A new mib allele with a chromosomal deletion covering foxc1a exhibits anterior somite specification defect

mibnn2002, found from an allele screen, showed early segmentation defect and severe cell death phenotypes, which are different from previously known mib mutants. Despite distinct morphological phenotypes, the typical mib molecular phenotypes: her4 down-regulation, neurogenic phenotype and cold sensitive dlc expression pattern, still remained. The linkage analysis also indicated that mibnn2002 is a new mib allele. Failure of specification in anterior 7-10 somites is likely due to lack of foxc1a expression in mibnn2002 homozygotes. Somites and somite markers gradually appeared after 7-10 somite stage, suggesting that foxc1a is only essential for the formation of anterior 7-10 somites. Apoptosis began around 16-somite stage with p53 up-regulation. To find the possible links of mib, foxc1a and apoptosis, transcriptome analysis was employed. About 140 genes, including wnt3a, foxc1a and mib, were not detected in the homozygotes. Overexpression of foxc1a mRNA in mibnn2002 homozygotes partially rescued the anterior somite specification. In the process of characterizing mibnn2002 mutation, we integrated the scaffolds containing mib locus into chromosome 2 (or linkage group 2, LG2) based on synteny comparison and transcriptome results. Genomic PCR analysis further supported the conclusion and showed that mibnn2002 has a chromosomal deletion with the size of about 9.6 Mbp.

Scientific RepoRts | 5:10673 | DOi: 10.1038/srep10673 membrane to induce downstream transcriptional activation, such as her4 that we used to show the Notch activation in this and previous studies 1 .
Ribs and vertebrae are originally derived from the paraxial mesoderm, where a group of cells locate in presomitic mesoderm (PSM) around the region of tail bud, propagate and be specified progressively into somites -the repeated epithelial segments. Somites give rise to dermatome, myotome and sclerotome, and further become dermis, skeletal muscle, ribs and vertebrae 16 . The segmentation defect has been found to correlate with the loss of cyclic expression of dlc and/or her1 in the PSM of different Notch-related zebrafish mutants, such as deltaC (bea), deltaD (aei), notch1a (des) and mib 1,[17][18][19][20][21][22] . In these mutants, dlc is desynchronizedly expressed with a salt-and-pepper-like, instead of cyclic, expression pattern, and the posterior somite boundaries are indistinct as those cells are unable to be synchronized by Notch signaling 20,[23][24][25] . On the other hand, there are genes required for somitogenesis in a Notch-independent manner. fss mutants with deficiency in tbx24 gene form no somites 26 . foxc1a, whose morphants exhibited a similar phenotype to fss mutants, acts in parallel to the fss/tbx24 pathway as an essential factor in zebrafish somitogenesis 27 .
To further dissect and explore the Mib functions, we carried out an allele screen to identify new mib alleles. Three new mib alleles, mib nn2000 , mib nn2001 and mib nn2002 , were identified. mib nn2000 and mib nn2001 , whose mutations lead to truncation of RF domains, showed segmentation phenotypes similar to previous mib alleles. Interestingly, mib nn2002 had an earlier segmentation defect with severe cell death. her4 and huc expression also suggested the loss of mib activity. We then focused on mib nn2002 mutant for its potential in identifying mib's novel function. The cell death was mainly distributed around the region of notochord and neural tube with an onset from 16-somite stage (ss) and an up-regulation of p53. While homozygous embryos lost anterior somite segmentation and failed to be specified into somite cell fate as shown by the expression of myoD, dlc and dystrophin, relatively normal-looking posterior somite segments formed with proper somite markers expressed gradually after 14 hours post fertilization (hpf). foxc1a expression was not detected at all stages. The expression of her1 and dlc were down-regulated but still cyclically expressed in the PSM during all developmental stages checked. By linkage analysis, we found a close linkage near the mib locus. Unexpectedly, we uncovered there is a large deletion covering the loci of foxc1a and mib on LG2.

Results
Three new mib alleles, showing similar Notch molecular phenotypes to other mib alleles, were found from allele screen. In order to further understand mib gene functions in the vertebrate development, we carried out an allele screen by non-complementation between F1 progeny of ENU-mutagenized founder fish and heterozygous carriers of mib ta52b , a well-known and widely-used mib allele 1,17 . 3849 potential F1 carriers for new mib alleles were successfully screened by single pair matings to fish that were mib ta52b heterozygous carriers. The resulting egg lays were checked around 48 hpf for non-complementation of the mib mutation. Three new alleles: mib nn2000 , mib nn2001 and mib nn2002 were identified.
The expression of a Notch downstream gene, her4, and a neuronal marker, huc, were used to indicate the compromise of Notch signaling pathway in mib mutants in previous studies 1,7 . Besides, the disruption of cyclically expressing dlc was also demonstrated in mib and other Notch-related mutants [19][20][21][22] . Therefore, the expression patterns of her4, huc and dlc were examined in these new alleles. At 22 °C, the typical down-regulation of her4 and up-regulation of huc were observed in mib nn2000 , mib nn2001 and mib nn2002 mutants ( Fig. 1A-H). Under the same condition, the dlc expression exhibited a salt-and-pepper (desynchronized) pattern in the anterior PSM and nascent somite region ( Fig.1I-L and supplementary Fig. 1). All the analysis suggested that mib nn2000 , mib nn2001 and mib nn2002 were new mib alleles that lost the E3 ubiquitin ligase activity 7 . mib nn2002 mutants showed distinct morphological phenotypes. Both mib nn2000 and mib nn2001 mutants showed typical mib segmentation defects at 22 °C: showing 11-14 and 14-17 visible somites, respectively, compared to about 20 somites (equivalent to 19 hpf at 28.5 °C) in wild-type at 30 hpf ( Fig. 2A-C). Besides, mib nn2000 and mib nn2001 exhibited morphological phenotypes similar to those of mib ta52b and mib m132 1,17 , including brain and pigmentation defects, curly tail and hemorrhage at 3 dpf ( Fig. 2D-F). Compared with the later boundary formation defect in mib nn2000 , mib nn2001 and previously-analyzed mib mutants, mib nn2002 showed somite segmentation defect at an earlier stage (Fig. 3). The V-shaped somite boundaries were not observed in mib nn2002 before 14 hpf (Fig. 3G). The somite-like structures could be observed in mib nn2002 mutants after 17 hpf, though the somites were not V-shaped and boundaries did not form sharply (Fig. 3H and Fig. 5F). Necrotic tissues appeared in mib nn2002 mutants after 24 hpf (Fig. 3I) and the phenotype became worse after 48 hpf with a great amount of cell death, edema and laterally curly tail with no observable pigmentation (Fig. 3J). Because of the distinct morphological phenotypes and similar Notch-related molecular phenotypes to other mib alleles, we then focused on analyzing mib nn2002 mutants as it might reveal a novel function of mib.
Noticeably, average 22% of embryos from incrosses of mib nn2002 carriers exhibited early segmentation phenotype (supplementary Table 1). The pre-selected embryos with early segmentation phenotype showed down-regulated her4, slightly up-regulated huc and severe cell death phenotypes (data not shown). Furthermore, about a quarter of transheterozygotes (mib nn2002 /mib ta52b ) exhibited typical mib   Table 2), such as late segmentation defect (supplementary Fig. 2), curly tail and less pigmentation (supplementary Fig. 3). The impact of maternal mib on the transheterozygotes also can be observed (supplementary Fig. 3, embryos laid from female mib ta52b carriers show a more severe mib phenotype, as observed previously 7 ). These results demonstrated that mib nn2002 mutation follows the patterns of autosomal recessive inheritance with high penetrance.
The cell death in mib nn2002 mutants began after 17 hpf by acridine orange staining and mainly distributed around the regions of neural tube and notochord (supplementary Fig. 4). It was demonstrated to be apoptosis by using TUNEL assay and the apoptotic cell number was significantly higher in mib nn2002 mutants than that in the control embryos (supplementary Fig. 5). The expression of p53 was further found to be highly up-regulated in mib nn2002 mutants (supplementary Fig. 6), which suggests that cell death might be related to genome instability.
Early somite specification defects were identified in mib nn2002 mutants. To discern if the segmentation defect is caused by a failure/delay in boundary formation or early somite specification, we checked the dlc expression first, because the disruption of cyclic expression pattern of Notch components was found highly correlated to the defect of somite boundary formation in Notch-related mutants [20][21][22][23][24][25][28][29][30][31] . The expression of dlc in mib nn2002 mutants under 28.5 °C incubation were examined from 12 hpf (~5 ss) to 19 hpf (20 ss). Unexpectedly, there was no obvious dlc desynchronization detected in the PSM, suggesting that the segmentation defect might be caused by a Notch-independent mechanism (supplementary Fig.  7). This result may be contradictory to the Notch role in somite segmentation at first glance. However, it is consistent with a previously-analyzed mib null allele, mib tfi91 , which shows regular boundary formation and normal dlc cycling at 28.5 °C but becomes cold-sensitive and exhibits somite boundary defect and desynchronized dlc expression pattern at 22 °C (compared to Fig. 1L; supplementary Fig. 1; reference 7).
We further used tbx24 and papc, two boundary markers in the region of nascent somites 26,32 . The expression of tbx24 was steadily expressed in the PSM, only lost in the anterior nascent somites in mib nn2002 mutants at 12 hpf ( Fig. 4A-B). Similar phenomenon was also observed by using papc: two anterior nascent somite boundaries were lost in mib nn2002 mutants ( Fig. 4C-D). Next, posterior somite marker, myoD 18 , anterior somite marker, fgf8a 33 , somite boundary marker, dystrophin 34 and tcf15 (paraxis 27 ) were used. The expression of myoD, fgf8a, dystrophin and tcf15 were lost or diminished in the region where somites normally form in mib nn2002 mutants at 14 hpf ( Fig. 4E-L). Among these markers, only myoD was  and papc were expressed in PSM, though the anterior segmental pattern was not observed in the region of nascent somites in mib nn2002 mutants. The expression of myoD, fgf8a, and dystrophin were not detected in the region where somites normally formed. tcf15 was weakly expressed in mib nn2002 mutants and no segmental pattern can be discerned. detected in the adaxial cells. These results indicate that the early somite segmentation defect in mib nn2002 is caused by a failure in cell fate specification. This also suggests that the early segmentation defect in mib nn2002 is mediated through a Notch-independent pathway. foxc1a is not essential in later somite segmentation and specification. Although visible somites were hardly observed in mib nn2002 at early stages, the somite-like structures did appear at later stages. To explore the phenotype, we analyzed the somite-like structures of mib nn2002 mutants in more details. Some small round-shaped structures in the region where somites were supposed to form at early stage were found in mib nn2002 mutants (Fig. 5A-B). The number of the small round-shaped structures increased with time in mib nn2002 mutants (Fig. 5A-D). The somites formed in mib nn2002 mutants at a relative late stage have a shape much similar to those observed in wild type (WT) (Fig. 5E-F). In order to understand whether the somites are really formed at later stages, we further examined the expression of dlc, tbx24, myoD and dystrophin at different stages. The expression of dlc, highly correlated to the formation of somite-like structures, was shown in the region of formed somites progressively in mib nn2002 mutants ( Fig. 5G-N). The cyclically expressed dlc in the PSM was not altered as described previously (supplementary Fig. 7). Its expression level in the PSM was gradually recovered at later stages, in contrast to the severe down-regulation at early stages ( Fig. 5G-N). Similarly, the anterior segmented tbx24 expression pattern was progressively restored (Fig. 5O-R). myoD was also expressed in the formed somites with sharp and striped pattern while the normal-looking somites started appearing after 14 hpf (Fig. 5S-X). The expression of dystrophin also suggested that somites develop into myotome with roughly normal somite boundaries at 22 hpf ( Fig. 5Y-Z). All the results suggest that somites are normally specified in mib nn2002 mutants at a relative late somitogenesis stage.
foxc1a has been shown to be an essential factor for somitogenesis and somite segmentation in a morpholino study 27 . Therefore its expression was examined. Surprisingly, the expression of foxc1a was not detected in mib nn2002 mutants at all stages examined ( Fig. 5a-f). These data indicated that foxc1a is only essential for anterior somitic cell fate determination, but not required for the later somite specification. mib nn2000 and mib nn2001 mutants both encode a Mib protein without RF3 domain. To identify the mutations in these three new alleles, the mib mRNA was reversely transcribed, amplified and examined by sequencing. mib nn2000 bore a T to A transversion mutation at the position 2694 from translational start site, resulting in a nonsense mutation C-terminal to RF2 ( Fig. 6A and supplementary Fig. 8A). mib nn2001 resulted in a different splicing isoform (mibΔ RF3) that has a 143 bp insert after position 2851 from translational start site, leading to a truncated protein without the RF3 domain ( Fig. 6A and supplementary Fig. 8B). The characterization and mutation identification of mib nn2000 and mib nn2001 are comparable to the results from mib m132 and mib ta52b , which have RF123 domain deletion and a non-functional RF3, respectively 1,7 . mib gene is deleted in mib nn2002 mutants. Difficulty was encountered while we were trying to amplify the coding sequence of mib from mib nn2002 mutants. Linkage analysis was then employed to assure the position of mib nn2002 mutation and it was demonstrated that mib nn2002 mutation is 3.32 cM away from the mib locus in chromosome 2 (LG2) with SSLP marker Z13620 (Fig. 6B). Therefore, mib nn2002 is indeed a new mib allele.
The phenocopy and rescue experiments were then used to further discern two hypotheses of the cause of mib nn2002 mutants. First, mib nn2002 mutation might encode a mutated Mib protein that leads to cell death and early segmentation defect in homozygous embryos. mib morpholino has been previously used in phenocopying mib tfi91 by knocking down mib ta52b transcript 7 . By using the same approach, morpholino can down-regulate the expression of the hypothetical mutated mib nn2002 transcript and phenocopy the mib null allele, mib tfi91 . Second, mib nn2002 might bear a mutation in the mib regulatory region that affects its expression level. If so, injecting mib mRNA into mib nn2002 should be able to rescue the mib nn2002 phenotypes to some extent. However, both hypotheses were proved to be incorrect. Although mRNA injection partially restored the expression level of her4 and huc and morpholino injection led to an increase in the number of embryos that exhibit mib ta52b -like phenotypes (supplementary Fig. 9), the early segmentation defects and cell death phenotypes in mib nn2002 homozygotes were neither rescued nor alleviated and the ratio of mib nn2002 homozygotes remained unchanged in both rescue and phenocopy experiments (supplementary Fig. 9 and data not shown).
By checking the expression of mib with in situ hybridization and real-time PCR, we found that the expression of mib was hardly detected and might be totally lost in mib nn2002 mutants (Fig. 6C-E) and, thereby, started to wonder whether mib was deleted in mib nn2002 mutants. To explore the possibility, genomic PCR was employed to check the existence of mib gene. However, the genomic sequence of mib was separated on two scaffolds, 3504 and 3540, and was not placed in chromosome 2 of Zebrafish Ensembl (Zv9, http://www.ensembl.org/Danio_rerio/Info/Index). Moreover, the 5' and 3' genomic sequences of mib were separated by other genes. To further clarify the assembly/continuity of the scaffolds, we first used existing mib mutants, mib tfi91 and mib nn2000 , to verify the continuity of the genomic sequence. The results showed that mib tfi91 and mib nn2000 mutations can be identified based on the genomic sequence of scaffolds 3504 and 3540, respectively (supplementary Fig. 10), suggesting that scaffolds 3504 and 3540 are continuous. We assembled the genomic mib sequence from scaffolds 3504 and 3540 and delimited the intron-exon boundaries with the help of cDNA sequence (supplementary File 11). According to the assembled mib genomic sequence, the primers were designed and used to examine the existence of mib in mib nn2002 with fih, a gene not located on LG2, as a positive control. Consistent with our speculation, mib could not be amplified from the genomic DNA of mib nn2002 mutants (Fig. 6F). mib nn2002 mutants have an arm deletion in chromosome 2, including the loci of mib and foxc1a. In order to dissect the link between mib, foxc1a and the cell death phenotype in mib nn2002 The results of mib genomic PCR were shown with fih (hif1an) as a positive control. AB represents AB WT samples; MU represents mib nn2002 homozygous samples; NE represents negative controls, where distilled water was used instead of genomic extracts as templates. The results showed that no mib genomic fragments are amplified.
Scientific RepoRts | 5:10673 | DOi: 10.1038/srep10673 mutants, we performed an RNA-seq analysis on mib nn2002 mutants and AB WT embryos using the Mi-Seq platform. The detailed data will be reported elsewhere. Here, we only present the genes in the chromosome 2. From the RNA-seq, an unexpected result was found. Around 140 transcripts that are located between 19233 and 9578708 in LG2 were not detected (mean expression value < 0.5 as set-point) only in the mutants. The full list of LG2 can be found in the supplementary Table 3. Selected transcripts were undergone genomic PCR for confirmation, which included both ends of the truncated region: cldn1 (88773-91785) and dvl3a (9490950-9532978) (Fig. 7, supplementary Fig. 12). The result (summarized in supplementary Table 4) indicated that probably there is an arm deletion in the chromosome 2 of mib nn2002 mutants and the breaking point is between dvl31 and ap2m1a (9532978-9578708) (OR 1133-1144 from zfin mib map ch2, http://zfin.org/cgi-bin/view_mapplet.cgi). To conclude, mib nn2002 mutants have a regional deletion in the genome, where part of the chromosome 2, including mib and foxc1a, is deleted.
The missing genes may play roles in somitogenesis and apoptosis. We further underwent DAVID analysis (see Materials and Methods) on the missing genes in mib nn2002 mutants. 91 genes had the DAVID ID, and 63 of them matched the zebrafish ID. The relatively low matching might be due to the present unknown genes in the list, such as those named as zebrafish gene collection (zgc) and CABZ sequences from clones or whole genome shotgun contigs. Nevertheless, the analysis identified ten enriched annotation clusters in mib nn2002 mutants. For example, somitogenesis and segmentation were identified in the analysis, which matched our molecular and phenotypic data in the mutants. In addition, signalings that are critical in early embryogenesis, such as Hedgehog and Wnt signaling pathways were deregulated (supplementary Table 5). All these results further support the phenotypic analysis of mib nn2002 mutants.
To confirm the causal consequence of foxc1a deficiency in anterior somite segmentation phenotype found in mib nn2002 mutants, foxc1a mRNA was injected into mib nn2002 mutants and found to be able to  Table 4 for a more complete set of data).
Scientific RepoRts | 5:10673 | DOi: 10.1038/srep10673 partially rescue the anterior myoD expression in mib nn2002 mutants at 16 hpf (Fig. 8). Two major rescued features of myoD expression in the anterior somites were segmental (Fig. 8B-E) and laterally-extended (prominently seen in Fig. 8D) myoD pattern, compared to Fig. 5V and Fig. 8A. In the first experiment, 111 embryos were collected for 16 hpf myoD expression pattern analysis. 22 (20% of total embryos) partially rescued or non-rescued like mib nn2002 homozygotes and 23 WT-like siblings were selected for imaging and genotyping. The 22 homozgyote-like embryos and the 23 WT-like embryos were all genotyped as mib nn2002 homozygotes and WT siblings, respectively. Among these 22 homozygotes, 6 foxc1a-overexpressed homozygotes showed a partially rescued myoD expression pattern as shown in Fig. 8B-E (27% of selected homozygotes). To better understand the rescue spectrum, we analyzed all the 38 injected embryos from the second set (supplementary Fig. 13 and data not shown). There were 9 embryos genotyped as homozygotes (supplementary Fig. 13B), 28 as WT and one as unknown (none of the primer sets worked, though the myoD expression pattern suggested that it is a homozygous mutant). Therefore, 24% (9/37) embryos are mib nn2002 homozygous. 4 (44%) homozygotes were partially rescued (supplementary Fig. 13Ac, d, h, i) and 3 (33%) homozygotes were almost fully rescued (supplementary Fig. 13Aa, e, f). These results strengthen the notion that foxc1a is only required for anterior somite segmentation.
selt1a and scp2a (related to oxidative damage) and smc6 (related to chromosome structure maintenance) were identified from the deleted genes in mib nn2002 mutants. This finding suggests that DNA damage and chromosome instability may be the cause of p53 up-regulation (supplementary Fig. 6) and further lead to cell death in mib nn2002 mutants. To explore this possibility, γ -H2AX staining was employed, because phosphorylated H2AX (γ -H2AX) can be spotted shortly after the induction of DNA double-stranded breaks (DSBs) and is essential for recruiting several DNA repair proteins to the DNA damage sites 35 . Indeed, many γ -H2AX foci can be detected in mib nn2002 mutants in contrast to control embryos (supplementary Fig. 14). Moreover, we injected p53 morpholino into mib nn2002 mutants and found that dying cells are decreased, suggesting that cell death in mib nn2002 mutants is p53-dependent (supplementary Fig. 15).

Discussion
Previously, mib was found to be an essential factor in Notch signaling pathway. The loss of mib function will lead to the failure of adding ubiquitin onto Notch ligands and then the down-regulation or loss of Notch downstream gene expression. To further explore the functions of mib, we carried out an allele screen and found three new mib alleles, mib nn2000 , mib nn2001 and mib nn2002 . mib nn2000 and mib nn2001 mutants exhibited typical mib phenotypes, including disruption in somite boundary formation, curly tail and down-regulation of Notch target genes. Interestingly, mib nn2002 mutants showed defects in early somite boundary formation and cell death. Although mib nn2002 mutants showed similar Notch molecular phenotypes as other mib mutants, their morphological phenotypes were demonstrated to be unrelated to the loss of mib in mib nn2002 mutants, because it was not rescued by mib or mib2 overexpression. Instead, the loss of foxc1a is more likely to be the cause of the anterior segmentation defect. In a previous morpholino study, the down-regulation of foxc1a specifically leads to the loss of myoD expression in the region where somites were supposed to form and does not affect the expression of myoD in the adaxial cells 27 . The failure of somite segmentation was not caused by the disruption of cyclic dlc expression in the PSM as the dlc is normally and cyclically expressed in the PSM of mib nn2002 mutants. Furthermore, mib nn2002 mutation contained an arm deletion and lost around 140 genes, including mib and foxc1a, on LG2. Therefore, it is possible that the failure of somite specification is caused by the loss of foxc1a. In other words, anterior somite segmentation defect found in mib nn2002 mutants is caused by a Notch-independent pathway. mib nn2002 is probably a spontaneous deletion allele that lacks mib and other genes, because ENU mutagenesis mainly induces random mutations and produces almost all possible nucleotide substitution changes (transversion and transition), including nonsense mutations (e.g. mib nn2000 ), mutations that affect transcript splicing (e.g. mib nn2001 ) and missense mutations 36 . Interestingly, syu t4 has been identified as a spontaneous deletion allele deleting shh gene. Similar to mib nn2002 mutants, syu t4 mutants also showed a brain necrosis phenotype, which may be due to other genes affected by the deletion that has not been delimited and characterized 37 .
In the transcriptome study, three genes related to somite development were found: mib, foxc1a and wnt3a. The participation of mib in the anterior somite formation is excluded, because the null allele (mib tfi91 ) did not exhibits early boundary formation defect, at least at 28.5 °C 7 . wnt3a is unlikely responsible for the somite defect, either, for two reasons. First, mouse Wnt3a regulates cyclic Notch signaling through one bHLH transcription factor, Mesogenin1 (Msgn1) 38 ; however, cyclic dlc expression is still observed in mib nn2002 mutants (supplementary Fig. 7). Second, wnt signaling does not appear involved in modulating segmentation clock but wavefront velocity in zebrafish 39 .
It was believed that foxc1a is crucial for the process of somite segmentation, because morpholino was able to inhibit somite formation before 7 ss but Foxc1a level was restored after 7 ss due to a negative feedback regulation of Foxc1a on transcription of its own gene 27 . Similar study was also carried out in compound Foxc1; Foxc2 homozygous mice (zebrafish did not have foxc2). It was found that mouse anterior somites (9.5 dpc with about 20 somites; with greater than 60 somites in total after 15 dpc) are not formed at 9.5 dpc; however, the formation of the posterior somite cannot be analyzed because the mice died far too early 40 . In our study, the expression of foxc1a was lost as it is deleted in the genome of mib nn2002 mutants. Nevertheless, the somite segmentation and boundary formation were relatively well processed without foxc1a at a relative late stage. Importantly, the anterior myoD expression pattern could be partially rescued in mib nn2002 mutants by overexpressing foxc1a. These results, therefore, demonstrate that foxc1a is only required for the formation of anterior somites but not for that of posterior somites.
As mentioned early, the genomic sequence of mib was placed on two scaffolds. Although the two scaffolds were proved to be the sequence encoded mib genomic DNA, they failed to be assembled into one as there are two genes, rab18b and Loc100331480 (PTCHD3 (3 of 3) located between the 5' and the 3' parts of mib genomic sequences (supplementary Fig. 10). Through comparing synteny of genomic DNA from different species, including Cavefish, Tilapia, Tetraodon, Stickleback, Platyfish, Fugu, Medaka, Cod, Xenopus, Mouse and Human, none of the mib genomic DNA is separated by rab18b in Ensembl database (data not shown). Besides, there is no overlap between the shotgun sequences, cu694527 and cu681854, which constitute scaffold 3504. Therefore, we propose that the shotgun sequence cu694527 should be removed and then the scaffold 3504 and scaffold 3540 can be assembled into one scaffold (supplementary Fig. 16).
On the other hand, the position of mib gene was suggested to be located on LG2 through different mappings, such as T51, LN54 and MGH genetic maps, based on the data in ZFIN genetic maps (http:// zfin.org/action/mapping/all-panels?record= JUMPTOREFCROSS). However, the genomic sequence of mib was not placed in any chromosome in zebrafish. To unravel the ambiguity, we analyzed and compared the synteny of genes near the mib locus and found that the synteny in zebrafish is much more similar to that in cavefish (supplementary Fig. 17), as in the case of hox gene clusters demonstrated previously 41 . Therefore, we used the information of these syntenies and the existed Ensembl sequences to rebuild a hypothetical genomic structure of LG2 near the locus of mib (supplementary Fig. 18). In the hypothetical genomic structure of LG2, nearby mib locus are the genes, such as mkxb, rab18b, ptchd3, yme1l1b, gata6 and cables1. The transcripts and genomic sequences of those nearby genes were also not detected in mib nn2002 mutants ( Fig. 7; supplementary Fig. 12; supplementary Table 4).

Materials and Methods
Zebrafish. Zebrafish was fed and maintained with the help of Taiwan Zebrafish Core Facility at NHRI (http://www.zebrafish-nthu-nhri.org/chinese/index.php). The maintenance followed the standard operation protocol of the core facility. Embryos were raised at 28.5 °C unless otherwise specified. The allele screen was performed according to a published protocol 37 . All experimental procedures on zebrafish were approved by the Institutional Animal Care and Use Committee, National Health Research Institutes, Taiwan (NHRI-IACUC-100028) and carried out in accordance with the approved guidelines.
Laid embryos that were used in the whole mount in situ hybridization experiment (WISH) and live imaging were collected within a 15 min interval after mating started. The collected embryos were selected from pooled quality eggs right after the collection and sorted into 50 ∼ 55 embryos per 10 cm dish with E3 egg water to avoid developmental delay. The embryos was staged as described 42 and phenotypically checked by comparing with WT or WT-like control embryos (the siblings) that were collected in the same time windows. The homozygotes were identified and collected into different vials according to their developmental stages. Embryos were then fixed in 4% paraformaldehyde and kept in fridge for 1 ∼ 3 days. Afterwards, embryos were dechorionated, dehydrated by methanol and stored in freezer.

Phenocopy, rescue and morpholino experiments.
For mib phenocopy and rescue experiments, 10 ng/μ l of mib exon1/intron morpholino 1 and 1 ng/μ l mib1 mRNA or mib2 mRNA were injected into embryos around 1 ∼ 2 cell stage, respectively. Embryos were placed in 28.5 °C incubator before the desired stage.
For foxc1a rescue or overexpression experiments, 20 pg zebrafish foxc1a mRNA was injected into embryos of one-cell stage. Embryos were raised at 28.5 °C before the desired stage.
For p53 knockdown experiments, 15.2 ng p53 morpholino 43 was injected into the embryos of one-cell stage. Embryos were raised at 28.5 °C before the desired stage.
Embryos for Tunel assay (Millipore, cat:S7160) were dehydrated in 70% ethanol overnight after fixation. 10 min ethanol/acetic acid treatment at − 20 degree was used to increase membrane permeability. After the pretreatment of equilibration buffer at room temperature for 10 min, embryos were placed in TdT labeling solution for 2 hours and then washed by phosphate buffered saline. Embryos were then mounted in 2% low-temperature agarose and captured by Nikon A1 confocal microscope.
The pixels of green fluorescence were counted and assayed by Image J. The statistic calculation of the pixels was carried out by Excel (Microsoft). γ-H2AX staining. Laid embryos were collected as mentioned above. The stage of embryos was staged as described 42 and doubly checked by WT or WT-like control embryos (the siblings) that were collected in the same time windows. Embryos were dechorionated and sorted into different dishes of mib nn2002 homozygotes and WT siblings at 14 hpf. WT and mib nn2002 homozygotes that did not exposed to UV were moved to room temperate for 15 min. At the same time, the WT-UV control was placed in laminar flow hood with UV on for 15 min. Afterwards, all dishes were moved to 28.5 °C incubator for 1 hour before collected. Collected embryos were dehydrated by − 20 °C methanol/acetone (1:1) and stored at − 20 °C. Embryos were then sorted into 10 embryos/vial and washed by 1% Triton/PBS several times on shaker. After overnight blocking, embryos were washed by 2% BSA/1% Triton/PBS at 4 °C on shaker, and 1:1000 rabbit anti-human γ -H2AX (Cell signaling, #9718) in 2% BSA/1% Triton/PBS was used for overnight reaction. After intensive wash, 1:1000 goat anti-rabbit Alexa-488 antibody (Molecular probe) in 2% BSA/1% Triton/PBS was used for overnight reaction. After intensive wash, embryos were fixed by 4% paraformaldehyde. Embryos were further deyolked and flat-mounted in 75% glycerol to avoid strong background fluorescence from yolk. Images were taken by Zeiss Axiovision Imager A1.
Genomic PCR and real-time RT-PCR. Genomic DNA was extracted by 50 mM NaOH and 1 M Tris-HCl pH 8.0. One PCR mix (GeneDireX, cat:MB203-0100) and 5x PCR Master Mix II (GeneMark, cat:RP02D) were used with primers (supplementary Table 6). For WISH samples, individual samples were washed repeatedly and sequentially by 1x PBST and double-distilled H 2 O to avoid glycerol contamination, cross contamination between samples and salt interference before extraction was carried out. PCR protocol followed the instruction of 5x PCR Master Mix II (GeneMark, cat:RP02D).
For the RT-PCR, embryos were collected and the total RNA was extracted using TRIzol (Invitrogen). Reverse transcription followed the instruction of SuperScript III FirstStrand (Invitrogen, cat:18080-051). After reverse transcription, PCRs were conducted using LightCycler 480 real-time PCR detection system using SYBR Green I Master (Roche). The data were normalized using the expression levels of β -actin mRNA. The occurrence of primer dimers and secondary products was inspected using melting curve analysis. Our data indicated that the amplification was specific. There was only one PCR product amplified for individual set of primers. Primer sequences were listed in supplementary Table 6.
Genomic DNA extraction of F2 embryos and linkage analysis. F1 heterozygous parents (nn2002/ WIK) were crossed to produce F2 embryos for genomic DNA extraction. mib nn2002 embryos were identified at 10 ss, lysed by 50 μ l lysis buffer (10 mM Tris-HCl (pH 8.3), 1.0 mM EDTA, 12.5 mM KCl, 0.3% Tween 20, 0.3% NP-40), heated at 98 °C for 10 min, and incubated at 55 °C overnight with 1 mg/ml proteinase K. Proteinase K was then inactivated by incubation at 98 °C for 10 min. Afterwards, typical PCR assays were performed using selected simple-sequence length polymorphisms (SSLPs) markers. Since the mutant has the typical mib mutant phenotype, we selected seven SSLPs markers at chromosome 2, where the mib is located. PCR products were then undergone gel electrophoresis, and recombinant band was counted in each PCR reaction.
RNA preparation and Illumina RNA-seq. Fish heterozygous for the mib nn2002 alleles were crossed to obtain mib nn2002 homozygotes. Total RNA from 15 pooled homozygous mutants and AB wild type at 16 ss was extracted using the mirVanaTM miRNA isolation kit (Applied Biosystems). RNA quality was assessed using the Agilent 2100 Bioanalyzer system and samples with a RNA Integrity Number (RIN) greater than 8 were used for RNA library construction.
Library construction was conducted according to the manufacturer's protocol (Illumina, TruSeq Stranded mRNA Sample Prep). Briefly, six RNA (cDNA) libraries were constructed (three AB wild-type and three mib nn2002 homozygotes), and each prepared from 300 ng of total RNA. Illumina adaptors (Integrated DNA technology) were ligated to their 5' and 3' ends to construct strand-specific cDNA libraries. Sequencing was performed on the Illumina MiSeq sequencer to generate 2.0 Gb of sequencing data per sample. Sequences were extracted from image files using the Illumina pipeline set at default parameters. Low quality, homopolymers and adaptor sequences were removed, and the resulting filtered reads were then aligned with the zebrafish reference genome Zv9 (GCA_000002035.2) and undergone transcriptome analysis by CLC Genomics Workbench 6. Only selected data on chromosome 2 are shown in this study to prove the arm deletion. Data deposit and further bioinformatics analysis will be reported and released in the future (unpublished data). Bioinformatics analysis. The arm deletion was found after the rearrangement of the chromosome two. Individual reads of wild-type AB fish and mib nn2002 mutant samples (n = 3 for each) were listed (supplementary Table 3). The Database for Annotation, Visualization and Integrated Discovery (DAVID) was used for functional annotation clustering analysis on the missing genes in mib nn2002 mutants with the highest classification stringency as default 51,52 . Note. Similar to what we have demonstrated here, we noticed that foxc1a plays an important role in early somitogenesis shown by examining two TALEN-induced foxc1a null alleles 52 .