Nkx genes establish second heart field cardiomyocyte progenitors at the arterial pole and pattern the venous pole through Isl1 repression

Cardiomyocytes (CMs) are derived from at least two distinct cardiac progenitor populations located in the splanchnic mesoderm. First heart field (FHF) cells differentiate early in the lateral plate mesoderm to give rise to the linear heart tube (Buckingham et al., 2005; Vincent and Buckingham, 2010). Subsequently, second heart field (SHF) cells migrate to the pharyngeal mesoderm, proliferate and differentiate into CMs at the arterial and venous poles. The anterior SHF (aSHF) subpopulation gives rise to the right ventricle and outflow tract (OFT), whereas the posterior SHF (pSHF) contributes primarily to the atria and inflow tract (IFT) (Buckingham et al., 2005; Cai et al., 2003; Dominguez et al., 2012; Dyer and Kirby, 2009; Galli et al., 2008; Rana et al., 2014; Vincent and Buckingham, 2010; Zaffran et al., 2004). Errors in specification and differentiation of the SHF cells under these precise spatial and temporal constraints can lead to a variety of congenital heart defects (CHDs), specifically conotruncal (OFT) and atrial (IFT) abnormalities. These malformations of the cardiac poles are found in 10% and 30%, respectively, of individuals with CHDs and result in significant neonatal morbidity and mortality (Gelb and Chung, 2014; Hoffman and Kaplan, 2002; Hoffman et al., 2004; Loffredo, 2000; Nemer, 2008; Payne et al., 1995; Pierpont et al., 2007; Supino et al., 2006). Despite this clinical impact, the mechanisms underlying the development of the arterial and venous poles of the heart are incompletely understood.

NKX2-5, a homeodomain transcription factor expressed in cells of the FHF and SHF (Stanley et al., 2002), and a regulator of cardiac development (Bruneau, 2008), is associated with a myriad of human CHDs. Moreover, individuals carrying NKX2-5 mutations demonstrate a propensity to develop malformations involving the poles of the heart, such as Tetralogy of Fallot, double outlet right ventricle and atrial septal defects (Abou Hassan et al., 2015; Benson et al., 1999; Chung and Rajakumar, 2016; Elliott et al., 2003; Goldmuntz et al., 2001; McElhinney et al., 2003; Nakajima, 2010; Schott et al., 1998; Srivastava and Olson, 2000). Investigation of the genetic basis of CHDs in model systems has yielded insights into functions of Nkx2-5 in progenitor specification in Drosophila, Xenopus and mouse (Azpiazu and Frasch, 1993; Bodmer, 1993; Grow and Krieg, 1998; Prall et al., 2007), and in cardiac morphogenesis in mouse and zebrafish (Barth et al., 2010; Lyons et al., 1995; Prall et al., 2007; Tanaka et al., 1999; Targoff et al., 2013, 2008; Tu et al., 2009). Furthermore, Nkx2-5 mutants, whether inside or outside the DNA-binding domains (Benson et al., 1999; Gutierrez-Roelens et al., 2002; Ikeda et al., 2002), regulate wild-type targets and also ‘off-targets’, destabilizing the cardiac transcriptional hierarchy in this context (Bouveret et al., 2015). Although previous work has revealed specific roles for Nkx2-5 in the SHF (Cambier et al., 2014; Clark et al., 2013; Dorn et al., 2014; George et al., 2015; Guner-Ataman et al., 2013; Prall et al., 2007; Watanabe et al., 2012; Witzel et al., 2012; Zhang et al., 2014), the distinct cellular and molecular mechanisms operating in the aSHF and pSHF subpopulations remain largely unknown.

Recent studies have revealed essential functions of Nkx2-5 transcriptional regulation in OFT development. In mouse, Nkx2-5 has been shown to drive arterial pole formation via a Bmp2/Smad1 negative-feedback loop (Prall et al., 2007) and through a direct repressive effect on a 1.7 kb Fgf10 enhancer element (Watanabe et al., 2012). In zebrafish, nkx2.5 plays a role in aSHF proliferation (Guner-Ataman et al., 2013) and stabilization of Nkx2.5 by Ajuba protein regulates progenitor specification in this population (Witzel et al., 2012). Moreover, the expression patterns of ltbp3 and mef2cb, aSHF markers in zebrafish, overlap with the nkx2.5-expressing region near the arterial pole (Lazic and Scott, 2011; Zhou et al., 2011). Although these data document the importance of Nkx genes in arterial pole development, it is unclear how Nkx2.5 controls differentiation and expansion of the aSHF progenitor pool to sculpt the precise dimensions of the OFT.

At the sinus venosus (SV), extensive studies have elucidated the function of Nkx2-5 in cardiac conduction system (CCS) development (Briggs et al., 2008; Jay et al., 2004; Moskowitz et al., 2007; Nakashima et al., 2009; Pashmforoush et al., 2004; Ye et al., 2015). Yet the mechanisms by which Nkx genes direct pSHF progenitors to pattern the venous pole have not been fully defined. It is evident from murine models that pSHF cells accumulate in the atrial chambers (Cai et al., 2003; Galli et al., 2008; Lescroart et al., 2012) and pSHF progenitors direct the molecular and cellular events contributing to atrioventricular (AV) septation (Goddeeris et al., 2008; Hoffmann et al., 2009, 2014; Mommersteeg et al., 2006; Snarr et al., 2007a,b; Xie et al., 2012). Interestingly, Nkx2-5 is important in establishing a balance between sinoatrial node (SAN) and atrial myocardial identities (Espinoza-Lewis et al., 2011; Mommersteeg et al., 2007; Nakashima et al., 2014). Another homeodomain transcription factor, Shox2, is also essential for SAN development, in part through repression of Nkx2-5 (Blaschke et al., 2007; Espinoza-Lewis et al., 2009). This Shox2-Nkx2-5 antagonistic mechanism inhibits pacemaker properties that define a ‘default’ primitive cellular phenotype (Ye et al., 2015). Furthermore, studies have shown that Shox2 directly regulates Isl1 (Hoffmann et al., 2013) and Nkx2-5 inhibits Isl1 (Dorn et al., 2014; Prall et al., 2007; Watanabe et al., 2012), a key cardiac transcription factor expressed in the pSHF (Cai et al., 2003; Snarr et al., 2007a; Sun et al., 2007; van den Berg et al., 2009). Variation in ISL1 confers genetic susceptibility for CHDs (Stevens et al., 2010), thus highlighting a crucial role in human disease. Moreover, data have illuminated a pivotal role for Isl1 in SAN development and function (Liang et al., 2015; Tessadori et al., 2012; Vedantham et al., 2015), and these findings offer explanations for CCS abnormalities in individuals harboring NKX2-5 mutations. Given embryonic stem cell work suggesting that direct Isl1 repression by Nkx2-5 is necessary to establish the ventricular lineage (Dorn et al., 2014) and our findings revealing essential roles of Nkx genes in ventricular identity maintenance (George et al., 2015; Targoff et al., 2013), we sought to uncover novel Nkx-Isl1 interactions that are required to pattern the IFT.

Our previous studies demonstrate that the Nkx2-5 homologs expressed in zebrafish CMs, nkx2.5 and nkx2.7 (Lee et al., 1996), act redundantly in cardiac morphogenesis (Targoff et al., 2008; Tu et al., 2009) and are crucial in maintaining ventricular identity by repressing atrial fate (Targoff et al., 2013). Furthermore, temporally controlled expression of nkx2.5 is essential to ensure chamber-specific identity maintenance in both the FHF and SHF during explicit developmental time windows (George et al., 2015). Despite these insights, little is known about the mechanisms by which Nkx genes control the specification and differentiation of SHF progenitors at the cardiac poles. In this study, we find that Nkx genes positively regulate aSHF progenitors at the arterial pole and restrict pSHF-derived CM differentiation at the venous pole. At the arterial pole, Nkx genes are required to promote the aSHF progenitors, within the proliferating pharyngeal region, to differentiate and assemble the OFT. In nkx2.5^−/−;nkx2.7^−/− embryos, unrestricted expression patterns of SAN genes, bmp4, tbx2b, hcn4 and shox2, highlight severe abnormalities in venous pole development. Furthermore, our findings reveal that Nkx genes normally restrict isl1 expression to the venous pole during heart tube formation, whereas overexpression of nkx2.5 is sufficient to inhibit isl1. Notably, electrophysiological chamber identity is dependent on Nkx gene dose with elimination of all heterogeneity of intercellular coupling in the nkx2.5^−/−;nkx2.7^−/− embryos. These findings support the innovative concept that CM differentiation is highly graded, with functional partitioning of the heart tube established and maintained through networks of overlapping regulators rather than through binary fate choices. This gradation is consistent with the pleiotropy of phenotypes observed in monogenic forms of CHD and the known redundancy and resilience of cardiac transcriptional networks. Taken together, our results suggest key functions of Nkx genes in SHF CM differentiation in mediating differential roles in anterior and posterior populations. Ultimately, our studies have the potential to shed light on the mechanisms underlying conotruncal and venous pole malformations, and conduction system defects in individuals carrying NKX2-5 mutations.

Although our previous studies revealed synergistic functions of nkx2.5 and nkx2.7 in OFT development (George et al., 2015; Targoff et al., 2013), the discrete cellular and temporal mechanisms operating downstream of Nkx genes in the aSHF remain unclear. Previous studies have identified SHF progenitors in the anterior lateral plate mesoderm (ALPM) that express nkx2.5 and contribute to the OFT (Guner-Ataman et al., 2013; Paffett-Lugassy et al., 2017). Thus, we examined the roles of nkx2.5 and nkx2.7 in regulating the spatiotemporal contribution of the aSHF to arterial pole formation. Employing a developmental timing assay to assess the differences in protein maturation time in embryos expressing Tg(myl7:EGFP) and Tg(-5.1myl7:nDsRed2) (de Pater et al., 2009; Huang et al., 2003; Mably et al., 2003) (Fig. 1A-I), we visualized the accretion of the aSHF-derived CMs in the OFT of wild-type and nkx2.5^−/−; nkx2.7^+/+, nkx2.5^−/−; nkx2.7^+/− and nkx2.5^−/−; nkx2.7^−/− (hereafter referred to as nkx mutant) embryos. Progressive loss of Nkx gene function highlights substantial reduction in the volume of OFT myocardium (Fig. 1D-I). Moreover, we detected a statistically significant reduction in aSHF-derived CMs, or EGFP⁺ DsRed⁻ cells, when comparing wild-type (24 cells±3 cells; n=12) with nkx2.5^−/−;nkx2.7^−/− (16 cells±4 cells; n=3) embryos (P<0.005). Taken together, these findings underscore that nkx2.5 and nkx2.7 positively regulate the contribution of aSHF progenitors at the zebrafish arterial pole.

Prior work has elucidated an important role for SHF CM proliferation in zebrafish arterial pole development (Guner-Ataman et al., 2013; Lazic and Scott, 2011; Nevis et al., 2013; Schindler et al., 2014; Zeng and Yelon, 2014; Zhou et al., 2011). Given the decreased number of aSHF-derived CMs in the nkx mutant OFT, we tested whether diminished proliferation is responsible for this observation. We pulsed embryos carrying Tg(-5.1myl7:nDsRed2) at 24 hpf, when aSHF-derived CMs proliferate, with the thymidine analog 5-ethynyl-2′-deoxyuridine (EdU) (Guner-Ataman et al., 2013; Schindler et al., 2014; Zeng and Yelon, 2014). At 48 hpf, we counted EdU⁺ DsRed⁺ MF20⁺ cells at the arterial pole to quantify proliferating nuclei in the myocardium (Fig. S1). We detected no statistically significant differences in the proliferation indices between wild-type (1.0%±1.3%; n=15) and nkx mutant (1.5%±1.2%; n=8) embryos, indicating that a reduced rate of OFT proliferation is not responsible for the decrease in CMs accrued at the arterial pole.

We then postulated that a shrunken arterial pole size in Nkx mutant embryos could represent an earlier influence of Nkx genes on the reservoir of undifferentiated aSHF progenitors. The SHF-derived population at the OFT can be divided into two cell types: differentiated CMs that strongly express myl7, vmhc and nkx2.5, and a progenitor pool that faintly expresses these same markers and strongly expresses mef2cb and ltbp3 (Lazic and Scott, 2011; Zeng and Yelon, 2014; Zhou et al., 2011). Indeed, our findings reveal diminished mef2cb expression in the arterial pole of wild-type compared with nkx mutant embryos (Fig. 2A-E). Specifically, mef2cb expression is slightly decreased in nkx2.5^−/−;nkx2.7^+/+ and nkx2.5^−/−;nkx2.7^+/− embryos (Fig. 2C,D), and further diminished in nkx2.5^−/−;nkx2.7^−/− embryos (Fig. 2E) when compared with wild-type and nkx2.5^+/+;nkx2.7^−/− embryos (Fig. 2A,B). These results corroborate our previously published studies illustrating a progressive decline in ltbp3 expression with successive reduction of Nkx gene dose (Targoff et al., 2013). Altogether, our findings suggest that Nkx gene function is required to establish the proper size of the aSHF progenitor pool. However, in assessing this population through gene expression profiles of SHF markers, we may fail to identify all cells destined to become aSHF-derived CMs.

Given the importance of the temporal regulation of aSHF precursor accretion to the arterial pole (Cambier et al., 2014; Jahangiri et al., 2016; Zeng and Yelon, 2014), we sought to quantify aSHF progenitors by tracking these late-differentiating cells during OFT recruitment. Recent studies have confirmed the contribution of distal Tg(nkx2.5:ZsYellow)-expressing progenitor cells to the OFT myocardium using Tg(nkx2.5:kaede) (Guner-Ataman et al., 2013; Zeng and Yelon, 2014). Employing Tg(nkx2.5:ZsYellow) and Tg(myl7:EGFP), we examined aSHF progenitor cells in the pharyngeal region distal to the arterial pole where they proliferate prior to differentiation and addition to heart (Hutson et al., 2010; Tirosh-Finkel et al., 2010; van den Berg et al., 2009; Zeng and Yelon, 2014). We observed dispersion and lower levels of expression of Tg(nkx2.5:ZsYellow) in the distal nkx2.5⁺ OFT progenitors in nkx mutant compared with wild-type embryos at 36 hpf (Fig. 2F,G). We quantified this alteration in expression patterns by counting ZsYellow⁺ EGFP⁻ cells distal to the arterial pole and between the branchial arches (Fig. 2H). Our results demonstrate a dramatic decrease in the number of distal nkx2.5⁺ progenitors contributing to the OFT in nkx mutant embryos. Together, these data suggest that loss of Nkx gene function limits the number of undifferentiated aSHF cells prior to their deployment to the arterial pole.

From our findings demonstrating that Nkx genes are required for OFT formation by early regulation of the aSHF progenitor pool size, as opposed to later influence on SHF-derived CM production, we postulated that the decreased number of OFT progenitors in Nkx mutants could be due to diminished progenitor proliferation. To investigate this hypothesis, we performed an EdU incorporation assay in embryos carrying Tg(nkx2.5:ZsYellow) at 18 hpf to identify dividing progenitor cells prior to their migration to the arterial pole. At 40 hpf, we performed MF20 immunostaining to delineate proliferative ZsYellow⁺ cells that had been incorporated into the OFT in wild-type (Fig. 3A,B) and nkx mutant (Fig. 3C,D) embryos. Surprisingly, we found no significant difference in the proliferation index of the total OFT progenitor population when comparing wild-type with nkx mutant embryos (Fig. 3E). We next tested whether there might be differences in OFT progenitor proliferation indices between wild-type and nkx mutant embryos in the proximal early-differentiating (ZsYellow⁺ MF20⁺) and distal late-differentiating (ZsYellow⁺ MF20⁻) populations. The proximal OFT progenitor population is recruited to the arterial pole while the distal OFT progenitor population remains in a proliferative state in the pharyngeal region (van den Berg et al., 2009). We found that our results from the EdU incorporation assay are comparable between wild-type and nkx mutant embryos in the proximal precursor population contributing to the arterial pole (Fig. 3F). However, there is a statistically significant increase in the proliferation index of those OFT progenitors that reside in the distal pharyngeal region when comparing wild-type and nkx mutant embryos (Fig. 3B compared with D; Fig. 3G). In summary, our data suggest that Nkx genes regulate OFT size by restricting the proliferation of aSHF progenitors in the distal pharyngeal region and promoting differentiation as they migrate to form the arterial pole.

In light of our evidence suggesting that Nkx genes positively regulate the aSHF OFT progenitor pool, we examined possible roles in pSHF development given the relatively high incidence of atrial septal defects in individuals carrying NKX2-5 mutations. In the absence of clearly defined regulators of the IFT in zebrafish, we investigated Isl1, a LIM homeodomain transcription factor, as a potential candidate downstream of Nkx genes at the venous pole. Extensive studies elucidated that isl1 is essential for normal cardiac development across species (Brade et al., 2007; Cai et al., 2003; de Pater et al., 2009; Pandur et al., 2013; Witzel et al., 2012), delineates the SHF at both arterial and venous poles (Cai et al., 2003; Hami et al., 2011; Snarr et al., 2007a; Stoyek et al., 2015; Witzel et al., 2012), and is required for proper pacemaker development and function (Liang et al., 2015; Tessadori et al., 2012; Vedantham et al., 2015). In zebrafish, using an antibody that recognizes both Isl1 and Isl2 proteins (D'Amico et al., 2007; Hutchinson and Eisen, 2006; Witzel et al., 2012), Isl1/2⁺ cells are found near the OFT and IFT (de Pater et al., 2009; Hami et al., 2011; Witzel et al., 2012), and co-expression of Nkx2.5 in this population is evident during somitogenesis (Witzel et al., 2012). However, Isl2b appears to play a role in arterial pole development (Witzel et al., 2017) and Isl1 in venous pole development with a loss of late-differentiating CMs in the IFT in isl1^−/− embryos (de Pater et al., 2009). Thus, we postulated that Nkx genes inhibit isl1 expression at the SV. In support of our model, we found that Isl1 immunostaining is localized specifically to the IFT of wild-type embryos (Fig. 4A), while Isl1⁺ CMs extend throughout the atrium (S46⁺ region) in nkx2.5^−/−;nkx2.7^+/+ embryos (Fig. 4C) and reach the OFT in nkx2.5^−/−;nkx2.7^−/− embryos (Fig. 4E). We quantified Isl1⁺ nuclei and detected a progressive increase in the number of cells expressing this transcription factor as Nkx gene dose is diminished (Fig. 4F). Notably, the number of Isl1⁺ CMs in the nkx2.5^−/−;nkx2.7^−/− heart is approximately equivalent to the total number of CMs in wild-type embryos (Targoff et al., 2013). Next, we investigated Nkx2.5 and Isl1 expression patterns employing immunohistochemistry in embryos carrying Tg(nkx2.5:ZsYellow) at 52 hpf. This transgene faithfully recapitulates the endogenous nkx2.5 expression observed by in situ hybridization (data not shown). We find that Nkx2.5 and Isl1 are virtually mutually exclusive at the venous pole in wild-type embryos (Fig. 5A-A″). Nkx2.5 is expressed in a decreasing gradient from the OFT to the inflow of the atrium, whereas Isl1 expression is confined to SV with only a few cells co-expressing Nkx2.5 and Isl1 (Fig. 5A″). Alternatively, in nkx mutants, Isl1 expression is derepressed in all atrial CMs that normally express Nkx2.5 (Fig. 5B-E″). Taken together, our studies strongly support the in vivo suppression of Isl1⁺ cell fate by Nkx genes in the SV during cardiac development.

We then explored the timing of the Nkx-Isl1 interaction in further detail. We evaluated isl1 expression during early stages of cardiac differentiation in the ALPM. Our data illustrate that specification of isl1⁺ progenitors occurs normally at 10 somites and 15 somites (Fig. 5F-I and data not shown), but isl1 expression is enhanced at 26 hpf in nkx mutant compared with wild-type embryos (Fig. 5J-M). These findings indicate that Nkx genes are first required to inhibit isl1 at the heart tube stage. Using a novel transgene generated in our lab for ubiquitous nkx2.5 overexpression, Tg(hsp70l:nkx2.5-EGFP) (George et al., 2015), we tested the sufficiency of nkx2.5 to inhibit isl1 expression (Fig. 5N-Q). Following heat-shock activation of this inducible transgene at 21 somites, isl1 expression was repressed specifically in the inflow region at both 24 hpf and 52 hpf (Fig. 5P,Q). Altogether, these results demonstrate that Nkx2.5 is not required for isl1⁺ progenitor specification, but activation of Nkx2.5 transcription prior to heart tube elongation is sufficient to repress isl1 at the venous pole during chamber emergence.

Although recent work has shown the contribution of late-differentiating CMs to the venous pole in zebrafish (de Pater et al., 2009; Mosimann et al., 2015), little is known about the transcriptional regulatory networks guiding accretion and differentiation of this pSHF progenitor population. Given the ability of nkx2.5 to restrict Isl1 to the IFT (Fig. 5N-Q), we hypothesize that Nkx genes are essential in patterning the pSHF-derived CMs in this region. To test this hypothesis, we first performed a developmental timing assay to delineate the kinetics of pSHF progenitor accumulation using embryos expressing Tg(myl7:EGFP) and Tg(-5.1myl7:nDsRed2) (as in Fig. 1A-I). Intriguingly, our data reveal a statistically significant increase in pSHF-derived CMs, or EGFP⁺ DsRed⁻ CMs, at the venous pole in nkx2.5^−/−;nkx2.7^+/+ embryos along with further enhancement in nkx2.5^−/−;nkx2.7^−/− embryos when compared with wild-type embryos (Fig. 6A-I,N). However, the source of the increased late-differentiating CMs at the IFT remains unknown. This finding may reflect either increased proliferation of pSHF progenitors or enhanced accrual of pSHF-derived CMs.

Consistent with our data illustrating normal CM production at the arterial pole following the loss of Nkx gene function (Fig. S1), we observed no difference in proliferation indices at the venous pole following EdU incorporation at 24 hpf between nkx mutant and wild-type embryos (data not shown). Thus, we hypothesized that Nkx genes are required for pSHF progenitor proliferation prior to accretion to the IFT. We performed an EdU incorporation assay at 18 hpf in embryos carrying Tg(-5.1myl7:nDsRed2) to detect dividing CMs at the venous pole. Quantification of the proliferation indices for EdU⁺ DsRed⁺ MF20⁺ cells in the IFT region at this time point also revealed no statistically significant difference between wild-type and nkx mutant embryos (Fig. 6J-M,O). These data support the conclusion that the IFT expansion in nkx mutant embryos is most likely due to enhanced pSHF-derived CM differentiation.

Although recent studies have elucidated the existence of specialized conduction tissue in the zebrafish heart (Arrenberg et al., 2010; Chi et al., 2008; Milan et al., 2006) and have identified isl1⁺ as molecular marker of this population (Tessadori et al., 2012), the pathways controlling the development of the sinoatrial node (SAN) are still under investigation. Therefore, from evidence of conduction system abnormalities in individuals with NKX2-5 mutations (Benson et al., 1999; Elliott et al., 2003; Gutierrez-Roelens et al., 2002; Schott et al., 1998) and from our findings revealing that Nkx genes regulate isl1 expression (Figs 4 and 5), we posited that Nkx genes are fundamental in patterning the pSHF-derived CMs that contribute to the specialized nodal tissue in the SV. To address this issue, we explored the expression of a number of pacemaker cell genes: bmp4, tbx2b, hcn4 and shox2 (Blaschke et al., 2007; Espinoza-Lewis et al., 2009; Tessadori et al., 2012). In wild-type embryos, bmp4, tbx2b and hcn4 expression is restricted to the OFT, atrioventricular canal (AVC) and IFT at 52 hpf (Fig. 7A,F,K). Intriguingly, all three genes are expressed throughout the myocardium in nkx mutant embryos (Fig. 7C-E,H-J,M-O). We observe a similar expression pattern for shox2 when comparing wild-type and nkx mutant embryos (Fig. S2). Taken together, our findings illustrate that Nkx genes regulate regionalization of SAN markers to the IFT.

Next, we postulated that Isl1 acting downstream of Nkx genes is responsible for patterning the SAN and establishing the pacemaker activity of the zebrafish heart. Using a previously validated morpholino (MO) to knockdown Isl1 function (Hutchinson and Eisen, 2006), we observed that bmp4 expression is absent in wild-type and nkx2.5^+/+;nkx2.7^−/− embryos specifically at the IFT following islet1E2 MO injection, whereas normal AVC and OFT expression of bmp4 persists (compare Fig. 8A,B with F,G). These findings are consistent with evidence of selective inhibition of bmp4 expression in the IFT of the isl1^sa29 mutant (de Pater et al., 2009). Interestingly, inhibition of isl1 splicing in nkx mutant embryos reveals specific downregulation of ectopic atrial and IFT bmp4 expression (compare Fig. 8H-J with C-E) in the context of otherwise normal development. Together, our results demonstrate that the misregulation of bmp4 expression in nkx mutant embryos is mediated by isl1 overexpression, ultimately placing isl1 downstream of Nkx genes in defining the pSHF-derived specialized conduction cells of the IFT.

Finally, we explored the effects of the derepression of SAN genes on the electrophysiological phenotype in nkx mutant hearts. The establishment of electrical gradients enables the transition from peristaltic to sequential contraction in the developing heart and this patterning is regulated by differential contributions of FHF and SHF progenitor populations (Mosimann et al., 2015). We hypothesized that the loss of Nkx gene function could lead to a disruption in regional patterns of intercellular coupling given both the absence of ventricular identity (Targoff et al., 2013, 2008) and the expanded SAN gene expression domain in nkx mutant compared with wild-type embryos (Fig. 7).

To explore these issues, we measured action potential characteristics and conduction properties in isolated wild-type and nkx mutant hearts using high-resolution optical mapping (Fig. 9, Fig. S3). We did not detect any electrophysiological phenotypes in embryos carrying single-allele deficiencies in one or both nkx genes (nkx2.5^+/−;nkx2.7^+/+, nkx2.5^+/+;nkx2.7^+/− and nkx2.5^+/−;nkx2.7^+/−). When comparing wild-type and nkx2.5^+/+;nkx2.7^−/− embryos, there is no significant difference in the detailed action potential morphology (Fig. 9A,B), maximum upstroke velocity, [dF/dt]_max (Fig. 9C), or atrial and ventricular action potential durations (Fig. 9D). Furthermore, there are no observed differences between the atrial and ventricular conduction velocities in wild-type and nkx2.5^+/+;nkx2.7^−/− embryos (Fig. 9E,F,I). In contrast, disruption of nkx2.5 gene function affects multiple components of the action potential with significant slowing of the upstroke phase in both chambers (Fig. 9A-C), marked diminution of the maximum slope of the action potential upstroke (Fig. 9C), as well as a substantial prolongation of atrial action potential duration and accentuation of phase 4 depolarization in atrial CMs (Fig. 9A,D). Moreover, homozygous loss of nkx2.5 function leads to severe slowing in action potential propagation, as demonstrated by closer spacing of the isochrones and significantly lower mean conduction velocities measured across both cardiac chambers (Fig. 9G-I). Interestingly, although there is failure of the normal specification of physiological AV delay in the hearts of nkx2.5^−/−;nkx2.7^+/+ embryos (Fig. 9K,L), there remain significant differences in action potential characteristics between the ventricle and atrium (Fig. 9J,K and Fig. S3A-C,E-G). Although the hearts from nkx2.5^+/+;nkx2.7^−/− embryos did not exhibit any detectable electrophysiological phenotype (Fig. 9F), combined loss of function of both nkx homologs in nkx2.5^−/−;nkx2.7^−/− embryos leads to complete loss of electrophysiological chamber identity and further slowing of the action potential upstroke phase in the atrium (Fig. 9A,C,J and Fig. S3D). The loss of action potential duration differences between chambers is also evident in nkx2.5^−/−;nkx2.7^−/− embryos (Fig. 9D,K and Fig. S3H), including the emergence of a pacemaker phenotype in ventricular cells (Fig. 9B) and slowing of the conduction velocity (Fig. 9H,I,L). Interestingly, there are subtle, but significant, paradoxical effects on absolute atrial and ventricular conduction velocities in nkx2.5^−/−;nkx2.7^−/− when compared with nkx2.5^−/−;nkx2.7^+/+ embryos. In the atrium, the nkx2.5^−/−;nkx2.7^−/− embryos exhibit even slower conduction than the nkx2.5^−/−;nkx2.7^+/+ embryos. However, in the ventricle, there is an increase in the conduction velocity in the nkx2.5^−/−;nkx2.7^−/− embryos, although conduction remains markedly slowed and chamber identity is lost from an electrophysiological standpoint. Although these electrophysiological attributes of the nkx mutant embryos reflect severe defects in intercellular coupling, the individual CM properties exhibit a phenotype that is more advanced developmentally than that observed in the primitive heart tube. Altogether, these data highlight the independent regulation of discrete CM parameters throughout early cardiac development and chamber identity maintenance.

Together, these findings offer new insights into the roles of Nkx genes in SHF development, highlighting the differential regulation at the arterial and venous poles. We show that Nkx genes promote the establishment of the SHF CM progenitor pool and differentiation of this population at the OFT. Our studies also reveal that Nkx genes are essential for the proper patterning of the venous pole by regulating the maintenance of working myocardial properties through repression of isl1. Following the loss of Nkx gene function, the entire myocardium adopts an Isl1⁺ identity associated with activation of the SAN genes bmp4, tbx2b, hcn4 and shox2. Misregulation of these SAN genes is associated with electrophysiological evidence of expansion of the pacemaker phenotype across both chambers, with cellular characteristics suggesting diminished sodium channel availability at the membranes, features of calcium-dependent action potentials, and diastolic phase 4 potentials. Together, our findings implicate Nkx genes in combinatorial roles for anterior-posterior patterning of chamber identity, regional integration of FHF and SHF territories, and maturation of discrete cassettes of physiological CM differentiation within the heart tube. Moreover, these cellular abnormalities induced by the loss of Nkx gene function in the two-chambered zebrafish heart offer a potential window into the morphological and conduction system defects in individuals with CHD that harbor NKX2-5 mutations.

Through illustration of the differential functions of Nkx factors in the aSHF and pSHF, our studies complement and extend work in other model organisms. Previous studies have revealed a role for Nkx2-5 in limiting cardiac progenitor specification and favoring SHF proliferation (Prall et al., 2007). Yet in zebrafish, early specification of SHF progenitors appears normal given the expression patterns of nkx2.5, nkx2.7, hand2 (Targoff et al. 2013) and isl1 (Fig. 5F-I) in nkx mutant embryos. However, nkx2.5 has been shown to regulate SHF progenitor proliferation at the proximal OFT via ltbp3 (Guner-Ataman et al., 2013). Through detailed dissection of the proliferative potential of the proximal and distal OFT progenitors, our results further elucidate the role that Nkx genes play in limiting the aSHF-derived CM population by regulating the timing of progenitor differentiation at the arterial pole. In view of studies illuminating a common progenitor of heart and head muscle derivatives in other models (Harel et al., 2012; Lescroart et al., 2015, 2010; Nagelberg et al., 2015; Wang et al., 2013), we postulate that aSHF progenitors could also adopt a head muscle fate in the absence of Nkx gene function. Future experiments would be required to test this hypothesis, which may offer additional insights into the diminished late-differentiating cells at the OFT of nkx mutant embryos. Taken together, although other mechanisms may play a role, our findings highlight a novel function of Nkx genes in establishing an aSHF progenitor pool necessary to assemble an arterial pole of the proper size and morphology.

Our data also illustrates that, in the absence of nkx2.5, the atrial myocardium adopts a venous pole identity characterized by misexpression of isl1 and SAN genes associated with failure of specification of the cardiac conduction system (CCS) tissue at the AV boundary. Further graded expansion of this abnormal CM differentiation, including manifestation of increased cellular automaticity and pacemaker-like behaviors throughout the entire ventricle, is observed following the combinatorial loss of nkx2.5 and nkx2.7. Knockdown of isl1 is sufficient to rescue the effects of nkx mutations on SAN gene expression, suggesting that Isl1 repression by Nkx genes patterns the venous pole of the heart and delineates the specialized CCS tissue. However, the molecular mechanisms by which Nkx genes control IFT development through refinement of isl1 expression have yet to be determined. Specifically, the role of bmp4 in establishing electrophysiological CM characteristics requires further investigation. Yet our results offer a potential explanation for the electrophysiological malfunction associated with the loss of fast-conducting properties and defects in specification of the ventricular conduction system lineage in Nkx2-5-deficient models (Moskowitz et al., 2007; Pashmforoush et al., 2004). Furthermore, we propose that the pivotal role for nkx genes in establishing a venous pole phenotype and SAN identity could explain progressive AV conduction defects into adulthood that are exhibited in individuals with NKX2-5 mutations (Jay et al., 2003).

We have previously shown that ventricular CMs transdifferentiate into atrial CMs, demonstrating plasticity of the differentiated myocardium in nkx mutant embryos (Targoff et al., 2013). It is possible that flexibility of the differentiated CMs results in sarcomeric rearrangement that allows for fate transformation in the nkx2.5^−/−;nkx2.7^−/− embryos. In support of this concept, a recent report demonstrates that ablation of ventricular CMs in the zebrafish embryos leads to atrial-to-ventricular transdifferentiation with associated sarcomeric restructuring (Zhang et al., 2013). Our electrophysiological studies further substantiate this hypothesis through their demonstration that graded loss of nkx alleles regulates multiple elements of the functional identity of the atrial and ventricular myocardium. Although atrial and ventricular action potential features and conduction properties are distinct in wild-type, nkx2.5^+/+;nkx2.7^−/− and nkx2.5^−/−;nkx2.7^+/+ hearts, each of the individual null alleles exhibits discrete effects on resting membrane potentials, CM activation, action potential morphology and conduction velocity. Moreover, the electrophysiological differences between atrial and ventricular CMs are completely abolished in nkx2.5^−/−;nkx2.7^−/− embryos. Despite these findings, there are discordances in the effects of individual genes on electrophysiology in the atrium and ventricle that implicate additional regulatory partners in the specification of the complex regional electrical function necessary for a vertebrate heart.

Taken together, these studies on the functions of Nkx genes in the aSHF and pSHF highlight important concepts in our understanding of CM differentiation. Changes in the expression of individual transcription factors, signaling molecules and ion channels reveal discrete physiological modules and can remodel nuanced functional characteristics in CMs. Given this malleability, the notion of a stable differentiated state does not fully reflect the resolution of modern cellular biology, which has revealed phenotypes that are a result of combinations of discrete and overlapping protein repertoires. These data parallel in vitro studies in iPS-derived differentiated cell types. Despite a physiological resemblance to mature CMs, these populations demonstrate remarkable cellular heterogeneity in detailed electrophysiological phenotypes and transcriptional profiles that more closely resemble one of the parent fibroblast than of the CM being studied. Our data suggest that individual proteins may be sufficient and necessary to induce and maintain plastic states where the regenerative potential in the heart might be realized. Thus, it would be valuable to address the fundamental nature of cardiomyocyte differentiation at the level of individual physiologic functions in order to obtain a better understanding of the different potential for cardiac regeneration between zebrafish and higher vertebrates.

We used zebrafish carrying the following previously described mutations and transgenes: nkx2.5^vu179 (Targoff et al., 2013), nkx2.7^vu413 (Targoff et al., 2013), Tg(-5.1myl7:nDsRed2)^f2 (Mably et al., 2003), Tg(myl7:EGFP)^twu277 (Huang et al., 2003), Tg(nkx2.5:ZsYellow)^fb7 (Zhou et al., 2011) and Tg(hsp70l:nkx2.5-EGFP) (Targoff et al., 2013). All zebrafish work followed Institutional Animal Care and Use Committee-approved protocols.

Embryos were injected at the one-cell stage with 6 ng of the validated morpholino (MO) islet1E2 (5′-TTAATCTGCGTTACCTGATGTAGTC-3′) (Hutchinson and Eisen, 2006). Uninjected sibling embryos were processed as controls.

Embryos from outcrosses of fish carrying Tg(hsp70l:nkx2.5-EGFP) were maintained at 28.5°C and exposed to heat shock at 37°C at desired stages (21 somites to 24 somites). To implement heat shock, 50 embryos were placed in 2.5 ml of embryo medium in a 35 mm Petri dish on top of a covered heat block for 1 h. Three hours after the initiation of heat shock, transgenic embryos were identified by visualization of ubiquitous EGFP expression. Non-transgenic sibling embryos exposed to heat shock served as controls.

When necessary, embryos were depigmented after overnight fixation in 4% paraformaldehyde at 4°C. They were washed twice in PBT (PBS-0.1%Tween) then incubated in 3% H₂O₂, 0.5% KOH and 0.1% Tween for 8 min at 26 h post fertilization (hpf) and 18 min at 52 hpf. Subsequently, the embryos were washed twice again in PBT and stored in methanol at −20°C.

We performed whole-mount in situ hybridization as previously described (Yelon et al., 1999) for mef2cb (ZDB-GENE-040901-7), isl1 (ZDB-GENE-980526-112), bmp4 (ZDB-GENE-980528-2059), tbx2b (ZDB-GENE-990726-27), hcn4 (ZDB-GENE-050420-360) and shox2 (ZDB-GENE-040426-1457).

Whole-mount double immunofluorescence was conducted using a protocol described previously (Alexander et al., 1998) for the following primary antibodies: anti-Islet1/2 (1:50; 39.4D5, Developmental Studies Hybridoma Bank), anti-atrial myosin heavy chain (1:20; S46, Developmental Studies Hybridoma Bank), anti-RCFP (1:50; Clontech, 632475) and anti-GFP (1:100; Life Technologies, A11122). When employing the following primary antibodies, anti-Eln2 (1:500; Miao et al., 2007) and anti-sarcomeric myosin heavy chain (1:20; MF20, Developmental Studies Hybridoma Bank), another published protocol was used (Zhou et al., 2011). The following secondary antibodies were employed: goat anti-mouse IgG2b Alexa Fluor 488 or 568, goat anti-mouse IgG1 Alexa Fluor 488 or 568, goat anti-mouse IgG2a Alexa Fluor 488, goat anti-rabbit IgG Alexa Fluor 488 or 568, and goat anti-mouse IgG Cy5 (all at 1:200; Invitrogen).

Cardiomyocyte proliferation in embryos expressing Tg(-5.1myl7:nDsRed2)^f2 or Tg(nkx2.5:ZsYellow)^fb7 was assessed using the Click-iT EdU Imaging Kit (Life Technologies) with reported protocols (Cheesman et al., 2011; Mahler et al., 2010; Schindler et al., 2014). Following EdU incubation for 30 min on ice, embryos were chased for 24 h. Embryos were then fixed in 2% formaldehyde, washed once in PBS and washed twice in PBS-0.2% saponin. Subsequently, immunostaining was performed as described above followed by the Click-iT staining reaction. EdU⁺ ZsYellow⁺ MF20^+/− and EdU⁺ DsRed⁺ MF20⁻ cells were counted using ImageJ software, going through all z-stacks. We calculated the proliferation indices as the percentage of cells that incorporated EdU. Student's t-test (homoscedastic, two-tailed distribution) determined statistical significance between the means of cell number data sets.

Images were captured with a Zeiss M2Bio microscope and a Zeiss AxioCam digital camera, prior to processing with Zeiss AxioVision and Adobe Creative Suite software. Confocal imaging was performed with a Nikon A1R MP confocal microscope and z-stacks were analyzed with ImageJ.

For the developmental timing assay, embryos carrying Tg(-5.1myl7:nDsRed2)^f2 and Tg(myl7:EGFP)^twu277 were fixed at 52 hpf for 1 h in 1% formaldehyde, washed once in PBS, washed three times in PBS-0.2% saponin, deyolked and flat-mounted. Newly differentiated CMs were identified as green, but not red cells (EGFP⁺ DsRed⁻ CMs) and were counted on z-stacks using ImageJ software. To count Isl1⁺ cells, we performed immunostaining for Isl1 and S46 (as described above) and quantified individual CMs that were positive for Isl1 using ImageJ software on reconstructed z-stacks. Student's t-test (homoscedastic, two-tailed distribution) determined statistical significance between the means of cell number data sets.

PCR genotyping was performed on genomic DNA extracted from individual embryos following in situ hybridization, immunofluorescence or live imaging. Detection of nkx2.5^vu179 was performed using primers 5′-CAAACTCACCTCCACACAGG-3′ and 5′-TTACCATCCCGAACCAAAAC-3′ to generate a 144 bp fragment. Digestion of the mutant PCR product with Hinf1 generates 30 bp and 114 bp fragments. Analysis of nkx2.5^vu179 in embryos carrying Tg(hsp70l:nkx2.5-EGFP) was performed as previously described (George et al., 2015). Detection of nkx2.7^vu413 was executed using primers 5′-TTGTTGCAAATGGACACGTT-3′ and 5′-GGAGTGTCTCGCCGTTTTTA-3′ to generate a 158 bp fragment. Digestion of the mutant PCR product with MseI creates 28 bp and 130 bp fragments.

Embryonic zebrafish hearts were isolated at 72 hpf and stained with the transmembrane potential-sensitive dye, di-8-ANEPPS (Invitrogen), to measure action potentials using high-speed optical mapping as previously described (Mosimann et al., 2015; Panáková et al., 2010). Fluorescence intensities were recorded with a high-speed CCD camera (CardioCCD-SMQ, Redshirt Imaging) at 2000 frames per second. Activation times were determined as the times at which action potentials reached 50% of their amplitude during the depolarization phase at every location. This criterion for defining electrical activation was found in computer simulations to correlate with the time at which the maximum depolarizing Na⁺ current occurs during propagation (Fast and Kleber, 1995). Action potentials were temporally (800 Hz low-pass cutoff) and spatially (3×3 pixel average) filtered to enhance signal-to-noise ratios. To quantify the slope of the action potential upstroke phase, the maximum time derivative of the normalized action potential, [dF/dt]_max, was calculated. Action potential duration (APD) was defined as the absolute time difference between the decay phase and the upstroke phase of the local action potential at 50% depolarization. By fitting a second-order polynomial hypersurface to the three-dimensional image plane-activation time relationship, conduction velocity vectors were subsequently derived using an established algorithm (Bayly et al., 1998) with custom scripts. Conduction velocity vector fields were estimated from the local action potentials (Panáková et al., 2010). All electrophysiological measurements were averaged across the relevant anatomic segments for each heart.

We are grateful to Joshua Barber for expert zebrafish care and members of the Targoff Laboratory for their constructive input. We appreciate the critical reading of the manuscript by Lilianna Solnica-Krezel and Joshua S. Waxman. We also thank C. Geoffrey Burns and Caroline E. Burns for sharing the Tg(nkx2.5:ZsYellow) line.