructs of Wwtr1 were tp1⁺, wwtr1^−/− cardiomyocytes that expressed the TBD* or ΔTBD constructs mostly remained tp1⁻ (Fig. S5C,D), suggesting that the Tead-binding domain of Wwtr1 is required for tp1 Notch reporter expression in cardiomyocytes.

Surprisingly, in the above-mentioned transplant experiments, we observed that wwtr1^−/− cardiomyocytes were able to contribute to the trabecular layer of wild-type hearts. In fact, we found that wwtr1^−/− cardiomyocytes were located more frequently than wwtr1^+/− cardiomyocytes in the trabecular layer of mosaic wild-type hearts (Fig. 4A, Fig. S6A). Conversely, wild-type cardiomyocytes were located less frequently in the trabecular layer of wwtr1^−/− than wwtr1^+/− hearts (Fig. 4B, Fig. S6B). As described earlier, Wwtr1 immunostaining appears weaker in the nuclei of trabecular cardiomyocytes than in those of compact layer cardiomyocytes (Fig. S1B), raising the possibility that Wwtr1 activity modulates the likelihood of a cardiomyocyte to leave the compact layer. We tested this model further using a gain-of-function approach with a constitutively active Wwtr1 (CAWwtr1) [Wwtr1 S79A, which is homologous to Yap1 S87A (Cox et al., 2016)] (Fig. 4C). The S79 residue of Wwtr1 is part of the HXRXXS motif, which, when phosphorylated by Lats1/2, leads to the nuclear export of Wwtr1 (Zhao et al., 2010). We mosaically expressed myl7:mKate (control) or myl7:mKate-2A-CAWwtr1 and found that cardiomyocytes expressing mKate-2A-CAWwtr1 ended up in the trabecular layer less frequently than control (Fig. 4C′,C″, Fig. S6C). Altogether, these data indicate that Wwtr1 can modulate the likelihood of a cardiomyocyte to egress the compact layer and seed the trabecular layer.

Wild-type cardiomyocytes were observed less frequently in the trabecular layer of mosaic wwtr1^−/− hearts, possibly indicating defects in the development of the compact wall in mutant hearts. Therefore, we investigated the cellular morphology and actomyosin network of cardiomyocytes prior to trabeculation. Compact wall cardiomyocytes in wild-type hearts are rich in basally localized cortical actin exhibiting striated patterns (Reischauer et al., 2014) (Fig. 5A,B). In contrast, compact wall cardiomyocytes in wwtr1^−/− hearts did not exhibit regular striations across their cortical actin bundles (Fig. 5B), and these bundles appeared thinner than wild type (Fig. 5C). Moreover, phospho-MRLC (pMRLC) immunostaining intensity appeared weaker in wwtr1^−/− than wwtr1^+/+ hearts (Fig. S7A). As these phenotypes could be secondary to defects in cardiac function, we examined the mutant hearts more closely, and although heart rate was reduced at 52 hpf, cardiac contraction was not obviously affected (Fig. S7B, Movie 1). We also examined actomyosin morphology shortly after Wwtr1 immunostaining in cardiomyocytes becomes evident (i.e. 30 hpf), and did not observe overt aberrations in mutant cardiomyocytes at this stage (Fig. S7C).

In 52 hpf wild-type hearts, N-cadherin-GFP is localized at the junctions between cardiomyocytes (Cherian et al., 2016); in wwtr1^−/− hearts, it appeared to distribute in a punctate manner over the apical and basal surfaces of some cardiomyocytes (Fig. 5D). However, analysis of optical cross-sections of regions showing aberrant N-cadherin-GFP localization revealed that it was present at the junctions of wwtr1^−/− cardiomyocytes, but that the lateral membranes were oblique to the apical surface and not perpendicular, as in wild type (Fig. 5E-G). Looking at earlier stages (30 hpf), N-cadherin-GFP localization did not appear to be affected in wwtr1^−/− hearts (Fig. S7D). Apicobasal polarity of compact layer cardiomyocytes in wwtr1^−/− hearts did not appear to be affected either, as assessed by EGFP-Podxl and Mark3a-tagRFPt transgene localization, at least at 52 hpf (Fig. S8). Altogether, these data indicate that Wwtr1 plays a role in compact wall morphogenesis prior to the emergence of trabeculae, specifically in the way cardiomyocytes organize their cortical actin networks and orient their lateral membranes in relation to the apicobasal axis.

To identify genes differentially regulated in wwtr1^−/− hearts, we compared their transcriptome with that of wild-type sibling hearts at 57-59 hpf, just prior to the initiation of trabeculation. A total of 33 genes were significantly differentially regulated (Fig. S9A, Table S1). These genes can be grouped into different categories, including muscle, actomyosin network, Wnt signaling, metabolism, extracellular matrix (ECM), ECM degradation and others (Fig. S9B). Because of the architectural phenotypes of 52 hpf wwtr1^−/− cardiomyocytes (Fig. 5), we focused our attention on differentially expressed genes that regulate the actomyosin network, including myh10, mybphb and abraa (Fig. S9C). These genes have been reported to be expressed in zebrafish hearts (Chong et al., 2012; Huang et al., 2013). Using whole-mount in situ hybridization (WISH), we found that mybphb is indeed expressed in most of the muscular tissues, including the heart, and that wwtr1^−/− embryos exhibited heart-specific downregulation of mybphb expression (Fig. S9D), consistent with our RNAseq data. Myh10 has been shown to modulate cardiomyocyte morphology (Chopra et al., 2018), and to play important roles in heart development (Tullio et al., 1997). However, mosaic expression of mybphb and mouse Myh10 in cardiomyocytes resulted in cardiomyocyte cell death (data not shown), precluding experiments using this strategy.

To further analyze the role of Wwtr1 in cardiac trabeculation, we investigated its regulation by two pathways implicated in this process: the neuregulin and Notch signaling pathways. Overexpression of Nrg2a in cardiomyocytes resulted in the loss of nuclear enrichment of Wwtr1 immunostaining in these cells (Fig. 6A); however, the loss of erbb2 did not modify the localization of Wwtr1 in cardiomyocytes (Fig. 6B). Additionally, loss of jag2b, which, as mentioned earlier, has been reported to be required for cardiomyocyte-specific tp1 Notch reporter expression (Han et al., 2016), did not appear to affect the localization of Wwtr1 in cardiomyocytes (Fig. S10A). These results suggest that neuregulin signaling can modulate the localization of Wwtr1 in cardiomyocytes.

Cardiac contractility/blood flow has been shown to modulate cardiac trabeculation (Chi et al., 2008; Peshkovsky et al., 2011; Samsa et al., 2015), and these mechanical cues were shown in other systems to regulate YAP1 nucleo-cytoplasmic localization (Esteves de Lima et al., 2016; Nakajima et al., 2017). We found that defects in cardiac contractility, by using tnnt2a (silent heart/sih) and myh6 (weak atrium/wea) morpholinos, resulted in aberrant accumulation of Wwtr1 in the cytoplasm of cardiomyocytes, although some nuclear localization could still be observed (Fig. S10B). Furthermore, we found that Wwtr1 immunostaining was now evident in the nuclei of atrial cardiomyocytes of tnnt2a and myh6 morphant hearts. In these morphant animals, nuclei of endocardial cells also exhibited Wwtr1 immunostaining, but limited to the affected cardiac chambers (i.e. both cardiac chambers of tnnt2a morphants, and the atrium of myh6 morphants) (Fig. S10B). Taken together, Wwtr1 localization appears to be modulated by neuregulin signaling and cardiac contractility.

The Hippo signaling pathway has been shown to regulate myocardial growth (Heallen et al., 2011; Xin et al., 2011; von Gise et al., 2012). In this study, we uncover novel functions for this pathway, especially for Wwtr1 during cardiac wall morphogenesis (Fig. 7).

Cardiac trabeculation initiates with cardiomyocyte delamination (Liu et al., 2010; Staudt et al., 2014) and, at least in mouse, proliferation (Chen et al., 2004; Grego-Bessa et al., 2007). In our study, we found that wwtr1^−/− hearts exhibit little to no appreciable mature trabecular ridges, fewer cardiomyocytes in the trabecular layer and a lower cardiomyocyte proliferation rate. In addition, we showed that wild-type cardiomyocytes are less frequently found in the trabecular layer of wwtr1^−/− than wwtr1^+/− mosaic hearts, indicating that the mutant cardiac wall cannot support trabeculation. Taken together, these results indicate that Wwtr1 has a role in cardiomyocyte development and proliferation during trabecular morphogenesis.

Signaling and mechanical cues regulate cardiomyocyte morphogenetic movements during trabeculation (Staudt et al., 2014). Neuregulin signaling is required to transform a single-layered ventricular wall into a mature trabeculated wall (Gassmann et al., 1995; Lee et al., 1995; Meyer and Birchmeier, 1995; Liu et al., 2010; Rasouli and Stainier, 2017). Here, we found that nuclei of cardiomyocytes in the trabecular layer exhibit weaker nuclear expression of Wwtr1, possibly in response to Nrg2a stimulation. Moreover, our mosaic studies found that wwtr1^−/− cardiomyocytes have a greater preference for the trabecular layer and, conversely, that CAWwtr1-expressing cardiomyocytes have a greater preference for the compact layer, indicating that Wwtr1 has a role in modulating the likelihood of a cardiomyocyte to enter the trabecular layer.

Although nrg2a appears to be expressed in all endocardial cells of the ventricular outer curvature (Rasouli and Stainier, 2017), only a subset of compact layer cardiomyocytes delaminate to become trabecular cardiomyocytes (Staudt et al., 2014). It is possible that, during this process, the delaminating cardiomyocytes, as they leave the epithelial sheet, mechanically impact their neighbors, a phenomenon that has been observed during cell extrusion and apoptosis in tissue development and morphogenesis (Toyama et al., 2008; Monier et al., 2015; Ambrosini et al., 2017). Specifically, single cells undergoing extrusion can affect the cellular tension of neighboring cells (Teng et al., 2017) and, moreover, these neighboring cells tend to exhibit nuclear localization of YAP1 (Saw et al., 2017), in keeping with the cytoskeletal tension-sensing property of YAP1/WWTR1 (Dupont et al., 2011). It is therefore possible that, during trabeculation, as individual cardiomyocytes delaminate to leave the compact wall, the mechanical influence on their neighbors induces Wwtr1 nuclear localization. In turn, Wwtr1 in compact layer cardiomyocytes cell-autonomously promotes tp1 Notch reporter expression, the activity of which has been shown to be restricted to compact wall cardiomyocytes by lineage tracing (Jiménez-Amilburu et al., 2016).

Recent studies have implicated YAP1/WWTR1 in the regulation of the cytoskeletal network and cell-cell junctions through the transcription of genes that modulate them (Kim et al., 2017; Kimelman et al., 2017). We observed a downregulation of myh10 and mybphb in 57-59 hpf wwtr1^−/− hearts, which may underlie the cardiomyocyte architecture phenotypes (Fig. 5). Indeed, MYH10 has been shown to play crucial roles in cardiomyocyte adherens junction maintenance, sarcomere organization, ventricular septum development and outflow tract orientation (Tullio et al., 1997; Ma et al., 2009; Li et al., 2015). MYBPH (the human ortholog of mybphb), on the other hand, is strongly expressed in muscle (Gilbert et al., 1999; Hayashi et al., 2016); together with sarcomeric myosin heavy chain, it can induce assembly of F-actin cables in non-muscle cells (Welikson and Fischman, 2002). These two genes are also downregulated in asb2b mutant hearts, and asb2b, which encodes an E3 ubiquitin ligase, has been implicated in cardiomyocyte maturation (Fukuda et al., 2017) (Table S2). Moreover, asb2b^−/− atrial cardiomyocytes, similar to wwtr1^−/− ventricular cardiomyocytes, exhibit aberrant N-cadherin localization. However, asb2b is not differentially regulated in wwtr1^−/− hearts, or vice versa (Table S2). These observations support the notion that Wwtr1, through the expression of myh10 and mybphb, modulates cardiomyocyte cytoskeletal maturation, and suggest a possible cross-talk between Asb2b and Wwtr1 in this process.

The Hippo signaling pathway is known for its role in modulating myocardial growth (Heallen et al., 2011; Xin et al., 2011; von Gise et al., 2012) and regeneration (Xin et al., 2013b; Heallen et al., 2013). In mouse, deletion of Yap1 in cardiac lineages (Nkx2.5-Cre) resulted in smaller hearts with thinner ventricular walls, but trabeculation did not appear to be affected (Xin et al., 2011). Although a Wwtr1 mutant was studied to address the role of WWTR1 in cardiac growth (Xin et al., 2013b), a careful investigation into its role in cardiac trabeculation was not carried out. Nevertheless, it is interesting to note that mouse Tead1 global mutants exhibit reduced cardiac trabeculation, which might be due to cardiomyocyte cell death, as indicated by the presence of ‘ghost’ particles detected by the periodic acid-Schiff staining (Chen et al., 1994). Additional studies will be required to determine the exact role of the Hippo pathway and its transcriptional effectors in cardiac trabeculation in mouse and in zebrafish.

All zebrafish husbandry was performed under standard conditions in accordance with institutional (Max Planck Society) and national ethical and animal welfare guidelines.

Transgenic lines used in this study were: Tg(myl7:mKate-CAAX)^sd11Tg (Lin et al., 2012); Tg(myl7:EGFP-Hsa.HRAS)^s883Tg (D'Amico et al., 2007); Tg(-5.1myl7:DsRed2-NLS)^f2Tg (Rottbauer et al., 2002), abbreviated Tg(myl7:nlsDsRed); Tg(EPV.Tp1-Mmu.Hbb:Venus-Mmu.Odc1)^s940Tg (Ninov et al., 2012), abbreviated Tg(tp1:Venus-PEST); Tg(myl7:BFP-CAAX)^bns193Tg (this study); Tg(myl7:LIFEACT-GFP)^s974Tg (Reischauer et al., 2014); TgBAC(cdh2:cdh2-EGFP,crybb1:ECFP)^zf517Tg (Revenu et al., 2014), abbreviated TgBac(cdh2:cdh2-EGFP); Tg(-0.2myl7:EGFP-podocalyxin)^bns103Tg (Jiménez-Amilburu et al., 2016), abbreviated Tg(-0.2myl7:EGFP-podxl); Tg(-0.2myl7:Mark3a-tagRFPt)^bns240Tg (Jiménez-Amilburu et al., 2016); and Tg(myl7:nrg2a202-p2a-tdTomato)^bns140Tg (Rasouli and Stainier, 2017), abbreviated Tg(myl7:nrg2a-2A-Tomato).

Mutant lines used in this study were: erbb2^st61 (Lyons et al., 2005) and jag2b^hu3425 (Han et al., 2016).

The wwtr1^bns35 allele was identified from a CRISPR/CAS9-mediated mutagenesis. Single-guide RNA (sgRNA) (5′-TCACCACATGGCACGACCCC-3′) sequence was designed through the CRISPR design tool from the Zhang Lab (crispr.mit.edu). sgRNA DNA oligos were cloned into the pT7-gRNA vector (Addgene plasmid #46759). The vector was linearized with BsmBI enzyme and followed by in vitro transcription with MEGAshortscript T7 transcription kit (Ambion) to obtain sgRNA for injections. CAS9 mRNA was obtained by in vitro transcription of linearized pT3TS-nCas9n vector (Addgene plasmid #46757) with MEGAscript T3 transcription kit (Ambion). sgRNA (100 pg) and 150 pg of CAS9 mRNA were co-injected into one-cell stage AB embryos. The wwtr1^bns35 allele contains a 29 bp insertion. To genotype this allele, we used the following primers, 5′-TTTGTTGTGCAGTCACATTGAG-3′ and 5′-GAGGGCGTCATGCTCTTC-3′, to generate 66 and 95 bp amplicons corresponding to wild-type and mutant alleles, respectively, which can be separated by gel electrophoresis on a 2% gel. When exogenous wwtr1 was present, the following intron-specific primer, 5′-CTGGAAGGGTGTCCACTGAT-3′, was paired with the above-mentioned reverse primer, 5′-GAGGGCGTCATGCTCTTC-3′, which leads to 169 and 198 bp amplicons corresponding to wild-type and mutant alleles, respectively. These amplicons can be resolved by gel electrophoresis on a 2% gel. Most experiments were performed on embryos/larvae from incrossing F3 heterozygous parents, which themselves were generated by serial outcrosses. Some wwtr1^bns35 mutants can survive to adulthood.

Morpholinos used in this study were: tnnt2a (silent heart or sih) morpholino (Sehnert et al., 2002), myh6 (weak atrium or wea) morpholino (Berdougo et al., 2003) and standard control morpholino (Gene Tools). A total of 1 ng of each morpholino was injected into one-cell stage embryos.

The pTol2 myl7:mKate plasmid was generated by cloning the NheI-CCACC-mKate-XhoI fragment into a multiple cloning site (MCS) of a Tol2 enabled vector that is downstream of a myl7 promoter. CCACC is the Kozak sequence. The pTol2 myl7:mKate-2A-wwtr1 plasmid was generated by ligating the NheI-CCACC-mKate-EcoRI, EcoRI-2A-ClaI and ClaI-wwtr1-XhoI fragments and cloned into the MCS of the aforementioned Tol2 enabled vector. The 2A peptide sequence is GSGEGRGSLLTCGDVEENPGP (Kim et al., 2011). To obtain the WW*, TBD* and ΔTBD mutant constructs, site-directed mutagenesis (SDM) was performed using a Q5 Site-Directed Mutagenesis Kit (NEB #E0552S) on the pTol2 myl7:mKate-2A-wwtr1 plasmid. The following primers were used for site-directed mutagenesis: WW*, 5′-GACgctAGGAAGAGCATGACGCCC-3′ and 5′-GTGagcTGTGGTGATCTTCTCAATGTGATTG-3′; TBD*, 5′-TATGCCGCAGgctTTCTTCCAGG-3′ and 5′-TCCTTGTTCCTCCAGGAGC-3′; ΔTBD, 5′-TCCGGTTCTCTCCCGCCG-3′ and 5′-CATATCCTTGTTCCTCCAGGAGCTC-3′. Lowercase letters are codons corresponding to amino acid substitutions to alanine. All reactions were carried out following the manufacturer's recommendation. The mutated wwtr1 constructs were confirmed by Sanger sequencing and re-cloned using the ClaI and XhoI cloning sites into the parent plasmid to avoid any off-target mutations. The CAWwtr1 construct was obtained by SDM via PCR with the following primers: 5′-GAGCCCGCCGGCAGCTGTAGGGACGCCGGAGAagcGCGCGAGCGAAAG-3′ and 5′-ACAGCTGCCGGCGGGCTCCGTG-3′ [the underlined sequence indicates the NaeI restriction sites and lowercase letters are the codon for the alanine substitution (S79A)]. The amplicon was digested and cloned into the pTol2 myl7:mKate-2A-wwtr1 plasmid. Correct orientation of the insert was selected by Sanger sequencing. pTol2 myl7:mKate-2A-mybphb plasmid was generated by replacing wwtr1 from pTol2 myl7:mKate-2A-wwtr1 with zebrafish mybphb. pTol2 myl7:mKate-2A-Myh10 was generated in a similar manner, but with mouse Myh10. The injection cocktail consisted of 15 pg of transposase mRNA and 8-12 pg of vector.

Embryos at 1 day post fertilization (dpf) were grown in embryo media containing phenylthiourea (PTU) to prevent pigmentation. Live embryos or larvae were embedded in 1% low-melt agarose (LMA) dissolved in egg media containing 0.2% (w/v) tricaine on glass-bottomed dishes. Zeiss Spinning Disk or LSM800 (inverted) confocal microscopes were used. Hearts were manually isolated from fixed animals and embedded in 1% LMA dissolved in 1× PBS on glass-bottomed dishes.

Embryos or larvae were first anesthetized with 0.2% tricaine for up to 5 min followed by fixation with 4% PFA. Samples were fixed for at least 2 h at room temperature or overnight at 4°C. For immunostaining, fixed samples were decapitated, leaving the hearts attached to the head and clearing out any residual yolk. The remaining body was used for genotyping, if applicable. Images for Fig. 5A,B, Fig. S7C,D were obtained from isolated fixed hearts.

Fixed samples were washed for 30 min with 1× PBSTTD [1× PBS, 0.1% Tween-20, 0.1% Triton X-100 and 1% DMSO (or 2% DMSO when using D24E4 primary antibody)]. This wash was followed by 1 h blocking with 2 mg ml⁻¹ BSA and 5% sheep serum added to 1× PBSTTD. Primary antibodies were diluted (see below) in blocking solution and incubated with samples overnight at 4°C. After incubation with primary antibody(ies), samples were rinsed three times with 1× PBSTTD and then washed every 30 min with 1×PBSTTD for at least 6 h. This step was followed by secondary antibody incubation overnight at 4°C and a similar washing regimen the next day. Samples were post-fixed with 4% PFA for at least 20 min prior to confocal imaging. Primary antibodies used and their dilutions were: Wwtr1 (Cell Signaling, D24E4), 1:200; pMRLC (Abcam, ab2480), 1:100; and MF20 (Developmental Studies Hybridoma Bank; DSHB), 1:200. All secondary antibodies were obtained from Life Technologies and used at 1:500. Phalloidin conjugated to Alexa-568 (Life Technologies) was used at 1:500 and DAPI (Life Technologies) at 1:10,000. Endogenous transgenic fluorescent signal perdured through the procedures and no further antibody staining was required. When staining mutant samples, wild-type and mutant samples were placed in the same reaction tube. For morpholino experiments, control and tnnt2a or myh6 morphants were placed in the same reaction tube. Control morphants retained most of the trunk (see ‘Fixed samples preparation’) for easy identification.

Standard whole-mount in situ hybridization protocol was carried out (Thisse and Thisse, 2008). Primers to make the mybphb antisense probe used in this study were: 5′-CATTAACCCTCACTAAAGGGAAATGCCGGCAAAACCAGC-3′ and 5′-TAATACGACTCACTATAGGGGATCTTCACGGAGGCAG-3′. cDNA from adult zebrafish hearts was used as a template for the PCR reaction. Purified amplicons were used as a template for the in vitro transcription (T7 promoter) reaction to obtain the antisense probe. At day 3 of the whole-mount in situ hybridization protocol, after development of the alkaline phosphatase (AP) staining, samples were fixed with 4% PFA for at least 20 min followed by three washes with 1× PBSTTD. This step was followed by the whole-mount immunostaining protocol with the MF20 primary antibody. Images were obtained with Nikon SMZ25 stereomicroscopes. Samples were mounted on a 1% LMA (dissolved in 1× PBS) scaffold. Mutants and wild-type siblings were placed in the same reaction tube.

Image analyses were carried out in Imaris (Bitplane), Zen Blue (Zeiss) or Fiji (Schindelin et al., 2012). Details for image analyses can be found in their individual subsections.

Analyses were carried out in R. The packages used for data management and plotting were ggplot2 (Wickham, 2010), beeswarm, RColorBrewer and plyr. Poisson regression statistical test was used on count data, otherwise standard linear models (i.e. Student's t-test, linear regression, etc.) were used on continuous data. Distribution of cardiomyocytes between compact and trabecular layer was tested with the Binomial test, setting the controls as the null hypothesis parameter. Data and scripts are available upon request from the authors.

Mutant and heterozygous embryos were obtained from a cross between wwtr1^−/−; Tg(myl7:mKate-CAAX) (♂) and wwtr1^+/−; Tg(tp1:Venus-PEST) (♀). Wild-type embryos were obtained from a cross between Tg(myl7:BFP-CAAX) and Tg(tp1:Venus-PEST). Donors were kept alive to screen for presence of the transgenes as well as for genotyping (when applicable). Both transgenes must be present in donors, but only the Tg(tp1:Venus-PEST) transgene was needed in hosts. When analyzing distribution of donor cardiomyocytes between compact and trabecular layer (Fig. 4A,B), only the myl7 transgenics were examined.

Images were acquired on a Spinning Disk microscope and analyzed with Imaris. Individual nuclei were selected using the Surface creation tool on the DAPI channel. Nuclei of trabecular cardiomyocytes were manually selected, as well as nuclei of adjacent compact layer cardiomyocytes. These cardiomyocytes could be identified with the aid of the Tg(myl7:mKate-CAAX) marker. The average intensity of Wwtr1 channel was extracted and analyzed in R.

Images were acquired on a Spinning Disk microscope and processed in Imaris. To obtain surface reconstruction of the Tg(myl7:EGFP-Hsa.HRAS) signal, we used the Normal Shading function in Volume tool. To determine the total number of ventricular cardiomyocytes, we used Tg(myl7:EGFP-Hsa.HRAS) in conjunction with Tg(myl7:nlsDsRed). The same animal was followed for all three time points: 58, 72 and 96 hpf. All nuclei were selected with the ‘Spot Creation’ tool. Cardiomyocytes in the atrium, atrio-ventricular canal (AVC) and outflow tract were manually excluded. For simplicity, a binary assignment of ventricular cardiomyocytes was made: ‘Compact wall cardiomyocytes’ was defined as the outermost layer of cardiomyocytes, otherwise cardiomyocytes were assigned to ‘Trabecular cardiomyocytes’.

Mosaic analyses of trabecular cardiomyocyte morphology were carried out by injecting the pTol2 myl7:mKate plasmid into one-cell stage embryos from an incross of wwtr1^+/− fish. Trabecular cardiomyocytes expressing mKate were selected with the ‘Surface Creation’ tool to apply channel masking on the LIFEACT-GFP channel.

Images were acquired on a Spinning Disk microscope and analyzed with Imaris. Maximum intensity projections were made to count the number of cardiomyocytes that express the tp1 Notch reporter. To test which domain of Wwtr1 was required for myocardial tp1 Notch reporter expression, the various Wwtr1 constructs were injected into one-cell stage embryos from an incross of wwtr1^+/− animals. tp1⁺ mutant hearts were hearts that showed at least one double-labeled tp1⁺, mKate⁺ cardiomyocyte, otherwise assigned as tp1⁻. For transplantation experiments, we relied on the donor/host cardiomyocyte marker to assign tp1⁺ donor/host cardiomyocytes. As donor animals were pre-screened for the presence of transgenes both marking cardiomyocytes and reporting tp1 activity, we could infer whether a tp1⁺ cardiomyocyte belonged to the host if the host cardiomyocyte marker was absent.

As the transplant and the mosaic CAWwtr1 experiments lacked a nuclear marker, the strategy to quantify the number of cardiomyocytes in the compact versus the trabecular layer had to be modified. As cardiomyocytes expressed transgenes that mark the membranes (EGFP-Hsa.HRAS, mKate-CAAX or BFP-CAAX), we carefully classified outer-curvature ventricular cardiomyocytes as ‘compact layer’ or ‘trabecular layer’ through the z-stacks. ‘Compact layer’ cardiomyocytes were those in the outermost layer, otherwise classified as ‘trabecular layer’.

Images were acquired with a Spinning Disk microscope and analyzed on Fiji. A background subtraction was applied on the LIFEACT-GFP channel using a 30 pixel rolling ball radius. A maximum intensity projection was generated followed by an automatic binary threshold of the gray values. We drew lines perpendicular to the cortical actin, sampling across the whole ventricular outer curvature wall, to generate a kymograph for calculation of cortical actin thickness.

Images were acquired with a Spinning Disk microscope and analyzed using Fiji. First, a maximum intensity projection of the mKate-CAAX (cardiomyocyte membranes) and N-cadherin-GFP channels was generated separately. Cardiomyocyte membranes were manually drawn on the mKate-CAAX channel to demarcate individual cells as a region of interest (ROI) and to calculate the ‘total area’. The ROI was then applied on the N-cadherin-GFP channel to guide selection of the ‘tilt area’ (see Fig. 5F for schematic).

Embryos from an incross of wwtr1^+/− fish were grown to 48 hpf. The fin-fold of these embryos were dissected out for genotyping. Wild-type and mutant siblings were identified and placed into separate dishes. Hearts were isolated starting from 57 hpf, for over a maximum of 2 h. Collections were made over five separate occasions to obtain 23 hearts per biological replicate (3 wild-type and mutant biological replicates each). Collection order was randomized over the genotype and biological replicates, and was never identical between collection sessions. Embryos were dissected in 1× Modified Barth's Saline (MBS). A 5× MBS solution consisted of: 17.6 ml 5 M NaCl, 1 ml 1 M KCl, 0.25 g MgSO₄, 1.2 g HEPES and 0.2 g NaHCO₃ to a final volume of 200 ml and adjusted to pH 7.8 with NaOH. The 1× MBS working solution consisted of: 20 ml of 5× MBS and 7 ml 0.1 M CaCl₂ adjusted to a final volume of 100 ml (protocol provided by David Kimelman, University of Washington, Seattle, WA, USA). Isolated hearts were snap frozen in liquid nitrogen and stored overnight at −80°C. Qiazol solution (Qiagen) was added followed by pipetting to mechanically dissociate the tissue. The samples were then passed through a QIAshredder homogenizer column (Qiagen) followed by RNA extraction with a miRNeasy micro kit (Qiagen). In-column DNase digestion was performed to eliminate DNA contamination. RNA and library preparation integrity were verified with a BioAnalyzer 2100 (Agilent) or LabChip G×Touch 24 (Perkin Elmer). Total RNA (6.5 ng) was used as input for Pico Input Mammalian (Takara Clontech). Sequencing was performed on the NextSeq500 instrument (Illumina) using v2 chemistry, resulting in a minimum of 30 M reads per library with a 1×75 bp single-end setup. The resulting raw reads were assessed for quality, adapter content and duplication rates with FastQC (www.bioinformatics.babraham.ac.uk/projects/fastqc). Reaper version 13-100 was employed to trim reads after a quality drop below a mean of Q20 in a window of 10 nucleotides (Davis et al., 2013). Only reads between 30 and 75 nucleotides in length were cleared for further analyses. Trimmed and filtered reads were aligned against the Ensembl Zebrafish genome version DanRer10 (GRCz10.87) using STAR 2.4.0a with the parameter ‘--outFilterMismatchNoverLmax 0.1’ to increase the maximum ratio of mismatches to mapped length to 10% (Dobin et al., 2013). The number of reads aligning to genes was counted with the featureCounts 1.4.5-p1 tool from the Subread package (Liao et al., 2014). Only reads mapping at least partially inside exons were admitted and aggregated per gene. Reads overlapping multiple genes or aligning to multiple regions were excluded. Differentially expressed genes were identified using DESeq2 version 1.62 (Love et al., 2014). Only genes with a combined minimum mean of five reads were used for subsequent analyses. The Ensembl annotation was enriched with UniProt data (release 24.03.2017) based on Ensembl gene identifiers [Activities at the Universal Protein Resource (UniProt)]. Genes were classified in four groups by the following thresholds: significantly differentially expressed, ±1.0 log₂FC (fold-change) and adjusted P<0.05; moderately differentially expressed, adjusted P<0.05; not significantly differentially expressed, ±1.0 log₂FC; and not differentially expressed otherwise.

Images were acquired on a LSM800 (inverted) microscope with fixed settings and analyzed using Fiji. A maximum intensity projection of pMRLC and phalloidin channels was generated using the same settings for all samples. The outline of the ventricular chamber was manually drawn to obtain the average and median gray values of pMRLC and phalloidin. The mean/median gray values of pMRLC was divided by the mean/median gray values of phalloidin for each heart. These normalized values were used for analyses.

Hearts were imaged with a Spinning Disk microscope and analyzed using Zen Blue software. We analyzed 20 µm of sagittal sections surrounding the mid-sagittal plane. Each section was 1 µm. The method to quantify apical constriction was performed as described previously (Jiménez-Amilburu et al., 2016). To quantify the number of polarized and depolarized trabecular cardiomyocytes, we analyzed 26 µm of sagittal sections surrounding the mid-sagittal plane. Each section was 1 µm. Trabecular cardiomyocytes must not be in contact with the abluminal surface throughout the sections. Only trabecular cardiomyocytes that exhibited clear expression of EGFP-Podxl were assessed. They were counted as polarized if EGFP-Podxl was localized to the apical surface only and depolarized if EGFP-Podxl was distributed around the cell.

Larvae (80 hpf) from wwtr1^+/− incrosses were incubated with 1 mM EdU for 16 h in egg water containing PTU and 1% DMSO to facilitate EdU incorporation. Larvae were rinsed twice in egg water, anaesthetized with tricaine 0.4% (w/v) and fixed overnight with 4% PFA. Whole-mount immunostaining was performed as described previously (Inoue and Wittbrodt, 2011). Anti-DsRed (Clontech 632496; 1:300) and anti-GFP (Aves GFP-1020; 1:500) antibodies were used to label the myl7:nlsDsRed and myl7:EGFP-Hsa.HRAS reporter lines, respectively. The CLICK-IT (Invitrogen) reaction for EdU staining was performed by following the manufacturer's instructions. Images were acquired with Zeiss LSM800 confocal microscopy using a 40× [1.1 numerical aperture (NA)] water-immersion objective. EdU⁺ cardiomyocytes and DsRed⁺ cells were quantified by hand in the outer curvature of the cardiac ventricle. A line along the long axis of the cardiac ventricle was drawn to define the outer curvature of the ventricle and 24 planes were analyzed around the mid-sagittal plane, as described previously (Jiménez-Amilburu et al., 2016). The total number of DsRed⁺ cells in the ventricle was quantified using Imaris. We performed this experiment twice independently.

Embryos (52 hpf) expressing the myl7:EGFP-Hsa.HRAS transgene were embedded in 1% LMA. Tricaine use was avoided throughout the procedure. The embryos were imaged live with a Spinning Disk confocal microscope with a 40× objective. The acquisition rate was set at 10 ms per frame. Movies of both GFP and bright-field channels were acquired (Movie 1). Movies from the GFP channel were used to quantify the heart rate using ImageJ software. First, the movies were loaded into ImageJ as a stack (time as the third dimension). The stack was ‘resliced’ with the following parameters: ‘Start at-Right’; ‘Rotate 90°’. Other parameters were left in their default settings. The resulting stack produces a kymograph that was converted into a binary image. The distance between two peaks is the duration of a single cardiac cycle. This distance was then converted into actual time scale (using the actual average number of frames per second recorded by Zen software during acquisition) in order to calculate the average number of heart beats per minute.

Movies from the bright-field channel were used to measure the ventricular fractional shortening with ImageJ (Hoage et al., 2012). Briefly, the maximum (end diastole) and minimum (end systole) short axis of the ventricle were measured from beating hearts to obtain the fractional shortening values=1-minimum/maximum.

We thank Felix Gunawan for an excellent immunostaining protocol; the trabeculation team, consisting of V.U., V.J.-A., Sven Reischauer, Rashmi Priya, Javad Rasouli and Pourya Savari, for discussions; Suchit Ahuja for injecting the myl7:BFP-CAAX construct; Hyun-Taek Kim for mouse Myh10 plasmid for cloning; Viola Graef for critical reading of the manuscript; Rita Retzloff and team for zebrafish care; Michael Potente (MPI-HLR), Ong Yu-Ting (MPI-HLR), Shuichi Watanabe (MPI-HLR), Anna Romão Stone (MPI-HLR), Virginie Lecaudey (Frankfurt), Chaitanya Dingare (Frankfurt), Brian Link (Medical College of Wisconsin) and Joel Miesfeld (UC Davis) for sharing reagents and discussions on the Hippo signaling pathway; and Evan Tan (Singapore) for encouraging J.K.H.L. to generate a zebrafish wwtr1 mutant. J.K.H.L. was a graduate student registered with the Faculty of Biological Sciences at Goethe University, Frankfurt am Main, Germany. Some data from this manuscript form part of J.K.H.L.’s PhD thesis submitted to Goethe University, Frankfurt am Main, Germany in 2018.