The transmembrane protein Crb2a regulates cardiomyocyte apicobasal polarity and adhesion in zebrafish

Trabeculation is key to cardiac wall maturation (Sedmera et al., 2000; Liu et al., 2010; Samsa et al., 2013), and defects in this process lead to congenital heart malformations (Jenni et al., 1999; Weiford et al., 2004; Zhang et al., 2013). In zebrafish, prior to trabeculation, cardiomyocytes (CMs) are organized in a monolayer. Complex coordination of different signaling pathways between the endocardium and myocardium, including Nrg/ErbB2 (Gassmann et al., 1995; Lee et al., 1995; Meyer and Birchmeier, 1995; Samsa et al., 2013, 2015; Rasouli and Stainier, 2017) and mechanical forces such as blood flow/contractility (Staudt et al., 2014; Li et al., 2016), promotes the delamination of a subset of CMs that subsequently seed the trabecular layer. Although recent studies have identified regulators of different steps in cardiac development, including trabeculation, much remains unknown.

During tissue morphogenesis, changes in the shape of individual cells, as well as modulation of the actomyosin cytoskeleton and positioning of tight and adherens junction (TJ and AJ, respectively) proteins, translate into changes across the entire tissue (Martin et al., 2010; Gorfinkiel, 2013; Heisenberg and Bellaïche, 2013). Modulation of apicobasal polarity is a fundamental feature of epithelial cells as they undergo proliferation, cell shape changes and migration (Martin-Belmonte and Mostov, 2008; Etienne-Manneville, 2013; Grifoni et al., 2013; Macara and McCaffrey, 2013), and these events also take place during cardiac trabeculation (Staudt et al., 2014; Uribe et al., 2018). Apicobasal polarity complexes, including PAR3/PAR6/aPKC (Kemphues et al., 1988), Scribble/DLG/LGL (Bilder et al., 2000) and crumbs/PALS1/PATJ (Tepass and Knüst, 1990; Tepass, 1996) are known to interact in order to maintain apicobasal identity, cell shape and tissue integrity. A recent study has shown that Prkci mouse mutants show trabeculation defects (Passer et al., 2016), further supporting the requirement of apicobasal polarity complexes during cardiac trabeculation. Crumbs (Crb) is a type I transmembrane protein composed of an extracellular domain necessary for homophilic interactions at cell-cell junctions (Letizia et al., 2013; Das and Knüst, 2018), and an intracellular domain required to maintain the identity of the apical domain (Bulgakova and Knüst, 2009). In flies, zebrafish and mammals, mutations in Crb affect tissue growth, cytoskeletal rearrangement, junction positioning and stability, as well as the establishment of apicobasal polarity (Grawe et al., 1996; Tepass, 1996; Izaddoost et al., 2002; Omori and Malicki, 2006; Chen et al., 2010; Zou et al., 2012; Alves et al., 2013).

We (Jiménez-Amilburu et al., 2016) and others (Li et al., 2016) have recently reported that CMs in zebrafish and mouse are polarized in the apicobasal axis and that, at least in zebrafish, they undergo apical constriction and depolarization as they delaminate to seed the trabecular layer (Jiménez-Amilburu et al., 2016). However, how apicobasal polarity is regulated in developing CMs remains unknown. In this study, we identified a new role for Crb2a, a member of the Crb complex, during cardiac trabeculation in zebrafish. Crb2a localization in compact layer CMs changes from junctional to apical at the onset of trabeculation. In crb2a mutants, we observed novel and severe defects in CM arrangement and morphology, as well as mislocalization of both TJ and AJ proteins. Interestingly, mutant CMs transplanted into wild-type animals were found exclusively in the trabecular layer. These data suggest that the compromised ability of mutant CMs to integrate into the wild-type compact layer makes them more likely to delaminate. Altogether, these data further emphasize the key role of apicobasal polarity in CMs and its tight regulation in the process of cardiac wall maturation.

Of the five Crb family genes in zebrafish, crb2a is the most highly expressed in the embryonic heart (data not shown). In order to explore the role of Crb2a during cardiac trabeculation, we first used the Tg(myl7:ras-EGFP) line (D'Amico et al., 2007) to label CM membranes and performed Crb2a immunostaining at several developmental stages (Fig. 1). At 50 hours post fertilization (hpf), before the onset of trabeculation, we observed that Crb2a was highly enriched at apical junctions in compact layer CMs (Fig. 1A-A″). At 74 hpf, once trabeculation has initiated, Crb2a has relocalized to the entire apical membrane of compact layer CMs (Fig. 1B-B″). This switch of Crb2a localization in CMs from junctional to apical was also observed in 3D maximum intensity projection images of hearts at 50 (Fig. 1C,E and Movie 1) and 74 (Fig. 1D,E and Movie 2) hpf. Notably, CMs in the trabecular layer do not appear to express Crb2a (Fig. 1B′,B″). We tested the specificity of the Crb2a antibody by staining crb2a^−/− hearts and detected no signal (Fig. S1A-B″). Next, to further investigate this junctional to apical relocalization, we performed immunostaining for Crb2a and the TJ marker ZO-1 at 50 and 79 hpf in the Tg(−0.2myl7:EGFP-podxl) line (Jiménez-Amilburu et al., 2016), in which the apical side of CMs is labeled by EGFP expression (Fig. S2). At 50 hpf, Crb2a colocalized with ZO-1 at the apical junctions between CMs (Fig. S2A′-A″″), but not with the apical marker podocalyxin. However, at 79 hpf, Crb2a immunostaining labeled the apical side of compact layer CMs, colocalizing with EGFP-Podocalyxin but no longer with the junctional protein ZO-1 (Fig. S2B′,B″). Overall, our results show that Crb2a in compact layer CMs transitions from a prominently junctional to an apical localization at the onset of trabeculation.

In order to identify regulators of Crb2a localization, we assessed the role of Nrg/Erbb2 signaling and blood flow. As previously reported, blood flow/contractility and Nrg/Erbb2 signaling are essential for cardiac trabeculation (Liu et al., 2010; Peshkovsky et al., 2011; Samsa et al., 2015; Lee et al., 2016; Rasouli and Stainier, 2017). First, to test whether blood flow regulates the localization of Crb2a in CMs, we stopped contractility and blood flow by injecting tnnt2a morpholinos (MOs) (Sehnert et al., 2002) into one-cell stage embryos. Uninjected larvae at 80 hpf displayed apical localization of Crb2a in compact layer CMs (Fig. 2A-A″), while tnnt2a morphants retained Crb2a localization at the junctions between compact layer CMs (Fig. 2B-B″). Next, we treated Tg(myl7:ras-EGFP) embryos at 54 hpf, i.e. before trabeculation starts, with 10 μM PD168393, an Erbb2 inhibitor, or with DMSO as a control. At 96 hpf, Crb2a immunostaining was observed along the entire apical membrane of compact layer CMs in DMSO-treated larvae (Fig. 2C-C″,E), while Erbb2 inhibitor-treated larvae displayed a significant increase in junctional Crb2a (average of 14.69 junctions between CMs with Crb2a enrichment, compared with 4.57 in DMSO-treated larvae) (Fig. 2D-D″,E). To test whether activation of Nrg/Erbb2 signaling caused the opposite phenotype, we overexpressed the Erbb2 ligand Nrg2a in CMs. CMs expressing the myl7:nrg2a-p2a-tdTomato transgene displayed increased apical localization of Crb2a at 52 hpf, a stage when Crb2a is junctionally located in control hearts (Fig. 2F-F″). Altogether, these data indicate that blood flow and Nrg2a/Erbb2 signaling regulate Crb2a localization in CMs.

To investigate the role of Crb2a during cardiac trabeculation, we studied the ome/crb2a mutant (Malicki et al., 1996; Omori and Malicki, 2006). First, to characterize CM morphology during trabeculation, we crossed the crb2a^−/− with the CM membrane reporter line Tg(myl7:MKATE-CAAX) (Lin et al., 2012). At 80 hpf, after the onset of cardiac trabeculation, crb2a^−/− larvae exhibit a more rounded and smaller ventricle compared with crb2a^+/+ siblings (Fig. 3A,B). At this stage, we observed that while crb2a^+/+ ventricle present a single layer of CMs in the compact wall with a few delaminated CMs (Fig. 3A′-A″,F,G), the crb2a^−/− cardiac wall was composed of two or three layers of CMs and appeared disorganized (Fig. 3B′-B‴,F′,G). Morphologically, CMs in the mutant ventricles appeared more elongated than those in wild-type siblings (Fig. 3A′,A″,B′,B″). To further investigate this phenotype, we measured the circularity of compact layer CMs in the outer curvature at 80 hpf, and observed a significant difference (Fig. 3E). Mutant larvae at 98 hpf did not exhibit myocardial protrusions into the ventricular lumen and failed to form trabeculae (Fig. 3D,D′), while wild-type siblings at this stage presented a complex trabecular network (Fig. 3C,C′). To examine whether the multilayering phenotype was present before the onset of trabeculation, we imaged 52 hpf embryos, and found that, although crb2a^+/+ ventricles consist of a single layer of CMs at this stage (Fig. S3A,A′,D), 19 of 22 mutant ventricles examined displayed multilayering (Fig. S3B,B′,D) and the other three exhibited a single layer of CMs (Fig. S4C,C′,D). We also tested whether defects in cardiac jelly degradation might account for this phenotype by imaging crb2a^−/− in the Tg(kdrl:EGFP) (Jin et al., 2005); Tg(myl7:MKATE-CAAX) background to label, respectively, endocardial and myocardial cells. We did not observe obvious differences in cardiac jelly thickness between crb2a^+/+ (Fig. S4A-C′,E) and crb2a^−/− (Fig. S4B-D′,E) siblings at 80 or 100 hpf.

Next, we studied the early stages of cardiac development in crb2a^−/− animals. At 36 hpf, crb2a^+/+ and crb2a^+/− hearts have looped (Fig. S5A,B), whereas crb2a^−/− hearts remained in the midline or exhibited delayed looping (Fig. S5C,C′). We also analyzed circulation at 48 hpf, and found that 40% of the mutants exhibited few or no circulating blood cells, while the other 60% exhibited wild-type-like circulation (Fig. S6D). To evaluate cardiac performance in crb2a^−/− hearts, we measured atrial fractional shortening (FS) (Fig. S6A-C′). Bright-field images of the beating heart were acquired at 48 hpf and FS was calculated in crb2a^+/+ and crb2a^+/− animals, as well as in crb2a homozygous mutants with and without circulation (Movies 3-6). This analysis showed that crb2a^−/− animals without circulation exhibit a significant decrease in atrial FS compared with crb2a^+/+ or crb2a^+/−, while mutants with circulation exhibit a slight, but not significant, reduction in atrial FS (Fig. S6E). In order to test whether this reduction in cardiac function was due to sarcomere defects, we performed α-actinin immunostaining in Tg(myl7:MKATE-CAAX); crb2a^−/− hearts at 80 hpf, and observed no obvious differences in sarcomeric structure between crb2a^+/+ and crb2a^−/− animals (Fig. S6F-G′). To avoid any confounding effects, all subsequent experiments in this manuscript were performed exclusively with crb2a^−/− animals exhibiting wild-type-like circulation. Taken together, these data show that Crb2a plays an important role during cardiac trabeculation.

To gain further insight into CM polarization in crb2a^−/− animals, we crossed the mutation into the Tg(−0.2myl7:EGFP-podxl) line. As previously reported (Jiménez-Amilburu et al., 2016), EGFP-Podocalyxin localizes to the apical side of compact layer CMs and covers the entire cortex of delaminated CMs in wild-type ventricles at 80 hpf (Fig. 4A-A′,C). In crb2a^−/− ventricles, CMs forming the inner layer of the wall were also polarized, as per their apical localization of EGFP-Podocalyxin (Fig. 4B,B′,C). These observations were confirmed by assessing EGFP-Podocalyxin colocalization with Tg(myl7:MKATE-CAAX) expression and quantifying the number of depolarized CMs in crb2a^+/+ and crb2a^−/− ventricles (Fig. 4C). In order to test whether the multilayering phenotype of crb2a^−/− CMs was due to an increase in CM number, we assessed cell proliferation in mutant hearts. To quantify proliferating CMs, we imaged embryos expressing the CM nuclear transgene Tg(myl7:nlsDsRED) (Rottbauer et al., 2002), as well as the Tg(myl7:VenusGeminin) line (Jiménez-Amilburu et al., 2016), which marks actively proliferating CMs, and examined maximum intensity projection images at 80 hpf (Fig. S7A-C). We found that the number of total CMs and proliferating CMs was similar in crb2a^+/+, crb2a^+/− and crb2a^−/− animals (Fig. S7D,E). Taken together, these data show that Crb2a plays a role in modulating CM polarity and specifically prevent CM delamination from the compact layer in the absence of Nrg2a/ErbB2 signaling.

The Crb complex regulates the localization of TJ and AJ proteins during cellular rearrangements in fly epithelia (Grawe et al., 1996; Izaddoost et al., 2002). To assess whether this regulation also takes place during cardiac development, we immunostained the AJ reporter line TgBAC(cdh2:Cdh2-EGFP) (Revenu et al., 2014) for Crb2a and ZO-1. At 54 hpf, before the onset of trabeculation, Crb2a colocalizes with ZO-1 at CM junctions (Fig. S8A-A″″,C). At this stage, N-Cadherin localizes laterally in CMs showing no overlap with Crb2a or ZO-1 (Fig. S8A-A″″,C). Once trabeculation initiates and Crb2a becomes apically localized in compact layer CMs, ZO-1 remains junctional and N-Cadherin extends basolaterally (Fig. S8B-B″″,C). Next, we analyzed whether TJ and AJ protein localization was affected in crb2a^−/− ventricles. First, we performed immunostaining for ZO-1 in Tg(myl7:MKATE-CAAX); crb2a^−/− at 79 hpf. Maximum intensity projections and single-plane images of crb2a^+/+ larvae show junctional localization of ZO-1 between compact layer CMs (Fig. S9A-B′). However, at the same stage, ZO-1 was mislocalized in both compact and inner layer crb2a^−/− CMs (Fig. 5A-A″″,B-B″″). In order to be able to study in vivo CM TJ protein dynamics more closely, we generated a Tg(-0.2myl7:ZO-1-EGFP) line. Using this line, we were able to reproduce the endogenous localization of ZO-1 during cardiac trabeculation in both mutant and wild-type siblings (Fig. S10). We also performed immunostaining for N-Cadherin in Tg(myl7:MKATE-CAAX); crb2a^−/− larvae at 79 hpf. Analysis of maximum intensity projection images showed that the lateral localization of N-Cadherin between compact layer CMs was lost in mutant hearts (Fig. S9C-D′), appearing punctate in both apical and basal membranes (Fig. 5C-C″″,D,D″″). These data indicate that junctional reorganization in CMs is likely to play a role during CM delamination, and suggest a role for Crb2a in regulating TJ and AJ protein localization in developing CMs.

To gain further insight into Crb2a function during cardiac trabeculation, we generated chimeric embryos by transplanting at the mid-blastula stage Tg(myl7:MKATE-CAAX); crb2a^+/+, crb2a^+/− or crb2a^−/− cells into Tg(myl7:ras-EGFP); crb2a^+/+ animals. After transplantation, larvae were imaged at 100 hpf and donors were subsequently genotyped. We found that crb2a^+/+ and crb2a^+/− CMs in crb2a^+/+ ventricles localized in both the compact and trabecular layers. Strikingly, crb2a^−/− CMs in crb2a^+/+ ventricles were found exclusively in the trabecular layer (Fig. 6A-C). To extend these observations, we quantified the number of donor-derived CMs in the compact and trabecular layers in chimeric larvae at 100 hpf, and found that 100% of the crb2a^−/− CMs populated the trabecular layer of crb2a^+/+ ventricles (Fig. 6D). These data show that CMs leave the compact layer in the absence of Crb2a function, indicating that Crb2a functions cell-autonomously in CMs to regulate this behavior.

Cardiac trabeculation in zebrafish has been proposed as an EMT-like process whereby a subset of CMs undergo apical constriction and lose apicobasal polarity as they delaminate to seed the trabecular layer (Liu et al., 2010; Staudt et al., 2014; Cherian et al., 2016; Jiménez-Amilburu et al., 2016). However, it remains unclear how CM apicobasal polarity is regulated. Here, we set out to investigate whether and how Crb2a, a member of the Crb polarity complex, regulates cardiac trabeculation. In this study, we present in vivo evidence that Crb2a has a key role in maintaining CM apicobasal polarity, as well as in regulating junctional rearrangements during early cardiac development.

The coordinated action of different proteins that modulate apicobasal identity, together with cytoskeletal and junctional rearrangements, maintains cell shape and tissue integrity (Hurd et al., 2003; Kempkens et al., 2006). For example, Crb has a conserved role in a number of epithelial tissues, regulating apicobasal polarity and maintaining junctional integrity during morphogenetic processes (Grawe et al., 1996; Izaddoost et al., 2002). Here, we found that the Crb polarity protein Crb2a shifts its localization in CMs from junctional to apical just prior to the onset of trabeculation. Accordingly, CMs have been reported to exhibit substantial cellular plasticity during trabeculation, losing their cobblestone-like shape once they start delaminating (Liu et al., 2010; Staudt et al., 2014). Our observations are consistent with previous reports in Drosophila where during follicular morphogenesis as cells change shape, Crb is lost from the marginal zone (the most apical surface), relocalizes to the AJ and subsequently reappears at the marginal zone (Sherrard and Fehon, 2015; Wu et al., 2016). We found that blocking Nrg/ErbB2 signaling leads to relocalization of Crb2a from the junctions to the apical surface of CMs. Interestingly, in Drosophila, depletion of cytoskeletal proteins required for Crb localization leads to relocalization of apical Crb to the junctions (Sherrard and Fehon, 2015). These data indicate that Crb2a localization is very dynamic and that Nrg/ErbB2 signaling, which is essential for cardiac trabeculation, is required for its junctional to apical shift.

Mutations in CRB2 in humans and Crb2 in mouse cause cardiac malformations (Xiao et al., 2011; Slavotinek et al., 2015; Jaron et al., 2016). Similarly, we found that zebrafish lacking Crb2a display a compact ventricular wall composed of multiple disorganized layers of CMs. This phenotype is reminiscent of that observed during Drosophila oogenesis, where early loss of Crb leads to multilayering (Tanentzapf et al., 2000; Pénalva and Mirouse, 2012). While delaminating CMs in wild-type ventricles lose their apicobasal identity (Jiménez-Amilburu et al., 2016), most CMs in crb2a mutants remain polarized, including those in the inner layers. These data suggest that inner layer CMs in crb2a mutants are not trabecular CMs.

During epithelial morphogenesis, changes in cell shape and the localization of junctions require the tight regulation of junctional proteins and the cytoskeleton (Costa et al., 1998; Raich et al., 1999; Martin et al., 2010). We observed that crb2a mutant CMs display mislocalization of TJ and AJ proteins. Similarly, loss of Crb in epithelial cells in Drosophila causes defects in cell-cell adhesion and tissue integrity (Tepass and Knüst, 1990; Grawe et al., 1996; van de Pavert et al., 2004; Letizia et al., 2013; Flores-Benitez and Knüst, 2015). Certainly, the polarity complex protein Crumbs regulates not only the stability but also the localization of junctions in Drosophila salivary glands (Xu et al., 2008) and Drosophila eyes (Fan et al., 2003). Therefore, we hypothesize that the lack of Crb2a in zebrafish mutant hearts leads to mislocalization of CM junction proteins, thereby affecting compact layer architecture and in turn CM delamination/migration. Moreover, during tissue morphogenesis, mechanical tension plays a key role in controlling actomyosin networks and cell-cell junction remodeling to maintain cell shape (Lecuit et al., 2011; Lecuit and Yap, 2015). Compact layer CMs are subjected to mechanical forces necessary for ventricular chamber maturation (Cherian et al., 2016). In view of these data, we propose that loss of cell-cell adhesions in crb2a mutant hearts is at least partially responsible for the CM multilayering phenotype coming from the collapse of the CM monolayer prior to the onset of delamination. Indeed, multilayering can be observed in crb2a mutants at early stages of development. As a consequence, ventricular shape is significantly affected in crb2a^−/−, and CMs remaining in the compact layer flatten out. However, how the actin cytoskeleton is affected in crb2a^−/− CMs remains to be investigated.

In mosaic embryos, we observed that all crb2a mutant CMs in wild-type ventricles become part of the trabecular layer. Similarly, wwtr1 mutant CMs in wild-type ventricles exhibit a preference to enter this layer (Lai et al., 2018). The cardiac phenotypes of wwtr1 and crb2a mutants are similar in the sense that they both exhibit CM multilayering, although it is more pronounced and is observed at an earlier stage in crb2a mutants. Interestingly, we found that Crb2a immunostaining was hardly detectable in wwtr1 mutant hearts (Fig. S11), suggesting that a reduction in Crb2a expression leads to the mislocalization of AJ proteins in wwtr1 mutant hearts (Lai et al., 2018). In summary, the lack of Crb2a in zebrafish CMs affects their ability to remain in a monolayer, subsequently leading to the lack of trabeculae.

All experiments using zebrafish were performed in accordance with German animal protection laws and approved by the local governmental animal protection committee. Zebrafish were maintained according to standard protocols (wiki.zfin.org/display/prot/General+Methods+for+Zebrafish+Care).

Transgenic lines used in this study are as follows: Tg(−0.2myl7:EGFP-podocalyxin)^bns103 (Jiménez-Amilburu et al., 2016), abbreviated Tg(−0.2myl7:EGFP-podxl); Tg(myl7-MKATE-CAAX)^sd11 (Lin et al., 2012), abbreviated Tg(myl7:MKATE-CAAX); Tg(myl7:EGFP-Has.HRAS)^s883 (D'Amico et al., 2007), abbreviated Tg(myl7:ras-EGFP); Tg(myl7:mVenus-gmnn)^ncv43Tg (Jiménez-Amilburu et al., 2016), abbreviated Tg(myl7:mVenus-gmnn); Tg(kdrl:EGFP)^s843 (Jin et al., 2005), abbreviated Tg(kdrl:EGFP); TgBAC(cdh2:Cdh2-EGFP,crybb1:ECFP)^zf517 (Revenu et al., 2014), abbreviated TgBAC(cdh2:Cdh2-ECFP); Tg(-5.1myl7:DsRed2-NLS)^f2Tg (Rottbauer et al., 2002) and abbreviated Tg(myl7:nlsDsRed).

The mutant allele used in this study is crb2a^m598 (Omori and Malicki, 2006). To screen for adult carriers or discriminate between wild-type, crb2a^+/− and crb2a^−/− at embryonic/larval stages, we used genomic DNA extracted from clipped fins. PCR1 was performed using the following primers: 5′-TCAGGCTTGTCCTTCAAGTC-3′ and 3′-TTACTTGGCTCAGGTGTGTC-5′. The product from PCR1 was then used to perform PCR2 using the following primers: 5′-TGTAAAACGACGGCCAGTTCAAGCATGCAGA GTTGAAG-3′ and 3′-AGGAAACAGCTATGACCATTATGCAAGACACTGGCACTC-5′. The product from PCR2 was sent for sequencing with an M13 universal primer (5′-TTACTTGGCTCAGGTGTGTC-3′). Chromatograms were used to distinguish between wild type (TGACTGTTACAGACCCC), crb2a^+/− (TGACTGTTACA/TGACCCC) and crb2a^−/− (TGACTGTTACTGACCCC) (differing nucleotides are indicated in bold).

The pTol2−0.2myl:EGFP-ZO-1 plasmid was generated using the Cold Fusion Cloning Kit (System Biosciences, MC101A-SBI). EGFP-ZO-1 was amplified using the Phusion Tag HF Polymerase and the following primers: forward, 5′-CAAAGCAGACAGTGAGCTAGCATGGTGAGCAAGGGCGAGGA-3′; and reverse, 5′-TCCCCCGGGCTGCAGGAATTCCTAGAGGCTCGAGATGGGAA-3′. The PCR product was subsequently cloned into a Tol2 enabled vector with the −0.2 kb myl7 promoter. After Cold Fusion Cloning, colonies containing the construct were sent for sequencing for validation.

Embryos were treated with 10 μM of the ErbB2 inhibitor PD168393 at 54 hpf. For all treatments, ten embryos per well were placed in 1 ml PTU egg water in 12-well plates containing the corresponding drug. DMSO was used as a control. In the case of PD168393, drugs were added into egg water containing PTU and 1% DMSO to help with solubilization.

For mosaic expression of nrg2a in CMs, 15 pg of myl7-nrg2a-p2a-tdTomato plasmid DNA was injected together with 10 pg of Tol2 mRNA at the one-cell stage into the cell.

Embryos were treated with PTU to avoid pigmentation and dechorionated using pronase at 1 dpf. Embryos/larvae were fixed at the stage of interest in 2 ml Fish fix buffer [1×PBS, 120 µl 1 M CaCl₂, 40 g sucrose with PFA (pH 7.35)] overnight at 4°C on a nutator. The next day, the Fish fix buffer was removed and embryos were washed three times for 15 min with PBS/0.1%Tween20 followed by manual removal of the yolk of the embryos/larvae manually using forceps. Deyolking was followed by permeabilization with proteinase K (3 μm/ml) in 1 ml PBS/0.1%Tween20 (50-54 hpf for 20 min; 72 hpf for 50 min; 79-80 hpf for 1 h). After permeabilization embryos/larvae were rinsed quickly with PBDT to stop the proteinase K reaction. Then, embryos/larvae were washed three times for 15 min in PBDT. Blocking was carried out using blocking buffer [PBS/0.1% Tween + 10% sheep serum + 4% (20% Triton-X) + 5% (20% BSA)] for 2-3 h. Embryos/larvae were then incubated with the corresponding primary antibodies with incubation buffer overnight at 4°C in the rotator (shaking gently). The following primary antibodies with their corresponding concentrations were used: Crb2a, 1:500 (Zou et al., 2012) was used in Figs 1 and 2; Crb2a, 1:50 (ZIRC, zs-4) was used in the rest of the figures; ZO1, 1:2000 (Invitrogen, 33-9100); EGFP, 1:500 (Aves Lab, gfp-1020); α-actinin, 1:1000 (Sigma, A7811); N-Cadherin, 1:250 (Abcam, ab12221); and tRFP, 1:500 (Evrogen, AB233). The next day, following incubation, the primary antibodies were removed and embryos/larvae were washed four times for 20 min with PBDT. Embryos/larvae were incubated with the secondary antibodies in incubation buffer: PBS/0.1% Tween + 1% sheep serum + 4% (20%Triton-X) + 5% (20% BSA) for 4 h at room temperature in the rotator (shaking gently). All secondary antibodies [Alexa Fluor 488 anti-chicken (A11039), Alexa Fluor 488 anti-rabbit (A11034), Alexa Fluor 488 anti-mouse (A11029), Alexa Fluor 568 anti mouse (A11004) and Alexa Fluor 647 anti-mouse (A21236)] were obtained from Life Technologies and used at 1:500. Next, secondary antibodies were removed and embryos/larvae were washed first with DAPI in PBS/Tween20 for 10 min, followed by 6×15 min washes in PBS/Tween20. All these steps were performed in the dark. DAPI (Life Technologies) was used at 1:10,000. Once the washes were finished, embryos/larvae were kept in PBS/0.1%Tween20 at 4°C in the dark until the day of imaging.

Embryos were treated with PTU to avoid pigmentation and dechorionated using pronase at 1 dpf. At 36 hpf, embryos were fixed with 4% PFA overnight at 4°C. PFA was removed the next day and embryos were washed with PBS/0.1%Tween20 twice for 10 min. After PFA removal, the heads of the embryos were cut using forceps and a blade. The cut was performed right under the head in an angle of 15° with respect to the base of the head. Removing the head at early stages helps to visualize the heart. Embryos were kept in PBS/0.1%Tween20 at 4°C until the day of imaging.

Donor cells for transplantation were obtained from embryos from Tg(myl7:MKATE-CAAX) crb2a^+/− incrosses. Wild-type host embryos were obtained from crossing Tg(myl7:ras-EGFP) animals. For transplantation, embryos were first dechorionated by treating with 1 mg/ml pronase in Danieau buffer and then maintained on agarose-coated petri dishes until blastomere stages. Embryos were then transferred into 12-well plates where each well was filled with agarose. Donor embryos were kept in Ringer's buffer containing penicillin/streptomycin, and donor cells were transplanted into two or three different wild-type host embryos along the blastoderm margin at mid-blastula stages. Donor larvae were maintained alive for genotyping. Prior to imaging, selection of mosaic larvae was carried out by screening for those expressing both the MKATE-CAAX and ras-EGFP transgenes. Imaging was carried out using an LSM800 confocal microscope (Zeiss) and the contribution of the donor-derived cells to the trabecular or compact layer was quantified by analyzing the whole ventricle plane by plane using the ZEN software (Zeiss).

Confocal images were processed and analyzed using the ZEN Black Software (Zeiss). Quantification was started at the ventricular mid-sagittal plane and counted 12 planes up and 12 planes down at an increment of 1 μm per plane, and this set of 24 planes was used for analysis. The number of compact layer and trabecular CMs was counted in this set of 24 planes. The percentage of trabecular CMs was calculated by dividing the number of trabecular CMs by the total number of CMs.

Bright-field images of wild-type, crb2a^+/− and crb2a^−/− beating hearts were acquired at 48 hpf using a spinning disc confocal microscope. Zebrafish embryos were mounted in low melting agarose containing 0.01% (w/v) tricaine in a glass-bottomed dish. Z-stacks of beating hearts were imaged at 50 frames/s. The width of the atrium at the maximum systolic and diastolic phases was measured in three different non-consecutives frames and the FS was calculated as (diastole length−sistole length/diastole length)×100.

Four different measurements of cardiac jelly thickness were taken at the mid-sagittal ventricular plane of 80 and 100 hpf hearts. The ZEN software was used to draw a line from the basal side of CMs to the endocardium.

Circularity of compact layer cardiomyocytes in the outer curvature of the ventricle at the mid-sagittal plane was measured using ImageJ as described previously (Rasouli and Stainier, 2017).

Confocal images (mid-sagittal ventricular plane) were used to count the number of polarized and depolarized CMs in the outer curvature of crb2a^+/+ and crb2a^−/− hearts using the ZEN software.

Embryos and larvae were mounted in 1% low melt agarose in glass-bottomed dishes. In order to stop the heart beat before imaging, 0.2% (w/v) tricaine was added to the agarose. Z-plane images were taken using a spinning disc confocal microscope (Zeiss, CSU-X1 Yokogawa), an LSM800 or an LSM880 (Zeiss). A 40× [1.1 numerical aperture (NA)] water-immersion objective was used to acquire the images. The optical sections were 1 μm thick.

Three-dimensional images were processed using the Imaris ×64 (Bitplane, 7.9.0) software. To quantify the number of CMs in the trabecular versus compact layer, the ZEN software was used. The Fiji or ZEN software was used for the rest of the 2D image processing. Images and schemes were prepared using Adobe Photoshop and PowerPoint.

Data are expressed as mean±s.e.m. Differences between groups were compared using a two-tailed Student's t distribution test or ANOVA. All tests were performed with a confidence level of 95% (*P<0.05; **P<0.01; ***P<0.001; ****P<0.0001; ns, no significant changes observed).

We thank Xiangyun Wei and Chuanyu Guo (University of Pittsburgh, PA, USA) for sharing the Crb2a antibody; Elizabeth Knüst and Cátia Crespo for providing the crb2a mutant and the genotyping protocol; Rubén Marín-Juez, Rashmi Priya, Jason Lai, Felix Gunawan, Sophie Ramas, Chi-Chung Wu and Sally Horne-Badovinac for helpful discussions and sharing reagents and expertise; Rubén Marín-Juez, Rashmi Priya and Michelle Collins for critical reading of the manuscript and feedback; Hans-Martin Maischein and Simon Howard for help with zebrafish transplantations and injections; Felix Gunawan for the immunostaining protocol; and Rita Retzloff, Martin Laszczyk and their teams for zebrafish care. V.J.-A. is 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 V.J.-A.’s PhD thesis entitled ‘Cardiomyocyte apicobasal polarity during cardiac trabeculation in zebrafish’, which was submitted to Goethe University, Frankfurt am Main, Germany, in June 2018.