All zebrafish (Danio rerio) were handled as described previously (Kimmel et al., 1995) and approved by the Gyeongsang National University Institutional Animal Care and Use Committee. Published lines include Tg(her5:mCherryCAAX) ^el72 (Choe et al., 2013) and Tg(sox10:EGFP) ^ba2 (Carney et al., 2006). Tg(sox17:EGFP), Tg(sox17:Gal4VP16), Tg(UAS:Pax1a), Tg(UAS:Efnb2a), Tg(UAS:EGFP), Tg(hsp70I:Pax1a) and Tg(hsp70I:Efnb2a) transgenic constructs were generated using the Gateway (Invitrogen) Tol2kit (Kwan et al., 2007). For p5E-sox17, PCR amplicons covering approximately 5 kb of upstream regulatory regions were acquired with primers sox17-B4F and sox17-B1R from multi-stage zebrafish genomic DNA. For pME-Pax1a, the coding sequence of pax1a was amplified using primers Pax1a-B1F and Pax1a-B2R from multi-stage zebrafish cDNA. pME-Efnb2a was published (Choe and Crump, 2015). Three independent transgenic lines for Tg(sox17:Gal4VP16:pA) were secured based on cmlc2:GFP heart fluorescence. Two independent transgenic lines for Tg(UAS:Pax1a:pA), two for Tg(UAS:Efnb2a:pA), three for Tg(hsp70I:Pax1a:pA), and two for Tg(hsp70I:Efnb2a:pA) were established based on α-crystallin:Cerulean eye fluorescence. Tg(sox17:Gal4VP16:pA) ^GNU99 , Tg(UAS:Pax1a:pA) ^GNU102 , Tg(UAS:Efnb2a:pA) ^GNU100 , Tg(hsp70I:Pax1a:pA) ^GNU111 , and Tg(hsp70I:Efnb2a:pA) ^GNU112 were used for this study.

Mutant lines for pax1a, pax1b, and efnb2a were generated with CRISPR/Cas9 system in wild-type Tübingen embryos. One mutant allele for pax1a (pax1a ^GNU25 ), one allele for pax1b (pax1b ^GNU28 ), and two alleles for efnb2a (efnb2a ^GNU89 and efnb2a ^GNU90 ) were established. For genotyping of pax1a ^GNU25 , PCR amplicons produced by primers pax1a-GT_F and pax1a-GT_R were treated with BsrI; while the wild-type allele produced 362 bp, the pax1a mutant allele generated 112 and 240 bp. For genotyping of pax1b ^GNU28 , PCR fragments obtained using primers pax1b-GT_F and pax1b-GT_R were digested with MspA1I; the wild-type allele generated 264 bp, whereas the pax1b mutant allele produced 117 and 142 bp. Genotyping of efnb2a ^GNU89 and efnb2a ^GNU90 was performed using PCR with primers efnb2a-GT_F and efnb2a-GT_R, followed by digestion with BsrBI and MspI, respectively. The wild-type allele remains uncut, whereas the efnb2a ^GNU89 allele is cleaved into 64 bp and 282 bp fragments, and the efnb2a ^GNU90 allele into 62 bp and 275 bp fragments. See Supplementary Materials and Methods for primers.

Total RNA was extracted from the wild type and each gene mutant at 48 hpf using the NucleoSpin RNA Plus kit (MACHEREY-NAGEL). Approximately 30–40 embryos were used for each RNA extraction. RNA quality and quantity were assessed using a NanoDrop spectrophotometer to ensure A260/A280 ratios between 1.8 and 2.0. cDNA was synthesized with 1 µg of total RNA using ReverTra Ace™ High Efficient Reverse Transcriptase (TOYOBO). Quantitative real-time PCR (qRT-PCR) was conducted using the SsoFast EvaGreen Supermix (Bio-Rad) on a Rotor-Gene Q (QIAGEN) PCR machine with the following cycling conditions: an initial denaturation at 95°C for 5 min, followed by 40 cycles of 95°C for 10 s and 60°C for 30 s. Each reaction was performed in triplicate to ensure technical accuracy. Ct values and relative gene expression were calculated using the ^ΔΔ Ct Relative Quantification method (QIAGEN). β-actin was used as an internal control to normalize the expression levels, with β-actin real-time primer used previously (Jeon et al., 2022). Statistical analysis was performed using GraphPad Prism. Data were analyzed using an unpaired t-test, with significance determined at p < 0.05.

Heat shock was performed as described for 40 min (Scott et al., 2017). After the heat shock, tubes were removed from the water bath and placed into a 28.5°C incubator to cool gradually. Following recovery, embryos were fixed in 4% paraformaldehyde for subsequent analysis. No developmental or morphological defects were observed in heat-shocked embryos.

Alcian blue and alizarin red staining, GFP immunohistochemistry (Torrey Pines Biolabs, 1:1000), fluorescent double in situ hybridizations, and Lysotracker and BrdU staining were carried out as described (Zuniga et al., 2011; Choe and Crump, 2014). TUNEL assay was performed according to the protocol described by Cole and Ross (2001). Riboprobes to tbx1, fgf3, itg-α-5, efnb2a, and efnb3b were published (Crump et al., 2004b; Choe and Crump, 2014; 2015). For riboprobes to pax1a, pax1b, barx1, and sox9a, PCR amplicons were cloned into pGEM^®-T Easy Vector Systems (Promega), linearized, and then digoxigenin- or dinitrophenol-labeled RNAs were synthesized using T7 or SP6 RNA polymerase (Roche). See Supplementary Materials and Methods for primers.

Craniofacial cartilages were dissected manually with fine insect pins, flat-mounted, and photographed on an Olympus BX50 upright microscope using mosaic V2.1 software. Fluorescent images were taken on an Olympus FV3000 confocal microscope using FV31S-SW software. After capturing approximately 80 μm Z-stacks at 3.0 and 1.5 μm intervals with Olympus UPLXAPO ×20 and ×40 objective lenses, respectively, maximum intensity projections encompassing static confocal sections were assembled using FV31S-SW software. Any adjustments were applied to all panels using Adobe Photoshop.

The numbers of HM cells and BrdU-, Lysotracker Red-, and TUNEL Red-positive cells were manually counted to quantify HM defects and cell proliferation and death. Three independent individuals cross-verified the counting of HM cells. We employed the multiple comparison tests of Tukey–Kramer to quantify HM defects and cell proliferation and death and a chi-square test to compare the frequency of groups quantitatively.

Previously, it was reported that pax1a and pax1b are expressed in the pharyngeal endoderm at the stage of arch segmentation, with their redundant role in the morphogenesis of pharyngeal pouches (Liu et al., 2020). Although it was shown that pax1a and pax1b expression continues in the embryonic head by 96 hpf, their expression domain has yet to be clearly determined (Liu et al., 2020). To get an insight into a late role for Pax1 in facial skeletal development, we examined the expression of pax1a and pax1b at 36 and 48 hpf. When arch segmentation was completed by the pouches at 36 hpf, pax1a and pax1b were still expressed in the mature pouches, with their expression continued at 48 hpf (Figure 1). Considering 36 and 48 hpf are critical time points for the cranial neural crest-derived mesenchyme in the arches to differentiate into the chondrocytes to form facial cartilages (Barske et al., 2016), the late expression of pax1a and pax1b in the mature pouches suggests a distinct role of Pax1 in facial skeletal development that differs from its early role in pouch formation.

To access a role for pax1a and pax1b in the development of facial cartilages, we generated loss-of-function mutations in the pax1a and pax1b genes with the CRISPR/Cas9 system. For pax1a mutants, we established a ten-nucleotide deletion allele in the pax1a gene (pax1a ^GNU25 ) compared to the wild-type allele (Supplementary Figure S1A). While the wild-type allele produces 359 amino acids, the mutant allele was predicted to produce only 15 normal amino acids and an extra 11 amino acids, with most of the paired box domain (PBD) missing due to a premature stop codon (Supplementary Figure S1A). Moreover, the mRNA levels of pax1a were significantly reduced in pax1a mutants compared to the wild type (Supplementary Figure S1C). Thus, pax1a ^GNU25 is a loss-of-function allele. For pax1b mutants, we secured a five-nucleotide deletion allele in the pax1b gene (pax1b ^GNU28 ) (Supplementary Figure S1B). Compared to the wild-type pax1b allele, producing 340 amino acids, the pax1b ^GNU28 allele was predicted to make 17 normal amino acids and an extra 57 amino acids, with most PBD missing (Supplementary Figure S1B). The mRNA levels of pax1b were also reduced by about 50% in pax1b mutants compared to the wild type (Supplementary Figure S1C). Thus, we suggest that pax1b ^GNU28 is a loss-of-function allele.

In zebrafish, single mutants for pax1a or pax1b did not affect the development of facial cartilages, whereas double mutants for pax1a and pax1b showed severe defects in the HM plate and ceratobranchial (CB) cartilages whose development relied on the pouches (Liu et al., 2020). However, in pax1a mutants, we observed defects in the HM plate, but with other facial cartilages, including CBs, being normal (Figures 2C–E). In wild-type animals, the HM plate is shaped like a triangle bearing a foramen in which the fifth facial nerve passes (Iwasaki et al., 2022; Figures 2A, B). The HM plate can be divided into anterior and posterior parts based on the foramen position (Iwasaki et al., 2022), with multi-layered cells composing the anterior part adjacent to the foramen (arrowhead in Figure 2B). In pax1a mutants, the HM plate bore the foramen, but only mono-layered cells formed the anterior part adjacent to the foramen (arrowhead in Figure 2D). These defects were classified as weak phenotypes. Eventually, the foramen and the anterior part were missing from the deformed HM plate in pax1a mutants (arrowhead in Figure 2E), and these defects were classified as severe phenotypes. Consistently, the number of cells in the HM plates was significantly decreased in pax1a mutants compared to wild-type animals (Figure 2K; Supplementary Table S1).

Since it has been shown that in/del mutation generated by the CRISPR/Cas9 system can lead to phenotypes that are weaker than either point mutation or morpholino-mediated knockdown due to a genetic compensation (Rossi et al., 2015; El-Brolosy et al., 2019), we examined pax1b expression in pax1a mutants with qRT-PCR. Indeed, we observed that the mRNA levels of pax1b increased by about 1.5 folds in pax1a mutants (Supplementary Figure S1D). Nonetheless, the HM defects in pax1a mutants suggest a distinct role of Pax1a in HM development, which Pax1b cannot entirely compensate. In contrast, as previously reported, any defects in facial cartilages, including the HM plate and CBs, were not seen in pax1b mutants (Figures 2F, G), in which the mRNA levels of pax1a increased in about 12 folds (Supplementary Figure S1E). This result implies that the Pax1b function can be replaced with Pax1a in the development of HM plate and CB cartilages.

Confirming a high degree of genetic redundancy of Pax1a and Pax1b in facial cartilage formation, the HM was distorted as a rod, with most CBs missing, in double mutants for pax1a and pax1b (Figures 2H, I), in which the mRNA levels of pax1a and pax1b were significantly reduced (Supplementary Figure S1C). In particular, in the double mutants, HM defects were more severe, with the number of cells in the HM plates significantly decreasing, and the genetic penetrance for the phenotype increased compared to pax1a mutants (Figures 2J, K). Despite a degree of genetic redundancy of Pax1a and Pax1b in the development of the HM plate, our data indicate a distinct role of Pax1a in the development of the HM plate that cannot be replaced with Pax1b.

Previously, it was reported that the first pouch is essential for HM plate development in the hyoid arch (Crump et al., 2004b), with Pax1a and Pax1b acting redundantly to form pouches, including the first pouch (Liu et al., 2020). Since we observed defects in the HM plate in pax1a mutants, we examined the first pouch in the mutants. To do so, we analyzed the first pouch and HM plate simultaneously in the same individual, using wild-type and mutant animals bearing Tg(her5:mCherryCAAX) and Tg(sox10:GFP) transgenes that visualize the pouches and arches in real-time. pax1a mutant individuals with first pouch defects at 34 hpf displayed defects in the HM plate at 4 dpf (Supplementary Figure S2). This result indicates that Pax1a is required to develop the first pouch to control the development of the HM plate.

Next, we addressed whether Pax1a-dependent first pouch formation is sufficient for the HM plate development. To do so, we tried to rescue the defects in the HM plate by rescuing the first pouch formation in double mutants for pax1a and pax1b, which shows almost complete penetrance of HM defects (Figure 2J). Since the first pouch formation begins at 16 hpf and is completed at 18 hpf (Choe et al., 2013), we expressed the wild-type pax1a gene at 16 hpf for 40 min with a heat-shock approach in double mutants for pax1a and pax1b. At 17 hpf, pax1a expression was nearly absent in double mutants for pax1a and pax1b compared to wild types, and it was not induced by the heat-shock treatment (Supplementary Figures S3A–C). However, in double mutants for pax1a and pax1b carrying Tg(hsp70I:Pax1a) transgene, pax1a was overexpressed in all areas of the embryos, including the pharyngeal regions, by the heat-shock treatment (Supplementary Figure S3D).

After the heat-shock treatment at 16 hpf for 40 min, all double mutant siblings carrying no Tg(hsp70I:Pax1a) transgene displayed defects in the first pouch at 37 hpf (arrowhead in Figure 3A), as shown by ZN5 staining, a molecular marker for pharyngeal pouches (Piotrowski and Nüsslein-Volhard, 2000). However, about 60.6% (n = 20/33) of double mutant siblings carrying Tg(hsp70I:Pax1a) transgene showed normal development of the first pouch at 37 hpf (arrowhead in Figure 3C), verifying the role of Pax1a in the first pouch formation. After the heat-shock treatment, about 6.3% (n = 4/63) of double mutant siblings bearing no Tg(hsp70I:Pax1a) transgene showed weak defects in the HM plate in which the foramen was seen with mono-layered cells composing the anterior part adjacent to the foramen. However, about 93.6% (n = 59/63) of animals showed intermediate to severe defects in the HM plate, which was characterized by the absence of the foramen and anterior part, resulting in a plate-like or rod-shaped HM, respectively (Figure 3B). In contrast, after the heat-shock treatment, about 34.4% (n = 21/61) of double mutant siblings bearing Tg(hsp70I:Pax1a) transgene showed wild-type (n = 11/21) or weak defects in the HM plate (n = 10/21), and about 65.6% (n = 40/61) animals showed intermediate or severe defects in the HM plate (Figure 3D). The increase in the degree of the weakly defective or wild-type HM plate in the double mutants by the heat shock was statistically significant and was due to the increase in the number of cells composing the HM plate (Figures 3E, F; Supplementary Table S1). This result suggests a rescue of the first pouches and HM plate in the double mutants by inducing pax1a expression at 16 hpf, which confirms the critical role of Pax1a in the HM development through the first pouch formation. Interestingly, however, we observed that in the double mutants carrying Tg(hsp70I:Pax1a) transgene treated by the heat shock, about 60.6% (n = 20/33) of animals showed rescued first pouch at 37 hpf, but only about 34.4% (n = 21/61) of animals showed rescued HM plate at 5 dpf. This discrepancy in rescuing the first pouch and HM plate may suggest that, in the HM plate development, Pax1a-dependent first pouch formation is essential, but more is needed.

The late expression of pax1a in the pouches at 36 and 48 hpf and the discrepancy in rescuing the first pouch and HM plate suggest that Pax1a has a distinct role in HM plate development independent of the first pouch formation. To test this hypothesis, we addressed whether the late expression of pax1a at 36 hpf after pouch formation can rescue the defects in the HM plate in the double mutants for pax1a and pax1b. Compared to wild types, in double mutants for pax1a and pax1b at 37 hpf, pax1a expression was significantly reduced, and its expression was not induced by the heat-shock treatment at 36 hpf for 40 min (Supplementary Figures S3E–G). After the heat-shock treatment at 36 hpf for 40 min, pax1a was overexpressed in most areas of the embryos, including the pharyngeal regions in double mutants for pax1a and pax1b carrying Tg(hsp70I:Pax1a) transgene (Supplementary Figure S3H).

In pax1a and pax1b double mutants bearing no transgene and their siblings carrying Tg(hsp70I:Pax1a) transgene, all analyzed animals showed the first pouch was malformed or barely seen after the heat-shock treatment, indicating the induction of pax1a expression at 36 hpf in the double mutants could not rescue the first pouch formation (Figures 4A, C). After the heat-shock treatment, about 3.9% (n = 4/103) of double mutant siblings bearing no Tg(hsp70I:Pax1a) transgene displayed weak defects in the HM plate in which the foramen was seen with mono-layered cells composing the anterior part adjacent to the foramen, with about 96.1% (n = 99/103) animals showing intermediate or severe defects in the HM plate in which the foramen and anterior part were missing (Figure 4B). Interestingly, however, after the heat-shock treatment, about 12.8% (n = 19/149) of double mutant siblings bearing Tg(hsp70I:Pax1a) transgene showed weak defects in the HM plate, and about 87.2% (n = 130/149) animals showed intermediate or severe defects in the HM plate (Figure 4D). We did not observe the complete wild-type HM plate by the heat-shock treatment in both pax1a and pax1b double mutants carrying Tg(hsp70I:Pax1a) transgene and their siblings bearing no transgene. Although the increase in the degree of the weakly defective HM plate in the double mutants by the heat shock at 36 hpf was not as dramatic as that by the heat shock at 16 hpf, it was still statistically significant (Figure 4E). In addition, after the heat shock, the number of cells composing the HM plate increased in the double mutants carrying Tg(hsp70I:Pax1a) transgene compared to the double mutant siblings carrying no transgene (Figure 4F; Supplementary Table S1). This result suggests a partial rescue of the HM plate in the double mutants by inducing pax1a expression at 36 hpf, indicating a distinct role of Pax1a in the HM development, independent of the first pouch formation.

The decrease in the number of cells in the HM plate of pax1a mutants, with the increase in cell numbers by inducing pax1a expression at 36 hpf in the double mutants carrying Tg(hsp70I:Pax1a) transgene suggests the role of Pax1a in the HM plate development would be involved in the survival or proliferation of cells contributing to the HM cartilages. To understand the cellular requirements of Pax1a at 36 and 48 hpf in the development of the HM plate, we analyzed the developmental processes of HM cartilage with molecular markers in pax1a mutants. In wild types, barx1 expression at 36 hpf marks condensed osteochondral progenitors in the dorsal region of the hyoid arch, which will further differentiate into the chondrocytes contributing to the HM plate (Barske et al., 2016). Compared to wild types, barx1 expression at 36 hpf in the dorsal regions of the hyoid arch was unaffected in pax1a mutants, indicating that Pax1a is not essential for the condensation of osteochondral progenitors in HM plate formation (arrows in Figures 5A, B). In wild types, barx1 expression continues in mesenchymal cells that are partially overlapped with sox9a-expressing chondrocytes in the dorsal region of the hyoid arch at 48 hpf (Paudel et al., 2022; arrow in Figure 5C). In pax1a mutants, while sox9a expression marking chondrocytes at 48 hpf was fairly normal in the dorsal region of the hyoid arch, barx1 expression marking mesenchyme was almost abolished, with chondrocytes co-expressing barx1 and sox9a not seen anymore (arrowhead in Figure 5D). This result suggests that Pax1a is required to form barx1-positive mesenchyme and barx1-and sox9a-co-expressing chondrocytes at 48 hpf.

Next, we analyzed the cellular responses of mesenchymal cells in the dorsal region of the hyoid arch during the specification of chondrocytes. Since it was reported that cell proliferation is barely seen in the hyoid arch at 48 hpf during facial skeletal development (Barske et al., 2016; Paudel et al., 2022), we examined cell proliferation in the hyoid arch at 36 and 42 hpf with BrdU staining. In wild types and pax1a mutants, we observed a similar level of proliferating cells in the hyoid arch at 36 and 42 hpf (Figures 5E–I; Supplementary Table S1). Then, we focused on analyzing cell survival in the hyoid arch in pax1a mutants with Lysotracker and TUNEL staining. In wild types, cell death was barely observed in the condensations of osteochondral progenitors and during the specification of chondrocytes (boxed in Figures 5J, L; boxed in Supplementary Figures S4A, C; Supplementary Table S1). In pax1a mutants, little cell death was seen in the condensations (boxed in Figure 5K; boxed in Supplementary Figure S4B; Supplementary Table S1), but it increased during chondrocyte specification compared to wild types (boxed in Figures 5M, N; boxed in Supplementary Figures S4D, E; Supplementary Table S1). This result suggests that Pax1a is required to survive osteochondral progenitors during chondrocyte specification.

Considering that Pax1a is a transcription factor, the limited expression of pax1a in the first pouch, not in the hyoid arch, indicates the presence of a mediator acting downstream of Pax1a to provide the survival signal to the HM progenitors in the hyoid arch. To identify the mediator, we first examined the expression of tbx1, fgf3, itg-α-5, efnb2a, and efnb3b that are expressed in the pouch endoderm for pouch formation (Crump et al., 2004b; Herzog et al., 2004; Choe and Crump, 2014; 2015). At 36 hpf, when the HM progenitors are condensing, tbx1, fgf3, and itg-α-5 were still expressed in the posterior pouches but not in the first pouch (arrows in Supplementary Figures S5A–C), with efnb3b expression not seen in the pouches (Supplementary Figure S5F). The expression of efnb2a was observed in the pouches, with particularly strong expression in the first pouch at 36 hpf, and the expression patterns continued to 40 hpf (Supplementary Figures S5D, E).

Next, we analyzed whether the efnb2a expression in the first pouch at 36 and 40 hpf requires Pax1a. Compared to wild types (arrows in Figures 6A, C), efnb2a expression in the first pouch was unaffected at 36 hpf (arrow in Figure 6B), but it was downregulated at 40 hpf (arrowhead in Figure 6D), in pax1a mutants. The reduction in efnb2a expression observed in pax1a mutants with in situ hybridization was confirmed to be significant only at 40 hpf by qRT-PCR (Figure 6G). At 40 hpf, efnb2a expression in the first pouch was also significantly reduced in pax1a and pax1b double mutants (Figures 6F, G), whereas it was barely affected in pax1b mutants (Figures 6E, G). In particular, it was further reduced in the double mutants compared to pax1a mutants (Figure 6G). Considering the first pouch is adjacent to the dorsal region of the hyoid arch, this result implies that EfnB2a could act downstream of Pax1a in the first pouch to transmit a signal to adjacent skeletogenic mesenchymal cells in the dorsal region of the hyoid arch during HM plate formation. To test this hypothesis, we first determined the function of efnb2a in HM development. To do so, we established two loss-of-function alleles of the efnb2a gene (efnb2a ^GNU89 and efnb2a ^GNU90 ) with the CRISPR/Cas9 system (Supplementary Figures S5G, H). Although defects in the HM plate were not appreciated in efnb2a ^hu3393 mutants in our previous study (Choe and Crump, 2015), we observed HM defects in efnb2a ^GNU89 and efnb2a ^GNU90 mutants with overall 12.8% (n = 6/47) genetic penetrance of the phenotype (Figures 6H, I; Supplementary Figures S5I, J). In part, the HM defects were associated with the malformed first pouches, as evidenced by the simultaneous analysis of pouches and facial cartilages in the same individual for efnb2a mutants (Figures 6H, I). Interestingly, the HM defects in efnb2a mutants phenocopied those in pax1a mutants in that the foramen and the anterior part adjacent are missing or defective. This result suggests a genetic linkage of Pax1a and EfnB2a in the development of the HM plate.

Indeed, double mutants for pax1a and efnb2a increased the severity of defects in the HM plate compared to single mutants for pax1a or efnb2a (Supplementary Figure S6). While 6.9% (n = 2/29) of the pax1a single mutant siblings and 2.6% (n = 1/38) of the efnb2a single mutant siblings showed severe defects in the HM plate, 44.5% (n = 8/18) of the double mutant siblings displayed severe defects in the HM plate (Supplementary Figure S6A). The increased severity observed in the double mutants was statistically significant, further supporting the genetic linkage of Pax1a and EfnB2a in the development of the HM plate (Supplementary Figure S6B). However, EfnB2a did not act upstream of Pax1a in the first pouch at 36 and 40 hpf, as pax1a expression in the first pouch was unaffected in efnb2a mutants (arrows in Figures 6J–M; Supplementary Figure S5H). Taken together with the downregulation of efnb2a expression in the first pouch of pax1a mutants at 40 hpf, this result suggests that Pax1a could control HM plate formation through Efnb2a.

To determine the genetic mechanism of Pax1a to control HM plate formation through EfnB2a, we tried to rescue the HM defects in the double mutants for pax1a and pax1b that showed complete penetrance of the phenotypes, either by spatially recovering efnb2a expression in the first pouch and by temporally inducing efnb2a expression at 39 hpf after pouch formation. We utilized the UAS/Gal4 system to induce EfnB2a expression in the pouches with the Tg(sox17:Gal4VP16) driver. The Tg(sox17:Gal4VP16) driver was able to induce expression of a target gene in the pouch endoderm, including the first pouch, at 36 and 48 hpf (Supplementary Figures S7A, B). While about 70.4 (n = 19/27) to 55.8% (n = 24/43) of double mutant siblings without transgenes displayed severe defects in the HM plate, with about 29.6 (n = 8/27) to 44.2% (n = 19/43) of animals showing intermediate or weak defects in the HM plate, about 25.4 (n = 17/67) to 30.2% (n = 16/53) of double mutant siblings expressing Pax1a or EfnB2a in sox17-positive endoderm showed severe defects in the HM plate, with about 74.6 (n = 50/67) to 69.8% (n = 37/53) of animals showing intermediate or weak defects in the HM plate (Figures 7A–C). The significant reduction in the severity of the defective HM plate in the double mutants supports the partial rescue by the induction of Pax1a or Efnb2a expression in sox17-positive endoderm in the double mutants (Figures 7D, E). The rescue was due to the increased number of cells composing the HM plate (Figures 7F, G; Supplementary Table S1). Importantly, this rescue effect was specific, as the induction of Pax1a or EfnB2a expression in the sox17-positive endoderm in wild types did not affect the development of facial skeletons (Supplementary Figures S7C–E). The partial rescue of the HM defects in the double mutants by the Efnb2a expression in the sox17-positive endoderm suggests that Efnb2a acts downstream of Pax1a in the first pouch for HM plate development.

Similarly, the pax1-dependent HM defects were partially rescued by inducing efnb2a expression after pouch formation with heat shock in the double mutants for pax1a and pax1b carrying Tg(hsp70I:EfnB2a) transgene (Figure 8). Compared to wild types, efnb2a expression in the pharyngeal region was significantly reduced in the double mutants for pax1a and pax1b at 40 hpf, with its expression not induced ectopically with heat-shock treatment at 39 hpf (Supplementary Figures S7G–I). However, heat-shock treatment at 39 hpf induced ectopic efnb2a expression in most areas of the double mutants for pax1a and pax1b carrying Tg(hsp70I:EfnB2a) transgene at 40 hpf (Supplementary Figure S7J), but with the first pouch still being malformed or hypoplastic (Figures 8A, C). We also verified that heat-shock treatment at 39 hpf in wild types bearing Tg(hsp70I:EfnB2a) transgene did not affect the facial skeletons at 5 dpf (Supplementary Figure S7F). After heat-shocking at 39 hpf, about 9.6% (n = 8/83) of double mutant siblings bearing no Tg(hsp70I:EfnB2a) transgene displayed weak defects in the HM plate, with about 90.4% (n = 75/83) animals showing intermediate or severe defects in the HM plate (Figure 8B). In contrast, after the heat-shock treatment, about 27.5% (n = 22/80) of double mutant siblings bearing Tg(hsp70I:EfnB2a) transgene showed weak defects in the HM plate, and about 72.5% (n = 58/80) animals showed intermediate or severe defects in the HM plate (Figure 8D). The increase in the degree of the weakly defective HM plate in the double mutants by the heat shock at 39 hpf was statistically significant (Figure 8E). In addition, after the heat shock, the number of cells composing the HM plate increased in the double mutants carrying Tg(hsp70I:EfnB2a) transgene compared to the double mutant siblings carrying no transgene (Figure 8F). The partial rescue of the HM defects in the double mutants by the Efnb2a expression after pouch formation suggests that EfnB2a acts downstream of Pax1a in HM plate development after pouch formation. Based on the rescues of pax1-dependent defects in the HM plate either by spatial or temporal recovering efnb2a expression, we propose that Pax1a controls HM plate development through Efnb2a in the first pouch.

Our work supports the role of Pax1a in the HM plate development as well as its role in the first pouch formation. Although the HM plate development largely relies on the first pouch (Crump et al., 2004b), our rescue analysis in the double mutants for pax1 with the heat-shock treatment at 16 and 36 hpf supports the late role of Pax1a in the HM plate development along with its early role in the first pouch formation. Considering the expression patterns of pax1a and pax1b in the pharyngeal endoderm throughout the morphogenesis of pouches (Liu et al., 2020), the early expression of pax1a in the pharyngeal endoderm would be required for all pouch formation, including the first pouch, but with the first pouch formation requiring a Pax1a function distinct from Pax1b partially (Figure 9); in the pax1a mutants, the first pouch is distorted, with the other pouches being normal due to a high degree of genetic redundancy with pax1b (Figure 9). The continuous expression of pax1a in the mature first pouch adjacent to the dorsal region of the hyoid arch at 36 and 48 hpf would promote the survival of osteochondral progenitors differentiating into the chondrocytes that contribute to the HM plate, probably through the efnb2a expression in the first pouch (Figure 9); in the single mutants for pax1a and efnb2a, the differentiating barx1-positive osteochondral progenitors in the dorsal region of the hyoid arch adjacent to the mature first pouch could not receive a survival signal from the first pouch through the Pax1a-Efnb2a pathway, which would result in the loss of the chondrocytes contributing to the anterior part of the HM plate, consequently causing defects in the anterior part of the HM plate (Figure 9). The model for the role of Pax1a helps us better understand the genetic and cellular mechanisms by which this transcription factor orchestrates multiple stages of facial skeletal formation. Furthermore, although the Pax1a-Efnb2a pathway promoting cell survival has not been reported in the development of other tissues and organs, both Pax1 and EphrinB signaling have been implicated in the survival of thymic epithelial cells during embryonic development of the thymus in mice, with the epistasis between them undetermined, yet (Su et al., 2001; García-Ceca et al., 2013). Given that the thymus develops from the endoderm of the third pouch, it is likely that the Pax1a-Efnb2a pathway is recurrently used in the pouches for their developmental roles in the facial skeletons as well as in the glands forming from the pouches, including the thymus and parathyroid.

Similar to previous studies on the genetic compensations in mutants generated with the CRISPR/Cas9 system (Rossi et al., 2015; El-Brolosy et al., 2019), in the single mutants for pax1a and pax1b, we observed a genetic compensation in which the transcriptions of pax1b increase in pax1a mutants and vice versa, compared to wild types. Despite the genetic compensation, the defects in the first pouch and HM plate in pax1a mutants suggest that the increased transcription of pax1b is not able to compensate entirely for the role of Pax1a in the development of the first pouch and HM plate. In contrast, the normal development of the facial cartilages in pax1b mutants suggests that the increased expression of pax1a can replace the role of Pax1b. Moreover, the downregulation of efnb2a in the first pouch at 40 hpf in pax1a mutants in which pax1b expression increases suggests that the molecular function of Pax1a and Pax1b to regulate efnb2a expression would be different. Thus, Pax1a seems to have a distinct role from Pax1b in the first pouch and HM plate development. Alternatively, although Pax1a and Pax1b are redundant to each other in the first pouch and HM plate development, as previously reported (Liu et al., 2020), about a 1.5-fold increase in pax1b expression in pax1a mutants might not be sufficient to compensate for the role of Pax1a. In contrast, about a 12-fold increase in pax1a expression in the pax1b mutant might be enough to replace the role of Pax1b. Indeed, the increased severity and genetic penetrance of the HM defect in double mutants for pax1a and pax1b, in which both transcripts are significantly reduced, indicate redundant roles for these genes in developing the first pouch and HM plate. To determine the distinct role of Pax1a from Pax1b in the development of the first pouch and HM plate, it is necessary to analyze whether the defects in the double mutants for pax1a and pax1b can be rescued ultimately by recovering pax1b expression in the pharyngeal endoderm before the first pouch formation and after pouch formation. A positive result would support a redundant role of Pax1a and Pax1b, whereas a negative result would more likely support a distinct role of Pax1a from Pax1b in developing the first pouch and HM plate.

Our study provides an interesting scenario for the expanding role of Pax1 in craniofacial development during vertebrate jaw evolution. In hemichordates, gill slits first emerge, and the expression of pax1/9 is associated with the development of these structures (Ogasawara et al., 1999). In prevertebrate chordates, a series of repetitive pharyngeal segments driven by endodermal pouches appear, and in jawless fish, these segmental structures give rise to cartilaginous gill baskets, with pax1 and pax9 expressed in the pouches, being associated with pouch formation and the segmentation of the pharynx (DeLaurier, 2019). Additional genes, such as tbx1, fgfs, and efns, expressed in the pouches, were likely newly employed to form pouches in prevertebrate chordates or early vertebrate lineages (Jandzik et al., 2014; Koop et al., 2014). In the evolution of jawed vertebrates, these segments further developed into the complex architecture of the upper and lower jaws, as well as the supporting structures that keep them anchored (Donoghue et al., 2006), for which pax1 and pax9 that are initially involved in pouch formation, would have adapted to shape the skeletal elements in the adjacent arches by re-utilizing the signaling molecules, including the efn genes, that were utilized in pouch formation. Given that a general role of Fgf signal in specifying skeletogenic fate has been implicated in zebrafish through the expression of dominant-negative Fgf receptors and treatment with the FGFR inhibitor SU5402 (Walshe and Mason, 2003; Sperber and Dawid, 2008; Das and Crump, 2012), a straightforward way to test this scenario would be to examine the direct role of fgf3, which acts downstream of Pax1 in pouch formation in medaka and zebrafish, in the development of facial skeletal elements. Indeed, although the genetic interaction between Fgf3 and Pax1 has not yet been elucidated, we have recently found a role of Fgf3 in the development of epibranchial cartilages, a dorsal skeletal element that arises in the branchial arches, in zebrafish. Our analysis of the role of Pax1a reveals a previously unappreciated role in the development of HM plate and provides a novel insight into a genetic mechanism of the first pouch controlling the HM plate formation through the Pax1a-Efnb2a pathway, as well as its significance in the evolution of vertebrate jaws.