A list of all the genes identified in the Hopkins lab by insertional mutagenesis screening has been published 6. The insertional mutant collection can also be viewed online 37. To identify genes required for craniofacial development we screened the insertional mutant collection at 5 days post fertilization using the cartilage stain Alcian blue. We inspected 306 recessive-lethal insertional mutant loci for defects in the craniofacial cartilages (Figure 1A, C). Briefly, the 1^st arch NC cells give rise to the zebrafish jaw that consists of a proximal palatoquadrate (PQ) that articulates with the distal Meckel's (M) cartilage. The 2^nd arch NC cells give rise to the hyosymplectic and ceratohyal (CH). The hyosymplectic is composed of a proximal hyomandibular (HM) region and a distal symplectic (SY) region. The hyosymplectic articulates with the distal CH and together these 2^nd arch-derived structures support the more anterior jaw by anchoring it to the neurocranium 89. Through inspecting these cartilages, we identified 3 specific cartilage phenotype classes (Classes 1–3 listed in Table 1). To identify specific craniofacial development mutants, we excluded mutants that, in addition to anterior arch defects, also showed progressive loss of the posterior arches.

The anterior arch patterning class contains 6 mutant loci with specific patterning defects in the arch 1 derivatives, the arch 2 derivatives or both (Table 1, Figure 1). Studies on lazarus/pbx4 mutants revealed essential roles in neural tube development and maintaining separate neural crest streams 38. Previous characterization of the nos/foxI1 mutant revealed the importance of cell survival signals in HM formation and additional screening identified two additional loci required for HM development, word of mouth (wom) and sec61a (Figure 1 C–H) 14. We identified 2 alleles of sec61a and although the sec61a^hi1058 allele shows general defects that complicate analysis, the sec61a^hi2839B allele lacks the anterior portion of the HM (aHM) region but clearly retains the posterior HM (pHM) region (compare regions separated by red dotted lines in Figure 1A, H). Notably, the wom mutant also appears to lack the aHM region and retain the pHM region. In contrast, the nos/foxI1 mutant lacks both the aHM and pHM regions

The chondrocyte differentiation class contains 7 mutant loci with generally impaired overall chondrogenesis (Table 1). Among these mutants is the previously characterized jellyfish/sox9a (jef) mutant that is a zebrafish disease model for Campomelic Dysplasia 3940. We also identified a mutation in the glycosaminoglycan biosynthetic enzyme b4galt7 that represents a new zebrafish disease model for the progeroid form of Ehlers-Danlos Syndrome [see Additional file 1].

The cell morphology class contains 3 mutant loci that are required for proper chondrocyte morphology (Table 1). The gonzo and knypek mutants have been previously described 4142. We have not yet identified the gene disrupted in the old skool (odk) mutant [see Additional file 1].

We performed blast searches against the current human protein database using the predicted protein sequences for the 15 identified zebrafish genes. In 15/15 cases, a human homolog to the zebrafish protein was readily identified. The average amino acid sequence identity between the fish and human homologs was 77% and ranged from 47% to 99% identity (Table 2). Database analysis also revealed diverse cellular functions and protein domain architectures for the identified genes. Notably, 4 of the 7 chondrocyte differentiation mutants function in the glycosaminoglycan biosynthesis pathway highlighting the importance of this pathway in chondrogenesis and related disease states. Also, 2 of the 15 genes function in the ER protein secretory pathway, 2 genes function in Wnt signaling, 3 encode transcription factors and 2 encode WD40-repeat domain proteins. Thus, the genes we identified as essential for craniofacial development in the zebrafish appear to function in diverse cellular processes and are highly conserved in humans suggesting they may also be essential for human craniofacial development.

The dirty south (dys) mutants display variably reduced M and PQ cartilages with variable reductions in the CH as well (Figure 1K, L). The mutant was named in reference to the variable lower jaw cartilage defects. Through inspecting large numbers of Alcian blue stained dys mutant animals we found that most dys mutants displayed the 'strong' phenotype of severe reductions in both the M and PQ with small clusters of cartilage cells possibly derived from either the M or PQ or both often present (M? and PQ in Figure 1L). Although in most dys mutant animals the CH was present, we occasionally observed dys mutant animals that lacked one of the two bilaterally symmetric CH cartilages (data not shown). In fact, we often observed asymmetric defects where cartilages on one side would appear more severely reduced than on the contra-lateral side (data not shown). Also, less than 5% of observed dys mutants showed mild defects instead of the more severe losses described above. These 'mild' defect animals tended to show only simple joint fusions between otherwise relatively normal M and PQ as well as CH and HM cartilages (Figure 1M, red asterisk indicates 1^st arch joint region, black asterisk indicates 2^nd arch joint region). The PQ is composed of two distinct regions. The ptergoid process (PTP) is a thin rod of chondrocytes that extends anteriorly from the planar body of the PQ. A small percentage of dys mutants showed restrictions of the planar mid-region of the PQ with what appears to be an identifiable anterior PTP region and posterior planar PQ body region (Figure 1N). The distal M cartilage was always observed in these animals, albeit fused to the anterior region of what might be the PTP. Occasionally, we observed what appears to be complete splitting of the PQ planar body (Figure 1O).

The dys locus encodes the 342 amino acid protein Wdr68 6. Genome database searches revealed only one wdr68 gene in the human, mouse, and fish although we cannot rule out the possibility that the genome databases are incomplete. The insertion is in the first exon of the wdr68 gene, 131 nucleotides downstream from the predicted translation initation site (Figure 2A). RT-PCR analysis revealed that dys^hi3812 is a null mutant (Figure 2B). Notably, wdr68 transcripts are readily detected by RT-PCR in wild type animals at all developmental stages including unfertilized oocytes (Figure 2C). As a negative control, edn1 transcripts, which are not detectable by ISH until late somitogenesis stages, could not be detected by RT-PCR in unfertilized oocytes or in samples earlier than 10 hpf (Figure 2C). Although the observed cartilage phenotype variation outlined in figures 1 L–O might be due to residual maternal wdr68 activity in the dys mutants, we cannot rule out the possibility that this observed phenotypic variation may result from the independent segregation of an unknown number of modifier loci.

Consistent with the RT-PCR data indicating the presence of wdr68 transcripts at all early stages of development, we found by whole-mount in situ hybridization (ISH) ubiquitous expression of wdr68 at sphere, tailbud, 5 somites and 10 somites stages (Figure 3A, B, C, D). By the 20 somites stage, some enrichment of transcripts in the forebrain and hindbrain regions could be detected and patterned expression in the developing somites was also apparent (Figure 3E). Also, transcripts could be observed lateral to the neural tube in a region that would correspond to the post-migratory NC cells at this stage (Figure 3F). By 24 hpf, wdr68 transcripts were significantly enriched in the head region relative to the tail region (Figure 3G). By 28 hpf, restricted expression of wdr68 was evident in the developing somites (Figure 3H). By 34 hpf, wdr68 transcripts were expressed strongly in almost all head structures and only expressed at very low levels in the tail region (Figure 3I, K). ISH analysis on dys mutant animals was not able to detect any wdr68 expression consistent with the RT-PCR results indicating that the dys^hi3812 allele encodes a null mutation of the wdr68 gene (Figure 3J). Notably, wdr68 transcripts were detected from sphere stage through to 34 hpf in all regions of the embryo currently known to play roles in craniofacial development including the pharyngeal pouches.

To further confirm that the dys mutant phenotype is caused by the proviral insertion disrupting the wdr68 locus, we designed two translation blocking antisense morpholino oligonucleotides (wdr68MO and wdr68MO2) to the wdr68 transcript. We found that both gave similar results and that both gave the similar range of variable M, PQ, SY and CH defects as were observed for the dys mutant (wildtype controls in Figure 4A, B and mutants or morphants in Figure 4C, D and data not shown). The images in Figure 4 depict the craniofacial apparatus in whole-mount embryos in contrast to the dissected flat-mounted cartilages shown in Figure 1.

In performing titrations of the wdr68MO we noted that low concentrations tended to cause only mild cartilage joint fusion defects while higher concentrations were more similar to the defects observed in dys^hi3812/dys^hi3812 homozygotes and occasionally yielded more severe reductions in the PQ and CH regions. Since RT-PCR analysis indicated that wdr68 transcripts are maternally supplied we reasoned that zygotic dys mutants probably harbor some residual amount of maternally supplied wdr68 activity. Likewise, antisense morpholino knockdowns are not 100% efficient at blocking translation and therefore residual wdr68 activity would likely also be present in the knockdown animals. Thus, we injected clutches of embryos from dys^hi3812/+ parental crosses with the wdr68MO (dys+wdr68MO) and stained the cartilages with Alcian blue in an attempt to reveal more clearly what might be the phenotype for a more complete loss of wdr68 activity. We found that approximately 3/4 of the embryos were similar to wild type animals injected with the wdr68MO and showed the range of phenotypes typically observed for dys/dys homozygotes leading us to suspect they represent the expected mendelian 3/4 of +/+ and dys/+ animals (Figure 4C, D and data not shown). In contrast, we found that approximately 1/4 of the injected animals nearly completely lacked the M, PQ, SY and CH cartilages (Figure 4E, F). A negative control morpholino had no effect in wild type embryos and did not significantly exacerbate the cartilage defects in clutches of embryos from dys^hi3812/+ parental crosses (data not shown). We found similar results using wdr68MO2 (data not shown).

We compared the dys+wdr68MO animals to animals injected with a previously characterized edn1 translation blocking morpholino (edn1MO) that is reported to nearly fully phenocopy the defects present in edn1 mutant zebrafish 43. Similar to previous reports we found that edn1 is required for the formation of the M, SY and CH cartilages but not for the PQ or HM regions (Figure 4G, H). In contrast, we found that dys mutants also present defects in the PQ that seem to affect the anterior PTP region more strongly than the posterior planar PQ region (Figure 1L). Attempts to reduce wdr68 activity from maternally supplied transcripts through generating dys+wdr68MO animals showed almost complete loss for the M, PQ, SY and CH (Figure 4E, F). Thus, wdr68 activity appears to be required for all edn1-dependent cartilages plus the PTP and PQ regions.

Since defects in NC cell migration might lead to the loss or malformation of craniofacial elements, we used a dlx2 probe in whole-mount ISH to determine whether cranial NC cell migration occurs normally in dys mutant animals. We could not detect any alterations in the migration of the 3 cranial NC cell streams that fill the 1^st, 2^nd and posterior arches at the 12 somites stage suggesting that neither cranial NC cell induction or migration is severely altered in dys mutants (Figure 5A).

Because loss of wdr68 activity caused defects in all edn1-dependent cartilage elements plus the PQ, we examined whether wdr68 might act upstream of the edn1 pathway 43. Using an edn1 probe in whole-mount ISH, we found that while edn1 was readily detected in wild type siblings at the 20 somites stage, dys^hi3812/dys^hi3812 animals had severely reduced edn1 transcript levels (Figure 5B, C). To rule out the possibility that there might be a mild delay in edn1 expression, we also examined expression at 28 hpf and found expression to also be severely reduced at this later stage (Figure 5D, E). We left the 28 hpf animals unbleached to indicate the relatively normal development of the neural crest that give rise to melanocytes to further indicate the absence of a general developmental delay in dys mutant animals (Figure 5D, E). Notably, low and slightly variable levels of edn1 transcripts were detected in the dys mutants at both the 20 somites stage and later at the 28 hpf stage. Variation in the residual levels of edn1 transcripts may account for some of the observed phenotypic variation (Figure 1L–O). For additional evidence that wdr68 is required for edn1 pathway activity we also examined several downstream components of the edn1 pathway. bapx1 is a transcription factor required for induction of the jaw joint that is edn1-dependent 19. In situ hybridization with bapx1 probe showed that while readily detected in wild type siblings, dys^hi3812/dys^hi3812 animals had severely reduced bapx1 transcript levels (Figure 5F, G). The transcription factors hand2 and dlx3b are also edn1-dependent and in situ hybridization likewise showed that dys^hi3812/dys^hi3812 animals had reduced transcript levels (Figure 5H, I and Figure 5L, M). Notably, only the arch-specific expression domains for both hand2 and dlx3b were affected indicating that a general delay in embryonic development cannot explain the absence of expression in the arches (Figure 5H, I, J, K the red dotted outline indicates the unaffected expression of hand2 in developing heart. Figure 5L, M, N, O the red and black asterisks indicate the unaffected olfactory and otic expression domains of dlx3b, respectively). Also, the 1^st arch domains for both markers was more reduced than the 2^nd arch domains perhaps explaining the relatively milder 2^nd arch derived CH cartilage reductions. (red arrowheads indicate 1^st arch, blue arrowheads indicate 2^nd arch, Figures 5H, I and 5L, M). Examining edn1 expression at this relatively late developmental stage, 28 hpf, revealed a similar difference with 1^st arch edn1 expression more reduced than 2^nd arch edn1 expression (red arrowheads indicate 1^st arch, blue arrowheads indicate 2^nd arch, Figure 5D, E).

Analyzing the cartilages of dys+wdr68MO animals suggested that wdr68 activity is required for all edn1-dependent cartilage development. To more precisely explore this possibility, we generated dys+wdr68MO animals and processed them for ISH analysis to determine whether the residual expression of hand2 and dlx3b we observed in the 2^nd arch is wdr68-dependent. We found that 2^nd arch expression of hand2 and dlx3b was consistently more reduced in approximately 1/4 of the dys+wdr68MO animals. The further reduction of 2^nd arch expression is not due to a non-specific general developmental delay because expression of hand2 in the developing heart was not affected and development of the NC cell-derived melanocytes was not affected in the animals analyzed for hand2 expression (dotted outlined structure is heart expression, Figure 5H–K). Likewise, expression of dlx3b in the developing olfactory and otic domains was not substantially affected in the dys+wdr68MO animals (red and black asterisks, respectively, Figure 5L–O). We also compared the expression of hand2 and dlx3b to edn1MO animals and found that dys+wdr68MO animals more closely resemble edn1MO animals than do the dys mutants but that directly blocking edn1 expression does yield more complete reductions suggesting that dys+wdr68MO animals still harbor some residual edn1 signaling (Figure 5K, O). We also found that wdr68 activity is similarly required for the arch-specific expression domains of dlx4a, dlx4b, dlx5a and dlx6a (data not shown). Expression of the dlx genes was also edn1-dependent (data not shown). Collectively, these data indicate that wdr68 activity is required for edn1 expression and its downstream targets.

In zebrafish, as in mammals, chondrocyte differentiation requires the activity of a sox9 transcription factor 3940. To determine whether wdr68 activity is also required for chondrocyte differentiation, we examined expression of the zebrafish sox9a gene. Like the other arch markers, we found that 1^st arch expression was more reduced than 2^nd arch expression in dys mutants (1^st arch red underline, 2^nd arch black underline, Figure 5P, Q, R). Consistent with edn1MO animals having a relatively normal PQ, we detected some residual sox9a expression in the 1^st arch of those animals (red underline, Figure 5S). In all cases, residual sox9a expression in the 2^nd arch region likely reflects the relatively normal formation of the HM region (Figure 1L).

The tbx1 gene has been shown to regulate edn1 expression in the AB* zebrafish line 21. However, loss of tbx1 causes less severe defects in the Tubingen zebrafish line, perhaps due to differences in modifier loci between the AB* and Tubingen lines 44. Since the insertional mutant lines were generated and are maintained on TAB5 and TAB14, two lines derived from a cross between Tubingen and AB* 45, we cannot readily predict the role for tbx1 in the lines used in this study. Regardless, we examined the expression of tbx1 in dys animals and found normal expression at 24 hpf and at most only modest reduction of tbx1 expression in the pharyngeal region at 30 hpf (data not shown). Therefore, wdr68 does not appear to be essential for tbx1 expression and tbx1 expression alone is not sufficient for edn1 expression in our lines.

The observation that the 1^st arch derived M and PQ cartilages were consistently more severely reduced than the 2^nd arch derived CH and HM cartilages in dys^hi3812/dys^hi3812 animals suggested that perhaps 1^st arch patterning requires more wdr68 activity than 2^nd arch patterning. ISH analysis of hand2 and dlx3b expression also suggests that 1^st arch patterning requires more wdr68 activity than 2^nd arch patterning. Consistently, injecting animals with antisense translation blocking morpholinos increased the severity of 2^nd arch cartilage and gene expression phenotypes. To test whether wdr68 activity is differentially required between the 1^st and 2^nd arches, we hypothesized that if the 1^st arch structures require higher levels of wdr68 activity than the 2^nd arch structures, then transformation of the 2^nd arch into a duplication of the 1^st arch would result in near complete absence of all jaw cartilages. The NC cells within the 1^st arch do not express the hox transcription factor genes that are expressed by NC cells within the 2^nd arch. Loss of hox transcription factor activity in the 2^nd arch causes transformation into 1^st arch identity. Likewise, it has been shown that loss of the hox transcription factor co-activator moz/myst3 causes transformation of the 2^nd arch into a duplication of the 1^st arch 464748495051. Therefore, we injected clutches of embryos from dys^hi3812/+ parental crosses with an antisense morpholino against moz/myst3 (mozMO3) and stained for cartilages. In approximately 3/4 of the animals, we observed transformation of the 2^nd arch derivatives into a duplication of the 1^st arch cartilages (Figure 6A, B). These animals were phenotypically indistinguishable from wild type TAB5 and TAB14 animals injected with mozMO3 consistent with the notion that these animals represent the +/+ and dys^hi3812/+ siblings as expected based on mendelian genetics. The slightly less complete nature of these transformations relative to results previously reported might be due to differences between the zebrafish lines used in this study versus the previously reported study 51. In contrast, approximately 1/4 of the animals lacked virtually any jaw cartilages (Figure 6C, D). We reason, based on mendelian ratios, that these animals are likely to be the dys^hi3812/dys^hi3812 animals. The strong reduction of the transformed 2^nd arch derivatives is consistent with a differential requirement for wdr68 activity along the A/P axis. These results together with the ISH data on edn1, hand2 and dlx3b support a model in which the level of wdr68 activity required for 1^st arch-associated edn1 expression is higher than that required for 2^nd arch-associated edn1 expression explaining why dys/dys animals display relatively milder 2^nd arch defects than would be predicted for a gene required upstream of edn1.

The only proteins known to physically interact with Wdr68 are Dyrk1a and Dyrk1b 30. Dyrk1a has been observed to localize within the nucleus 52. To determine whether Wdr68 also localizes to the cell nucleus we constructed mRFP1-Dyrk1a and GFP-Wdr68 fusion protein vectors and transiently transfected HEK-293FT cells with various combinations of the vectors 53. As controls, empty GFP and mRFP1 vectors were both detected throughout the cell (Figure 7A–D). Consistent with previous reports on Dyrk1a localization, the mRFP1-Dyrk1a fusion protein localized to the nucleus when co-expressed with an empty GFP vector (data not shown). The GFP-Wdr68 fusion protein also localized to the nucleus when co-expressed with an empty mRFP1 vector (data not shown). When expressed together, GFP-Wdr68 and mRFP1-Dyrk1a co-localized to the nucleus (Figure 7E–H). In all cases, the locations of nuclei were determined using DAPI.

In Arabidopsis, there are at least 3 wdr68 homologs with Clustal W sequence similarity ranging from 54% to 45%. Among these are the protein TTG1, which is essential for root and shoot patterning events including axis patterning of the leaf. Among several strong Arabidopsis ttg1 mutant alleles, ttg1-9 encodes a point mutation that changes a single serine to a phenylalanine, S282F 28. This position is within a region of the protein that is highly conserved across all organisms examined and all multicellular animals code for the chemically similar threonine at this position (highlighted in red in Figure 8A). Thus we engineered the corresponding putative null T284F mutation into the zebrafish wdr68 gene. To determine whether the engineered putative null T284F can still localize with nuclear Dyrk1a, we constructed a GFP-T284F fusion. The T284F fusion protein failed to co-localize with mRFP1-Dyrk1a (Figure 7I–L).

BLAST analysis with wdr68 revealed the presence of wdr68 homologs in all eukaryotic organisms for which complete genome sequence is available including non-vertebrate animals like Drosophila 54. Using Clustal W to determine sequence similarities 55, the Drosophila homolog, CG14614, is 84% similar to the zebrafish gene. To determine whether CG14614 is a functional homolog to the vertebrate protein, we developed an RNA rescue assay for the dys mutant cartilage phenotype. We injected dys mutant clutches with FLAG-tagged zebrafish wdr68 transcripts and compared them to either uninjected animals or animals injected with ef1α transcripts as a negative control (Table 3 and data not shown). While ef1α transcripts were unable to rescue the mutant phenotype, zebrafish wdr68 rescued the strong cartilage patterning defects causing a concomitant increase in the number of observed mild joint fusion animals (Figure 8F). We also tested the engineered T284F mutant in the RNA rescue assay. The T284F transcript was unable to rescue the cartilage defects consistent with the prediction that this encodes a null allele similar to the plant homolog (live in Figure 8B, Alcian blue in Figure 8D, graph in Figure 8F). We then tested CG14614 for the ability to rescue the cartilage defects present in the dys mutant zebrafish. Remarkably, CG14614 rescued the strong cartilage defects as efficiently as the zebrafish wdr68 transcripts (live in Figure 8C, Alcian blue in Figure 8E, graph in Figure 8F). Thus, wdr68 activity is functionally conserved from animals that lack NC cells.

As the zebrafish insertional mutant collection represents approximately 25% of the genes essential during the first 5 days and this screen identified 16 loci, we can extrapolate that at least 80 genes essential during the first 5 days of development are also required for craniofacial development in the zebrafish 6. Genes essential for very early developmental events such as gastrulation may have been missed in this screen. Similarly, not all defects in craniofacial patterning will necessarily result in lethality by day 5 in the zebrafish. Thus, the range of approximately 80 genes establishes a lower limit on the total number of genes required for craniofacial development.

Previous large-scale ENU screens identified a group of anterior arch mutants with defects in the distal cartilages and most appear to function within the edn1 pathway 356. Notably, the edn1 pathway controls distal cartilage fate in both the 1^st and 2^nd arches 1013. Neither previous screens nor our screening identified a complementary pathway controlling all proximal cartilage fate. However, 3 of the 6 identified patterning mutants have specific reductions in the proximal 2^nd arch derived HM region without affecting the proximal 1^st arch PQ implying that distinct pathways control the patterning of the HM region.

Previous studies on the nos/foxI1 mutant indicated that HM formation requires cell survival signals 14. Analysis of a zebrafish integrin-alpha-5 mutant that displays a reduced aHM region also revealed an essential role for NC cell survival signals 15. Interestingly, the wom and sec61a genes might be required for aHM formation due to roles in mediating cell survival through processes typically viewed as general 'housekeeping' activities. At 29% sequence similarity, the closest yeast homolog to wom is Swd2p, an essential subunit of the Set1 histone methyltransferase (HMT) complex required for transcription of many genes. Interestingly, differential HMT substrate specificity by different regulatory complexes containing different SET domains is one feature of the proposed 'histone code' for epigenetic mechanisms of gene regulation 57. Also, Swd2p is a required subunit in the cleavage and polyadenylation factor complex 585960. The sec61 complex is essential for protein translocation into the ER. Thus, HM formation may require cell survival signals together with high expression levels of a relatively limited number of apparent housekeeping genes that perhaps mediate rate-limiting steps in survival pathways. Alternatively, paralogs with partial functional redundancy for these presumably cell-essential genes may compensate in other tissues.

We showed that wdr68 activity is required for edn1 expression. We also provided evidence that higher levels of wdr68 activity are required for 1^st arch edn1 expression and expression of downstream edn1-dependent genes such as hand2 and dlx3b than are required for 2^nd arch edn1 expression, providing an explanation as to why defects in the 2^nd arch appear less severe in dys mutants than in edn1MO animals. The 2^nd arch likely requires lower levels of wdr68 activity for induction of the edn1 pathway. We also found that tbx1 expression was not substantially reduced in dys mutants or wdr68MO animals suggesting that wdr68 either acts downstream of tbx1 or in a parallel pathway also required for edn1 expression.

Unlike edn1 mutants, which retain an identifiable and intact PQ, dys mutants display severe disruptions of the PTP and PQ. Thus, wdr68 activity is required for all edn1-dependent cartilages plus the 1^st arch proximal edn1-independent PQ region indicating that wdr68 activity is required upstream of a putative PQ formation pathway. Future studies will be needed to identify the putative wdr68-dependent PQ formation pathway genes. Also, the PTP of the PQ appears to serve as the maxilla region at this early developmental stage. Therefore, it would be very informative to generate and analyze a wdr68^-/- mouse for defects in the pharyngeal arch derivatives. Such an analysis would also shed substantial light on the evolution of craniofacial structures.

The nuclear localization of mammalian Dyrk1a has been previously reported and the biochemical purification of a large protein complex from rabbit skeletal muscle that contains both Wdr68 and Dyrk1a suggests that Wdr68 might also localize to the cell nucleus 30. Our observation that a zebrafish GFP-Wdr68 fusion indeed localizes to the nucleus is consistent with the formation of an intra-nuclear Wdr68-Dyrk1 protein complex (Figure 7). Since the Wdr68 T284F mutant was not able to complement the craniofacial defects of dys mutant animals in the RNA rescue assay, the failure of the Wdr68 T284F mutant to localize to the nucleus may suggest that nuclear localization of Wdr68 is required for craniofacial development. The change from a nuclear to a punctate cytoplasmic localization may also indicate that nuclear localization of Wdr68 requires some additional feature associated with having a hydroxyl group at position 284 such as the ability to accept phosphorylation. However, punctate cytoplasmic expression patterns can also indicate misfolding or destabilization and degradation of a fusion protein. Thus, a more detailed biochemical analysis of the Wdr68-Dyrk1 interaction will be required to determine the true significance of these observed changes.

The fly homolog CG14614 rescued the strong cartilage defects as efficiently as the zebrafish wdr68 transcripts, indicating that the molecular functions of wdr68 required by arthopods is the same as that required by vertebrates. Notably, the Wdr68 interaction partners Dyrk1a and Dyrk1b are homologs to the Drosophila minibrain (mnb) gene, suggesting conservation of a protein complex 61. Although the expression pattern for dyrk1b has not been reported, in mice, ISH analysis revealed expression of dyrk1a in the pharyngeal arches and the dyrk1a^-/- animals showed severe growth delay and were lethal by E13.5 31. Although the dys mutants showed only a modest reduction in overall body size, we found that wdr68 co-localized in the nucleus with Dyrk1a (Figure 7). Thus, it will be worthwhile to investigate whether the zebrafish dyrk1 genes might have roles in craniofacial development. The observation that Dyrk1a intracellular localization undergoes dynamic changes during brain development in the chick suggests that the Wdr68-Dyrk1 complexes may function as signaling intermediates 62. Likewise, the finding that Dyrk1b is required for the differentiation switch in myotubes suggests that the Wdr68-Dyrk1 protein complexes might function in diverse organisms and tissues to regulate various developmental events 34.

High throughput mass spectrometric identification of protein complexes in yeast identified the Wdr68 homolog YPL247C in 2 complexes, containing 6 and 13 proteins each, and both complexes contain the Dyrk1 homolog YAK1 63. Thus, Wdr68-Dyrk1 protein complexes are conserved from yeast to mammals.

Detailed studies of YAK1 indicate that it is required for pseudohyphal differentiation and functions in growth control 6465. YAK1 may act within the Target of Rapamycin (TOR) pathway as rapamycin induces nuclear localization of YAK1 and YAK1 deletion confers rapamycin resistance in certain strains 66. Thus, the presence of wdr68 homologs in unicellular eukaryotes such as S. cerevisiae further suggests that Wdr68-Dyrk1 protein complexes participate in switches from growth to differentiation. Further studies will be required to determine how the Wdr68-Dyrk1 protein complexes function and what role they might play in craniofacial development.

Animals were maintained as described 45. The Alcian blue staining method used in this study has been previously described 14.

The primers used to compare wild type to mutant animals by RT-PCR and to create plasmids for generating anti-sense probes for ISH are 3812-f1 5'-ATCTTTCAGACCAAATGCGCCGTTG and 3812-r1 5'-CCTCATTCTCCCAGACGAAAGAAG. Total RNA was isolated from equal numbers of either wild type or mutant animals by the TRIZOL method essentially as described by the manufacturer (Invitrogen). First strand cDNA synthesis using a poly dT primer and PCR from this material was performed as previously described using actin primers as controls 14. The corresponding PCR fragments were subcloned into the pCR-TOPO bluntII (Invitrogen) vector to yield the plasmid pCR-3812. To make the 1^st strand cDNA samples for comparison of the various developmental stages, animals were harvested at the appropriate time points and processed by the TRIZOL method. Unfertilized oocytes were harvested by squeezing an anesthetized female and eggs were processed by the TRIZOL method. The primers used to compare wdr68 across the various developmental stages are 3812-f2 5'-GATCGCCATTTGCTACAACAACTGC and 3812-r1 5'-CCTCATTCTCCCAGACGAAAGAAG. The primers used for edn1 are edn1-start1 5'-ATGCATTTGAGGATTATTTTCCCAGTTCTGAC-3' and edn1-stop1 5'-CTATGAGTTTTCAGAAATCCACGCTTGG-3'. The primers to compare rpL35 are 258-RTF2 5'-GCTGCTTCCAAGCTCTCAAAAATCC-3' and 258-RTR 5'-TGCCTTGACGGCGAACTTGCGAATG-3'. The ISH method was as previously described with all probes at a final concentration of 200ng/mL 14. The edn1 plasmid has been previously described 13. The bapx1 plasmid has been previously described 19. The hand2 plasmid has been previously described 67. The dlx3b plasmid has been previously described 68. The sox9a plasmid has been previously described 40.

The morpholino sequences used in this study for the wdr68 gene are 3812-1 5'-GAAGC

ATG

CAT

CAT

Wild type and dys mutant clutches were injected with morpholino concentrations ranging from 100 uM to 1 mM for the wdr68 sequences or 200 uM for the edn1-MO using pulled glass needles and a picospritzer as previously described 14. The dys+wdr68MO animals received an approximate 1 nL injection of 1 mM wdr68MO. The wdr68MO2 sequence was fully active at 300 uM. Comparable injections with the 3812-1 control sequence had no effect on cartilage formation.

The wdr68 open reading frame was amplified using primers 3812-FLAGf 5'-taatagaattccaccATGGATTACAAGGATGACGACGATAAGggtATGTCactcCAtGGcAAgCGAAAAGAGATCTACAAATACGAGGCG-3' and 3812-stop1 5'-ctcgagCTACACCCGCAGGATCTCCAGG-3' and subcloned into the pCS2+ vector as an EcoRI-XhoI fragment. The T284F mutation was created by nested amplification of primer T284F-f1 5'-CACCTCATTCCTCCTGCCACATATGTtttGCAGTAGCGGACGATCACCAG -3'with 3812-stop1 and 3812-FLAGf with T284F-r1 5'-CTGGTGATCGTCCGCTACTGCaaaACATATGTGGCAGGAGGAATGAGGTG -3' followed by mixing of the products and further amplification with the outer primers 3812-FLAG-f1 and 3812-stop1. Sequencing confirmed the presence of the T284F mutation subcloned into the pCS2+ vector. The Drosophila homolog CG14614 was amplified from 1^st strand Drosophila cDNA using primers dmWDR68-f2 5'-ttcttcgaattccaccATGTCCTCGACCGCCGGAAAGC-3' and dmWDR68-r2 5'-ttcttctcgtcgacTTAGACCCGCAGGATCTCGCACG-3', digested with EcoRI and SalI and subcloned into the EcoRI and XhoI sites of the pCS2+ vector. All rescue constructs were linearized using NotI and capped mRNA was synthesized using the mMessage mMachine kit essentially as described by the manufacturer (Ambion). The xenopus ef1a control transcript was supplied as the control template in the kit. All mRNA were quantified and ran on an agarose gel to confirm synthesis. Mutant clutches were injected as described for the morpholinos with approximately 1 nL of a 5 ng/uL solution of each RNA containing 0.1% phenol red. Animals were processed for Alcian blue staining at 4dpf.

The GFP open reading frame from pEGFP-C2 (Clontech) was subcloned as a Eco47III-EcoRI fragment into the ClaI(blunted)-EcoRI sites of the pCS2+WDR68 vectors to yield pCS2+GFP-WDR68 and pCS2+GFP-T284F. The mRFP1 open reading frame was PCR amplified using primers mRFP1-start 5'-ttcttgaattccaccATGGCCTCCTCCGAGGACG-3' and mRFP1-nonstop 5'-ttcttcttctcgagagatctgatatcGGCGCCGGTGGAGTGGCGGCCCT-3' from a bacterial expression vector 53. The mRFP1 PCR fragment was subcloned into pCS2+ as an EcoRI-XhoI fragment. The mouse DYRK1A coding sequence was amplified from E17 stage mouse embryo total RNA using primers mh-DYRK1A-f1 5'-ttcttcttagatctATGCATACAGGAGGAGAGACTTCAGC -3' and mh-DYRK1A-r1 5'-ttcttcttctcgagTCACGAGCTAGCTACAGGACTCTGTTG-3'. The DYRK1A PCR fragment was subcloned into pCS2+mRFP1 as a BglII-XhoI fragment.

Transient transfection of the fluorescent protein fusion plasmids into HEK-293FT cells grown in glass chamber slides was performed using Lipofectamine 2000 (Invitrogen) as per the manufacturers protocol. Cells were fixed and visualized approximately 20–24 hours post-transfection.