Unique developmental trajectories and genetic regulation of ventricular and outflow tract progenitors in the zebrafish second heart field

During mammalian embryogenesis, cardiac progenitor cells in the anterior second heart field (SHF) produce the right ventricle and embryonic outflow tract (OFT) through late differentiation and accretion from pharyngeal mesoderm to the arterial pole of the heart tube (reviewed by Abu-Issa and Kirby, 2007; Vincent and Buckingham, 2010; Zaffran and Kelly, 2012). Before localizing to pharyngeal mesoderm, SHF progenitors are specified in lateral plate mesoderm when they initiate expression of several cardiac transcription factors, including nkx2.5 (Stanley et al., 2002; Zhang et al., 2014).

As conduits for blood exiting the heart, OFTs share common structural features across vertebrate species. Specifically, they comprise proximal (relative to blood flow) myocardial segments, middle segments with overlapping myocardium and smooth muscle, and distal segments composed exclusively of smooth muscle, all of which encase lumenal endocardium (Grimes et al., 2010). In higher vertebrates, the embryonic OFT is a relatively long transient structure with extensive myocardium (Waldo et al., 2005), which becomes intricately remodeled through wall rotation (Bajolle et al., 2006), tubular compaction (Watanabe et al., 1998), septation (Kirby, 2007) and aortic wedging (Kirby, 2007). These morphogenetic events serve to create an exclusive connection between each great artery and the appropriate ventricle. Defects in OFT morphogenesis or remodeling cause a class of common congenital heart defects (Benjamin et al., 2017) (i.e. conotruncal malformations) characterized by misalignment of the great arteries with the ventricles or a common arterial trunk, respectively (Kirby, 2007). In both cases, deoxygenated blood mixes with oxygenated blood, which necessitates surgical repair of the malformation. Therefore, deciphering the molecular and cellular determinants of OFT morphogenesis remains an important area of investigation for preventing and managing congenital heart defects.

Zebrafish hearts contain a single ventricle and OFT composed of a relatively short myocardial base and taller elastin2⁺ (Eln2⁺) smooth muscle collar known as the bulbus arteriosus, both of which surround OFT endocardium (Grimes et al., 2006; Miao et al., 2007). Unlike in higher vertebrates, the zebrafish OFT does not become remodeled or septated. Recent studies have demonstrated that half of the ventricular myocardium and three OFT lineages in zebrafish derive from cardiac progenitors with defining features of the anterior SHF (de Pater et al., 2009; Guner-Ataman et al., 2013; Hami et al., 2011; Lazic and Scott, 2011; Zhou et al., 2011). Specifically, SHF progenitors for the distal ventricular myocardium are specified bilaterally in nkx2.5⁺ anterior lateral plate mesoderm (ALPM) before migrating to the arterial pole of the linear heart tube, where they undergo late myocardial differentiation and accretion to the heart. Evidence for myocardial accretion was uncovered when dye-labeled pharyngeal mesodermal cells were subsequently localized to the distal ventricle (Hami et al., 2011; Zhou et al., 2011).

Based on genetic lineage-tracing and mutant analyses, it was assumed that SHF progenitors for the OFT, also specified as nkx2.5⁺ cells in the ALPM (Guner-Ataman et al., 2013; Hami et al., 2011), were either one and the same with SHF ventricular progenitors or followed the same trajectory to the arterial pole of the linear heart tube. This assumption was based on co-labeling of the distal ventricle and OFT in the lineage trace of ltbp3⁺ SHF cells and the observation that ltbp3 expression initiates in SHF progenitors as they reach the midline (Zhou et al., 2011). Moreover, four loss-of-function experimental groups – ltbp3 morphants (Zhou et al., 2011), mef2cab morphants (Lazic and Scott, 2011), smoothened null animals (Hami et al., 2011) and tbx1 mutants (Hami et al., 2011; Nevis et al., 2013) – exhibited reductions or deficiencies in both SHF-derived ventricular and OFT lineages. These findings suggested that a single progenitor population, compromised in these animals, might be giving rise to both the distal ventricle and OFT. In this scenario, after SHF progenitors complete myocardial accretion to the ventricle, the same progenitor population would produce OFT lineages. Despite the plausibility of this hypothesis, evidence to support or refute this notion has yet to be provided.

Here, we report that, contrary to the previous assumption, SHF ventricular and OFT progenitors take unique trajectories to the arterial pole of the heart. Whereas SHF ventricular precursors migrate directly to the midline and maintain nkx2.5 expression, OFT progenitors remain lateral and become temporarily sequestered in pharyngeal arch (PA) 2, where they downregulate nkx2.5. While there, OFT progenitors are in close apposition to precursors for PA2-derived head muscles (HMs) and hypobranchial artery (HA) endothelium, both of which are co-specified in the nkx2.5⁺ ALPM. Last, whereas mutations in fgf8a compromise SHF ventricular development, the OFT and PA2-derived head muscles are unaffected, demonstrating that progenitor compartmentalization in PA2 confers unique genetic regulatory mechanisms to this population. Taken together, our data illuminate previously unappreciated aspects of SHF, OFT and HM development in zebrafish that will become instrumental for guiding future studies designed to understand the cellular mechanisms of cardiopharyngeal phenotypes in zebrafish models of human congenital craniofacial and cardiovascular malformations.

We have reported previously that transgenic zebrafish embryos expressing ZsYellow from nkx2.5 cis-regulatory sequences display reporter fluorescence in the atrium, ventricle, OFT, ventral aorta and pharyngeal arch arteries (PAAs) by 60 hours post-fertilization (hpf) (Nevis et al., 2013; Paffett-Lugassy et al., 2013). Upon closer examination of 72 hpf Tg(nkx2.5:ZsYellow) larvae, immunostained for ZsYellow, we observed the reporter in a subset of HMs, including the ventral and middle PA1 (mandibular) muscles, as well as all five PA2 (hyoid) muscles (Fig. 1A-F) (Schilling and Kimmel, 1997). Additionally, ZsYellow was colocalized with endothelial cells composing the HAs (Fig. 1G-I). Last, ZsYellow was observed in the outermost layer of the pericardial sac (Fig. 1A-C). These data are consistent with, and provide higher resolution for, a previous description of ZsYellow expression in Tg(nkx2.5:ZsYellow) animals (Nagelberg et al., 2015).

To determine whether the presence of ZsYellow in HMs, HA endothelium and the heart mirrored ongoing nkx2.5 expression, we localized endogenous transcripts in 72 hpf wild-type animals by in situ hybridization. Although nkx2.5 transcripts were abundant in the cardiac chambers, they were absent from the HMs, HA endothelium and the smooth muscle collar of the OFT and underlying endocardium (Fig. S1A). The same structures were also devoid of ZsYellow transcripts in Tg(nkx2.5:ZsYellow) larvae (Fig. S1B-F), suggesting that their ZsYellow positivity (Fig. 1A-I, Fig. S1D-F) stems from residual protein produced earlier in development by progenitor cells for these lineages.

We sought to visualize and lineage trace HM and HA endothelial progenitor cells at progressively earlier developmental stages to identify when and where they express nkx2.5. A previous study reported that myoblasts for PA1 and PA2 muscles initiate MyoD⁺ expression within the mesodermal cores of their respective PAs by 55 hpf (Schilling and Kimmel, 1997). The arches themselves arise through anterior-to-posterior segmentation of the pharynx, a process that is largely complete by 30 hpf (Schilling and Kimmel, 1994). Therefore, we examined PA1 and PA2 in Tg(nkx2.5:ZsYellow) animals at two stages prior to the initiation of MyoD expression but after pharyngeal segmentation in an attempt to visualize ZsYellow⁺ myoblast precursors. To unequivocally identify the mesodermal cores of the arches, we used a second transgene, Tg(fli1a:GFP), that drives GFP expression in cranial neural crest-derived cells surrounding the cores (Lawson and Weinstein, 2002).

At a developmental stage corresponding to the end of pharyngeal segmentation (28 hpf), ZsYellow⁺ cells were observed in PA1 and PA2 surrounded by, but not overlapping with, GFP fluorescence (Fig. 2A-D, Fig. S2A-C), indicative of mesodermal core localization. At 48 hpf, clusters of ZsYellow⁺ cells, also non-overlapping with surrounding GFP fluorescence, were located ventrally in both arches and dorsally in PA2 (Fig. 2E-H, Fig. S2D-F), all of which represent locations where myoblasts initiate MyoD expression by 55 hpf (Schilling and Kimmel, 1997). Notably absent was a cluster of ZsYellow⁺ cells in a dorsal location within PA1 (Fig. S2D-I) where progenitors reside for the two dorsal PA1 muscles, which are negative for ZsYellow in Tg(nkx2.5:ZsYellow) animals (Fig. 1D-F). Last, we corroborated that ZsYellow⁺ cells reside in the mesodermal cores of PA2 using the Tg(sox10:GFP) reporter for neural crest-derived cells (Carney et al., 2006) (Fig. S2J-L).

To determine whether ZsYellow reporter fluorescence in PA1 and PA2 (Fig. 2A-H, Fig. S2A-F) reflected ongoing nkx2.5 expression, we performed in situ hybridization for endogenous transcripts in wild-type embryos at the same developmental stages. At 28 hpf, we detected nkx2.5 expression in PA2 (Fig. S1G). We reported previously that nkx2.5 expression becomes undetectable in the pharynx by 48 hpf (Paffett-Lugassy et al., 2013), which demonstrates that nkx2.5 messages disappear from PA2 between 28 hpf and 48 hpf. We did not detect nkx2.5 expression in PA1 at 28 hpf (Fig. S1G). Based on lineage-tracing experiments detailed below, we predict that nkx2.5 expression occurs transiently in PA1 between the 14-somite stage (ss; 16 hpf) and 28 hpf. Consistent with this conclusion, we detected ZsYellow transcripts in PA1 of Tg(nkx2.5:ZsYellow) animals at 28 hpf (Fig. S1H), likely because ZsYellow mRNAs produced transiently in PA1 from nkx2.5 cis-regulatory elements persist longer than endogenous nkx2.5 transcripts.

Next, we prospectively lineage traced the presumptive ZsYellow⁺ myoblast precursors in PA1 and PA2 to uncover their fates. To achieve this, we used a previously validated transgenic strain, Tg(nkx2.5:Kaede) (Guner-Ataman et al., 2013; Paffett-Lugassy et al., 2013), that expresses the photoconvertible Kaede protein from nkx2.5 cis-regulatory elements. A confocal laser scanning microscope was used to achieve targeted photoconversion of Kaede⁺ cells in PA1 or PA2 at 28 hpf. The same animals were evaluated at 72 hpf for labeling of HMs with the photoconverted Kaede protein (Kaede^PC). Following unilateral photoconversion of Kaede⁺ cells in PA1 (Fig. 2I), we observed Kaede^PC signals in the ventral and middle PA1 muscles on the same side of the embryo (Fig. 2J-L). Unilateral photoconversion of Kaede⁺ cells in PA2 (Fig. 2M) caused all five PA2 muscles on the same side of the embryo to become positive for Kaede^PC (Fig. 2N-P). These data demonstrate that nkx2.5 cis-regulatory elements label myoblast precursors for all PA1 and PA2 muscles, except the two dorsal PA1 muscles. Moreover, to our knowledge, this study represents the first prospective lineage tracing of myoblast precursors localized in the anterior PAs.

Unilateral photoconversion of Kaede⁺ cells in PA1 or PA2 (Fig. 2I,M) also caused Kaede^PC labeling of HA endothelium (Fig. 2J,K,N,O), indicating that HM and endothelial progenitors are one and the same or closely apposed in the PAs. Interestingly, although myoblast precursors give rise to muscles exclusively on the same side of the embryo, HA endothelial precursors cross the midline because unilateral photoconversion of Kaede⁺ cells resulted in bilateral Kaede^PC labeling of HA endothelium (Fig. 2J,K,N,O). Unexpectedly, photoconverting Kaede⁺ cells in PA2 caused cells within the OFT on the photoconverted side to become labeled with Kaede^PC (Fig. 2M-O; see below), suggesting that SHF OFT progenitors are also among HM and HA endothelial precursors in PA2.

nkx2.5 transcripts become detectable initially in zebrafish embryos by 6 ss (12 hpf) (Lee et al., 1996; Schoenebeck et al., 2007) in a bilateral population of ALPM cells comprising progenitors for ventricular myocardium (Guner-Ataman et al., 2013; Hami et al., 2011; Schoenebeck et al., 2007; Serbedzija et al., 1998), OFT lineages (Guner-Ataman et al., 2013; Hami et al., 2011) and endothelial cells that make up the VA and PAAs (Paffett-Lugassy et al., 2013). We hypothesized that the HM, HA endothelial and OFT progenitors residing in PA1 and/or PA2 (Fig. 2I,M) initiate nkx2.5 expression in the ALPM.

To test this hypothesis, we unilaterally photoconverted Kaede⁺ cells in the anterior half of the ALPM in 14 ss (16 hpf) Tg(nkx2.5:Kaede) embryos (Fig. 3A-C) and analyzed the same embryos at 28 hpf for Kaede^PC expression in PA1 and PA2. As anticipated, we detected Kaede^PC fluorescence in the ventricular segment of the heart tube and in SHF progenitors at its arterial pole (Fig. 3D-F) (Zhou et al., 2011). We also observed Kaede^PC+ cells in PA2 (Fig. 3D-F), demonstrating that these progenitors are specified in the ALPM. PA1, although negative for Kaede^PC+ cells, harbored a cluster of cells expressing the native Kaede protein (Fig. 3D-F), indicating that they initiated Kaede expression after 14 ss.

Next, we used a complementary experimental approach to test the hypothesis that HM and endothelial progenitors in PA2, but not PA1, derive from nkx2.5⁺ cells in the ALPM. We employed a previously characterized two-part transgenic system for prospective lineage tracing of nkx2.5⁺ cells (Guner-Ataman et al., 2013; Paffett-Lugassy et al., 2013). The system comprises a Tg(nkx2.5:CreER^T2) driver transgene, to achieve 4-OHT-induced activation of Cre recombinase, and a ubiquitous reporter transgene termed ubi:Switch (Mosimann et al., 2011). In double transgenic embryos, we activated Cre transiently between tail bud and 12-16 ss (10 to 15-17 hpf), a developmental window that overlaps with the initiation of nkx2.5 expression in the ALPM (Lee et al., 1996; Schoenebeck et al., 2007), and analyzed embryos for reporter expression at 72 hpf. As documented previously, we observed reporter fluorescence in the ventricle, OFT and PAAs (Fig. 3G-I and data not shown) (Guner-Ataman et al., 2013; Paffett-Lugassy et al., 2013). Furthermore, in these animals, all five PA2 muscles (Fig. 3G-K) and HA endothelium (Fig. 3L) were labeled, demonstrating that HM and HA endothelial cells are newly appreciated fates of nkx2.5⁺ cells in the ALPM. We never observed PA1 muscles labeled when 4-OHT exposure ceased at 12 ss. When 4-OHT was removed at 16 ss, a minority of animals showed labeling in PA1 muscles (data not shown), which provides further evidence that PA1 muscle progenitors are specified subsequent to those for PA2.

Finally, we monitored the behavior of ZsYellow⁺ cells in the ALPM of nkx2.5:ZsYellow animals using two-photon excitation time-lapse microscopy. This analysis suggests that ventricular myocardial progenitors are specified in anterior and medial positions relative to pharyngeal progenitors, which reside more laterally and posteriorly (Movies 1 and 2), consistent with the conclusion that progenitor subpopulations are specified in locations approximating their future destinations.

During our prospective lineage tracing of Kaede⁺ cells in PA2, we unexpectedly observed Kaede^PC+ cells in the OFT of the heart (Fig. 2N,O), a structure known to descend from nkx2.5⁺ cells in the zebrafish ALPM (Guner-Ataman et al., 2013) but not by way of PA2. To obtain a qualitative measure of the OFT contributions made by progenitors in PA2, we bilaterally photoconverted them in 28 hpf Tg(nkx2.5:Kaede) embryos (Fig. 4A) and analyzed the same animals at 72 hpf for the breadth of Kaede^PC labeling in the OFT. The vast majority of cells carrying the native Kaede protein also carried Kaede^PC (Fig. 4B-D), indicating that most, if not all, of these OFT cells derive from progenitors in PA2.

It was presumed previously, based on ltbp3 localization and lineage tracing (Zhou et al., 2011), that SHF OFT progenitors were among the nkx2.5⁺ cells that migrate directly to the midline to produce ventricular myocardium (see Introduction). Knowing that the OFT is derived, in large part, from progenitors temporarily sequestered in PA2, we sought to learn whether any OFT cells derive from nkx2.5⁺ cells that migrate directly to the midline after specification. To achieve this, we photoconverted Kaede⁺ cells in the cardiac cone (Fig. 4E,F), an intermediate midline structure that becomes the heart tube (Stainier et al., 1993). Consistent with previous studies (Guner-Ataman et al., 2013; Hami et al., 2011; Schoenebeck et al., 2007; Serbedzija et al., 1998), Kaede^PC labeling was observed throughout the ventricular myocardium (Fig. 4G,H). However, the OFT was devoid of Kaede^PC (Fig. 4G,H), instead carrying the native Kaede protein, demonstrating that the earliest population of nkx2.5⁺ cells migrating to the midline does not contribute cells to the OFT. This observation suggests that the OFT labeling we observed previously in Tg(ltbp3:TagRFP2Acre) animals resulted from ltbp3 expression in the OFT proper, or its PA2 progenitors, rather than from the midline population overlapping the cardiac cone at 23 ss (Zhou et al., 2011).

To determine which OFT lineages derive from nkx2.5⁺ progenitors in PA2, we attempted to colocalize Kaede^PC+ OFT cells with transgenic reporters in live embryos. However, this approach proved technically impossible because of spectral overlap between the native or photoconverted Kaede protein and available reporters. Moreover, Kaede^PC becomes undetectable after fixation, which prevents colocalization by double immunofluorescence staining or in situ hybridization. Nonetheless, based on the broad localization of lineage-traced cells in the OFT (Fig. 4C,D), we speculate that all three OFT lineages are labeled. Specifically, Kaede^PC+ cells were observed at the base of the OFT (Fig. 5C-E,I-K,O-Q) where cardiomyocytes are localized [Fig. 5A-B′; (Grimes et al., 2006)]. They also resided within the Eln2⁺ smooth muscle wall of the OFT (Fig. 5C-E,F-H,L-N, and O-Q), which extends from the myocardial base to the ventral aorta (Fig. 5A-B′) (Grimes et al., 2006; Miao et al., 2007). Last, Kaede^PC+ cells were observed internally (Fig. 5I-Q), where OFT endocardium forms the lumen (Fig. 5A-B′). Taken together, our data identify PA2 as a novel source of three OFT lineages during zebrafish embryogenesis. A small number of neural crest-derived cells have been reported to colonize the OFT but subsequent to the stages we analyzed (Cavanaugh et al., 2015).

Finally, to determine if the observed partitioning of SHF ventricular and OFT progenitors confers each population with unique genetic sensitivities, we analyzed acerebellar (ace) mutant embryos, which harbor a mutation in fgf8a (Reifers et al., 1998), for defects in HM and OFT morphogenesis. Previous studies have demonstrated that inactivation of fgf8a, or chemical inhibition of FGF receptor activity, results in a deficiency of SHF progenitors at the arterial pole (Lazic and Scott, 2011; Zeng and Yelon, 2014) and reductions in SHF-derived ventricular cardiomyocytes (de Pater et al., 2009; Lazic and Scott, 2011). We sought to determine if ace embryos also harbor defects in OFT and HM morphogenesis. First, we analyzed 28 hpf ace animals carrying the nkx2.5:ZsYellow transgene, which labels extra-cardiac SHF progenitors at the arterial pole of the linear heart tube (Zhou et al., 2011) and precursors for HM, HA endothelium and OFT lineages in PA2 (Fig. 2M-P, Fig. 4A-D). Consistent with previous studies (Reifers et al., 1998; Zeng and Yelon, 2014; Zhou et al., 2011), extra-cardiac SHF ventricular progenitors were reduced or absent (Fig. 6A,B). However, the ZsYellow⁺ progenitor population in PA2 was grossly unaffected (Fig. 6A,B). Next, we analyzed 72 hpf ace animals double immunostained to visualize ventricular, OFT and head muscle (MF20 antibody) as well as Eln2⁺ OFT smooth muscle cells. Consistent with previous reports, the ventricles in ace mutants were smaller than those in siblings (Fig. 6C,D) owing to a reported combination of ventricular progenitor specification and cardiomyocyte accretion defects (de Pater et al., 2009; Marques et al., 2008). By contrast, the OFTs in ace animals appeared grossly unperturbed (Fig. 6C,D), containing normal numbers of Eln2⁺ smooth muscle cells (Fig. 6E-G), indicating that the mutation preferentially impacts ventricular morphogenesis. In greater than half of the mutant animals (14/25), the PA1 and PA2 HMs were indistinguishable from siblings (Fig. 6H-K). In the rest, HMs were largely present but partially mispatterned (Fig. S3A,B), perhaps attributable to pharyngeal laterality defects, which have been reported to affect a similar fraction of ace animals (Albertson and Yelick, 2005). Taken together, these data demonstrate that the physical partitioning of nkx2.5⁺ SHF ventricular and OFT progenitors, which occurs soon after their specification, confers each population with unique genetic requirements during their maturation.

Our study highlights previously unappreciated aspects of SHF, OFT and HM development in zebrafish that appear to be consistent with findings in higher vertebrates. The nkx2.5 lineage in mouse embryos gives rise to head muscles derived from PA1 and PA2 (Harel et al., 2009). Furthermore, dye-tracing experiments in avian and rodent embryos suggest that progenitor cells in the anterior PAs contribute myocardium to the embryonic outflow tract (Kelly et al., 2001; Nathan et al., 2008; Tirosh-Finkel et al., 2006). In addition, retrospective clonal analysis in the mouse uncovered a clonal relationship between OFT cardiomyocytes and head muscles that derive from PA2 (Lescroart et al., 2010), suggesting that OFT progenitors are also temporarily sequestered in PA2 in mammals. Despite these precedents in higher vertebrates, our study is the first to achieve spatiotemporally localized prospective lineage tracing of a genetically labeled HM and OFT progenitor population in PA2. Although our studies cannot reveal that, as in higher vertebrates (Lescroart et al., 2010), HM and OFT progenitors share a common progenitor, it is tempting to speculate that one exists based on the similarities in specification and shared trajectories to PA2. Taken together, our findings uncover additional similarities between SHF, OFT and HM development in zebrafish and higher vertebrates that validate the use of zebrafish in deciphering SHF biology in the context of human congenital heart disease models.

Zebrafish adults and embryos were maintained as previously described (Westerfield, 2007), which was approved by the Massachusetts General Hospital Institutional Animal Care and Use Committee. The strains used in experimental studies included: wild-type TuAB; ace mutants (Reifers et al., 1998); Tg(ubi:loxP-EGFP-loxP-mCherry)^cz1701 [Tg(ubi:Switch)] (Mosimann et al., 2015); Tg(fli1a:EGFP)^y1 (Roman et al., 2002); Tg(fli1a:nEGFP)^y7 (Roman et al., 2002); Tg(-36nkx2.5:ZsYellow)^fb7 (Zhou et al., 2011); TgBAC(-36nkx2.5:ERCreERt2)^fb8 (Guner-Ataman et al., 2013); TgBAC(-36nkx2.5:Kaede)^fb9 (Guner-Ataman et al., 2013); and Tg(-4.9sox10:EGFP)^ba2 (Carney et al., 2006). Unless specified, live embryos were kept at 28.5°C and anesthetized in standard embryo media containing 0.4% tricaine (ethyl 3-aminobenzoate methanesulfonate, MS222; Sigma).

Adult zebrafish carrying the nkx2.5:ERCreER^t2 driver transgene were crossed with Tg(ubi:loxP-EGFP-loxP-mCherry)^cz1701 animals. The resulting progeny were collected according to standard zebrafish protocols (Westerfield, 2007) and raised to tailbud stage. For experimental treatments, tailbud stage embryos were incubated in 10 μM 4-hydroxytamoxifen (4-OHT) in embryo medium (E3) for a total of 5-7 h. For mock treatments, 4-OHT was replaced with an equivalent volume of 100% ethanol, which is the standard solvent for the 10 mM 4-OHT stock solution. After 5-7 h, the 12-16 ss embryos were washed with 50 ml E3 three times and raised to 72 hpf for subsequent antibody staining and imaging. The presence of the ubi:loxP-EGFP-loxP-mCherry allele was identified by screening for ubiquitous EGFP activity. The presence of the nkx2.5:ERCreER^t2 allele was inferred from evidence of Cre activity through mCherry expression.

Single whole-mount in situ hybridizations were performed in glass vials as previously described (Thisse and Thisse, 2008). Digoxygenin-labeled antisense RNA probes to nkx2.5 (Reiter et al., 1999) and ZsYellow were synthesized using a DIG RNA Labeling Kit (SP6/T7; Roche). To generate the ZsYellow riboprobe, a plasmid containing the full-length ZsYellow sequence (pCRBIITOPO ZsYellow BamHI/ClaI rev) was digested with NotI and transcribed with SP6 polymerase. A blue (NBT/BCIP) chromogenic substrate was used (Roche). For immunohistochemistry, manipulations were performed at room temperature unless otherwise stated. Embryos were fixed overnight in 4% PFA at 4°C and then washed at room temperature with phosphate-buffered saline (pH 7.4), 0.1% Tween-20 (PBST) four times for 10 min. Embryos were incubated in a bleach solution (0.5% KOH/3% H₂O₂/0.1% Tween-20) in the dark until pigment was removed and washed four times for 10 min with PBST. Permeabilization was achieved by incubating embryos in PBS/0.5% Triton X-100 (PBSTx) for 2 h followed by incubation in block solution (5% bovine serum albumin/5% goat serum in PBST) for 2 h. Embryos were incubated with primary antibodies diluted in block solution overnight at 4°C. Embryos were washed six times for 20 min in PBST and incubated with secondary antibodies diluted in block (1:500) overnight at 4°C. Secondary antibodies were removed by washing with PBST six times for 20 min prior to imaging or storage at 4°C in PBST. Primary antibodies specific for GFP (1:50, A11122, Invitrogen), MF20 (1:50, University of Iowa), Eln2 (1:1000) (Miao et al., 2007), mCherry (1:50, ab167453, Abcam) and ZsYellow (1:50, 632475, Clontech).

Mutant ace embryos were identified by PCR amplification and sequencing of the mutation-containing region of the genome (Reifers et al., 1998) using the following primers: 5′-ACG GAC ACA TTT GGG AGT CG-3′ and 5′-GGT GGG ATT CTT CTC ATG CAT TC-3′. The latter primer was used to sequence the purified PCR product.

Live TgBAC(-36nkx2.5:Kaede)^fb9 embryos were positioned and embedded in 0.9% low melting point agarose in 35 mm glass-bottom Petri dishes, covered with embryo media and 0.4% Tricaine, and photoconverted with either a Zeiss LSM5 Pascal Laser-Scanning Microscope or a Nikon Eclipse Ti Confocal Microscope, as previously described (Guner-Ataman et al., 2013). Animals were pre-screened for signs of spontaneous photoconversion, which was never observed. Photoconversion was then achieved by manually setting the desired exposure area on the animal and performing a brief z-stack under UV light. Animals with imprecise photoconversion were excluded from further analysis. Post-photoconversion images were collected with the 488 and 546 lasers. Embryos were carefully extracted from the agarose gel and raised individually until subsequent analysis. Extreme care was taken to minimize light exposure for the duration of the experiment.

Embryos processed for in situ hybridization were mounted on glass slides in 100% glycerol (Thisse and Thisse, 2008). For fluorescent imaging, embryos were embedded in 0.9% low melting point agarose (NuSieve GTG Agarose, Lonza) in 35 mm glass-bottom Petri dishes (MatTek Corporation). Fixed embryos were covered with PBST. Live embryos were covered with embryo media containing 0.4% tricaine. To arrest cardiac activity, 80 µM 2,3-butanedione monoxime (Sigma Aldrich) was also added. Live or processed embryos were imaged on a Nikon 80i compound microscope. Images of embryos processed by in situ hybridization were captured using a mounted Excelis AU600HDS HD Camera (Accu-Scope) and the resulting images were focus-stacked with Zerene Stacker Software (Build T201412212230). Confocal microscopy was performed using either a Zeiss LSM5 Pascal Laser-Scanning Microscope or a Nikon A1SiR Confocal Microscope. Time-lapse imaging was performed using a 514 nm argon laser and 25× objective on a Nikon A1R Multiphoton Confocal Microscope. Embryos were embedded in low-melt agarose to prevent movement. Z-stack images and time-lapse movies were viewed and analyzed with ImageJ (Schneider et al., 2012) and Fiji (Schindelin et al., 2012) software.

Eln2⁺ OFT smooth muscle cell nuclei were quantified in Fiji by generating RGB images of confocal z-stacks in which the DAPI signals were pseudocolored blue and the Eln2 signals were pseudocolored green. Nuclei in cells surrounded by Eln2 were identified while scrolling through the z-stacks and marked with dots in the red channel using the paintbrush tool to avoid double counting. After splitting the three channels, the 3D object counter tool was used to quantify the number of dots present in the red channel, which corresponded to the total number of Eln2⁺ cell nuclei in the OFT.

We thank J. M. González-Rosa from the Burns laboratory for useful discussions and advice on immunohistochemistry, confocal imaging and ace genotyping; J. Galloway (Massachusetts General Hospital) for providing expertise on muscle populations and early muscle development; and T. Schilling (University of California, Irvine) for providing expertise on pharyngeal segmentation and muscles.