An atlas of seven zebrafish hox cluster mutants provides insights into sub/neofunctionalization of vertebrate Hox clusters

In bilaterian animals, Hox genes encoding the homeodomain transcription factors specify the positional identity of cells along the body axis. As one of the characteristic features, Hox genes are arranged in a cluster (Hox cluster) in most cases. The clustering of Hox genes in the chromosome has functional relevance. The positional order of Hox genes in each cluster corresponds to spatially-organized expression with a distinct anterior limit of the expression domain. Through this combinatorial collinear expression, Hox genes instruct the development of distinct morphology dependent on the positional information along the body axis.

In the early lineage of vertebrate evolution, the Hox cluster underwent drastic changes in terms of quality and quantity (Duboule, 2007). Two rounds of whole-genome duplication (WGD) quadrupled a single ancestral Hox cluster (Dehal and Boore, 2005). Subsequently in the teleost lineage, Hox clusters were further duplicated by additional WGD. At present, zebrafish possess seven hox clusters, following the loss of the hoxdb cluster (Amores et al., 1998; Woltering and Durston, 2006), and tetrapods such as mice and humans retain four Hox clusters. Concomitant with the amplification of the Hox cluster number, the structure of vertebrate Hox clusters has been drastically changed to an organized and constrained structure (Duboule, 2007). For example, non-Hox genes are absent within the vertebrate Hox clusters, and the transcription of all Hox genes in each cluster is oriented from the same strand (Darbellay et al., 2019). The genomic size of the vertebrate Hox clusters is highly compacted compared with that of the invertebrate counterparts, partially by the elimination of repetitive DNA sequences. It has been proposed that these drastic changes in vertebrate Hox clusters contribute to the rapid vertebrate-specific morphological innovation (Duboule, 2007; Wagner et al., 2003).

Genetic studies using knockout mice have significantly expanded our understanding of the function of Hox genes in the vertebrate body plan. Several lines of evidence indicate that Hox genes play important roles in the specification of positional identity in vertebrate-specific morphologies such as the vertebral column (Fromental-Ramain et al., 1996; Horan et al., 1995; McIntyre et al., 2007; Wellik and Capecchi, 2003), limbs (Kmita et al., 2005; Zakany and Duboule, 2007) and internal organs such as the kidney (Wellik et al., 2002). However, it remains to be elucidated whether the roles of Hox genes that have been revealed by a number of studies in mice are conserved among vertebrates. From an evolutionary point of view, fish are distantly related to mice among vertebrates. Thus, zebrafish offer a plausible opportunity to elucidate the functional comparisons of vertebrate Hox genes. Indeed, functional conservation of homologous HoxA/D genes has been shown in the development of forelimbs in mice and the homologous pectoral fins in zebrafish (Kmita et al., 2005; Nakamura et al., 2016). On the other hand, functional discrepancies of Hox genes between mice and zebrafish such as hindbrain patterning and pharyngeal arch formation have been pointed out (McClintock et al., 2001, 2002). Thus, whether the functional roles of Hox genes are conserved or different among vertebrates at the levels of whole Hox clusters remains to be determined. Functional studies have so far been restricted to a small number of hox genes among the 48 hox genes in zebrafish (Breau et al., 2013; Hunter and Prince, 2002; McClintock et al., 2001, 2002; Nakamura et al., 2016; Waxman et al., 2008; Weicksel et al., 2014; Ye and Kimelman, 2020). A large number of zebrafish hox genes makes comprehensive comparisons with homologous mouse Hox genes difficult. Instead, the existence of four well-characterized Hox cluster-deleted knockout mice (for the murine HoxB cluster, Hoxb1-Hoxb9 except Hoxb13 are deleted) (Kmita et al., 2005; Medina-Martinez et al., 2000; Spitz et al., 2001; Suemori and Noguchi, 2000) prompted us to compare the functional roles of homologous Hox clusters between mice and zebrafish by generating hox cluster mutants in zebrafish. In this study, by using the CRISPR-Cas9 system, we isolated all of the seven individual hox cluster deficiencies in zebrafish. We examined the phenotypes of the seven hox cluster mutants during embryogenesis such as hindbrain patterning, jaw cartilage formation, pectoral fins and lateral line system. For adult fish of the five viable hox cluster homozygous mutants, the whole-body skeletons and soft tissues were analyzed by X-ray computed tomography (CT) scans. Although functional conservations between homologous Hox clusters were present in some cases, functional discrepancies of homologous Hox clusters between mice and zebrafish were widely observed. As zebrafish and mice share common ancestors that possessed four Hox clusters, our results suggest different subfunctionalization and neofunctionalization of vertebrate Hox clusters after the second round of WGD.

Vertebrate Hox clusters generally do not contain non-Hox genes (Duboule, 2007). In the genome database, we confirmed that the seven hox clusters in zebrafish also do not possess non-hox genes (Fig. 1A). To elucidate the function of each hox cluster in zebrafish development, we tried to delete the corresponding genomic regions using the CRISPR-Cas9 system. We employed crRNA;tracrRNA duplex guideRNA (dgRNA), which was recently shown to be highly effective for introducing a genomic deletion in zebrafish (Hoshijima et al., 2019). For the generation of six hox cluster-deleted mutants (all of the clusters except for hoxcb), we used two dgRNAs that are targeted to both ends of hox genes in each cluster (Fig. 1B). As a representative, the isolation of hoxaa cluster mutants is shown in Fig. 1C-F. Two dgRNAs specific to the exons of hoxa1a and hoxa13a were injected with Cas9 protein into fertilized embryos (Fig. 1C). Among the F1 offspring of the injected embryos, we identified fish with deletion of the genomic region corresponding to the hoxaa cluster (Fig. 1D). We also verified that the introduced genomic deletion resulted in a frameshift mutation so that an aberrant fusion protein was not produced. Furthermore, we confirmed that the targeted genomic region of the hoxaa cluster was specifically deleted in homozygous mutants (Fig. 1E,F), as transposition of the targeted genomic region was sometimes detected in the course of the generation of mutants. Likewise, the other five hox cluster mutants (hoxab, hoxba, hoxbb, hoxca and hoxda) were isolated and confirmed (Figs S1-S4,S6). As we could not isolate the hoxcb cluster-deleted mutant owing to technical difficulties, we instead introduced frameshift mutations in all of the four hoxcb genes in the same allele (Fig. S5). Taken together, we successfully isolated all of the seven individual hox cluster deficiencies in zebrafish.

Using these seven zebrafish hox cluster deficiencies, we carried out phenotype analysis during embryogenesis, in which vertebrate Hox genes have been shown to play essential roles. Hox genes have been shown to play crucial roles in the patterning of the hindbrain in vertebrate embryos (Trainor and Krumlauf, 2000), and segmental expression patterns of zebrafish hox genes, which are located in different hox clusters such as hoxab, hoxba and hoxda, in the rhombomeres have been characterized in detail (Prince et al., 1998b). To examine the role of each hox cluster in hindbrain patterning, we performed immunostaining for reticulospinal interneurons, which exhibit rhombomere-specific cell body shapes and axonal projection patterns (Metcalfe et al., 1986). In wild-type embryos at 48 hours post fertilization (hpf), contralaterally projecting r4-specific Mauthner neurons were evident on both sides (asterisks in Fig. 2A). In contrast, in hoxba and hoxbb cluster mutants, Mauthner neurons were rarely detected, but instead smaller projecting neurons were observed (Fig. 2D,E). The hindbrain defects observed in hoxba and hoxbb cluster mutants show significant similarity to the individual loss-of-function phenotypes of zebrafish hoxb1a and hoxb1b, respectively (McClintock et al., 2002; Weicksel et al., 2014), suggesting that other hoxba genes (such as hoxb2a, hoxb3a and hoxb4a, which are segmentally expressed in rhombomeres; Prince et al., 1998b) or hoxbb genes other than hoxb1b do not significantly contribute to the hindbrain patterning in zebrafish. In addition, no apparent morphological abnormality in reticulospinal neurons was observed in other hox cluster deficiencies (Fig. 2B,C,F-H).

Cranial neural crest cells derived from the hindbrain migrate to form pharyngeal arches that give rise to most of the cartilage and bone in the developing head (Minoux and Rijli, 2010). In mice, Hox genes have been shown to establish the positional identity of cranial neural crest-derived cells in pharyngeal arches (Trainor and Krumlauf, 2001). Thus, we characterized the contributions of each hox cluster to the formation of craniofacial cartilage in zebrafish (Fig. 2I-P). At 5 days post fertilization (dpf), we found that abnormal morphology is evident only in hoxba cluster mutants. In hoxba cluster mutants, cartilages consisting of the basihyal and ceratohyal, which are derived from the second pharyngeal arch, were misshapen, and the connection between the basihyal and Meckel's cartilage was improper (Fig. 2L,L′). In addition, we observed that some hoxba cluster-deficient embryos at 7 dpf lacked the basihyal (Fig. S7). In zebrafish, knockdown of hoxb2a alone does not result in any defects in the pharyngeal arch, whereas simultaneous knockdown of hoxa2b and hoxb2a leads to major defects in the second pharyngeal arch (Hunter and Prince, 2002). As hoxba genes such as hoxb2a, hoxb3a, hoxb4a and hoxb5a are specifically expressed in the pharyngeal arches in zebrafish (Miller et al., 2004), abnormal morphology of pharyngeal arches in hoxba cluster mutants suggests that other hoxba genes cooperate with hoxb2a in the proper formation of pharyngeal arches. Despite severe defects in hindbrain patterning in hoxbb cluster mutants, craniofacial cartilage appeared to be unaffected (Fig. 2M,M′).

It is well known that Hox genes in HoxA and HoxD clusters regulate the proximal-distal patterning of developing limbs in mice (Kmita et al., 2005; Zakany and Duboule, 2007). The formation of the pectoral fin, which is a teleost homolog of the forelimb, is also regulated by the combinatorial expression of hox genes in zebrafish (Ahn and Ho, 2008; Nakamura et al., 2016; Sordino et al., 1995). To investigate the role of each hox cluster in the paired pectoral fin development, we examined the phenotype of homozygous embryos. In wild-type zebrafish, pectoral fins were present on both sides of larvae at 3 dpf (Fig. 3A). In contrast, hoxba cluster mutants apparently exhibited misshapen pectoral fins (Fig. 3D). At this stage, morphological abnormalities were not detectable in other hox cluster mutants including hoxaa, hoxab or hoxda cluster mutants (Fig. 3B,C,E-H). To examine the pectoral fin growth in more detail, we compared the lengths of the endoskeletal disc (cartilage) and fin-fold. Using cartilage-stained pectoral fins at 5 dpf, we found that the endoskeletal disc in hoxba cluster mutants was shortened, whereas those in the other hox cluster mutants appeared normal (Fig. 3A′-H′,Q,S). On the other hand, the length of the fin-fold along the proximal-distal axis was decreased in hoxaa and hoxab cluster mutants (Fig. 3A′-C′,R). The differences in phenotypes between hoxba and hoxaa/ab cluster mutants suggest their different functional roles in pectoral fin development. The other hox cluster mutants did not exhibit significant abnormal lengths of the fin-fold (Fig. 3R). Despite the specific expression of hoxc6a and hoxda genes in pectoral fin buds (Ahn and Ho, 2008; Nakamura et al., 2016; Neumann et al., 1999; Sordino et al., 1995), apparent abnormalities of pectoral fin development were not observed in hoxca or hoxda cluster mutants.

We also investigated the lateral lines that comprise a number of individual sensory organs, called neuromasts. In the posterior lateral line, collective migration of primordial cells from anterior to posterior results in the deposition of several groups of neuromasts at regular intervals on the surface of larvae (Ledent, 2002) (Fig. 3I). In zebrafish, hoxb8a was shown to control this cell migration (Breau et al., 2013). As in the case of hoxb8a knockdown embryos, neuromasts in hoxba cluster mutants were located more anteriorly than those in wild-type zebrafish (Fig. 3L,T). The number of neuromasts was not affected in hoxba cluster mutants (Fig. S8), suggesting retarded cell migration. Although several hox genes are expressed in the migrating lateral line primordial cells (Thisse et al., 2001), the deposition of posterior lateral line neuromasts in other hox cluster mutants appeared to be normal (Fig. 3J,K,M-P,T).

We proceeded to investigate the roles of each hox cluster in later development. By intercross between each hox cluster-deficient hemizygous fish, the offspring were raised to examine whether homozygous mutants are viable or lethal. Among intercrossed juvenile fish at ∼1-2 months of age, hoxab cluster-deleted and hoxba cluster-deleted homozygous mutants were not identified (Table 1), indicating that deletion of the hoxab or hoxba cluster causes lethality. The hoxba cluster mutants exhibited a severe expansion of the pericardial cavity in addition to the several defects described above. The hoxab cluster mutants showed a relatively mild abnormality, such as slightly short pectoral fins during embryogenesis; however, we could not identify viable hoxab cluster homozygous fish. In contrast, homozygous fish deficient in the other five hox clusters (hoxaa, hoxbb, hoxca, hoxcb and hoxda) survived to adulthood, although the survival rate of homozygous mutants was lower than that of siblings (Table 1).

Using these viable homozygous mutants, we performed X-ray micro-CT scan analysis to investigate the whole-body skeletons in detail. As is often observed in Hox knockout mice (Fromental-Ramain et al., 1996; Horan et al., 1995; McIntyre et al., 2007; Wellik and Capecchi, 2003), anterior homeotic transformation of the vertebral column was observed in zebrafish hox cluster mutant fish (Fig. 4; Movies 1-10). The Weberian apparatus, which consists of four ossicles to transmit the vibration of the swim bladder to the inner ear, is a specialized structure in the anteriormost region of the vertebral column in fish of the Ostariophysian, including zebrafish (Grande and Young, 2004; Sanger and McCune, 2002). In wild-type zebrafish, the tripus, which exhibits a fan-shaped bone of the Weberian apparatus, was located on both sides of the 3rd vertebra (Fig. 4A-C). In contrast, such a unique structure of the tripus was severely misshaped and turned out to be laterally-extending bone in hoxca cluster mutants (Fig. 4J-L). In addition, the altered bone in hoxca cluster mutants was attached to the anterior portion of the 3rd vertebra, whereas the tripus was attached to the center of the 3rd vertebra in wild-type zebrafish. These abnormal morphologies resemble the lateral process, which is laterally extending bone attached to the anterior portion of the 2nd vertebra, suggesting anterior homeotic transformation in the Weberian apparatus in hoxca^−/− fish. Besides, the os suspensorium, which is ventrally extending bone from the 4th vertebra, was found to be completely missing in hoxca cluster mutants, although the transverse process of vertebra 4 (tp4), which also extends lateral-ventrally from the 4th vertebra, was formed (Fig. 4J-L). Furthermore, hoxcb cluster mutants exhibited a slight morphological abnormality of shortening of the anterior portion of the tripus (asterisks in Fig. 4M,N). In hoxbb cluster mutants, the width of the tripus was slightly shortened, although the patterning of the tripus appeared normal (Fig. 4G-I). The other viable hox cluster mutants (hoxaa and hoxda) showed indistinguishable morphology of the Weberian apparatus in comparison with that of wild-type zebrafish (Fig. 4A-F,P-R).

In addition, anterior homeotic transformation was detected in the pleural-to-caudal transition regions of the vertebral column in hox cluster mutants (Fig. 4S-X). Although the identity of the vertebral column has been reported to vary in zebrafish (Bird and Mabee, 2003), our wild-type zebrafish generally possessed nine pairs of pleural ribs on the 5th-13th vertebrae and short ventrally extending bones from the 14th vertebra (Fig. 4S; Table S1). In contrast, hoxaa, hoxca and hoxda cluster mutants exhibited long pleural ribs from the 14th vertebra (Fig. 4T,V,X; Table S1), which show similarities to the 13th vertebra. In wild-type zebrafish, the ventrally extending bone (hemal arch) from the 15th vertebra reached the dorsal tip of the first radial attached to the anal fin (Fig. 4S; Table S1). However, in hoxca and hoxda cluster mutants, the 15th vertebra exhibited morphology similar to that of the 14th vertebra in wild-type zebrafish, and the 16th vertebra of the mutants resembled the 15th vertebra of wild-type zebrafish (Fig. 4V,X; Table S1). On the other hand, hoxbb and hoxcb cluster mutants did not show such anterior transformation (Fig. 4U,W; Table S1). These results provide the first genetic evidence that anterior transformation of the vertebral column indeed occurs in zebrafish hox cluster mutants.

We also found that the number of vertebrae was increased in hox cluster mutants. Our wild-type zebrafish usually possessed 31-33 vertebrae in total (Fig. 5A,O). However, hoxca cluster mutants exhibited an increased number of vertebrae, 34-37, although the other four viable hox cluster mutants did not exhibit noticeable differences (Fig. 5A-F,O). As the number of somites is linked to the number of vertebrae (Morin-Kensicki et al., 2002), we visualized the somite boundaries of all hox cluster mutants at 2 dpf (Fig. 5G-N). Consistent with the increased number of vertebrae, we found that the number of somite boundaries was increased in hoxca cluster mutants (Fig. 5L,P). An increase of the body length along the anterior-posterior axis in hoxca cluster mutants was not detected, mainly due to the slight increase of small somites. Among the other hox cluster mutants, hoxab cluster mutants also exhibited an increase of somite number (Fig. 5I,P). Although hoxab cluster mutants were not viable (Table 1), these results suggest that the hoxab cluster, as well as the hoxca cluster, contributes to the regulation of somite/vertebra number in zebrafish. Furthermore, we also confirmed that the position of the pelvic fins along the anterior-posterior axis was not significantly affected in the five viable hox cluster-deficient mutants (Fig. S9).

Finally, we examined the whole-body tissues of surviving hox cluster mutant fish using X-ray micro-CT scans (Fig. 6; Movies 11-20). In wild-type zebrafish, anterior and posterior swim bladder lobes are present in the abdomen of adult zebrafish (Parichy et al., 2009) (Fig. 6A). However, we found that hoxca cluster mutant fish possessed a single swim bladder lobe (Fig. 6D). This morphological abnormality was specifically evident in hoxca cluster mutants, and apparent morphological defects in the swim bladder and other tissues were not observed in other hox cluster mutants (Fig. 6A-F).

In this study, we isolated loss-of-function alleles for all of the seven hox clusters in zebrafish. Using hox cluster mutants, we performed not only embryonic phenotype analysis but also X-ray micro-CT scan analysis for the whole-body skeletons and tissues in viable homozygous fish. In mice, four individual Hox cluster-deleted mutants have been generated and characterized in detail (Kmita et al., 2005; Medina-Martinez et al., 2000; Minoux et al., 2009; Soshnikova et al., 2013; Spitz et al., 2001; Suemori and Noguchi, 2000; Vieux-Rochas et al., 2013). Hereafter, we discuss the phenotypes of zebrafish hox cluster mutants in comparison with those of mouse Hox cluster mutants.

The present study revealed that hoxba cluster mutants show the most severe defects during embryogenesis among the seven hox cluster mutants in zebrafish. Several defects were observed in hindbrain patterning, jaw formation, pectoral fin formation and positioning of posterior lateral lines, although another HoxB-derived hoxbb cluster is intact. These results suggest that the zebrafish hoxba cluster contains several hox genes which play crucial roles in the early embryogenesis. Besides, the zebrafish hoxba cluster is the only cluster that contains the paralogous group 7 (PG7), hoxb7a (Fig. 1A). Therefore, the deletion of the hoxba cluster results in complete lack of PG7 function, which cannot be compensated by other clusters. In contrast to zebrafish hoxba cluster mutants, Hoxb1-Hoxb9-deleted mutant mice exhibit only slight defects: anterior transformation along the cervical and thoracic vertebral column and some cranial nerve defects (Medina-Martinez et al., 2000). Although cranial nerve defects are commonly observed, morphological abnormalities in jaw and forelimbs have not been reported in Hoxb1-Hoxb9-deleted mutant mice (Medina-Martinez et al., 2000). In mice, morphological defects in the cranial facial skeleton are caused by deletion of the HoxA cluster (Minoux et al., 2009), although additional deletion of Hoxb1-Hoxb9 enhances the phenotype of HoxA cluster mutants (Vieux-Rochas et al., 2013). On the other hand, knockout of mouse Hoxb13 increases the number of tail vertebrae (Economides et al., 2003). However, in zebrafish hoxba and hoxbb cluster mutants, we did not detect any increase in the number of somite boundaries (Fig. 5P), which is associated with the number of vertebrae. It is unlikely that hoxba and hoxbb clusters redundantly control the number of somite boundaries because the hox13 paralogous gene is only present in the hoxba cluster (Fig. 1A). Instead, such an increase is rather prominent in other hox cluster mutants such as hoxca and hoxab clusters, which possess hoxc13a and hoxa13b, respectively (Fig. 5P). These results suggest that the responsible Hox genes are not always located in homologous Hox clusters between zebrafish and mice, although the function of responsible Hox genes is conserved, suggesting different subfunctionalization of Hox clusters in vertebrate evolution.

HoxC cluster mutant mice do not show any drastic embryonic abnormalities but show a series of anterior transformations of the vertebral column (Suemori and Noguchi, 2000). Similarly, in zebrafish, hoxca and hoxcb cluster mutants do not show severe embryonic defects, but hoxca cluster mutants also show the anterior transformation of vertebral column (Fig. 4V). In hoxca cluster mutants, unique defects were observed in teleost-specific structures such as the Weberian apparatus and swim bladder (Figs 4J-L and 6D). The hoxcb cluster mutants also exhibit weak defects in the Weberian apparatus (Fig. 4M-O). As the anterior expression boundaries of hoxc6a and hoxc6b correspond to the 5th somite, which gives rise to the anterior vertebrae in zebrafish (Morin-Kensicki et al., 2002; Prince et al., 1998a), we presume that hoxc6a and hoxc6b are plausible candidate genes responsible for the defects in formation of the Weberian apparatus. In addition, zebrafish hoxc4a, hoxc6a and hoxc8a were shown to be specifically expressed in the anterior portion of the primitive gut, which gives rise to swim bladder primordium (Zheng et al., 2011). Thus, the swim bladder defects in hoxca cluster mutants may be caused by the combined loss-of-function of these hoxca genes. These results suggest that HoxC-derived hoxca and hoxcb clusters acquire new functions to contribute to the development of teleost-specific structures in the teleost lineage. In addition, hoxca cluster mutants show an increased number of somites and axial vertebrae. Elongation of the posterior body has been shown to be regulated by vertebrate Hox genes (Denans et al., 2015). Our results suggest that hox genes in the hoxca cluster are preferentially responsible for this process in zebrafish.

In mice, deletion of the entire HoxD cluster results in malformations and fusions of anterior vertebrae and truncation of the limbs (Spitz et al., 2001; Zakany et al., 2001). In contrast, zebrafish hoxda cluster mutants did not show anterior transformation of the anteriormost vertebrae but exhibited anterior transformation in the pleural-caudal region. In addition, zebrafish hoxda cluster mutants did not show any malformation of pectoral fins during embryogenesis and even in adult fish. The absence of pectoral fin defects in zebrafish hoxda cluster mutants is not due to compensation by the HoxD-related hoxdb cluster, because zebrafish have lost the hoxdb cluster except for microRNA (Amores et al., 1998; Woltering and Durston, 2006). In mice, HoxA/D clusters redundantly function to pattern the distal forelimb (Kmita et al., 2005). In this study, we have shown that the pectoral fins in zebrafish HoxA/D-related hoxaa, hoxab and hoxda cluster mutants were not significantly affected at 3 dpf. However, the pectoral fins in individual hoxaa and hoxab cluster mutants at 5 dpf were shortened, whereas those in hoxda cluster mutants appeared to be normal (Fig. 3). During the early stage of pectoral fin development, hox9-13 genes in hoxab and hoxda clusters are overlappingly expressed in the fin bud, and relatively weak expression of hoxa9a/a11a is confined to the fin mesenchymal cells (Ahn and Ho, 2008). After the transition of the embryonic fin buds into larval pectoral fins (48-60 hpf), the expression of posterior hox genes in the hoxda cluster and hoxa9a/a11a is gradually weakened, whereas the expression of hoxa13a is upregulated in the fin-fold and the expression of posterior hox genes in the hoxab cluster persists (Ahn and Ho, 2008). The differences of phenotypes in the pectoral fin formation in individual hoxaa, hoxab and hoxda cluster mutants may be caused by differential expression patterns of zebrafish hoxaa, hoxab and hoxda genes. Taken together, our results suggest that the contribution of the zebrafish hoxda cluster to appendage formation is less than that of the mouse HoxD cluster and that the degree of the contribution of each Hox cluster has diversified after the quadruplication of Hox clusters.

Our study revealed phenotypic differences of homologous Hox cluster mutants between mice and zebrafish. As both are derived from common primitive vertebrates that are thought to possess quadrupled Hox clusters, our results suggest that Hox clusters in mice and zebrafish underwent discrete subfunctionalization and neofunctionalization in their respective lineages. Functional discrepancies of Hox genes between mice and zebrafish have been pointed out (Hunter and Prince, 2002; McClintock et al., 2002). In zebrafish, hoxb1a and hoxb1b play roles similar to those of mouse Hoxa1 and Hoxb1, respectively, in hindbrain patterning (McClintock et al., 2001, 2002). However, zebrafish hoxa1a, an ortholog of mouse Hoxa1, is unlikely to possess such a function because segmental expression of hoxa1a was not detected in the hindbrain as is observed in mouse Hoxa1 (McClintock et al., 2001). In another case, zebrafish hoxa2b and hoxb2a have a function similar to that of mouse Hoxa2 in second pharyngeal arch formation (Hunter and Prince, 2002). In contrast, it appears that mouse Hoxb2, an ortholog of zebrafish hoxb2a, does not play a major role in this process. In this study, by analyzing the phenotypes of seven individual hox cluster mutants, we also found that responsible Hox genes are not always located in homologous Hox clusters between mice and zebrafish, suggesting that functional discrepancies of Hox genes between mice and zebrafish are widely present. Changes in cis-regulatory elements have been recognized as an important molecular mechanism underlying phenotypic evolution. As was previously proposed to explain the functional discrepancies of paralog 1 genes in hindbrain patterning (McClintock et al., 2001, 2002; Prince, 2002), evolutionary changes of cis-regulatory modules of Hox genes may occur in the teleost lineage and change the responsible Hox genes between zebrafish and mice. Alternatively, it is possible that such an event also took place in the tetrapod lineage or both lineages. We could not identify which zebrafish hox gene(s) is responsible for the specific defects observed in hox cluster mutants. Thus, deciphering the zebrafish Hox code in comparison with the mouse Hox code will provide significant insights not only into the ancestral states of Hox gene function in common primitive vertebrates but also different sub/neofunctionalization of Hox genes between zebrafish and mice.

Finally, our collection of all of the zebrafish hox cluster mutants provides valuable genetic resources not only for the elucidation of hox gene functions in zebrafish development but also for comparative studies to evaluate the evolution of vertebrate Hox genes. As functional redundancy between teleost-specific duplicated hox clusters is expected, it would be interesting to compare the phenotypes of zebrafish double mutants of these clusters with those of homologous Hox cluster mutant mice with respect to the evaluation of differential contribution of Hox genes in patterning the appendicular skeleton and the vertebral column in fish and mice in a future study.

All of the experiments were performed using Riken WT (RW) fish, which were maintained at 27°C with a 14 h light/10 h dark cycle. Embryos were obtained from natural spawning and developmental stages were determined based on the hours or days post fertilization (hpf, dpf) or the morphological features as previously described (Kimmel et al., 1995). All of the experiments using live zebrafish were approved by the Committee for Animal Care and Use of Saitama University and conducted under the regulations of Saitama University.

Genomic information including the genomic structures of hox clusters is based on DNA sequences deposited in the zebrafish Ensembl database (GRCz11, release 91).

For the generation of hox cluster mutants except for the hoxcb cluster, the genomic region encompassing each hox cluster was deleted using the Alt-R CRISPR-Cas9 system (Integrated DNA Technologies). Briefly, two crRNAs, which are targeted to hox genes located at both ends of each cluster, were selected by using the web program CRISPR Design (http://CRISPR.mit.edu) or CHOPCHOP (http://chopchop.cbu.uib.no/index.php). The sequences of synthesized crRNAs used in this study are listed in Table S2. Target-specific crRNA was incubated with common tracrRNA to prepare the crRNA:tracrRNA duplex. Two crRNA:tracrRNA duplexes, which recognized both ends of each hox cluster, were incubated with Cas9 nuclease according to the manufacturer's protocol. Approximately 1 nl of crRNA:tracrRNA-Cas9 solution was injected into zebrafish fertilized embryos. After sexual maturation of the injected fish, candidate founder fish were mated with wild-type fish to generate hemizygous F1 offspring. F1 juvenile fish carrying the deletion of each hox cluster were identified by PCR using the genomic DNA derived from the partially dissected tailfin. Adult fish carrying hemizygous deficiency for each hox cluster are externally normal and fertile. By intercross between male and female fish possessing the same mutation, homozygous mutants were obtained. Using the genomic DNA extracted from homozygous mutant embryos, the deletion of each hox cluster was confirmed by PCR using primers specific to hox genes located within a cluster and flanking genomic regions (Table S4). As we observed translocation of the deleted hox cluster into another genomic region in some cases, the absence of the target locus was confirmed by performing PCR for hox genes in each cluster.

For generation of the hoxcb cluster mutant, frameshift mutations were introduced in all of the four hoxcb genes. Deletion of the hoxcb cluster by two gRNAs targeted at both ends of the hoxcb cluster was not accomplished due to poor efficiency. Frameshift mutations were first introduced in hoxc11b and hoxc12b using the Alt-R CRISPR-Cas9 system (Integrated DNA Technologies). Subsequently, frameshift mutations of hoxc6b and hoxc13b were introduced using the CRISPR-Cas9 system in the same allele of mutated hoxc11b and hoxc12b. The introduced mutations were confirmed by DNA sequencing.

The alleles of mutants described in this study are as follows: hoxaa cluster mutant, sud111; hoxab cluster mutant, sud112; hoxba cluster mutant, sud113; hoxbb cluster mutant, sud114; hoxca cluster mutant, sud115; hoxcb cluster mutant, sud124; and hoxda cluster mutant, sud116. Frozen sperms derived from all of the mutants have been deposited in the National BioResource Project Zebrafish in Japan (https://shigen.nig.ac.jp/zebra/).

Genotyping was carried out for the maintenance of mutant fish lines or the analysis of stained embryos. Genomic DNA was extracted from live or stained embryos or dissected tailfins using the NaOH method (Meeker et al., 2007) and was used as a template for PCR-based genotyping. For six hox cluster mutants excluding hoxcb, three primers were used to distinguish the genotype of each mutant. For hoxcb cluster mutants, the genotyping was determined using primers to amplify the genomic regions surrounding the mutations in hoxc6b, hoxc11b, hoxc12b or hoxc13b, respectively. The PCR conditions and the sequences of the primers are listed in Table S3.

Whole-mount in situ hybridization was performed as described previously (Thisse and Thisse, 2014). Stained embryos mounted in 80% glycerol were photographed under a stereomicroscope (Leica M205 FA) with a digital camera (Leica DFC350F). After taking the photos, the genotype from the stained embryos was determined by PCR as described above.

Neuromast hair cells in the zebrafish lateral line were stained by 4-Di-2Asp [4-(4-diethylaminostyryl)-N-methylpyridinium iodide; Sigma-Aldrich] as previously described (Sapede et al., 2002). Briefly, live larvae at 3 dpf were incubated in 100 µM of 4-Di-2Asp in 1/3 Ringer's solution for 10 min. After washing several times with 1/3 Ringer's solution, the anesthetized larvae were mounted in 2% methylcellulose and photographed under a fluorescent stereomicroscope (Leica M205 FA). Subsequently, genotyping of the stained larvae was carried out by PCR as described above. By using merged images of fluorescence and brightfield, the position of neuromasts on both sides was determined on the basis of the somite number. For neuromasts located posterior to the 30th somite, their position was regarded as >30 owing to the difficulty in distinguishing the small somites in the tail region.

Reticulospinal neurons in hindbrain at 48 hpf were stained by anti-neurofilament RMO-44 monoclonal antibody (Thermo Fisher Scientific, 13-0500, 1:50) as previously described (Banote et al., 2016). As a secondary antibody, goat anti-mouse IgG-Alexa 488-conjugated antibody (Abcam, ab150113, 1:500) was used. Subsequently, images of flat-mounted specimens were taken using a confocal microscope (Olympus FV1000).

For the staining of cartilage in jaws, 5 dpf larvae were fixed with 4% paraformaldehyde in PBS and stained by acid-conditioned Alcian Blue solution as previously described (Neuhauss et al., 1996). Before the staining, the posterior body of the larvae was manually dissected, and PCR-based genotyping was carried out using extracted genomic DNA. After the staining, the surrounding tissues of the jaws were digested in 0.25% trypsin-EDTA solution. The cartilage of zebrafish pectoral fins at 5 dpf was stained by acid-free Alcian Blue solution as previously described (Walker and Kimmel, 2007). After the staining, the pectoral fins were dissected. Images were obtained under a stereomicroscope (Leica M205 FA). The lengths of the endoskeletal disc and fin-fold were measured using ImageJ software.

Skeletal structures of zebrafish were observed using X-ray micro-CT scanning as previously described (Akama et al., 2020). Adult zebrafish were fixed with 4% paraformaldehyde in PBS at 4°C overnight and were transferred to 70% ethanol. Using an X-ray micro-CT (ScanXmate-E090S105; Comscantechno), the fixed specimens were scanned at a tube voltage peak of 85 kV and a tube current of 90 µA. For scanning of the Weberian apparatus, samples were rotated 360° in steps of 0.24-0.30°, generating 1200-1500 projection images of 992×992 pixels. These micro-CT data were reconstructed using coneCTexpress software (Comscantechno) at an isotropic resolution of 4.8-5.0 μm. For the whole scanning of a zebrafish with higher resolution, the entire body was scanned in four parts for each specimen. For each part, samples were rotated 360° in steps of 0.3°, generating 1200 projection images of 992×992 pixels. The micro-CT data of each part were reconstructed and stored as a dataset with an isotropic resolution of 10 µm. Finally, the data of four sites were combined to generate a dataset of the whole specimen. Three-dimensional image analysis was performed using OsiriX MD software (Pixmeo).

After observation of the skeletal structures, soft tissues in the specimens were analyzed using X-ray micro-CT scanning. The specimens stored in 70% ethanol were gradually hydrated in distilled water and then stained overnight with 20% Lugol's solution as previously described (Maeno et al., 2019; Metscher, 2009). For the whole scanning of a zebrafish with higher resolution, the entire body was scanned in four parts for one specimen. For each part, samples were rotated 360° in steps of 0.3°, generating 1200 projection images of 992×992 pixels. The micro-CT data of each part were reconstructed and stored as a dataset with an isotropic resolution of 9-11 µm. Subsequently, the data of four sites were combined and three-dimensional images were constructed using OsiriX MD software (Pixmeo). Finally, movies were edited using Adobe Premiere Pro (Adobe).

We thank Dr Bernard Thisse for providing the xirp2a/cb1045 plasmid.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.198325