Pharyngeal pouches provide a niche microenvironment for arch artery progenitor specification

During vertebrate development, the pharyngeal arch arteries (PAAs), also known as aortic arch arteries, are transient embryonic blood vessels that connect the heart to the dorsal aorta (Hiruma et al., 2002). These arteries form in a cranial-to-caudal sequence, and in mammals (or amniotes) are followed by the regression of the first and second PAAs, whereas the PAA3, PAA4 and PAA6 undergo asymmetric remodeling and contribute to the carotid arteries and great vessels of the heart, including the aorta and pulmonary arteries (Congdon, 1922; Hiruma et al., 2002). Improper embryonic development of the PAAs may cause life-threating congenital cardiovascular defects that are frequently of unknown etiology (Abrial et al., 2017; Srivastava, 2001). The regulatory mechanisms involved in PAA remodeling have been studied extensively (Abu-Issa et al., 2002; Kameda, 2009; Liu et al., 2004; Papangeli and Scambler, 2013; Watanabe et al., 2010); however, the cellular events and genetic control of PAA formation are just beginning to be unveiled.

Embryos of the model vertebrate zebrafish exhibit similar processes of PAA morphogenesis, despite the absence of aortic arch remodeling (Anderson et al., 2008; Isogai et al., 2001). During zebrafish mid-somitogenesis, a common mesodermal progenitor population, which is segregated from the cardiac precursors in the heart field, can give rise to PAAs, head muscles (HM) and cardiac outflow tract (OFT) (Guner-Ataman et al., 2018). These common progenitors are housed in different pharyngeal arches and exhibit distinct gene expression profiles prior to the morphogenesis of PAAs, HM and cardiac OFT (Nagelberg et al., 2015; Paffett-Lugassy et al., 2017). In particular, the common progenitors located in PAAs 3-6 condensed into nkx2.5⁺ clusters in a craniocaudal sequence and then gave rise to PAA endothelium, implying the progressive emergence of PAA progenitors (Paffett-Lugassy et al., 2013). Specifically, cell lineage-tracing experiments showed that progenitors for PAA5 and PAA6 were specified from the pharyngeal mesoderm after 30 hpf, a time point much later than that of the segregation of the common progenitors (Paffett-Lugassy et al., 2013). Thus, these common progenitors might undergo further specification in the pharyngeal region, a hypothesis that remains to be evaluated experimentally.

Interestingly, the homeodomain transcription factor nkx2.5 is expressed in presumptive PAA endothelial progenitors; however, it is dispensable for PAA progenitor specification (Paffett-Lugassy et al., 2013). Subsequently, chemical inhibition experiments suggest that transforming growth factor β (TGFβ) signaling is required for the differentiation of PAA progenitors into angioblasts (Abrial et al., 2017). In addition, transcription factors etv2 and scl have been shown to be essential for the initiation of the angioblast program (Sumanas and Lin, 2006). However, the molecular mechanism underlying PAA progenitor specification in the pharyngeal region has yet to be fully investigated.

Endodermal pouches are a series of outpocketings budding from the developing foregut (Graham and Smith, 2001). Interestingly, affected arch arteries often occur simultaneously with pouch defects, possibly because of their close physical relation and potential interactions during development (Li et al., 2012; Wendling et al., 2000). Because the pharyngeal pouches express several signaling molecules that participate in the patterning of the pharyngeal skeleton and in the specification of the arch-associated ganglia, their roles in aortic arch morphogenesis have been traditionally considered as secondary (Crump et al., 2004; Holzschuh et al., 2005; Ning et al., 2013). Intriguingly, our recent study indicated an indispensable role for PDGF signaling from pharyngeal pouches in the PAA angioblast proliferation (Mao et al., 2019), but whether pouch endoderm directly functions in PAA progenitor specification remains unknown. In this work, we further found that pouch-derived BMP signaling is necessary for the specification of PAA progenitors.

PAAs originate from a fraction of nkx2.5⁺ cells within the heart field (Guner-Ataman et al., 2018; Nagelberg et al., 2015; Paffett-Lugassy et al., 2013). To meticulously observe cell behaviors during the formation of these PAAs, time-lapse confocal imaging studies were performed in Tg(nkx2.5:ZsYellow) embryos from 22 hpf. At this time point, some cells in the ZsYellow⁺ pharyngeal mesoderm started to pile up in the ventral root of the prospective third aortic arch, and then sprouted dorsally by 24 hpf (Fig. S1A). This process was repeated for PAA4-PAA6 in a cranial-to-caudal sequence from 28 to 42 hpf (Fig. S1A), which is consistent with previous observations (Paffett-Lugassy et al., 2013).

During somitogenesis, the common progenitors of PAAs, HM and cardiac OFT are specialized and remain lateral with consecutive nkx2.5 expression when cardiac precursors migrate medially (Guner-Ataman et al., 2018; Paffett-Lugassy et al., 2013). Interestingly, the PAA progenitor clusters sequentially emerged at discrete positions in the pharyngeal mesoderm to form aortic arches, whereas the ZsYellow⁺ cells between the PAA progenitor clusters seemed to preserve their locations and would not contribute to PAAs (Fig. S1A,B). These observations support that the pharyngeal mesoderm within pharyngeal arches 3-6 might be further specified into different subpopulations.

To test this hypothesis, we evaluated the expression pattern of nkx2.5, the specific marker of PAA progenitors from 28 hpf to 38 hpf (Guner-Ataman et al., 2018; Nagelberg et al., 2015; Paffett-Lugassy et al., 2013). As described in previous report (Paffett-Lugassy et al., 2013), ZsYellow transcripts derived from nkx2.5 cis-regulatory sequences in Tg(nkx2.5:ZsYellow) embryos gradually appeared in the progenitor clusters (Fig. 1A). The endogenous nkx2.5 transcripts were also sequentially observed, but eventually decreased when the PAA progenitors differentiated into angioblasts (Fig. 1A). Importantly, the transcripts of both ZsYellow and nkx2.5 were enriched in the PAA progenitor clusters, showing a discontinuous distribution (Fig. 1A). We further combined immunofluorescence and fluorescence in situ hybridization experiments, and found that most of the nkx2.5⁺ progenitors were restricted to the PAA clusters within ZsYellow⁺ pharyngeal mesoderm (Fig. 1B). The different expression patterns between endogenous nkx2.5 and ZsYellow may be due to the higher stability of reporter proteins from the transgene. Furthermore, the etv2⁺ and scl⁺ PAA angioblasts located in the ventral root of each sprout, and nkx2.5 transcripts were transiently enriched in the clusters undergoing progenitor-to-angioblast transition (Fig. 1C,D). Taken together, these results show that the pharyngeal mesoderm is composed of distinct subpopulations with or without nkx2.5 expression.

The above findings raised an interesting question about whether these subpopulations in the pharyngeal mesoderm undergo distinct cell fates. To answer this question, we performed lineage-tracing analysis in Tg(nkx2.5:kaede) embryos, where the pharyngeal mesodermal cells expressing photo-convertible Kaede proteins that could instantly switch from green to red fluorescence following ultraviolet light exposure (Guner-Ataman et al., 2013). In the first set of experiments, the Kaede proteins in the PAA progenitor cluster 3 and the subsequent posterior pharyngeal mesoderm located on the right-side of the embryo were photoconverted at 22 hpf, whereas the pharyngeal mesoderm on the left side remained unconverted as an internal control (Fig. 1E). As expected, the derivatives of the photoconverted cells were found in the sprouts of PAAs 3-5 at 36 hpf and in the endothelium of the aortic arches 3-6, as well as the ventral aorta at 60 hpf (Fig. 1F,G). Interestingly, less red fluorescence and more green fluorescence were observed in the cells of caudal PAAs 5 and 6, and the posterior region of ventral aorta (Fig. 1G). Nevertheless, these results indicate that the endothelial cells of PAAs and ventral aorta originate from the Kaede⁺ pharyngeal mesoderm.

Next, we specifically photoconverted the Kaede proteins in PAA cluster 5 at 36 hpf and found their red derivatives in PAA 5 at 54 hpf, but not in other PAAs (Fig. 1H,I). A few cells with red fluorescence were observed in the junction of PAA 5 and ventral aorta (Fig. 1I), suggesting the occurrence of endothelial cell rearrangements during blood vessel fusion (Herwig et al., 2011). In contrast, the photoconversion of Kaede⁺ cells located between PAA cluster 4 and 5 led to red derivatives housed specifically in ventral aorta (Fig. 1J,K). Based on these observations, we concluded that the pharyngeal mesoderm cells within PAAs 3-6 are specified into two vascular progenitor subpopulations: nkx2.5⁺ cells that give rise to PAAs and nkx2.5⁻ cells that generate the connective ventral aorta.

We next aimed to determine the requirement of pouch endoderm during PAA morphogenesis. First, time-lapse recordings of pouch development and PAA formation were performed in Tg(nkx2.3:mCherry;nkx2.5:ZsYellow) embryos, where pouches were labeled with red fluorescent protein mCherry (Choe et al., 2013). At around 24 hpf, the third pharyngeal pouch appeared to have fully formed and reached the sprouting ZsYellow⁺ cluster 3 (Fig. 2A). At later stages, the fourth, fifth and sixth pouches successively made contact with the developing ZsYellow⁺ clusters for PAAs 4-6 (Fig. 2A), indicating a close interaction between the endodermal pouches and the progenitor clusters. Depletion of sox32 in zebrafish embryos by injection antisense morpholinos (MOs) resulted in a lack of early endoderm and endoderm pouches (Fig. S2A,B) (Alexander et al., 1999). As previously described in endoderm-less bon mutants (Paffett-Lugassy et al., 2013), the ZsYellow⁺ cells remained in the pharyngeal region in Tg(nkx2.5:ZsYellow) embryos injected with sox32 MO, but the PAAs were completely absent (Fig. 2B).

In order to examine the specific function of pouch endoderm in the establishment of PAAs, we generated a NTR-mediated tissue-ablation system, Tg(nkx2.3:KalTA4-p2a-mCherry;UAS:NTR-mCherry), using an optimized Gal4-UAS system to drive NTR protein expression in nkx2.3⁺ cells (Curado et al., 2007, 2008; Distel et al., 2009). The Tg(nkx2.3:KalTA4-p2a-mCherry;UAS:NTR-mCherry) embryos were exposed to MTZ from the bud stage to 36 hpf. Live embryo imaging revealed that mCherry-expressing pouch endoderm was markedly reduced at 24 hpf and all the pouch structures were successfully ablated at 36 hpf (Fig. S3A). Intriguingly, in the pouch endoderm-ablated embryos, the expression of PAA endothelial cell marker tie1 was absent and the ZsYellow⁺ cells did not undergo sprouting nor give rise to PAAs (Fig. 2C,D). These results provide strong evidence that pharyngeal pouches are essential for PAA morphogenesis. In addition, pouch-depleted Tg(flk:EGFP) embryos also showed severe defects in these vessels (Fig. S3B), indicating the lack of plasticity in the formation of the PAAs (Nagelberg et al., 2015). Cxcl12b, a Cxcr4a ligand derived from the endoderm underlying the lateral dorsal aortae (LDA), has been reported to be required for the formation of LDA (Siekmann et al., 2009). Interestingly, the LDA in pouch endoderm-less embryos displayed no obvious malformations (Fig. S3B), suggesting the specificity of NTR-mediated pouch endoderm ablation in our related experiments.

To determine whether pharyngeal pouches function in PAA progenitor specification, the expression of putative earlier angioblast lineage markers scl and etv2 and more mature angioblast marker tie1 was first examined in MTZ-treated Tg(nkx2.3:KaTAa4-p2a-mCherry;UAS:NTR-mCherry) embryos at 38 hpf. We found that the formation of these angioblast clusters was abolished in embryos treated with MTZ (Fig. 2E,E′). We further generated a transgenic line, Tg(sox10:KalTA4-p2a-mCherry), in which the red fluorescence of mCherry proteins was expressed in the neural crest cells. We then crossed fishes to generate Tg(sox10:KalTA4-p2a-mCherry;UAS:NTR-mCherry) embryos. MTZ treatment from the bud stage induced obvious cell death in the pharyngeal neural crest cells at 36 hpf (Fig. S3C). However, the PAA angioblasts developed normally (Fig. S3D), implying that pharyngeal neural crest cells are not necessary for the specification and angioblast differentiation of PAA progenitors. We then analyzed the expression of the PAA progenitor marker nkx2.5 in pouch-depleted embryos and found a significant reduction of nkx2.5-expressing progenitors of PAAs 3-6 (Fig. 2F). Moreover, at 18 hpf, pouch depletion did not disrupt the segregation of the nkx2.5⁺ common progenitors (Fig. S4A,B) (Guner-Ataman et al., 2018). Taken together, these results suggest that pouches are essential for PAA progenitor specification in the pharyngeal region.

Several genes encoding BMP ligands, including bmp2a, bmp2b, bmp4 and bmp5, are expressed in the pouch endoderm during pharyngeal segmentation (Holzschuh et al., 2005). Indeed, when pouch endoderm was ablated, expression of these BMP genes were eliminated in the pharyngeal region (Fig. 3A). Immunostaining experiments revealed robust signals of phosphorylated Smad1, Smad5 and Smad8 (Smad1/5/8), the intracellular effectors for BMP signaling, in the forming ZsYellow⁺ clusters and neighboring cranial neural crest cells of Tg(nkx2.5:ZsYellow) embryos (Fig. 3B). In contrast, no detectable phosphorylation of Smad1/5/8 could be seen in the ZsYellow⁺ sprouts composed of the migrating PAA angioblasts (Fig. 3B), suggesting the activation of canonical BMP signaling occurs primarily in the early stage of PAA morphogenesis. Additionally, we crossed Tg(nkx2.5:ZsYellow) with Tg(BRE:EGFP), a BMP signal activity reporter line (Laux et al., 2011). We observed strong GFP protein expression in the ZsYellow⁺ clusters and the other nearby tissues (Fig. 3C). These results demonstrate that BMP/Smad signal is highly activated in the presumptive PAA progenitors.

We then ablated the pharyngeal pouches in Tg(nkx2.5:ZsYellow) and Tg(BRE:EGFP) embryos, respectively. Interestingly, the pouch depletion led to an evident decrease in BMP signal activity in both ZsYellow⁺ clusters and other pharyngeal tissues (Fig. 3D,E). These findings imply that pharyngeal pouches function as a niche for activating BMP signal in presumptive PAA progenitors. To further confirm this assumption, nkx2.3:noggin3-mCherry plasmids expressing a secreted BMP inhibitory protein Noggin3 (Ning et al., 2013), along with Tol2 transposase mRNA, were co-injected into Tg(sox17:GFP) embryos. Part of the injected embryos exhibited uneven, but abundant, mCherry fluorescence in the pouches (Fig. 3F). The same injections were then performed in Tg(nkx2.5:ZsYellow;BRE:EGFP) embryos. The embryos with strong mCherry fluorescence in the pharyngeal region were selected at 36 hpf for immunostaining analysis. The pouch-derived Noggin3 had no obvious effect on pouch endoderm development, as indicated by nkx2.3 expression, but significantly inhibited the BMP signal activity in pharyngeal region (Fig. 3G, Fig. S5), indicating that the secreted BMP ligands from pouches are the biochemical niche cues that trigger signal activation in presumptive PAA progenitors.

To determine whether BMP signaling is required for PAA progenitor specification, we first examined the expression of angioblast marker genes scl and etv2 in embryos treated with DMH1, a selective chemical inhibitor of the BMP pathway (Hao et al., 2010, 2014). Interestingly, embryos treated with DMH1 from 20 hpf exhibited impaired angioblast formation in PAA clusters 4-6, whereas the angioblasts in cluster 3 developed normally (Fig. S6A,A′). To further elucidate the role of the BMP pathway in angioblast formation, we treated embryos with DMH1 from 16 hpf, when the common progenitors of PAAs, HM and cardiac OFT have been specified in the ALPM (Guner-Ataman et al., 2018). Noticeably disturbed generation of angioblast clusters 3 and 4-6 was observed (Fig. S6B). However, blocking BMP signaling from such early stages (16 and 20 hpf) led to different severities of pouch defects (Fig. S6C), which contributed to the difficulties in distinguishing a direct role of BMP pathway in PAA development. Fortunately, we found that embryos treated with DMH1 from 24 hpf, preceding the formation of PAA cluster 4, showed normal pouch structures (Fig. S6C). However, the expression of scl and etv2 was decreased in cluster 4 and completely abolished in clusters 5 and 6 (Fig. 4A,A′). Hereafter, dorsomorphin, another small chemical inhibitor of BMP signaling (Yui et al., 2008), was applied to wild-type embryos from 24 hpf and resulted in similar angioblast phenotypes (Fig. S7A,A′). Consistent with these findings, blocking BMP signaling from 24 hpf greatly reduced the expression of endothelial cell marker tie1 in the caudal PAAs at 60 hpf (Fig. 4B). Similar angioblast and PAA defects were observed in Tg(hsp70l:dnBmpr1a-GFP) embryos that were heat-shocked at 24 hpf (Fig. S7B,C), excluding the potential off-target effects of the pharmacological treatments.

Next, we analyzed the expression of PAA progenitor marker nkx2.5 in BMP signal-suppressed embryos at 36 hpf. BMP signal inhibition from 24 hpf significantly repressed the nkx2.5 expression (Fig. 4C), indicating a serious imperfection of PAA progenitor specification. Unexpectedly, when Tg(hsp70l:caBmpr1b-GFP) embryos were heat shocked at 24 hpf to induce the expression of constitutively active BMP receptor 1b (caBmpr1b), the phosphorylation of Smad1/5/8 was evidently elevated, while PAA progenitor specification and angioblast differentiation remained unchanged (Fig. S7D-G). These results suggest that BMP signal activation is necessary, but not sufficient, for PAA progenitor specification. Furthermore, the pouch depletion-induced deficiencies in PAA progenitor specification and subsequent angioblast differentiation were recovered in heat-shocked Tg(hsp70l:caBmpr1b-GFP) embryos (Fig. 4D-E′). However, these defects in MTZ-treated embryos without the hsp70l:caBmpr1b-GFP transgene were not alleviated by heat-shock treatment (Fig. S8A,B), ruling out the inactivating effects of heat-shock on MTZ in the pouch-depletion experiments. These results imply that pharyngeal pouches induce PAA progenitor specification via activation of BMP signaling in the pharyngeal mesoderm.

To explore whether BMP signaling has a role in angioblast differentiation, DMH1 treatment was performed in Tg(nkx2.5:ZsYellow;gata1:DsRed) embryos between 30 and 60 hpf, a time window after the specification of progenitors for PAAs 3 and 4. Such DMH1 treatment abolished the formation of PAAs 5 and 6, and led to a lack of blood flow in these caudal PAAs, but had no obvious impact on PAAs 3 and 4 (Fig. 5A). When the DMH1 treatment was carried out from 38 hpf, a time point when most of the nkx2.5⁺ progenitors had accomplished angioblast transition, no obvious defects in PAA development were observed in the resulting embryos (Fig. 5A). These results indicate that BMP signaling is crucial for progenitor specification, while dispensable for angioblast differentiation, dorsal migration, endothelial maturation and lumen formation during PAA morphogenesis. It is interesting that if the DMH1 treatment was performed from 30 hpf and then terminated 8 h later, nkx2.5⁺ progenitors for the caudal PAAs reappeared at 48 hpf and went on to develop into growing sprouts at 60 hpf (Fig. 5A,B). These observations imply that, when the BMP inhibition is removed, BMP signal might be reactivated in the pharyngeal mesoderm cells and restore the formation of PAA progenitors.

To explore which BMP ligands are specifically required for PAA progenitor specification, knockdown experiments were performed using previously validated antisense MOs targeting bmp2a, bmp2b, bmp4 and bmp5 (Chocron et al., 2007; Li et al., 2019; Naye et al., 2012; Shih et al., 2017). As expected, injection of these MOs into wild-type embryos caused clear defects in the development of the hepatic bud, pharyngeal pouches, presumptive cloaca and neural crest cells, respectively (Fig. S9A-D) (Li et al., 2019; Naye et al., 2012; Shih et al., 2017; Stickney et al., 2007), indicating a satisfactory level of efficiency and specificity of these MOs. We observed that the expression of etv2 was not obviously changed in embryos injected with bmp4 MO (Fig. 6A,A′). However, etv2 expression was almost abolished in bmp2b morphants (Fig. 6A,A′). We have previously reported that bmp2b is essential for pharyngeal pouch progenitor specification (Li et al., 2019). In fact, bmp2b morphants showed no pharyngeal pouches at 36 hpf, as indicated by the expression of the pouch epithelium marker nkx2.3 (Fig. S9B). Therefore, although we cannot rule out the possibility that bmp2b plays a direct role in PAA development, the loss of PAA angioblast in bmp2b morphants is due mainly to the deficiency of pharyngeal pouches. Importantly, knockdown of bmp2a or bmp5 resulted in a steady reduction in the number of etv2⁺ clusters (Fig. 6A,A′). Furthermore, the expression of nkx2.5 was evidently decreased in the pharyngeal region of embryos injected with bmp2a and bmp5 MOs (Fig. 6B). Together, these data suggest that bmp2a and bmp5 may play an important role in PAA progenitor specification.

To examine the direct function of pouch-expressed BMP ligands, we performed tissue-specific knockdown experiments using a KalTA4-UAS system to drive the expression of miR30-based short hairpin RNAs (shRNAs), which is widely used for gene silencing in eukaryotic organisms (Fig. S10A) (Li et al., 2018; Stegmeier et al., 2005; Zeng et al., 2005). We first generated an UAS:EGFP-shRNA plasmid expressing a shRNA targeting bmp2a (named shRNA-bmp2a). KalTA4-p2a-mCherry mRNA and shRNA-bmp2a were co-injected into one-cell stage embryos, and the expression of bmp2a was examined by whole-mount in situ hybridization and quantitative real-time PCR. KalTA4-mediated expression of shRNA-bmp2a clearly knocked down endogenous bmp2a expression (Fig. S10B,B′). Similarly, we found that shRNA-bmp2b, shRNA-bmp4 and shRNA-bmp5 could evidently silence genes when employed independently (Fig. S10C-E′). Furthermore, these shRNA-mediated gene knockdowns led to similar defects to those found in the related mutants or morphants (Fig. S10F-I) (Li et al., 2019; Naye et al., 2012; Shih et al., 2017; Stickney et al., 2007). These analyses provide further evidence for the efficiency of these shRNAs.

Next, these UAS:EGFP-shRNA plasmids and Tol2 transposase mRNA were injected into Tg(nkx2.3:KalTA4-p2a-mCherry) embryos. A subset of the resulting embryos showed bright-green fluorescence in the pharyngeal pouches at 36 hpf as previously reported (Fig. S11A) (Li et al., 2018). Such embryos were collected to examine the developmental consequences of BMP gene deficiency. We found that these shRNAs selectively disturbed the expression of their target genes in the pouches (Fig. S11B-E). Interestingly, although the depletion of bmp2b or bmp4 expression in pouches had no effect on angioblast formation, silencing of bmp2a or bmp5 eliminated the expression of etv2 in the pharyngeal region (Fig. S11F,F′). Moreover, knockdown of both bmp2a and bmp5 reduced the expression of nkx2.5 in the PAA clusters, but did not affect the formation of pharyngeal pouches (Fig. S11G,H). These findings support the idea that pouch-derived BMP2a and BMP5 are responsible for PAA progenitor specification.

To further substantiate the function of bmp2a and bmp5 in PAA development, we generated one genetic mutant line for each gene using CRISPR/CAS9 technology. The mutant allele of bmp2a or bmp5 carries a DNA deletion near the gRNA targeting sequence in the first exon, resulting in a premature stop codon and presumably a truncated protein lacking the prodomain and C-terminal mature peptide (Fig. S12A,B). In situ hybridization results revealed that about 50% of bmp2a^−/− and bmp5^−/− mutants, which were confirmed by genotyping, showed defective development of the hepatic bud or neural crest cells (Fig. S12C,D). However, there was no compensational increase in the expression of bmp2a or bmp5 in relevant mutants (Fig. S12E,F), suggesting that the incomplete penetrance is not due to compensatory functions between these two genes.

The bmp2a^−/− and bmp5^−/− embryos exhibited no apparent morphological defects and could live to adulthood. We then generated the bmp2a^−/−;bmp5^−/− double mutant by incrossing bmp2a^−/− and bmp5^−/− fishes. A small proportion (about 20%) of bmp2a^−/−;bmp5^−/− embryos displayed evident pericardial edema and died before 7 days post-fertilization, but the rest showed no gross morphological and survival differences compared with wild-type embryos. To confirm the role of bmp2a and bmp5 in PAA progenitor specification, immunostaining analysis was first performed in these mutants at 36 hpf. As shown in Fig. 6C, compared with wild-type embryos, the phosphorylation level of Smad1/5/8 in the pharyngeal region was decreased in bmp2a^−/− and bmp5^−/− mutants, and almost abolished in bmp2a^−/−;bmp5^−/− embryos. Besides, these mutants exhibited obviously impaired formation of PAA progenitors (Fig. 6D). Finally, when compared with control animals and bmp2a^−/− or bmp5^−/− embryos, a significant reduction in etv2⁺ angioblast clusters was observed in bmp2a^−/−;bmp5^−/− double mutants, while the pharyngeal pouches were normally developed (Fig. 6E,E′, Fig. S12G).

Collectively, these data suggest that, among the BMPs expressed in the pouch endoderm, BMP2a and BMP5 are crucial for BMP pathway activation in the pharyngeal mesoderm, thereby promoting PAA progenitor specification.

The progenitors for PAA, HM and cardiac OFT are all marked by nkx2.5 expression (Guner-Ataman et al., 2018; Paffett-Lugassy et al., 2013). Future identification of specific biomolecular markers for PAA progenitors can provide new avenues to investigate the cellular and molecular events in PAA progenitor specification. A recent study identified npas4l, which encodes a PAS-domain-containing bHLH transcription factor, as the gene defective in the cloche mutant that lacks most endothelial as well as hematopoietic cells (Reischauer et al., 2016). Henceforth, npas4l is also called cloche. Therefore, we speculate that npas4l may be expressed in PAA progenitors and be crucial for PAA development. To verify this hypothesis, the expression of npas4l in the pharyngeal region was analyzed by in situ hybridization. We found that npas4l was not expressed in the pharynx at 20 hpf (Fig. 7A). However, npas4l transcripts were then detected in the presumptive PAA progenitor cluster 3 at 24 hpf, ∼2 h later than the initial expression of nkx2.5 in the same PAA cluster (Fig. 7A). Over the next 14 h, npas4l transcripts gradually appeared in a craniocaudal sequence in the PAA clusters (Fig. 7B). Moreover, the expression of npas4l in the PAA clusters was further confirmed by the colocation of npas4l and nkx2.5 transcripts (Fig. 7C).

A previous study has shown that the expression of nkx2.5 is reduced following the differentiation of PAA progenitors into angioblasts (Paffett-Lugassy et al., 2013). Intriguingly, npas4l transcripts persisted during PAA progenitor differentiation (Fig. 7B). These observations raised the possibility that npas4l is not only expressed in the progenitors but also in the angioblasts of PAAs. It has been shown that injection of nkx2.5 MO into zebrafish embryos can disrupt the angioblast differentiation and result in an accumulation of PAA progenitors (Paffett-Lugassy et al., 2013). We then examined the expression of npas4l in nkx2.5 morphants. If npas4l is specifically expressed in PAA progenitors, we would expect a clear increase of npas4l expression in nkx2.5 morphants. Indeed, compared with control animals, embryos injected with nkx2.5 MO showed much higher levels of nkx2.5 expression (Fig. 7D). By contrast, the expression levels of npas4l were not obviously changed in the pharynx upon nkx2.5 MO injection (Fig. 7E). These results may imply that, although the inhibition of nkx2.5 function led to excess PAA progenitors at the expense of angioblasts, the total number of cells with endothelial potential was unchanged. Thus, npas4l is expressed in both PAA progenitors and angioblasts.

We next examined whether npas4l plays a role in PAA development and observed that, in comparison with wild-type and heterozygous siblings, cloche homozygous (cloche^−/−) mutants in Tg(nkx2.5:ZsYellow) background showed almost normal formation of ZsYellow⁺ clusters at 36 hpf (Fig. 8A). However, the PAA vascular channels were absent in cloche^−/− embryos at 60 hpf (Fig. 8A). Interestingly, slightly different from our results, some residual PAA vasculatures were found in cloche^−/− mutants expressing the Tg(flk:EGFP) transgene (Reischauer et al., 2016). As endothelial cells from the LDA could compensate for the loss of PAA vessels under certain conditions (Nagelberg et al., 2015), the flk⁺ PAA endothelial cells in cloche^−/− mutants might be due to the plasticity during PAA development.

Our in situ hybridization analyses further revealed that, in comparison with wild-type and heterozygous siblings, cloche homozygous (cloche^−/−) mutants exhibited normal nkx2.5 expression in pharyngeal clusters 3-5 at 32 hpf (Fig. 8B). To our surprise, the expression of etv2, the PAA angioblast marker, was completely missing in the cloche^−/− mutants at 38 hpf (Fig. 8C), suggesting an unsuccessful differentiation of nkx2.5⁺ progenitors. This phenomenon thus raised an interesting question about the cell fate of the nkx2.5⁺ progenitors in cloche^−/− mutants. It has been suggested that the nkx2.5⁺ progenitors in the lateral plate mesoderm can differentiate into various pharyngeal tissues, including PAA, HM and cardiac OFT (Guner-Ataman et al., 2018; Paffett-Lugassy et al., 2013). Therefore, we investigated whether the nkx2.5⁺ progenitors within presumptive PAA clusters in cloche^−/− mutants altered their cell fate to give rise to cardiac OFT and/or to become muscle cells. We found no distinct difference in the expression of mef2cb and ltbp3, both of which label the outflow pole of the heart tube (Zeng and Yelon, 2014; Zhou et al., 2011), between cloche^−/− mutants and their siblings (Fig. S13A,B). On the contrary, the transcripts of the head muscle precursor marker myod1 and the pharyngeal musculature marker actn3b were unexpectedly expressed in the presumptive PAA structures of cloche^−/− mutants (Fig. 8D,E) (Holterhoff et al., 2009; Lin et al., 2006), suggesting a muscle cell fate transformation of the nkx2.5⁺ progenitors. Therefore, before npas4l expression, the pharyngeal mesoderm seems to have multilineage differentiation potential. Together, these data suggest that npas4l plays a pivotal role in the specification of PAA progenitors from pharyngeal mesoderm.

To determine whether pharyngeal pouches are required for npas4l expression, Tg(nkx2.3:KalTA4-p2a-mCherry;UAS:NTR-mCherry) embryos were exposed to MTZ from the bud stage. Ablation of pouch endoderm completely abolished npas4l expression in the pharynx at 38 hpf (Fig. 8F). Moreover, both DMH1 treatment and injection with MOs targeting bmp2a and bmp5 induced a dramatic reduction in npas4l transcripts (Fig. 8F). We also found a steady decrease in the number of npas4l⁺ PAA clusters in bmp2a^−/− or bmp5^−/− single mutants and bmp2a^−/−;bmp5^−/− embryos (Fig. 8G,G′). Taken together, these findings support the idea that the pharyngeal pouches provide a niche microenvironment for the commitment of multipotent pharyngeal mesoderm toward PAA progenitors through expressing BMP2a and BMP5 (Fig. 8H).

Improper embryonic development of the PAAs may cause life-threating congenital cardiovascular defects (Abrial et al., 2017; Srivastava, 2001). Malformations of the aortic arch system are often accompanied by anomalies of endodermal pouches, which lead to compromised pharyngeal segmentation (Jerome and Papaioannou, 2001; Lindsay et al., 2001; Piotrowski et al., 2003). The effects of pouches on aortic arch development have traditionally been considered secondary to pharyngeal patterning defects (Kopinke et al., 2006; Matt et al., 2003; Wendling et al., 2000). In this study, our results support a model in which the pharyngeal pouches provide a niche microenvironment for PAA progenitor specification via the expression of BMP proteins. Our findings suggest that the segmentation of pharyngeal pouches coincides spatiotemporally with the emergence of PAA progenitor clusters. Furthermore, depletion of pouch endoderm in zebrafish embryos by an MTZ-NTR system resulted in a remarkable reduction of BMP signal activity in the pharyngeal mesoderm and the complete absence of PAA structures attributed to impaired progenitor specification. Most importantly, the PAA progenitor specification is directly regulated by pouches and their derived signal molecules, as the ablation of pharyngeal neural crest cells shows no effect on the emergence of PAA angioblasts, which are differentiated from vascular progenitors. These data, combined with our recent findings that pharyngeal pouches regulate PAA angioblast proliferation by expressing PDGF ligands (Mao et al., 2019), suggest multiple distinct roles of pouch endoderm during PAA development.

It has been reported that a common progenitor population for PAAs, HM and cardiac OFT is specified in zebrafish ALPM during mid-somitogenesis (Guner-Ataman et al., 2018). At later stages, these progenitors, which are located within pharyngeal arches 3-6, contribute to the endothelium of their respective PAAs (Paffett-Lugassy et al., 2013). Interestingly, our data indicate that, at later stages, this pharyngeal mesoderm lineage within PAs 3-6 comprises two subpopulations: one of which is nkx2.5⁺, in which cells are restricted to different domains; the other is nkx2.5⁻ , in which cells are located between the nkx2.5⁺ clusters. Consistent with previous findings, our cell-lineage tracing analysis reveals that the nkx2.5⁺ clusters sprout out and contribute to corresponding PAAs (Paffett-Lugassy et al., 2013). Previous reports have also suggested that, in zebrafish, the ventral parts of PAAs merge to form the bilateral ventral aortae (Anderson et al., 2008; Paffett-Lugassy et al., 2013). Unexpectedly, we found that the nkx2.5⁻ subpopulation does not migrate dorsally and ultimately gives rise to ventral aortae. Therefore, PAAs are sequentially generated from nkx2.5⁺ progenitors within the developing ventral aortae (the pharyngeal mesoderm). Interestingly, resin filling of mouse embryonic vasculature has shown that the PAA endothelium arises by branching off the aortic sac – the mammalian homolog of the ventral aorta of gill-bearing vertebrates (Berta, 2006; Hiruma et al., 2002). Based on these results, it is likely that the process of PAA morphogenesis is evolutionarily conserved across vertebrate classes.

Several lines of evidence support the idea that the pharyngeal mesoderm lineage is further specified into PAA progenitors in a niche microenvironment provided by pouches. First, the pharyngeal mesoderm within PAs 3-6 contains nkx2.5⁺ progenitors that give rise to PAAs and nkx2.5⁻ progenitors that generate ventral aorta. Second, the PAA progenitor clusters emerge in a craniocaudal sequence following pharyngeal pouch segmentation. Third, depletion of pouch endoderm eliminates the PAA progenitors without disrupting the segregation of pharyngeal mesoderm lineage from cardiac precursors. Fourth, BMP signal inhibition produced by pharmacological treatments and tissue-specific knockdown or genetic depletion of bmp2a/5 results in remarkably reduced PAA progenitors. Finally, yet most importantly, head muscle markers are ectopically expressed in the presumptive PAA structures of cloche^−/− mutants, implying that the pharyngeal mesoderm has multilineage differentiation potential. This idea is supported by a previous observation that, in cloche^−/− mutants, the rostral mesoderm undergoes a fate transformation and generates ectopic cardiomyocytes (Schoenebeck et al., 2007).

Our study demonstrates that npas4l is essential for PAA progenitor specification and its expression is tightly controlled by BMP signaling. Moreover, mutation of the cloche locus results in a cell fate transformation: rather than producing progenitors of PAAs, the pharyngeal mesoderm produces ectopic head muscle progenitors. Interestingly, when BMP signal inhibition is relieved, PAA progenitors are capable of reappearing in the pharyngeal mesoderm and develop into growing sprouts. Previous studies have revealed that npas4l functions upstream of etv2, and etv2-deficient vascular progenitors can acquire a skeletal muscle fate, whereas overexpression of etv2 induces vascular gene expression and converts skeletal muscle cells into functional endothelial cells (Chestnut et al., 2020; Reischauer et al., 2016; Veldman et al., 2013; Yan et al., 2019a). Therefore, when BMP signal is reactivated after the washout of BMP inhibitors, the expression of npas4l and its downstream gene etv2 might be reinduced in the muscle progenitors located in the presumptive PAA clusters and then transdifferentiated into endothelial cells. Additional studies will be required to learn whether BMP-Npas4l-Etv2 pathway is necessary and sufficient to switch the fate of muscle cells into the vascular lineage in the pharyngeal region.

During vertebrate embryonic development, pharyngeal pouches play a central role in organization of the head through expressing signaling molecules such as FGFs and BMPs (Crump et al., 2004; David et al., 2002; Graham, 2008; Holzschuh et al., 2005; Ning et al., 2013). Our data further indicate that pharyngeal pouches induce PAA progenitor specification by expressing BMP ligands. Stem cells or progenitor populations are established in niches where niche factors function to maintain their quiescent state or to induce their proliferation and differentiation for fetal development (Birbrair and Frenette, 2016; Jhala and Vasita, 2015; Scadden, 2006). As endoderm pouches are in close contact with the pharyngeal mesoderm at discrete locations, establishing a physicochemical environment for cell fate determination through the activation of BMP signaling, it is reasonable to hypothesize that pharyngeal pouches provide a niche microenvironment for PAA progenitor specification.

Pouch endoderm expresses several BMP genes including bmp2a, bmp2b, bmp4 and bmp5 (Holzschuh et al., 2005). Using the KalTA4-UAS system to drive pouch-specific expression of miR30-based short hairpin RNAs, we find that both bmp2a and bmp5 are responsible for progenitor specification. This conclusion is in agreement with the fact that bmp2a^−/−;bmp5^−/− double mutants also exhibit clear defects in PAA formation. Interestingly, the phenotypes seem to be a little more pronounced in the shRNA-expressing embryos. This observation might be due to some unexpected off-target side effects of shRNA-mediated gene silencing. Furthermore, the compensatory expression of other BMP ligands and related components of the BMP signaling pathway in bmp2a^−/−;bmp5^−/− double mutants may also contribute to this phenomenon. In addition, as most of bmp2a^−/−;bmp5^−/− double mutants are viable, it will be interesting to investigate whether and how the morphogenesis of the PAAs is recovered in the mutants at later developmental stages in the future.

Our zebrafish experiments were all approved and carried out in accordance with the Animal Care Committee at the Institute of Zoology, Chinese Academy of Sciences (permission number IOZ-13048).

Our zebrafish experiments were performed by using the following mutant and transgenic lines: Tg(nkx2.5:ZsYellow) (Paffett-Lugassy et al., 2013), Tg(nkx2.5:kaede) (Paffett-Lugassy et al., 2013), Tg(nkx2.3:mCherry) (Li et al., 2018), Tg(nkx2.3:KalTA4-p2a-mCherry) (Li et al., 2018), Tg(flk:EGFP), Tg(gata1:DsRed), Tg(sox17:GFP) and Tg(BRE:EGFP) (Laux et al., 2011), Tg(hsp70l:dnBmpr1a-GFP) (Pyati et al., 2005), Tg(hsp70l:caBmpr1b-GFP) (Row and Kimelman, 2009), Tg(sox10:KalTA4-p2a-mCherry), and Tg(UAS:NTR-mCherry) and cloche^m378 (Stainier et al., 1995). The Tg(sox10:KalTA4-p2a-mCherry) transgenic line was generated by our lab with the sox10 upstream regulatory sequence as previously described (Carney et al., 2006). The Tg(UAS:NTR-mCherry) transgenic line was obtained from China Zebrafish Resource Center. Unless otherwise specified, live embryos were kept at 28.5°C in Holtfreter's solution, and staged based on morphology as previously described (Kimmel et al., 1995).

Digoxigenin-UTP-labeled antisense RNA probes for scl, etv2, nkx2.5, ZsYellow, tie1, sox17, bmp2a, bmp2b, bmp4, bmp5, nkx2.3, dlx2a, hhex, evx1, myod1, actn3b, mef2cb and npas4l were transcribed using MEGAscript Kit (Ambion) according to the manufacturer's instructions. Whole-mount in situ hybridization with these RNA probes were performed using the NBT-BCIP substrate.

Morpholino oligonucleotides (MOs) were purchased from Gene Tools. The standard control MO (5′-CCTCTTACCTCAGTTACAATTTATA-3′) (Dickmeis et al., 2001), sox32 MO (5′-CAGGGAGCATCCGGTCGAGATACAT-3′) (Dickmeis et al., 2001), bmp2a MO (5′-AGTAAACACTTGCTTACCATCATGG-3′) (Naye et al., 2012), bmp2b MO (5′-CGCGGACCACGGCGACCATGATC-3′) (Li et al., 2019), bmp4 MO (5′-GTCTCGACAGAAAATAAAGCATGGG-3′) (Chocron et al., 2007), bmp5 MO (5′-TTGACCAGGATGATGATGCTTTCAG-3′) (Shih et al., 2017) and nkx2.5 MO (5′-TGTCAAGGCTCACCTTTTTTCTCTT-3′) (Paffett-Lugassy et al., 2013) were used as previously described. All the MOs were injected at one-cell stage into zebrafish embryos.

The miR30-based shRNAs were designed according to previously published methods (Dow et al., 2012). The target sequences are shown in Table S1. Plasmids expressing shRNAs were microinjected into fertilized eggs at the one-cell stage at the indicated concentrations. The injected embryos were cultured at 28.5°C until needed.

We generated bmp2a and bmp5 mutants using the CRISPR/CAS9 technology. We designed the gRNAs of bmp2a and bmp5 using the CRISPR Design website http://crispor.tefor.net/. The Cas9 mRNA and gRNAs were prepared as described previously (Wei et al., 2017), and co-injected into wild-type embryos at the one-cell stage. Embryos were collected to make genomic DNA for genotyping at 24 hpf. For screening of the F1 fish with mutant alleles, genomic DNA was isolated from the tail of individual fish. The forward primer 5′-AAAGACTCGCAATGGCTCG-3′ and reverse primer 5′-TCCCTGTCAGGCATGAAG-3′ were used to amplify bmp2a gRNA targeted sequence. And the forward primer 5′-GACTTCTGTGGAGCTGTTTAG-3′ and reverse primer 5′-TGCGTGACCTCTTTACACCAT-3′ were used to amplify bmp5 gRNA targeted sequence. The amplified fragments were identified with Sanger DNA sequencing for genotyping. F2 embryos were generated by incrossing F1 mutant fishes and genotyped by digesting PCR products with BtsI (NEB, R0667S) and SmaI (NEB, R0141V), respectively.

Real-time quantitative PCR was performed as previously described (Yan et al., 2019b). The target genes were amplified with the primers listed in Table S2. The expression level of each sample was normalized to β-actin. Statistical analysis was carried out with an unpaired Student's t-test.

Live fluorescent embryos were mounted in 1% low-melting agarose in glass-bottomed dish (Solarbio; D35-10-1-N) at indicated stages. Tg(nkx2.5: ZsYellow) embryos were imaged and analyzed for the formation of PAAs using a Nikon A1R+ confocal microscope (20× objective). For cell lineage tracing, photoconversion in Tg(nkx2.5:kaede) embryos was achieved using a DAPI filter, and the converted embryos were immediately imaged, removed from the agarose and raised in dark conditions until subsequent evaluation. All confocal stack images were processed using Nikon NIS-Elements AR 4.13.00 software.

Immunofluorescence staining was performed as previously described (Ning et al., 2013). Embryos were stained with the following affinity-purified antibodies: anti-GFP (1:1000; A111201, Invitrogen), anti-ZsYellow (1:200; 632475, Clontech), anti-ZsYellow (1:400; TA180004, Origene) and anti-p-Smad1/5/8 (1:200; 9511, Cell Signaling Technology). Fluorescence in situ hybridization was performed as previously described (Schoenebeck et al., 2007). Anti-DIG HRP-conjugated Fab fragments (1:400; Roche) were used to detect the digoxigenin (DIG)-labeled probes. Embryos were then incubated with fluorescein (FLU) tyramide (1:100; PerkinElmer) for 3 h at 28.5°C. Next, the embryos were subjected to immunofluorescence after removal of HRP activity.

For tissue-specific ablation, Tg(nkx2.3:KalTA4-p2a-mCherry;UAS:NTR-mCherry) or Tg(sox10:KalTA4-p2a-mCherry;UAS:NTR-mCherry) embryos were raised in Holtfreter's solution containing 10 mM MTZ (M1547, Sigma) from bud stage, and then harvested for live imaging or in situ hybridization at the indicated stages. To block BMP signaling, embryos were treated with DMH1 (10 μM; D8946, Sigma) or dorsomorphin (10 μM; P5499, Sigma) under dark conditions. Tg(hsp70l:dnBmpr1a-GFP) and Tg(hsp70l:caBmpr1b-GFP) embryos were subjected to heat shock at 40°C for 20 min at 24 hpf, and then incubated at 28.5°C until harvest.

Statistical analysis was performed with GraphPad Prism software version 5.00 for Macintosh (GraphPad). The numbers of PAA angioblast clusters were counted based on PAA3-PAA6 per embryo. All results are expressed as mean±s.d. Differences between control and treated groups were analyzed with unpaired two-tailed Student's t-test. Results were considered statistically significant at P<0.05.

We are grateful to Dr Jingwei Xiong (Peking University, China) for the Tg(nkx2.5:ZsYellow) fish line and to Dr Caroline E. Burns (Massachusetts General Hospital, USA) for the Tg(nkx2.5:kaede) fish line.