Endodermal pouch-expressed dmrt2b is important for pharyngeal cartilage formation

ABSTRACT Pharyngeal pouches, a series of outpocketings derived from the foregut endoderm, are essential for craniofacial skeleton formation. However, the molecular mechanisms underlying endodermal pouch-regulated head cartilage development are not fully understood. In this study, we find that zebrafish dmrt2b, a gene encoding Doublesex- and Mab-3-related transcription factor, is specifically expressed in endodermal pouches and required for normal pharyngeal cartilage development. Loss of dmrt2b doesn't affect cranial neural crest (CNC) specification and migration, but leads to prechondrogenic condensation defects by reducing cxcl12b expression after CNC cell movement into the pharyngeal arches. Moreover, dmrt2b inactivation results in reduced proliferation and impaired differentiation of CNC cells. We also show that dmrt2b suppresses crossveinless 2 expression in endodermal pouches to maintain BMP/Smad signaling in the arches, thereby facilitating CNC cell proliferation and chondrogenic differentiation. This work provides insight into how transcription factors expressed in endodermal pouches regulate pharyngeal skeleton development through tissue–tissue interactions.


INTRODUCTION
Craniofacial malformations, which occur due to developmental issues of the head, face and neck, account for approximately onethird of congenital birth defects (DeLuke, 2014). Owing to the ceaseless efforts of scientists, more than 700 distinct craniofacial syndromes have been described (Johnson and Wilkie, 2011;Lisa and Elden, 2014;Schutte and Murray, 1999). The neurocranium is derived from both the cranial neural crest (CNC) and mesoderm, while the pharyngeal skeleton, including the jaw and branchial arches, is solely derived from CNC cells (Yelick and Schilling, 2002). CNC cells emerge from the dorsal and lateral regions of the neural ectoderm when the epidermal ectoderm interacts with the neural plate to induce formation of the neural plate border (Donoghue et al., 2008). Subsequently, bilateral CNC cells migrate medially with the developing head and then from the midbrain and hindbrain as three streams of collective cell populations (mandibular, hyoid and branchial) into the pharyngeal arches to form the pharyngeal cartilages (Couly et al., 1993;Köntges and Lumsden, 1996;Lumsden et al., 1991;Schilling and Kimmel, 1994).
The craniofacial complex comprises cells from all three germ layer origins: ectodermal, endodermal and mesodermal, and craniofacial morphogenesis requires continuous and reciprocal tissue-tissue interactions (Chai and Maxson, 2006). In particular, in the pharyngeal arches, the CNC with mesoderm core is separated with endodermal pouch inner and covered with epidermal ectoderm outer (Noden, 1988). Endodermal pouches are a series of outpocketings developed in an anterior-posterior wave from the pharyngeal endoderm. Interestingly, although CNC cells are not required for the formation of endodermal pouches (Veitch et al., 1999), these pouches have signaling functions important for the development of the pharyngeal skeleton. Zebrafish mutants, such as casanova/sox32 and faust/gata5, which lack early endoderm and the pharyngeal pouches, exhibit severe defects in craniofacial chondrogenesis, suggesting the endodermal requirements in pharyngeal skeletal development (Dickmeis et al., 2001;Kikuchi et al., 2001;Reiter et al., 1999). Endodermal pouch-derived FGF and BMP ligands have been shown to be required for the survival, proliferation and differentiation of postmigratory CNC cells that give rise to branchial cartilages (Crump et al., 2004;David et al., 2002;Holzschuh et al., 2005;Ning et al., 2013;Nissen et al., 2003). In addition, the T-box transcription factor Tbx1 is a key molecule in the regulation of tissue-tissue interactions (Choe and Crump, 2014;Huh and Ornitz, 2010;Kopinke et al., 2006;Okada et al., 2016;Okubo et al., 2011;Piotrowski et al., 2003). tbx1 is expressed in the endodermal pouches as well as the mesodermal core of the pharyngeal arches (Piotrowski et al., 2003). In vgo/tbx1 mutants, the pharyngeal pouches are largely absent, and the pharyngeal cartilages are misshapen and fused together (Piotrowski and Nüsslein-Volhard, 2000). Yet while mesodermal Tbx1 has been shown to function in shaping the lower jaw (Aggarwal et al., 2010), transplantation of wild-type endoderm into vgo/tbx1 mutants partial rescue the formation of pharyngeal cartilages, indicating that tbx1 acts non-autonomously in the endoderm (Piotrowski et al., 2003). The identification of new chondrogenic regulators with endodermal expression will promote our understanding of the tissue-tissue interactions during craniofacial skeleton development.
The doublesex/mab-3 related (Dmrt) gene family consists of transcription factors with a DSX/MAB-3 (DM) domain, which is a zinc finger-like DNA binding motif first identified in the sexual regulatory proteins Doublesex (DSX) and MAB-3 (Erdman and Burtis, 1993;Raymond et al., 1998). There are multiple dmrt paralogs in the animal kingdom, but most Dmrt proteins display little similarity with the exception of their DM domain (Volff et al., 2003). These dmrt genes have different spatial-temporal expression, suggesting they could have additional functions besides sex determination (Hodgkin, 2002;Hong et al., 2007;Lints and Emmons, 2002;Volff et al., 2003). There are five dmrt genes, designated drmt1, dmrt2a, dmrt2b, dmrt3 and dmrt5, in zebrafish. The dmrt2a and dmrt2b genes originated from the second round of genome duplication, and dmrt2a is the homolog of Dmrt2 that is involved in somitogenesis in vertebrates (Lu et al., 2017;Matsui et al., 2012;Meng et al., 1999;Sato et al., 2010;Saúde et al., 2005;Seo et al., 2006). Interestingly, dmrt2b is expressed in the pharyngeal region (Johnsen and Andersen, 2012;Zhou et al., 2008), indicating its potential role in the development of the branchial skeleton.
In this study, we find that zebrafish dmrt2b is uniquely expressed in endodermal pouches and reveal a function for this gene in regulating endodermal expression of cxcl12b and crossveinless 2 (cv2) to promote pharyngeal CNC cell condensation, proliferation and differentiation. Therefore, this study uncovers a molecular mechanism for regulation of craniofacial cartilage development through tissue-tissue interactions mediated by endodermal pouchexpressed transcription factor Dmrt2b.

Zebrafish dmrt2b is specifically expressed in pharyngeal pouches
To evaluate the developmental functions of dmrt2b, we first examined its spatiotemporal expression during zebrafish embryonic development using in situ hybridizations with an anti-sense probe targeting the cDNA sequence downstream the coding region of DM domain. We found that dmrt2b was uniquely expressed in the pharyngeal region as early as 18 hours post-fertilization (hpf ), when the first endodermal pouch budded (Fig. 1A). During later stages, dmrt2b expression spread to the bilateral side of the head in a thread-like manner ( Fig. 1B-D), suggesting that dmrt2b is expressed in the endodermal pouches. To examine this, 8 ng sox32 morpholino (MO) was injected into embryos at the one-cell stage, which resulted in the elimination of the entire endoderm and endoderm-derived pouches as indicated by sox17 and nkx2.3 expression (Fig. 1E). As expected, dmrt2b expression disappeared from the sox32 morphants (Fig. 1F). Furthermore, RNAscope in situ hybridization combined with immunofluorescence was employed to figure out the exact expression pattern of dmrt2b in the Tg(sox17: GFP) transgenic fish embryos (Chung and Stainier, 2008). As shown in Fig. 1G, dmrt2b transcripts co-localized with GFP-expressing endodermal pouches at 36 hpf. These observations strongly suggest that dmrt2b is expressed specifically in the pharyngeal pouches.

Loss of dmrt2b causes malformation of pharyngeal cartilages
To investigate whether dmrt2b has functions in pharyngeal pouch formation and craniofacial cartilage development, we mutated the dmrt2b gene using the CRISPR-Cas9 system. Because the DM domain enables dmrt2b to act as a transcription factor, we targeted this domain and obtained one mutant with a four base frameshift deletion in dmrt2b, which led to a premature stop codon ( Fig. 2A; Fig. S1A). Furthermore, the obvious reduction of dmrt2b transcripts in dmrt2b homozygous mutants confirmed the loss of function of this gene (Fig. S1B). In comparison to wild-type and heterozygous siblings, dmrt2b homozygous mutants had shrunken heads, pericardial edema and smaller jaws (Fig. 2B,C). In addition, the dmrt2b mutants had restricted protruding jaws due to poor pharyngeal arch outgrowth (Fig. 2D). Alcian Blue staining revealed severe reduced and dysmorphic neurocranium cartilages and pharyngeal arch-derived chondrogenic elements in these mutants, suggesting chondrogenic differentiation defects (Fig. 2E). Importantly, these cartilages were recovered by injection of 475 pg dmrt2b mRNA into dmrt2b mutants (Fig. 2E).
To further confirm the role of dmrt2b in these pharyngeal cartilage defects, knockdown experiments were performed using antisense MOs. Specifically, 4 ng of dmrt2b MO targeting the intron-exon boundary of the first intron and the second exon of the dmrt2b gene was injected into one-cell stage embryos. This resulted in the elimination of endogenous mature dmrt2b mRNA and the emergence of interfered mRNA products in the morphants (Fig. S2A), indicating the dmrt2b MO is specific and effective. Similar to the dmrt2b mutants, the knockdown morphants had obvious defects in head cartilage formation, which were abrogated by co-injection of dmrt2b mRNA ( Fig. 2F; Fig. S2B). Interestingly, the pericardial edema in the morphants was also alleviated by dmrt2b mRNA injection, indicating that dmrt2b might function in cardiac development (Fig. S2B). miR30-based short hairpin RNAs (shRNAs) from tissue specific promoters displayed very efficient knockdown of gene expression in eucaryotic organisms (Stegmeier et al., 2005;Zeng et al., 2005). To explore tissue-specific roles of dmrt2b, we utilized the KalTA4-UAS system to drive the expression of miR30-based shRNAs (dmrt2b-shRNA1 and dmrt2b-shRNA2) against two different regions of dmrt2b transcripts. By using the Tol2 transposon, we generated a Tg(nkx2.3:KalTA4-p2a-mCherry) transgenic zebrafish line with a 5.5 kb nkx2.3 promoter that could specifically drive the expression of KalTA4 activators and red fluorescent proteins in endodermal pouches (Choe et al., 2013), which were indicated by Tg(sox17:GFP) embryos at 36 hpf ( Fig. S3A). As shown in Fig. S3B, co-injection of 50 pg UAS:dmrt2b-shRNA1 plasmid with 100 pg Tol2 transposase mRNA into one-cell stage Tg(nkx2.3:KalTA4-p2a-mCherry) embryos led to an obvious decrease of dmrt2b expression in endodermal pouches compared with control embryos (Fig. S3B). Importantly, the inactivation of dmrt2b in pouches resulted in obvious defects in head cartilage formation (Fig. 2G,H). Thus, endodermal pouch-expressed dmrt2b is important for craniofacial cartilage development. In addition, tissue specific depletion of dmrt2b by injection of UAS:dmrt2b-shRNA1 plasmid into Tg(nkx2.3: KalTA4-p2a-mCherry) embryos resulted in obvious pericardial edema, implying a non-cell autonomous role of pharyngeal endodermal dmrt2b during heart development ( Fig. 2G). Interestingly, more severe pharyngeal cartilage defects were observed in dmrt2b MO or UAS:dmrt2b-shRNA1 plasmid injected embryos than dmrt2b mutants, indicating that a compensatory protective response against the loss of dmrt2b may be to some extent activated in the mutants (Rossi et al., 2015;Wei et al., 2017).

Inactivation of dmrt2b results in disorganized pharyngeal arches
To delineate the mechanisms underlying pharyngeal cartilage defects in the absence of functional dmrt2b, we examined the expression of several different markers over the course of embryonic development. In situ hybridization revealed the CNC specification marker foxd3 was expressed in a similar manner in control embryos and dmrt2b mutants at the five somite stage (Fig. S4A). This demonstrates that dmrt2b is not required for the specification of the CNC. In control embryos, dlx2a was expressed in the three CNC groups at 18 and 24 hpf. A slight decrease of dlx2a expression was observed at 18 hpf in dmrt2b mutants, but subsequently recovered at 24 hpf, indicating that dmrt2b is not essential for CNC cell migration into the pharyngeal arches ( Fig. S4B,C).
Subsequently, the pharyngeal endoderm migrates laterally to form pouches that interdigitate each of the pharyngeal arches (Crump et al., 2004), then the postmigratory CNC cells in the arches start condensation and proliferation processes and finally differentiate into chondrocytes (Clouthier and Schilling, 2004;Hall and Miyake, 1995;Hall and Miyake, 2000). As dmrt2b continues to express in the endodermal pouches after CNC cells reach their destination in the arches, we then examined the arch morphology in dmrt2b depleted embryos via the expression of two markers of postmigratory crest, dlx2a and hand2, at 36 hpf. Compared with control embryos, there were notably fewer CNC cells in dmrt2b mutants and morphants (Fig. 3A,B). Similar phenotypes were observed in dmrt2b MO-injected Tg(fli1:GFP) embryos expressing GFP in CNC derivatives at 28 hpf ( Fig. 3C) (Lawson and Weinstein, 2002). Furthermore, dmrt2b morphants exhibited loose and disorganized anterior arch structures and the CNC cells failed to aggregate toward certain centers (Fig. 3C). Moreover, injection of 4 ng dmrt2b MO into Tg(nkx2.3:mCherry; sox10:EGFP) embryos (Carney et al., 2006) resulted in no obvious defects in mCherry-expressing pharyngeal pouches at 36 hpf, and the endodermal pouch marker nkx2.3 was normally expressed in both dmrt2b mutants and morphants (Fig. 3D,E). In contrast, dmtr2b depletion gave rise to a similar disorganized arch phenotype and notably fewer GFP-positive CNC cells (Fig. 3D). Moreover, the CNC cells in the pharyngeal region were flattened and elongated, reflecting condensation defects in 4 ng dmrt2b MO injected Tg(sox10:mCherry-CAAX) embryos, in which cell shape was outlined through plasma membrane-bound mCherry (Fig. 3F,G). Taken together, these findings reveal that dmrt2b is dispensable for endodermal pouch morphogenesis, but plays an important role in controlling prechondrogenic condensation in the pharyngeal arches.
Because Dmrt2b is a transcription factor specifically expressed in the pharyngeal endoderm, the non-autonomous activity of dmrt2b should be mediated by some secretory molecules that are derived from endodermal pouches and able to regulate pharyngeal arch development. It has been shown that chemokine Cxcl12b signaling from the endodermal pouches is required for the proper condensation of Cxcr4a expressing CNC cells in pharyngeal arches (Boer et al., 2015;Olesnicky Killian et al., 2009). Therefore, we speculate that dmrt2b might regulate the expression of cxcl12b in pouches to control the prechondrogenic condensation. In support of this hypothesis, after CNC cells migrating from the brain into the pharyngeal arches, both genetic depletion and knockdown of dmrt2b resulted in a significant reduction of cxcl12b expression in the pharyngeal region ( Fig. 3H-J). In addition, the condensation defects of GFP-positive CNC cells were partially rescued by co-injection of 20 pg cxcl12b mRNA into dmrt2b morphants (Fig. 3K). Therefore, we conclude that dmrt2b is a positive regulator of cxcl12b in endodermal pouches and thereby drives CNC cell compaction in the pharyngeal arches.

The proliferation and chondrogenic differentiation of CNC cells require dmrt2b
After migrating into the pharyngeal arches, CNC cells proliferate and differentiate into chondrocytes (Hall and Miyake, 2000;Ning et al., 2013). To dynamically observe pharyngeal cartilage defects induced by dmrt2b inactivation, in vivo time-lapse imaging of CNC cells was performed on Tg(fli1:EGFP) embryos. In the control MO (cMO)-injected embryos, GFP-positive CNC cells aggregated as prechondrogenic condensations at 48 hpf, differentiated into chondrocytes at 60 hpf, and organized into chondrocyte stacks from 72 to 84 hpf (Fig. S5). However, in addition to the condensation defects in the first and second pharyngeal arches in dmrt2b morphants, there were significantly fewer CNC cells at 48 hpf (Fig. S5). Moreover, only a few CNC cells were observed in pharyngeal arches 3-7 (Fig. S5), consistent with the previously noted cell number reduction in the dmrt2b morphants (Fig. 3C,D). These posterior segments emerged at 60 hpf, but were much smaller (Fig. S5). At 72 and 84 hpf, the palatoquadrate and ceratohyal cartilages were shorter and the chondrocytes failed to stack (Fig. S5). These observations are consistent with the Alcian Blue staining results (Fig. 2E).
The reduction in the size of the pre-chondrogenic segments in the dmrt2b morphants raises the possibility that cell proliferation and/or survival are inhibited. Therefore, Tg(fli1:EGFP) embryos were immunostained for phosphorylated histone 3 ( pH3) to assess CNC cell proliferation at 28, 36 and 48 hpf. We found notably fewer mitotic pH3-positive CNC cells in dmrt2b morphants than control embryos (Fig. 4A-E). Conversely, TUNEL revealed a lack of apoptotic cells in the pharyngeal region in dmrt2b mutants and morphants (Fig. S6A,B). These findings support that loss of dmrt2b impairs CNC cell proliferation, which contributes to the reduction in chondrocyte number within the pharyngeal arches.
Due to the disordered arrangement of the chondrocytes in the absence of dmrt2b, we inferred that dmrt2b also has an essential role in CNC cell differentiation into chondrocytes. To test this, we assessed the expression of col2a1, the gene encoding type II collagen, the primary cartilage matrix protein produced by mature chondrocytes (Vandenberg et al., 1991;Yan et al., 2002). There were dramatically fewer col2a1 transcripts in the pharyngeal region of dmrt2b mutants (Fig. 4F). Meanwhile, Col2 protein levels were also reduced and displayed discontinuous distribution in dmrt2bdepleted embryos (Fig. 4G,H). Therefore, loss of dmrt2b disrupts chondrogenic differentiation of CNC cells.
dmrt2b maintains BMP signaling through inhibiting crossveinless 2 to facilitate CNC cell proliferation and chondrogenic differentiation BMP signaling is essential for CNC cell proliferation and chondrogenic differentiation during pharyngeal cartilage development (Ning et al., 2013;Retting et al., 2009;Yoon et al., 2005). Therefore, we mated Tg(sox10:mCherry-CAAX) with Tg(BRE:EGFP), a BMP signaling reporter transgenic line (Laux et al., 2011), to examine whether inactivation of dmrt2b affects BMP signaling. In the dmrt2b morphants, there was decreased fluorescence intensity in the pharyngeal arches, implying loss of dmrt2b attenuates BMP signaling in CNC cells (Fig. 5A). To further confirm this result, we examined the expression level of endogenous phosphorylated Smad1/5/8 ( p-Smad1/5/8), the intracellular effectors of BMP signaling, in dmrt2b mutants and Tg(fli1: EGFP) embryos injected with 4 ng dmrt2b MO, and found a significantly decrease of p-Smad1/5/8 level in pharyngeal region when dmrt2b was depleted (Fig. 5B,C). Thus, we established that dmrt2b is required to maintain BMP signaling in CNC cells.
BMP genes, including bmp2a, bmp2b, bmp4 and bmp5, are expressed in the endodermal pouches during pharyngeal arch development (Holzschuh et al., 2005). However, the expression of these BMP genes was not obviously decreased in dmrt2b-depleted embryos (data not shown). Another possibility is that dmrt2b promotes BMP signaling in the pharyngeal region by inhibiting the expression of some BMP antagonists. In zebrafish embryos, several BMP antagonist genes like noggin3, follistatin, chordin and chordin-like 2, have been found to be expressed in the pharyngeal region, but only crossveinless 2 (cv2) is specifically expressed in the endodermal pouches during cartilage development (Ning et al., 2013;Rentzsch et al., 2006). It has been reported that Cv2 functions as a secreted BMP inhibitory protein during human chondrogenic and osteogenic differentiation (Binnerts et al., 2004). In the loss-offunction experiments, we observed that the expression of cv2 in the pharyngeal pouches was significantly higher in dmrt2b morphants compared to control embryos (Fig. 5D,E). Importantly, BMP signaling in the CNC cells was partially restored by co-injecting 100 pg cv2 MO into the dmrt2b morphants or mutants (Fig. 5F,G). Together, these data indicate that dmrt2b inhibits cv2 expression in the pharyngeal pouches, thereby maintaining BMP signaling and facilitating CNC cell proliferation and differentiation.

DISCUSSION
The interaction of different tissues plays vital roles in organogenesis during embryo developemt (Zorn and Wells, 2009). For example, in zebrafish, wnt2bb is expressed in restricted bilateral domains in the lateral plate mesoderm and directly induces the adjacent endoderm to form liver anlage (Ober et al., 2006). Recent evidence suggests that trachea-derived Decapentaplegic, the main bone morphogenetic protein ligand in Drosophila, is required for adult midgut homeostasis (Li et al., 2013). During craniofacial cartilage development, once the CNC cells reach their destination in the arches, signals from surrounding endodermal pouches such as FGF, BMP and CXCL12, direct the final cell fate (Crump et al., 2004;David et al., 2002;Holzschuh et al., 2005;Ning et al., 2013;Nissen et al., 2003;Boer et al., 2015;Olesnicky Killian et al., 2009). In this study, we find that dmrt2b is expressed in the endodermal pouches and loss of dmrt2b impairs pharyngeal cartilage formation. dmrt2b plays a critical role in the condensation of postmigratory CNC cells by promoting cxcl12b expression. We also provide evidence that dmrt2b is required for CNC cell proliferation and chondrogenic differentiation due to its ability to suppress cv2 expression and, thus, maintain BMP signaling in pharyngeal regions. Therefore, this study demonstrates dmrt2b-mediated tissue-tissue interactions are essential for pharyngeal skeleton development. dmrt2b mutants also exhibit severe neurocranial defects and pericardial edema, indicating the possibility that dmrt2b is expressed at relatively low levels in other regions beyond endodermal pouches.
Members of the Dmrt family are generally associated with sex determination, but mouse Dmrt2 is not essential for sexual differentiation (Seo et al., 2006). Mouse Dmrt2 and its homologue gene, zebrafish terra/dmrt2a, have shown to be expressed specifically in developing smites and function in somitogenesis (Meng et al., 1999;Saúde et al., 2005;Seo et al., 2006). Zebrafish dmrt2a is also required for left-right asymmetric organ positioning (Matsui et al., 2012;Saúde et al., 2005), while this left-right function is not conserved in mouse (Lourenço et al., 2010). Interestingly, in mouse Dmrt2 mutants, the somite patterning defects were gradually recovered during embryonic development, but the axial skeletal and rib malformations were evidently induced by the lacking of Fgf4 and Fgf6 expression in the myotome, suggesting a non-cell autonomous role of Dmrt2 in controlling skeletal development (Seo et al., 2006). In our experiments, dmrt2b is found to be essential for pharyngeal skeleton development. Multiple lines of evidence support the idea that dmrt2b functions in a similar non-cell-autonomous manner by transferring developmental signals from pharyngeal endoderm to postmigratory CNC cells. (1) The expression of dmrt2b can be specifically detected in endodermal pouches by in situ hybridization experiments during pharyngeal cartilage development.
(2) The endodermal pouches in dmrt2b depleted embryos are normally All data are presented as the mean of three independent experiments. Error bars represent s.d. Significance was analyzed using unpaired t-tests. **P<0.01. (F) Co-injection of 100 pg cv2 MO with 4 ng dmrt2b MO partially rescued the EGFP expression in Tg(BRE:EGFP;sox10:mCherry-CAAX) transgenic embryos. (G) p-Smad1/5/8 levels were partially rescued in the pharyngeal region of dmrt2b mutants co-injected with 100 pg cv2 MO at 36 hpf. Embryos were stained with the indicated antibodies. Nuclei were counterstained with DAPI (blue). ov, otic vesicle; p-S1/5/8, phosphorylated Smad1/5/8. Scale bars: 50 µm (A-C,F), 20 µm (G). developed, indicating that the skeletal defects in the pharyngeal region are not the secondary effects of abnormalities of endodermal pouch development. (3) The experimental inhibition of dmrt2b function gene in pouches gives rise to obvious defects in head cartilage formation. (4) Inactivation of dmrt2b leads to obviously altered expression of cxcl12b and cv2 in endodermal pouches, which results in disorganized arches and proliferation and differentiation defects in CNC cells. (5) Importantly, co-injection of cxcl12b mRNA or cv2 MO into the dmrt2b morphants can partially recover the condensation defects or the decrease of BMP signaling in CNC cells. All these observations imply that dmrt2b regulates craniofacial skeleton development through tissue-tissue interactions. Although mouse Dmrt2 mutant does not exhibit obvious craniofacial abnormalities (Seo et al., 2006), our findings will help to understand the developmental differences of craniofacial skeletons between lower vertebrates and mammals.
CXCL12, also known as stromal derived factor 1 (SDF-1), signals via its cognate receptor CXCR4 and plays a key role in many cellular processes including hematopoiesis, organogenesis and vascularization (Cheng et al., 2014;Teicher and Fricker, 2010). In mouse and chick embryos, Cxcl12 is expressed in the lateral ectoderm and pharyngeal endoderm at early stages of CNC cell migration, while Cxcr4 is expressed in migrating pharyngeal NC cells (Escot et al., 2016). Defective CXCR4 signaling impedes the migration of CNC cells into pharyngeal arches and leads to anomalies of the lower jaw and hyoid bone (Escot et al., 2016). In zebrafish embryos, cxcl12b, but not cxcl12a, is expressed within the domain of CNC cell migration from 14-17 hpf and in the endodermal pouches during pharyngeal arch morphogenesis (Olesnicky Killian et al., 2009). Unlike mouse mutants, cxcl12b or cxcr4a morphants display only a mild migration phenotype as most CNC cells arrive at the pharyngeal arches. Disruption of Cxcl12b/Cxcr4a signaling in zebrafish also results in the failure of CNC cells to fully condense within the pharyngeal arches, which is thought to be secondarily caused by the aberrant migration of CNC cells (Olesnicky Killian et al., 2009). Interestingly, loss of dmrt2b gives rise to decreased expression of cxcl12b and disorganized cells within the arches that resemble the defects observed in cxcl12b morphants, suggesting that cxcl12b is genetically downstream of dmrt2b during pharyngeal NC development. In addition, the expression of dmrt2b in the pharyngeal region is not detected until 18 hpf, and the expression of cxcl12b starts to decrease at 24 hpf, when the CNC cells have already migrated into pharyngeal arches. These observations would explain the lack of CNC cell migration defects in dmrt2b mutants, and raise the provocative idea that Cxcl12b/Cxcr4a signaling may function directly in pharyngeal NC condensation. Moreover, the disorganized arch phenotype could be rescued by injecting cxcl12b mRNA into dmrt2b morphants, suggesting that cxcl12b is the major downstream target of dmrt2b for regulating CNC cell condensation.
Cv2 displays opposing effects on BMP signaling depending on the biological context. Cv2 has been shown to potentiate BMP signaling during mouse organogenesis (Ikeya et al., 2006), and crossvein formation in the fly wing (Conley et al., 2000;O'Connor et al., 2006;Ralston and Blair, 2005), but functions as a BMP antagonist during endothelial cell differentiation (Moser et al., 2003), frog embryogenesis (Ambrosio et al., 2008;Coles et al., 2004) and human chondrogenic and osteogenic differentiation (Binnerts et al., 2004). In zebrafish, loss of cv2 via MO-mediated knockdown results in reduced BMP signaling and dorsalized phenotypes during gastrulation (Rentzsch et al., 2006). In contrast, in our experiments, co-injection of cv2 MO into dmrt2b morphants could partially recover the reduction of BMP activity in the pharyngeal arches, suggesting that Cv2 antagonizes BMP activity during lower jaw development. Indeed, the full-length zebrafish Cv2 protein acts as an inhibitor of BMP signaling and can be converted from an anti-to a pro-Bmp factor by proteolytic cleavage (Rentzsch et al., 2006). However, whether Cv2 protein can be cleaved in the pharyngeal region remains to be determined. BMP signaling has long been recognized as an essential signal for neural crest cell specification and migration (Kanzler et al., 2000;Nie et al., 2006;Tribulo et al., 2003). During early craniofacial development, BMP signaling is required for the dorsal-ventral (DV) patterning of the pharyngeal arches Bonilla-Claudio et al., 2012;Zuniga et al., 2011). Previous studies show that, in zebrafish, the requirement of BMP activity for ventral arch development only occurs within a narrow time window from 17 to 24 hpf, a period just after CNC cell migration and before the establishment of arch primordia . Not surprisingly, in dmrt2b mutants, no obvious defects of CNC cell specification, migration and DV patterning were found, as dmrt2b is expressed in the pharyngeal endoderm as early as 18 hpf and regulates cv2 expression after 24 hpf. By contrast, our studies reveal a significant role of dmtr2b in CNC cell proliferation and chondrogenic differentiation by maintaining BMP activity in pharyngeal arches via suppressing cv2 expression. This is supported by previous findings that, after arch primordia are established, inactivation of BMP signaling leads to poor proliferation and impaired differentiation of pharyngeal chondrogenic progenitors (Ning et al., 2013).

Generation of dmrt2b mutants
The zebrafish dmrt2b mutant was generated using the CRISPR/Cas9 system. The dmrt2b gRNA was designed using ZiFiT Targeter (http://zifit. partners.org/ZiFiT/ChoiceMenu.aspx) and the targeting sequence was 5′-GGTTGCAGACTGCGCCGGAC-3′. The Cas9 mRNA and gRNA were prepared as previously described (Wei et al., 2017) and co-injected into one-cell stage wild-type embryos. For genotyping analysis, the genomic DNA was isolated and used as template for amplification of gRNA targeted sequences with the forward primer 5′-CAATCACTGCTGCATTCCGAC-3′ and the reverse primer 5′-TGTCTCCGTAGGGCGACTTGA-3′. Then the amplified fragments were identified with Sanger DNA sequencing.
For rescue studies, the full length cDNAs of dmrt2b and cxcl12b were amplified by RT-PCR using the following primers and then cloned into the pCS2-Flag vector.
TUNEL assays were performed using the In Situ Cell Death Detection Kit, TMR red (12156792910, Roche) according to the manufacturer's instructions. DAPI was used to visualize nuclei.

Alcian Blue staining
Embryos were fixed in 4% paraformaldehyde overnight at 4°C. Immediately following, fixed embryos were washed in distilled water with 0.1% Tween-20 for 8 h. The embryos were then stained with Alcian Blue staining buffer (0.015% Alcian Blue, 80% ethanol, and 20% acetic acid) overnight at room temperature and then de-stained in 70% ethanol/30% acetic acid. Next, the embryos were rehydrated through a graded series of alcohols to distilled water and then treated with 0.5% trypsin (0458, AMRESCO) in supersaturated borax at room temperature until the tissues were soft enough to dissect. The embryos were then transferred to 1% KOH for 30 min and then washed in distilled water with 0.1% Tween-20 twice for 5 min each. Finally, the embryos were dehydrated with a graded series of glycerol solutions and dissected for imaging.

Time-lapse imaging
Embryos were anaesthetized and embedded in 0.8% low-melt agarose (0815, AMRESCO) at the indicated time points for live imaging with a Nikon A1R+ confocal microscope (20× dry, 40× dry or 60× oil objectives). All confocal stack pictures were processed using Nikon NIS-Elements AR 4.13.00 software.

Statistical analysis
The cell shape of CNC cells was indicated by length/width ratio and the front eight rows of CNC cells in the first and second pharyngeal arches were analyzed. ImageJ software was used to measure the distance. All experiments were performed in triplicate and unpaired t-test was employed to analyze all data sets. Results were considered statistically significant at P<0.05.