A comprehensive series of Irx cluster mutants reveals diverse roles in facial cartilage development

Development of the facial and limb skeletons follows a similar trajectory. Interactions of specialized epithelia with skeletogenic mesenchyme (neural crest-derived for the face, mesoderm-derived for the limb) promote proliferative outgrowth of mesenchyme followed by differentiation into a series of cartilage elements. These cartilages function primarily as transient templates for later bone differentiation, with permanent cartilage remaining at flexible joints between bones and in the nose, ear and trachea. Compared with cartilage specification, we know less about how individual cartilage elements are kept separate, both in the embryonic mesenchyme and later at joints.

Members of the Iroquois-related (Irx) transcription factor family are good candidates for maintaining cartilage separation during embryogenesis. Irx genes are required for the development of numerous organs, including the nervous system, heart, kidney and skeleton, and are thought to function generally as transcriptional repressors (Cavodeassi et al., 2001; Gomez-Skarmeta et al., 2001). The tetrapod IrxA cluster consists of Irx1, Irx2 and Irx4, and the IrxB cluster consists of Irx3, Irx5 and Irx6. In chickens and mice, IrxA and IrxB genes are expressed in the mesenchyme and perichondrium between nascent digit cartilages and in joint-forming regions (McDonald et al., 2010). The Fused Toes mouse mutant involves heterozygous deletion of the IrxB cluster (and three other genes) and is characterized by fusions of phalangeal foot bones (Grotewold and Ruther, 2002; Peters et al., 2002). In zebrafish, the IrxB gene irx5a, along with the fish-specific non-clustered irx7, are expressed at and required for development of the hyoid joint in the face, in part by blocking Sox9 activation of target genes such as col2a1a (Askary et al., 2015). IrxB genes also have earlier roles in formation of the limbs in mice, with proximal and anterior limb cartilages reduced in Irx3^−/−; Irx5^−/− mice (Li et al., 2014). In addition, mice homozygous for the Fused Toes deletion, and hence lacking all IrxB genes, have hypoplastic pharyngeal arches, the embryonic structures populated by the neural crest-derived mesenchymal precursors of the facial skeleton (Anselme et al., 2007). In human Hamamy syndrome, dominant mutations in IRX5 cause distinctive facial defects, with work in frog suggesting that Irx5 functions with Gata3 and Trps1 to regulate migration of the neural crest-derived precursors of the facial skeleton (Bonnard et al., 2012). In mice, Irx3 and Irx5 are also redundantly required for osteoblast differentiation in the cranium (Cain et al., 2016) and long bones (Tan et al., 2020).

Challenges in addressing Irx function include not only their iterative roles during development, e.g. early limb patterning (Li et al., 2014) versus later digit separation (Peters et al., 2002), but also genetic redundancy. Within and between clusters, Irx genes display similar and overlapping expression patterns, explained in part by super-enhancers that co-regulate Irx gene expression within a cluster (Tena et al., 2011). Zebrafish have 11 Irx genes, compared with six in tetrapods, which are organized into two IrxA clusters (IrxAa, IrxAb), two IrxB clusters (IrxBa, IrxBb), and the single fish-specific irx7 gene, likely an evolutionary remnant of a third Irx cluster (Askary et al., 2015). We have taken advantage of the power of zebrafish genetics to mutate all Irx cluster genes and analyze the combinatorial requirements of individual Irx clusters. Phenotypic analysis reveals requirements of individual Irx clusters in body axis maintenance, heart development and skeletal development in the face and pectoral fin. We find a requirement for IrxAb and irx7 genes in formation of the anterior facial skeleton in part by promoting the timely formation of cranial neural crest-derived cells (CNCCs). In contrast, IrxBa genes are required for formation of the posterior gill support skeleton in part by promoting the formation of endodermal pouches. Combinatorial Irx mutants also display widespread fusions of cartilages throughout the face, with in vivo imaging showing that inappropriate fusions between the lower jaw Meckel's and jaw support ceratohyal cartilages are due to inappropriate chondrogenesis of the perichondrium. Our work highlights evolutionarily conserved and redundant roles for Irx genes in the formation and individuation of cartilages throughout the body.

In order to understand requirements for the four clusters of Irx genes, we used Cas9 and multiple single guide RNAs (sgRNAs) (Hwang et al., 2013) to make small insertions or deletions that disrupt the open reading frame of each Irx gene, thus leading to truncated proteins lacking the conserved DNA-binding domains (Fig. 1A). We did not delete intervening regions between Irx genes in a cluster. Starting with the previously published irx5a mutant (Askary et al., 2015), we injected sgRNAs targeting both irx3a and irx6a and identified linked alleles with out-of-frame mutations in irx3a, irx5a and irx6a (‘IrxBa’ allele). Of the 23 injected animals that were screened, 18 had germline mutations in at least one gene, four had germline mutations in all IrxBa genes, and, of these, one had out-of-frame mutations in each and was used for phenotypic analysis. To generate the ‘IrxBb’ allele, we injected an sgRNA for irx5b, identified a founder for an irx5b mutant allele, and subsequently injected its progeny with irx3b sgRNA to generate linked irx3b and irx5b out-of-frame mutations. IrxAa and IrxAb alleles were created by simultaneously targeting all genes within each cluster (irx1a, irx2a and irx4a for IrxAa; irx1b and irx4b for IrxAb) and identifying founders with linked out-of-frame mutations for each cluster gene. For the IrxAa cluster, 48% of 19 injected and screened animals had single germline mutations, 33% two gene mutations and 19% three gene mutations, with one out of ten of the triple mutant founders having out-of-frame indels in each IrxAa gene and used for further analysis. For the IrxAb cluster, 54% of injected and screened animals had single germline mutations and 46% had mutations in both genes; three out of four double mutants had out-of-frame indels in both genes, with one allele used for further analysis. Our findings reveal that multiple genes within a cluster can be simultaneously targeted at reasonably high efficiency, without large deletions, in a single generation.

Animals with homozygous mutations in IrxAa or IrxBa clusters failed to inflate swim bladders and did not survive beyond 7 days post-fertilization (dpf). For IrxBa^−/− mutants, this reflects a role of irx3a and/or irx6a in survival as single irx5a mutants survive to adulthood (Askary et al., 2015). Homozygous IrxAb or IrxBb cluster mutants developed normally and were adult viable. To resolve comprehensively the diverse requirements of Irx genes during embryonic development, we performed in-crosses of quintuple heterozygous zebrafish (IrxAa^+/−; IrxAb^+/−; IrxBa^+/−; IrxBb^+/− ; irx7^+/−). In total, we collected 2081 embryos from 13 clutches. Animals that failed to inflate their swim bladders by 4 dpf were collected, visually inspected for abnormalities, and processed for staining of cartilage (Alcian Blue) and bone (Alizarin Red). To determine the genetic drivers for gross and skeletal phenotypes, ∼500 embryos were categorized by phenotype and genotyped. A subset of embryos developed dorsal curvature of the body axis between 3 and 5 dpf, which correlated with abnormal circling behavior at 5 dpf. These phenotypes were driven primarily by IrxAa loss, although IrxAa^+/−; IrxAb^−/− mutants often displayed a similar abnormal circling behavior as adults (Fig. 1B,C, Fig. S1A, Movie 1). Circling behavior seems unlikely to be due to vestibular defects as semicircular canals were unaffected in IrxAa^−/−; IrxAb^−/− mutants at 6 dpf (Fig. S1B). IrxAa^−/−; IrxAb^−/− mutants also displayed an abnormal jaw gape (Fig. S1C, Movie 2). As we failed to detect any abnormalities in jaw cartilages, joint structures, muscles and isl1:GFP⁺ motor nerves innervating these muscles at 6 dpf (Fig. S1D-F), abnormal jaw gape might instead reflect the known role of Irx genes in patterning the central nervous system (Cavodeassi et al., 2001). Progeny of quintuple heterozygous in-crosses also developed cardiac edema (Fig. 1B), which was attributable to mutations in IrxAa, IrxBa and irx7 (Fig. 1B,C), consistent with published roles for IrxA and IrxB cluster genes in heart development (Kim et al., 2012).

Irx genes play crucial roles in limb development (Grotewold and Ruther, 2002), and the fish pectoral fin is homologous to the mammalian forelimb. We observed Irx mutants with severe reductions of the scapulocoracoid cartilage, a major component of the proximal pectoral fin (Fig. 2A). The postcoracoid process, the posterior element in the proximal fin, and the endochondral disc of the distal fin were relatively unaffected. This phenotype resembles loss of the scapula, the proximal-most portion of the forelimb, in mice lacking IrxB genes (Irx3, Irx5) (Li et al., 2014). Pectoral fin defects were solely attributable to IrxBa loss (Fig. 2B). This conservation of IrxB gene function further supports proximal-distal identity being an ancestral feature of the fish pectoral fin that pre-dates evolution of the tetrapod limb (Gehrke et al., 2015) (Fig. 2C).

In addition to fin defects, we identified mutant zebrafish with reduced or absent facial cartilages. Three major classes emerged (Fig. 3). The first lacked the anterior facial skeleton, including the Meckel's and palatoquadrate cartilages, which was driven by IrxAb and irx7 loss. The second lacked the posterior ceratobranchial cartilages, attributable to IrxBa and irx7 loss, with enhanced severity induced by IrxBb loss (Fig. 3A,B, Fig. S2A). The final group was characterized by a near-complete loss of facial cartilages, including a reduced/absent neurocranium (Fig. 3A,B, Fig. S2B,C), which was driven by mutations in IrxAb, IrxBa, irx7 and, to a modest degree, IrxBb; 14/15 zebrafish with 9/10 Irx cluster alleles deleted had widespread loss of facial cartilages. We failed to recover any embryos at 4 dpf completely lacking all 11 Irx genes (IrxAa^−/−; IrxAb^−/−; IrxBa^−/−; IrxBb^−/−; irx7^−/−). This could reflect the previously reported expression of irx7 during early gastrulation (Satija et al., 2015), and hence early lethality, or alternatively failure to screen enough progeny from all-heterozygous in-crosses (0/2081 analyzed versus expected frequency of 1/1024, P=0.07, one-tailed Student's t-test).

The jaw cartilages derive from the most anterior pharyngeal arch, the mandibular, and the ceratobranchial cartilages from the five most posterior arches, the branchials. To understand better the facial cartilage losses in Irx mutants, we assessed earlier pharyngeal arch development (Fig. 3C). In IrxAb^−/−; irx7^−/− mutants, we observed a near-complete loss of sox10:dsRed⁺ CNCCs of the mandibular arch, a reduction in CNCCs of the hyoid (second) arch, but little change in branchial CNCCs at 36 hours post-fertilization (hpf). Re-imaging of the same individuals at 4 dpf confirmed that these early arch patterning defects precede the specific loss of jaw cartilages. Reciprocally, IrxBa^−/−; IrxBb^−/− mutants displayed reduced branchial arches 3-7 yet relatively normal mandibular and hyoid arch CNCCs at 36 hpf, which correlated with severe loss of ceratobranchial cartilages when re-imaged at 4 dpf. Irx mutants with 9/10 alleles deleted displayed a near-complete loss of sox10:dsRed⁺ CNCCs in the mandibular arch, a poorly formed hyoid arch, and reduced arches 3-7, which correlated with large-scale loss of facial cartilages, with the exception of the ceratohyal, at 4 dpf. These findings reveal region-specific roles for IrxA and IrxB genes in the formation of the anterior and posterior pharyngeal arches, respectively.

In order to determine the origins of pharyngeal arch defects in Irx mutants, we examined a time course of CNCC development. In wild-type embryos at 11 hpf, expression of sox10 labels premigratory neural crest cells at the neural plate border. In IrxAb^−/−; irx7^−/− mutants, we observed a near-complete loss of sox10 expression in the presumptive neural crest domain at this stage (Fig. 4A). In wild-type embryos at 16.5 hpf, expression of dlx2a labels the neural crest-derived ectomesenchyme precursors of the craniofacial skeleton that have begun to migrate toward the pharyngeal arches in three streams. In IrxAb^−/−; irx7^−/− mutants, dlx2a expression, particularly in the first (mandibular) stream, was reduced (Fig. 4B). A similar and preferential loss of mandibular CNCCs in IrxAb^−/−; irx7^−/− mutants was seen with the pan-neural crest transgene sox10:dsRed at 16.5 hpf, showing a requirement for IrxAb and irx7 for the generation of CNCCs and not the ectomesenchyme identity of mandibular arch CNCCs (Fig. 4C, Fig. S3). In contrast, no notable differences in dlx2a expression or the number of sox10:dsRed⁺ cells were apparent in IrxBa^−/− mutants (which display reduced gill cartilages) at 16.5 hpf. By 24 hpf, quantification of sox10:dsRed⁺ cells revealed a 50% reduction in the number of mandibular arch CNCCs in IrxAb^−/−; irx7^−/− mutants, with no statistical differences in hyoid and branchial neural crest-derived cells (Fig. 4D,F). We also failed to detect any changes in phospho-histoneH3⁺ proliferative or Caspase-3⁺ apoptotic arch CNCCs in IrxAb^−/−; irx7^−/− or IrxBa^−/− mutants at 24 hpf (Fig. 4D,G,H). Consistent with an early role for IrxAb genes and irx7 for the formation of CNCCs, we observed irx1b and irx7 expression within sox10⁺ premigratory neural crest cells at 11 hpf, as well as stronger irx1b expression in the pre-placodal region (Fig. 4E). Our analysis points to roles for IrxAb and irx7 genes, but not IrxBa genes, in the timely specification of CNCCs, with defects in this process resulting in the later preferential loss of mandibular arch CNCCs and the anterior facial skeleton.

As we found that the loss of gill cartilages in IrxBa^−/− mutants was not the consequence of defects in the early specification, migration, proliferation or survival of posterior arch CNCCs, we next examined development of the pharyngeal pouches. Although formation of endodermal pouches is largely independent of CNCCs (Cox et al., 2012; Veitch et al., 1999), loss of pouches is known to secondarily disrupt the survival of posterior arch CNCCs and the formation of gill cartilages (David et al., 2002; Piotrowski et al., 2003). At 16.5 hpf, a stage prior to the onset of pouch morphogenesis, we observed no changes in her5:GFP⁺ endoderm and sox10:dsRed⁺ CNCCs of IrxBa^−/−; IrxBb^+/− mutants versus wild-type siblings (Fig. 5A). By 24 hpf, however, when wild types have developed three pouches, we observed only a rudimentary first and second pouch in IrxBa^−/−; IrxBb^+/− mutants, and this correlated with a failure to properly segment the posterior-most neural crest-derived stream (Fig. 5B). These phenotypes became more apparent by 36 hpf, with six pouches in wild types and only rudimentary first and second pouches in mutants, which correlated with a reduction in and failure to segment the posterior arch CNCCs (Fig. 5C). In addition, the lack of gill cartilage defects in irx5a single mutants (Askary et al., 2015) indicates roles for irx3a and/or irx6a in pouch formation. Consistent with such roles, we observed expression of irx3a in the posterior arches at 24 hpf, both within dlx2a⁺ CNCCs and dlx2a⁻ cells that could be endoderm and/or mesoderm (Fig. 5D). Although further studies are needed to determine in which tissues IrxBa genes are required, our data reveal roles for IrxBa genes in pouch formation that likely contribute to the selective loss of gill cartilages.

In addition to cartilage loss, we also uncovered Irx mutant combinations displaying widespread fusions of facial cartilages, both across joints and between neighboring cartilages (Fig. 6A-D). Consistent with our previous data that loss of the IrxBa cluster gene irx5a enhanced hyoid joint defects in irx7 mutants (Askary et al., 2015), loss of irx7, and to a lesser extent the IrxBa cluster, resulted in fusions across the hyoid joint and truncation of the adjacent symplectic cartilage (Fig. 6E,K). We also observed fusions across the jaw joint (Fig. 6F,L; attributable to IrxAb and irx7 mutations) and across the midline joint between the two lower jaw Meckel's cartilages (i.e. mandibular symphysis; Fig. 6G,M; attributable to IrxBa mutations). In more severe cases of Meckel's-Meckel's midline fusions, the tips of the lower jaw cartilages project posteriorly (Fig. S4A,B; attributable to IrxBa and irx7 loss). In addition to joint fusions, we also observed fusions of cartilages derived from neighboring arches, including Meckel's and ceratohyal (Fig. 6H,N; attributable to IrxAb and irx7 loss, and to a lesser extent IrxBa and IrxBb mutations), palatoquadrate and hyomandibula (Fig. 6I,O; attributable to irx7 mutations), ceratohyal and ceratobranchial 1 (Fig. 6J,P; attributable to irx7, and to a lesser extent IrxAa and IrxAb mutations), Meckel's and basibranchial (Fig. S4C-E; attributable to IrxBa, IrxBb and irx7 mutations), and, rarely, palatoquadrate and symplectic (Fig. S4F; genotypes not defined). In the most severe cases, we observed fusions of cartilages of the first three arches (Meckel's, palatoquadrate, ceratohyal, hyosympletic and the first ceratobranchial) into a single mass (Fig. 6B).

In order to understand the etiology of cartilage fusions, we examined earlier pharyngeal arch development in IrxBa^+/−; IrxBb^−/−; irx7^−/− mutants; these animals lack the confounding cardiac edema associated with non-specific reductions of the head skeleton observed in IrxBa homozygous mutants. Of three IrxBa^+/−; IrxBb^−/−; irx7^−/− mutants displaying normal pouches separating sox10:dsRed⁺ arch CNCCs at 36 hpf, re-imaging at 4 dpf confirmed that all individuals developed Meckel's-ceratohyal fusions (Fig. S4G). Serial live imaging of IrxBa^+/−; IrxBb^−/−; irx7^−/− mutants with Meckel's-ceratohyal fusions revealed that 8/8 displayed earlier ectopic outgrowths of sox10:dsRed⁺ cells from the ceratohyal cartilage at 60 hpf, with these outgrowths fusing with Meckel's cartilage between 72 and 96 hpf in all cases examined (Fig. 7A). Examination of an irx7:GFP knock-in reporter line revealed expression in the ceratohyal perichondrium where outgrowths and eventual fusions to Meckel's cartilage occur (Fig. S5). We also performed imaging of IrxBa; IrxBb; irx7 mutants double-positive for trps1:GFP, which is expressed at higher levels in the perichondrium versus cartilage at 5 dpf, and the cartilage marker sox10:dsRed. Among the seven mutants identified with ectopic cartilage along the ceratohyal, five displayed a severe Meckel's-ceratohyal fusion, one a mild fusion, and one an unfused outgrowth from the ceratohyal in the same region in which fusions to Meckel's typically occur in mutants (Fig. 7B). On the side of the ceratohyal that typically fuses to Meckel's, we observed a mixture of trps1:GFP^high; sox10:dsRed⁺ and trps1:GFP^low; sox10:dsRed⁺ cells in the mutant perichondrium, with many having a flattened appearance typical of wild-type trps1:GFP^high; sox10:dsRed⁻ perichondral cells. In the mild fusion example, cells at the fusion site were trps1:GFP^high; sox10:dsRed⁺. Our findings suggest that, in addition to preventing inappropriate cartilage maturation in the hyoid joint (Askary et al., 2015), Irx7 and IrxB family members also function to prevent inappropriate chondrogenesis in the perichondrium, although we cannot rule out the possibility that proliferative expansion of chondrocytes alternatively or also contributes to mutant fusions (Fig. 7C).

Here, we uncover diverse roles for Irx proteins in the development of the facial and fin skeletons of zebrafish. The common loss of the most proximal skeletal element in the IrxBa mutant pectoral fin (scapulocoracoid) and the Irx3/Irx5 mutant forelimb (scapula) indicate common requirements for IrxB genes in these homologous appendages. Interestingly, IrxA and IrxB clusters appear to play largely distinct roles in facial development. We find that loss of jaw cartilages in IrxAb^−/−; irx7^−/− mutants is due to preferential defects in the specification of the anterior-most skeletogenic CNCCs. In contrast, IrxB genes play a role in formation of the endodermal pouches required for gill cartilage formation.

The role of IrxA genes in zebrafish neural crest formation appear to be distinct from what has been reported in mouse and frog. In Fused toes mice with homozygous deletion of the IrxB cluster, the pharyngeal arches are hypoplastic at embryonic day 10.5, and this correlates with a reduction in expression of the neural crest marker Sox10 at earlier stages (Anselme et al., 2007). In Xenopus laevis, morpholino knockdown of the IrxB gene Irx5 was suggested to affect the migration rather than specification of CNCCs, although, as in zebrafish IrxAb^−/−; irx7^−/− mutants, neural crest defects were more pronounced in the anterior head (Bonnard et al., 2012). It will be interesting to investigate whether Irx proteins promote mesenchyme development through a similar mechanism in the face as reported for Irx3 and Irx5 in appendages, for example by interacting with Shh signaling (Li et al., 2014) and controlling chromatid segregation (Tao et al., 2020).

Our findings suggest that it is the expression domains rather than the protein functions of individual Irx clusters that account for their regional requirements in the skeleton, and that reductions in Irx genes impact facial skeleton formation by affecting the earlier development of both the skeletogenic ectomesenchyme and the pharyngeal pouches. Although we lacked the resources to analyze contributions of individual members of each Irx cluster, we note that the larval lethality and pouch defects observed in IrxBa^−/− mutants were not seen when only irx5a was mutated (Askary et al., 2015), indicating roles for the two other members of the IrxBa cluster (irx3a and irx6a) in these processes. Hence, the general co-regulation of genes within an Irx cluster, combined with our combinatorial analysis of all Irx cluster deletions, would support genetic redundancy both within and across Irx clusters for patterning the vertebrate skeleton.

By analyzing all possible combinations of Irx cluster loss, we were also able to uncover mutant animals that developed later fusions of cartilages as opposed to earlier losses of CNCCs and pouches. These fusions included those between neighboring facial cartilages, reminiscent of the fusions of the toe bones seen in the Fused toes mouse mutant. Whereas toe fusion had been proposed to occur through a loss of programmed cell death between digits (van der Hoeven et al., 1994), in vivo imaging in zebrafish revealed that fusion between the lower jaw Meckel's and the ceratohyal cartilages, which form from distinct pharyngeal arches, occurs through ectopic chondrogenesis. A requirement for irx7 mutations for this type of fusion is consistent with expression of irx7:GFP in the region of the ceratohyal perichondrium undergoing inappropriate chondrogenesis in mutants, and more generally with the expression of multiple Irx genes in the perichondrium and joint-forming regions of the avian and murine autopod (McDonald et al., 2010). Previous data had shown that Irx5a and Irx7 ensure formation of the hyoid joint by suppressing the ability of Sox9 to activate target genes and promote cartilage maturation (Askary et al., 2015). It will therefore be interesting to explore whether Irx proteins act in a similar manner to prevent cartilage formation in the perichondrium. In addition to loss of the hyoid joint and fusion of Meckel's and ceratohyal cartilages, we observed numerous other joint losses and cartilage fusions in Irx mutant combinations. It remains unclear which particular Irx members are expressed in these other zones of cartilage fusions. Moreover, it is possible that some of these other fusions are indirect consequences of earlier defects, such as reduced CNCCs. Indeed, reductions of mandibular arch CNCCs in zebrafish h3f3a mutants was also found to correlate with loss of the jaw joint in some cases (Cox et al., 2012).

A particular advantage of the zebrafish system is the higher complexity of Irx clusters, thus allowing us to obtain and study genotypic combinations with severe early defects, as well as milder combinations with later skeletal fusions. We find that co-injection of Cas9 and guide RNAs targeting multiple genes can result in independent mutations in up to three linked genes in a single injection, thus demonstrating feasibility of gene editing of complex gene loci in zebrafish. The comprehensive series of Irx cluster mutations we have created should also be a useful resource for disentangling requirements of Irx genes in a diversity of other organ systems.

All experiments on zebrafish (Danio rerio) were approved by the Institutional Animal Care and Use Committee at the University of Southern California (Protocol 20771). Previously published lines include Tg(sox10:dsRed)^el10Tg, Tg(trps1:GFP)^j127aGt (Talbot et al., 2010), SAGp11A (irx7:GFP), irx7^el540 and irx5a^el576 (Askary et al., 2015), Tg(∼3.4her5:EGFP)^ne1911 (Tallafuss and Bally-Cuif, 2003), and Tg(isl1:EGFPCAAX)^fh474 (Barsh et al., 2017). We generated mutant lines using CRISPR/Cas9 editing as previously published (Hwang et al., 2013). We designed sgRNAs within or upstream of the homeobox domain (Fig. 1A, Table S1). The XbaI-digested pT3TS-nCas9n plasmid (Addgene plasmid #46757, deposited by Wenbiao Chen) was used as a template to transcribe Cas9 mRNA with the T3 mMESSAGE kit (Invitrogen). We transcribed sgRNAs from PCR-generated templates using the MEGAscript T7 Kit (Thermo Fisher Scientific). One-cell-stage embryos were injected with a mix of 100 ng/µl Cas9 mRNA and 50 ng/µl of each gene-specific sgRNA. Up to three genes were targeted simultaneously. Injected embryos were raised and founders identified by genotyping progeny for deletions and/or insertions using PCR and restriction digestion (see Table S1 for genotyping assays).

To generate IrxAa alleles, 19 injected animals were screened and 18/19 (95%) P0 s carried germline mutations, with an average of 0.95 independent alleles for each P0. Of these, 48% carried single-mutant alleles, 33% carried double-mutant alleles, and 19% carried triple-mutant alleles. Of the triple-mutant alleles, 1/10 (10%) contained out-of-frame alleles for each gene and was used for further phenotypic analysis (irx4a^el832 irx2a^el831 irx1a^el830). The final founder frequency to create the IrxAa allele was 1/19 (5.3%).

To generate IrxAb alleles, four injected animals were screened and 4/4 (100%) P0 s carried germline mutations, with an average of 1.625 independent alleles for each P0. Of these, 54% carried single-mutant alleles and 46% carried double-mutant alleles. Of the double-mutant alleles, 3/4 (75%) contained out-of-frame alleles for each gene and one was selected for further phenotypic analysis (irx4b^el834 irx1b^el833). The final founder frequency to create the IrxAa alleles was 3/4 (75%).

To generate IrxBa alleles, irx3a and irx6a guides were injected into zebrafish heterozygous for the irx5a el576 mutant allele. Twenty-three injected animals were screened and 18/23 (78%) P0s carried germline mutations for irx3a and/or irx6a, with an average of 0.36 independent alleles for each P0. Of these, 52% carried single-mutant alleles in the absence of irx5a, 27% carried single-mutant alleles in the presence of irx5a, 9% carried double-mutant alleles in the absence of irx5a, and 12% carried triple-mutant alleles. Of the triple-mutant alleles, 1/4 (25%) contained out-of-frame alleles for each gene and was used for further phenotypic analysis (irx6a^el836 irx5a^el576 irx3a^el835). The final founder frequency to create the IrxBa allele was 1/23 (4.3%).

To generate IrxBb alleles, five injected animals were screened and 5/5 (100%) P0s carried germline mutations, with an average of one independent allele for each P0. Of these, 100% carried single-mutant alleles and 0% carried double-mutant alleles. To generate double-mutant alleles, progeny from the irx5b^el722 founder carrying an out-of-frame deletion was injected with guides targeting irx3b. One P0 was screened to identify a double-mutant allele (100%) containing out-of-frame alleles for each gene (irx5b^el722 irx3b^el837).

Quintuple heterozygous zebrafish were genotyped every generation for every mutation within each cluster to ensure that recombination had not taken place between loci. Across multiple generations, we did not detect any recombination events.

Alcian Blue and Alizarin Red staining of cartilage and bone was performed as previously described (Walker and Kimmel, 2007). Briefly, larvae were fixed for 1 h in 2% paraformaldehyde (PFA) in PBS at room temperature, washed, and stained overnight in an Alcian Blue solution. Larvae were then rehydrated, bleached with 3% hydrogen peroxide, washed, and stained with an Alizarin Red solution for 30 min. Stained embryos were washed and transferred to a 50% glycerol solution for long-term storage. For muscle staining, larvae were fixed overnight in 4% PFA, washed in PBS, permeabilized overnight in 1% Triton X-100 in PBS, stained with Alexa Fluor 674 phalloidin (Molecular Probes, A22287) at 1:100 overnight, and washed in PBS before imaging.

Whole-mount larval skeletons were imaged with a Leica S8APO microscope, and dissected flat-mount skeletons were imaged with a Leica DM2500 microscope. Fluorescent embryos were imaged on a Zeiss LSM800 confocal microscope. All fluorescence images are maximal intensity projections processed with ImageJ. Brightfield images and fluorescence images were further modified for size and level with Adobe Photoshop CS6. To image zebrafish from pharyngula to larval stages, zebrafish carrying transgenic lines were selected between 22 and 24 hpf and individually housed in 24-well plates. Sorted zebrafish were moved to a 22°C incubator at 24 hpf to slow development, and each animal was repeatedly imaged at 36 hpf, 60 hpf, 3 dpf and 4 dpf. To visualize the jaw gape phenotype, 6 dpf zebrafish were embedded in 2% methylcellulose under minimal tricaine dosage to slow zebrafish but enable jaw movement. Movies were taken with a Zeiss Stemi 305 microscope using Zeiss Labscope software. To score dorsal curvature and swimming defects, individually housed animals were imaged at 3 dpf and movies were taken at 4 and 5 dpf using an Apple iPhone 10. Videos were trimmed and annotated using iMovie (Apple).

Probes for sox10 (Dutton et al., 2008), dlx2a (Akimenko et al., 1994), and irx1b and irx7 (Askary et al, 2015) were previously published. The probe for irx3a was amplified from zebrafish cDNA using Phusion High-Fidelity DNA Polymerase (New England BioLabs) with primers GCCATTGTACACAGGACATGC (irx3a-F) and AGACTACTTGGCCTGACAGC (irx3a-R). The PCR product was cloned into pCR Blunt II Topo (Thermo Fisher Scientific), and RNA probe was synthesized from linearized plasmid using Sp6 polymerase and dinitrophenol (DNP)-labeled nucleotides (Roche). Colorimetric and fluorescence in situ hybridizations were performed as previously described (Zuniga et al., 2010). Whole-mount immunofluorescence was performed with a mouse anti-phospho-Histone H3 (Ser10) antibody (Millipore Sigma, 05-1336) and a rabbit anti-Active Caspase-3 antibody (BD Pharmingen, 559565). Briefly, dechorionated sox10:dsRed⁺ embryos were sorted, fixed overnight in 2% PFA in PBS, and washed in PBS. Embryos were permeabilized with ice-cold acetone for 7 min, blocked in 2% goat serum in PBDTx (PBS with 1% bovine serum albumin, 1% DMSO, 0.1% Triton X-100) for 4 h, and incubated overnight in primary antibody at 4°C. Embryos were washed several times with PBDTx and stained with secondary antibodies Alexa Fluor 488 goat anti-rabbit IgG (Thermo Fisher, A-11008) and Alexa Fluor 674 goat anti-mouse IgG1 (Thermo Fisher, A-21240) at 1:300 overnight and washed in PBS. Colorimetric in situ hybridizations were imaged on a Leica S8APO microscope, and fluorescence-stained embryos were imaged on a Zeiss LSM800 confocal microscope.

Except where indicated, all phenotypes were scored blindly from quintuple heterozygous in-crosses, followed by genotypic analysis. To analyze the distribution of genotypes within a given phenotype, Chi-squared analysis was performed using Excel. After Bonferroni correction, significance was defined as P<0.01 based on P<0.05 for each of the five genotype categories. Using GraphPad's Prism, an unpaired t-test was performed to determine whether there is a genetic interaction between IrxBa and IrxBb for ceratobranchial cartilage loss, and P<0.05 was assigned as significant. In all figures, exact P-values are listed given the arbitrary nature of the 0.05 cut-off value. The number of sox10:dsRed⁺, pHH3⁺ and Caspase-3⁺ cells within the first, second and posterior pharyngeal arches were manually counted using ImageJ. Control and mutant cell counts were compared using a two-tailed non-parametric Student's t-test.

We thank Megan Matsutani and Jennifer DeKoeyer Crump for fish care, and Cecilia Moens for sharing the Tg(isl1:eGFPCAAX)^fh474 zebrafish line.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.197244