In vertebrate embryos, streams of cranial neural crest (CNC) cells migrate to form segmental pharyngeal arches and differentiate into segment-specific parts of the facial skeleton. To identify genes involved in specifying segmental identity in the vertebrate head, we screened for mutations affecting cartilage patterning in the zebrafish larval pharynx. We present the positional cloning and initial phenotypic characterization of a homeotic locus discovered in this screen. We show that a zebrafish ortholog of the human oncogenic histone acetyltransferase MOZ (monocytic leukemia zinc finger) is required for specifying segmental identity in the second through fourth pharyngeal arches. In moz mutant zebrafish, the second pharyngeal arch is dramatically transformed into a mirror-image duplicated jaw. This phenotype resembles a similar but stronger transformation than that seen in hox2 morpholino oligo (hox2-MO) injected animals. In addition, mild anterior homeotic transformations are seen in the third and fourth pharyngeal arches of moz mutants. moz is required for maintenance of most hox1-4 expression domains and this requirement probably at least partially accounts for the moz mutant homeotic phenotypes. Homeosis and defective Hox gene expression in moz mutants is rescued by inhibiting histone deacetylase activity with Trichostatin A.
Although we find early patterning of the moz mutant hindbrain to be normal, we find a late defect in facial motoneuron migration in moz mutants. Pharyngeal musculature is transformed late, but not early, in moz mutants. We detect relatively minor defects in arch epithelia of moz mutants. Vital labeling of arch development reveals no detectable changes in CNC generation in moz mutants, but later prechondrogenic condensations are mispositioned and misshapen.
Mirror-image hox2-dependent gene expression changes in postmigratory CNC prefigure the homeotic phenotype in moz mutants. Early second arch ventral expression of goosecoid (gsc) in moz mutants and in animals injected with hox2-MOs shifts from lateral to medial, mirroring the first arch pattern. bapx1, which is normally expressed in first arch postmigratory CNC prefiguring the jaw joint, is ectopically expressed in second arch CNC of moz mutants and hox2-MO injected animals. Reduction of bapx1 function in wild types causes loss of the jaw joint. Reduction of bapx1 function in mozmutants causes loss of both first and second arch joints, providing functional genetic evidence that bapx1 contributes to the moz-deficient homeotic pattern. Together, our results reveal an essential embryonic role and a crucial histone acetyltransferase activity for Moz in regulating Hox expression and segmental identity, and provide two early targets, bapx1 and gsc, of moz and hox2 signaling in the second pharyngeal arch.
The jaw forms embryonically as hinged dorsal and ventral cartilages in the first (mandibular) pharyngeal arch. The jaw cartilages are classically considered segmentally homologous with the jaw-supporting cartilages in the second (hyoid) pharyngeal arch and the gill-supporting (branchial) cartilages in more posterior pharyngeal arches(Kimmel et al., 2001b). First arch cranial neural crest (CNC), which forms the jaw, lacks Hox gene expression. Second and more posterior arch CNC displays a nested pattern of Hox gene expression (Hunt et al.,1991). Hox2 genes act as selector genes for second arch segmental identity: mutation of Hoxa2 in the mouse or reduction of hoxa2b and hoxb2a function in zebrafish results in homeotic transformation of second pharyngeal arch skeletal elements into first arch identity (Gendron-Maguire et al.,1993; Rijli et al.,1993; Hunter and Prince,2002) (C. T. Miller, PhD Thesis, University of Oregon, 2001). Downregulation of Hoxa2 expression by ectopic sources of FGF8 results in a similar transformation of the second arch-derived skeletal elements in the chick (Trainor et al.,2002). Forcing ectopic expression of Hoxa2 in the first arch of Xenopus or chicks, or hoxa2b and hoxb2a in zebrafish, results in the converse phenotype, where the jaw segment adopts a second arch morphology (Grammatopoulos et al., 2000; Pasqualetti et al.,2000; Hunter and Prince,2002). Other Hox genes are proposed to specify segmental identity throughout the pharyngeal arches (Hunt et al., 1991). In support of this, ectopic expression of Hoxa2 can divert the Xenopus third pharyngeal arch (the first gill-bearing or branchial arch) towards hyoid fate(Pasqualetti et al., 2000). Furthermore, third arch CNC in valentino (val) mutant zebrafish inappropriately expresses hoxb2a, which is normally restricted to second arch CNC, perhaps contributing to a mild transformation of the val mutant third arch cartilages to a second arch pattern(Moens et al., 1998; Kimmel et al., 2001a).
Although it is still not clear why Hox2 dysfunction results in homeotic transformations, analyses in the mouse, chick and Xenopushave begun to unravel the Hox2-responsive genetic circuitry. A subtractive screen in mice discovered that Pitx1 is ectopically expressed in Hoxa2 mutant second arch primordia, and inactivating Pitx1 in a Hoxa2 mutant partially rescues the homeosis(Bobola et al., 2003). Expression analyses in mice reveal that at late stages Hoxa2represses expression of the chondrogenic factor Sox9 and the osteogenic factor Runx2 (Cbfa1)(Kanzler et al., 1998). Two other known Hoxa2 target genes are the homeobox genes bapx1and goosecoid (gsc). In Xenopus, ectopic expression of Hoxa2 in the first arch represses expression of Bapx1(Pasqualetti et al., 2000). In chicks, ectopic expression of Hoxa2 induces Gsc expression(Grammatopoulos et al., 2000),and in zebrafish early gsc expression is reported to be downregulated in hox2-MO injected animals (Hunter and Prince, 2002). Both Gsc and bapx1 are essential for craniofacial development, although reducing function of either gene does not result in homeosis (Rivera-Perez et al., 1995; Yamada et al.,1995; Miller et al.,2003).
Hox gene expression is maintained by trithorax group(trxG) activity, which involves chromatin remodeling, including histone acetylation (Simon and Tamkun, 1998). In humans, mutations in trxG members cause leukemia(Look, 1997; Ernst et al., 2002). The MYST family histone acetyltransferase MOZ (monocytic leukemia zinc finger protein;MYST3 - Human Gene Nomenclature Database) is mutated in human leukemias(Borrow et al., 1996). Human MOZ is a large protein of 2004 amino acids and biochemical analyses reveal MOZ to possess both histone acetyltransferase (HAT) and transcriptional activation activity (Champagne et al.,2001; Kitabayashi et al.,2001a). Targets of either of these activities in vivo are unknown and the function of MOZ during embryonic development has not been reported.
Previous screens in zebrafish have identified a large number of mutations causing craniofacial defects (Schilling et al., 1996a; Piotrowski et al., 1996; Neuhauss et al.,1996). The cloning of several of these mutations[endothelin1 (edn1 or sucker)(Miller et al., 2000); tbx1 (van gogh)(Piotrowski et al., 2003); tf2ap2a (lockjaw)(Knight et al., 2003)] reveals remarkable conservation in the genetic control of vertebrate craniofacial development, as each of these molecules is also required for patterning the mammalian pharyngeal arches (Kurihara et al., 1994; Jerome and Papaioannou, 2001; Schorle et al., 1996; Zhang et al.,1996). To identify genes required for segmental identity in the pharyngeal arches, we directly screened for mutations affecting cartilage patterning in zebrafish.
We present the molecular identification and phenotypic characterization of a zebrafish homeotic mutant discovered in this screen. Fine mapping,positional cloning, sequencing and morpholino phenocopy experiments reveal this homeotic locus to encode a zebrafish ortholog of the human oncogene MOZ,a MYST family HAT. Severely reduced hox2 expression in mozmutant zebrafish contributes to a mirror-image duplication of jaw cartilages in place of second arch cartilages. moz is also more broadly required for maintenance of most hox1-4 expression domains, probably resulting in the homeotic transformation of the third and fourth arch gill support cartilages. In the hindbrain, moz is required for maintenance, but not initiation, of Hox gene expression, and moz mutants display aberrant facial motoneuron migration. Inhibition of histone deacetylase activity with Trichostatin A rescues Hox maintenance defects and homeotic cartilage transformations in moz mutants, indicating that HAT activity is essential for moz function. Pharyngeal musculature appears transformed late but not early in moz mutants. We find little evidence for patterning defects in arch epithelia of moz mutants. However, striking gene expression changes in moz mutant postmigratory hyoid CNC are apparent. Expression of bapx1, which is normally restricted to the jaw joint (Miller et al., 2003) is robustly duplicated in second arch CNC of moz mutants. Although reduction of bapx1 function in wild-type embryos results in absence of the jaw joint(Miller et al., 2003),reduction of bapx1 function results in absence of both the first and second arch joints in moz mutants. Expression of gsc is profoundly reorganized in the moz mutant second arch, with lateral CNC expression shifting to medial, mirroring the wild-type first arch pattern. Together our results reveal that a zebrafish ortholog of the human oncogene MOZ regulates Hox gene expression and segmental identity in the vertebrate pharynx.
Materials and methods
Fish maintenance and Alcian screen
Fish were raised and staged as described(Westerfield, 1995; Kimmel et al., 1995). For the head cartilage screen, ENU-mutagenized F2 gynogenetic diploid clutches were generated by EP treatment (Streisinger et al., 1981). Mutagenesis was postmeiotic(Riley and Grunwald, 1995) for b719 and premeiotic(Solnica-Krezel et al., 1994)for b999. Survivors at day four were fixed, stained with Alcian Green, and bleached to remove pigmentation as described(Miller et al., 2003). Clutches were screened for cartilage morphology under a Zeiss STEMI SR dissecting microscope at 50× magnification. The recessive larval lethal mutations b719 and b999 were outcrossed to the AB strain. All detailed phenotypic analyses were carried out with the b719allele.
Mapping and positional cloning
Initial mapping was performed with mozb719 on an outbred wik background. Fine mapping was performed with mozb719 crossed onto the Islet1:GFP background(Higashijima et al., 2000),which was found to be highly polymorphic relative to AB within the z6371-z7351 interval. In these fish, primers 1 and 2(Table 1) were used to amplify the microsatellite z6371. Primers and enzymes were used to reveal co-dominant polymorphisms in the 5′ and 3′ UTRs, respectively, of fc32e05(3+4, MnlI) and fc15g12 (5+6, XmnI). All size polymorphisms were resolved on 1-4% agarose gels using standard techniques. The 3′ end of mki67l was not present on PAC74G4. The SP6 end of PAC 14P16 begins with the ninth nucleotide of the fc15g12 ORF. PAC ends were sequenced and the following primers and enzymes used to reveal codominant polymorphisms: 4T (T7 end of PAC 4O19, 7+8, BclI); 14T (T7 end of PAC 14P16, 9+10, ScfI); 114T (T7 end of PAC 114E16, 11+12, DraI). Accession numbers are: AY600370 (moz cDNA) and CL525848-CL525855 (PAC end sequences).
In situ hybridization was performed using standard techniques with PTU-treated animals to inhibit pigmentation(Westerfield, 1995). Genotypes of animals with gene expression defects were confirmed by PCR genotyping, using dCAPS (Neff et al.,1998) to turn the b719 lesion into a codominant polymorphism (using primers 13+14, followed by cutting with BstXI). Both alleles of moz segregate as Mendelian recessive mutations. For all gene expression defects reported for stages before 4 days, at which point the mutant phenotype is readily scorable by morphology, at least four moz mutants had their genotypes confirmed by PCR. Embryos and larvae were deyolked manually with an insect pin and eyebrow hair, cleared in 70%glycerol and photographed on a Zeiss Axiophot with Nomarski optics. For motor neuronal analyses, mozb719 heterozygotes were crossed to the Islet1:GFP strain(Higashijima et al., 2000). References for probes used are as follows. All Hox probes(Prince et al., 1998); eng2 (Ekker et al.,1992); myod(Weinberg et al., 1996);α -tropomyosin (Thisse et al., 1993); pea3(Brown et al., 1998); shh (Krauss et al.,1993); pitx2c (Essner et al., 2000); rag1(Willett et al., 1997); bapx1 (Miller et al.,2003); and gsc(Schulte-Merker et al.,1994).
Morpholino oligo injections
MOs were purchased from Gene Tools (Philomath, OR), and sequences are listed in Table 1. MOs were diluted to 25 mg/ml in 1× Danieau buffer. Subsequent dilutions were made in 0.2 M KCl and 0.2% Phenol Red. These dilutions were injected into the yolk of one- to four-cell zebrafish embryos, roughly 5 nl per embryo. hoxa2b-MO and hoxb2a-MO were each injected at 3 mg/ml, and bapx1-MO was injected at 3 mg/ml.
Trichostatin A treatment
Trichostatin A (TSA; Sigma) was dissolved in DMSO to make a 3 mM stock solution, which was stored at -20°C. For embryo incubations, this stock solution was diluted to 0.1 M in Embryo Medium (with 0.003% PTU and penicillin/streptomycin). This concentration of TSA has been shown to cause increased H4 acetylation in zebrafish(Collas et al., 1999);treatments at higher concentrations cause severe edema and severely reduced cartilage development in wild-type embryos (L.M., unpublished). Embryos in their chorions were incubated in 0.1 M TSA, 0.003% DMSO beginning at 15 hours postfertilization and were maintained in this treatment until fixation for in situ hybridization or Alcian staining. Control sibling embryos were incubated in 0.003% DMSO.
Vital imaging with the fluorescent dye BODIPY-ceramide was performed as described (Yan et al., 2002). Briefly, clutches were soaked in dye continually from late gastrulation onwards. Although animals continue to develop normally in this dye, it does slightly retard development, as does keeping the fish at room temperature,which we did while viewing repeatedly under a Zeiss LSM confocal microscope. Therefore stages given are the corresponding stages at 28°C, based on the head-trunk angle and other morphological criteria(Kimmel et al., 1995). A total of 31 fish were examined, eight mutants and 23 wild-type siblings. One side of the head was imaged from the outer surface to the midline with optical sections 3 μm apart.
A new screen reveals a homeotic locus
Although large scale zebrafish screens revealed over 100 mutations affecting craniofacial development(Schilling et al., 1996a; Piotrowski et al., 1996; Neuhauss et al., 1996), no mutants with clear homeotic pharyngeal arch phenotypes were found. Such phenotypes were possibly to be expected based on homeotic mutant phenotypes in the mouse (Rijli et al., 1993;Gendron-Macguire et al., 1993; Selleri et al., 2001). We reasoned that homeotic phenotypes in fish might not result in a severe overall morphological phenotype and that directly screening pharyngeal cartilage shapes might reveal homeotic loci. We designed a screen in which F2 ENU-mutagenized gynogenetic clutches were grown up to 4 days, and surviving larvae were fixed, stained with Alcian Green to visualize cartilage,and then bleached to remove pigmentation. This enabled rapid scoring of head cartilage shapes under a dissecting microscope. This screen revealed two non-complementing alleles, b719 and b999, of a homeotic locus. We initially named the b719 locus bimandibular, for the apparent homeotic duplication of the first arch (see below).
Positional cloning reveals the homeotic locus to encode moz
We mapped b719 using bulk segregant analysis and subsequent fine mapping to an ∼4 cM interval between z7351 and z6371 on LG5(Fig. 1A, data not shown). Two zebrafish ESTs, fc15g12 and fc32e05, mapped to this region(http://wwwmap.tuebingen.mpg.de; http://220.127.116.11/zonrhmapper/Maps.htm)were found to closely flank the b719 locus, by 0.11 cM and 0.05 cM,respectively (Fig. 1A). Two PACs for each of these two ESTs were isolated by PCR from DNA pools of an arrayed zebrafish PAC library (Amemiya and Zon, 1999). Mapping polymorphisms derived from the PAC ends revealed that ends of two of the PACs had crossed the recombinants(Fig. 1A). Sequencing the T7 end of PAC4O19 revealed an exon highly homologous to the human histone acetyltransferase MOZ, positioning moz as within the non-recombinant interval (Fig. 1A). By aligning vertebrate Moz sequences and using degenerate PCR, the rest of the predicted zebrafish moz ORF was isolated and is predicted to encode a 2246 amino acid protein. The first five exons were found to reside on PAC 114E16, whereas exons 8-16 were contained on PAC4O19.
Zebrafish Moz is highly conserved with human MOZ(Fig. 1B), a MYST family histone acetyltransferase (HAT, Fig. 1). In addition to activity of its HAT domain, human MOZ contains a transcriptional activation (TA) domain at its C terminus(Champagne et al., 2001; Kitabayashi et al., 2001a). A 1 bp deletion at position 2590 of 6738 in zebrafish mozb719 mutants results in an early frameshift, predicted to produce a truncated protein lacking this TA domain(Fig. 1A,B). A C-to-T nonsense mutation at position 4977 of 6738 in mozb999 mutants is predicted to also truncate the TA domain(Fig. 1A,B).
Expression of zebrafish moz at 28-48 hours postfertilization (hpf)appears ubiquitous in the head but expression is undetectable in the trunk and tail (Fig. 1C-E). The diffuse posterior boundary of moz expression during this time frame roughly coincides with the boundary between the hindbrain and spinal cord(Fig. 1C-E). Analyzing moz expression at 24, 28, 36 and 48 hpf in clutches of embryos from mozb719 heterozygotes yielded three clear classes of animals based on moz expression: strong, intermediate and faint. PCR genotyping revealed the strong class to be homozygous wild types, the intermediate class to be heterozygous for the mozb719mutation, and the faint class to be mozb719 homozygous mutants. The reduction of moz expression in both classes of moz mutants appeared to globally affect moz expression levels in all cranial tissues (data not shown).
Reducing moz function results in pharyngeal cartilage homeosis
In moz mutants, second arch cartilages adopt a mirror-image first arch pattern, forming an ectopic jaw (Fig. 2A,B). A novel large opening to the pharynx is present on either side of the head, resembling an ectopic mouth(Fig. 2B, but see below). Especially ventrally, this homeotic phenotype resembles the phenotype seen upon reducing function of both hoxa2b and hoxb2a (C. T. Miller, PhD Thesis, University of Oregon, 2001)(Hunter and Prince, 2002). Flat-mounting dissected cartilages from moz mutants reveals that the second arch cartilages adopt shapes characteristic of first arch cartilages(Fig. 2D,E). In the moz mutant second arch, the hyomandibular region of the dorsal second arch cartilage that normally articulates with the otic capsule is missing(Fig. 2D,E; Table 2), presenting a more complete homeotic transformation than observed in the earlier work (C. T. Miller, PhD Thesis, University of Oregon, 2001)(Hunter and Prince, 2002). The moz mutant dorsal second arch cartilage, in the more ventral position of the thin symplectic cartilage (Kimmel et al., 1998) is thicker than its wild-type counterpart,resembling the wild-type first arch dorsal (upper) jaw cartilage, the palatoquadrate. The moz mutant ventral second arch cartilage is shorter, thinner, contains fewer rows of chondrocytes, and forms a knob on its lateral end, resembling the wild-type first arch ventral (lower) jaw cartilage, Meckel's (Fig. 2;and see below). Furthermore, in the first two arches of moz mutants,the dorsal cartilages fuse to one another and the ventral cartilages fuse to one another. by contrast, dorsal/ventral fusions within either arch are only rarely seen (see below).
Cartilages in more posterior pharyngeal arches of moz mutants are also mispatterned. The third arch ventral cartilage in moz mutants appears slightly shorter and thicker and also has a distinctive knob on its lateral end (Fig. 2G,H),resembling the retroarticular process of Meckel's cartilage and thus suggesting a mild transformation of arch three to arch one fate. A similar phenotype is seen in the moz mutant fourth arch (data not shown).
To confirm that these homeotic phenotypes in moz mutants are due to reduction of Moz function, we injected moz morpholino antisense oligonucleotides (MOs). We have previously shown that MO injections can efficiently phenocopy severe phenotypes of larval head skeletal mutants, as well as reveal hypomorphic phenotypes at lower doses(Miller and Kimmel, 2001). Animals injected with any of three different moz MOs display dose-dependent homeotic phenotypes seen in moz mutants(Fig. 2C,F,I; Table 2), strongly supporting our conclusion that reduction of moz function causes the b719 and b999 homeotic phenotypes.
Injection of lower doses of each morpholino, as well as analyses of the slightly variable b719 and hypomorphic b999 phenotypes(Table 2), show that the several homeotic phenotypes described above are separable, and that homeosis is not an all-or-nothing phenomenon. Some mutant animals display shape changes of the dorsal hyoid cartilage (the HM cartilage) without having deletions of HM (Table 2), showing moz controls at least two processes, positioning and shaping, of dorsal second arch cartilage formation. Likewise, some mutant animals display homeotic shape changes of the ventral hyoid cartilage without displaying the inversion (Table 2), similarly arguing that these two processes are separable. Interestingly, shape changes are more frequently seen than fusions, regardless of MO or dose(Table 2). The dorsal deletion of the hyomandibular portion of the hyosymplectic was the least penetrant for both mutant alleles and with both doses of all three MOs(Table 2). Thus, sensitivity to reduction of moz function ranges from high for shape changes to intermediate for fusions to low for dorsal deletions.
moz is required for most Hox group 1-4 expression domains
As anterior transformation of arch two to one resembles the mouse Hoxa2 and zebrafish hoxa2b;hoxb2a loss-of-function phenotypes (Gendron-Maguire et al.,1993; Rijli et al.,1993; Hunter and Prince,2002) (C. T. Miller, PhD Thesis, University of Oregon, 2001), we asked if moz functions upstream of hox2 genes. Although zebrafish have at least seven Hox clusters, only two hox2 genes are retained, hoxa2b and hoxb2a(Amores et al., 1998). Expression of both hoxa2b and hoxb2a is broadly and severely downregulated in the moz mutant second arch primordia by 33 hpf(Fig. 3A-D, see below). Like the second pharyngeal arch expression and despite the separable regulation of pharyngeal arch and CNS Hox gene expression domains(Prince and Lumsden, 1994; Maconochie et al., 1999),hindbrain expression of hoxa2b is also severely reduced in moz mutants at 33 hpf (Fig. 3E,F). All rhombomeres that express hoxa2b (rhombomeres 2-5 or r2-r5) appear to do so at a lower level in moz mutants,although r2 appears to be the most strongly affected. Expression of hoxa2b is strikingly reduced in medial r2 and lateral r4(Fig. 3E,F). To determine if hoxa2b expression was simply delayed in moz mutants, we assayed a later time point, 48 hpf. Even with overdeveloped in situ hybridization and similar to expression at 33 hpf, arch expression of hoxa2b is undetectable (Fig. 3G,H), and hindbrain expression is still drastically reduced (data not shown).
To determine if moz is required for the initiation of expression of the Hox genes it regulates, we examined hoxa2b expression at 11 hpf, soon after it initiates embryonic expression(Prince et al., 1998). Initiation of hoxa2b expression in the hindbrain in mozmutants appears unaffected (Fig. 3I,J). However, initiation of hoxa2b in second arch CNC around 12-14 somites occurs but is substantially reduced in mozmutants (data not shown). Thus, moz regulates hoxa2bexpression in distinct manners in the hindbrain and CNC.
We next asked if moz is required for expression of other Hox genes. hoxa1a expression is normally not present in a typical Hox domain spanning one or multiple segments but instead is in clusters of cells in the ventral forebrain and midbrain and in scattered cells in the anterior hindbrain (McClintock et al.,2001; McClintock et al.,2003). In contrast to the moz requirement for later expression of hoxa2b and hoxb2a, hoxa1a expression is not appreciably affected in moz mutants(Fig. 3K,L).
For the hoxba cluster, expression of group 1-4 genes are affected in a graded fashion in moz mutants, with hoxb1a being the most severely affected and hoxb4a the most mildly affected. Severe Hox expression defects in moz mutants are also present in the embryonic hindbrain. At 36 hpf, the r4-restricted hindbrain expression of hoxb1a is nearly abolished in moz mutants(Fig. 3M,N). In addition to the missing second arch domain (see above), hindbrain expression of hoxb2a is reduced in moz mutants(Fig. 3O,P). hoxb3aexpression is reduced in the hindbrain and in the third to fifth pharyngeal arch primordia in moz mutants(Fig. 3Q,R), perhaps contributing to the third arch homeotic phenotype presented above. Expression of hoxb4a is mildly reduced in both the hindbrain and pharyngeal arches four through six of moz mutants(Fig. 3S,T). Expression of the other hox3-4 genes is similarly affected as their respective hoxb3a and hoxb4a paralogs (data not shown). Initiation of Hox genes b1a-b4a, like hoxa2b, appears unaffected in the hindbrain of moz mutants. Hence, similar to trithoraxgroup (trxG) genes (Simon and Tamkun, 2002; Ernst et al.,2002), moz appears to be required for the maintenance,but not initiation, of expression of particular Hox genes in the hindbrain.
Inhibition of histone deacetylase activity partially rescues the moz mutant phenotype
Because human MOZ has been shown to have histone acetyltransferase (HAT)activity (Champagne et al.,2001), and because trxG factors that maintain Hox gene expression are associated with HAT activity(Petruk et al., 2001; Milne et al., 2002), we wondered whether the inability of moz mutants to maintain Hox gene expression was due to hypoacetylation. The histone deacetylase inhibitor trichostatin A (TSA) has been shown to rescue defects caused by trxG mutations in Drosophila (Sollars et al., 2003) and human cells(Milne et al., 2002). We therefore asked whether TSA treatment could rescue the moz mutant phenotype. moz mutant embryos that are incubated in 0.1 M TSA starting at about 15 hpf show striking rescue of arch cartilage homeosis(Table 3; Fig. 4A-D) and rescue of Hox gene expression (Fig. 4E-H).
TSA-treatment partially rescues hoxa2b expression in the hindbrain and second arch CNC of moz mutants (n=12/12) compared with DMSO-treated moz mutant controls (n=6/6; Fig. 4E-H). TSA-treated moz mutants (n=5/5) have increased, but not wild-type levels, of r4 expression of hoxb1a compared with DMSO-treated moz mutant controls (n=8/8) (not shown). TSA-treated moz mutants have weakly rescued expression of hoxb2a in hyoid CNC (n=6/7) compared with DMSO-treated moz mutant controls (n=5/5, data not shown). Therefore, even though neither of our moz mutant alleles directly affect the Moz HAT domain, these results suggest that the moz homeotic phenotype is at least partially dependent on the function of Moz HAT activity.
moz mutants display late hindbrain neuronal phenotypes
Given the broad expression defects of group 1-4 Hox genes, we analyzed hindbrain neuronal development in moz mutants. Facial motoneurons differentiate in r4 and begin to migrate posteriorly towards r5-6 around 15 hpf (Chandrasekhar et al.,1997; Maves et al.,2002). At 48 hpf, some mutants display mispositioned facial motoneurons (Fig. 5A,B),resembling hoxb1 loss-of-function phenotypes seen in zebrafish and mice (McClintock et al., 2002; Studer et al., 1996; Goddard et al., 1996; Gavalas et al., 2003). In contrast to Mll/Trithorax mutant mice, which also fail to maintain Hox expression (Yu et al.,1998), cranial ganglia appear to innervate each pharyngeal arch(Fig. 5C,D). The early reticulospinal neurons, some of which are born by 10 hpf(Mendelson, 1986) and show anterior transformations upon reduced function of hoxb1(McClintock et al., 2002),display no detectable alterations in moz mutants (data not shown). Thus, early segmentation and neuronal specification of the moz mutant hindbrain occurs relatively normally, while later hindbrain phenotypes in moz mutants are consistent with a defect in maintenance, but not initiation, of Hox gene expression.
Late patterning defects in moz mutant arch mesoderm
Segmental identity in the vertebrate head periphery involves complex crosstalk between CNC and head mesoderm. In the mouse, Hox gene expression in CNC requires unknown factors emanating from cranial mesoderm(Trainor and Krumlauf, 2000). Conversely, in the chick, transplanting (presumably Hoxa2-negative)first arch CNC into the second arch results in second arch muscles non-autonomously adopting a first arch pattern of beak muscles(Noden, 1983a; Trainor et al., 2002). In zebrafish, mosaic analyses have revealed that CNC patterns arch mesoderm(Schilling et al., 1996b; Knight et al., 2003). Additionally, zebrafish injected with hox2-MOs and Hoxa2 mutant mice display altered head musculature (Rijli et al., 1993; Barrow and Capecchi,1999; Hunter and Prince,2002). Thus, we expected head musculature to be affected in moz mutants but wondered at what stage segmental identity defects in the head mesoderm occur.
Pharyngeal arch muscles are derived from paraxial mesoderm, which initially occupies central locations (arch `cores') in the pharyngeal arch, ensheathed by postmigratory CNC (reviewed by Kimmel et al., 2001b). Each arch mesodermal core subdivides into a discrete pattern of identified myogenic cores. The first and second arches display different sequences of mesodermal core subdivision. Although at intermediate stages in fish the first arch contains three myogenic cores[constrictor dorsalis (CD), adductor mandibulae (AM) and intermandibularis(IM)], the second arch contains only two [constrictor hyoideus dorsalis and ventralis (CHD and CHV)]. These myogenic cores subsequently subdivide into primordia for individual muscles(Edgeworth, 1935) (reviewed by Kimmel et al., 2001b). This early difference in the wild-type arch one and two intermediate myogenic core pattern precludes assigning segmental homology to subsequently-forming dorsal and intermediate muscles. However, this intermediate pattern serves as a segmental character distinguishing the first two arches. Thus, we wondered whether this aspect of segmental identity was transformed in mozmutants.
eng2 expression marks the dorsal first arch myogenic condensation,constrictor dorsalis (Hatta et al.,1990; Ekker et al.,1992) (reviewed by Kimmel et al., 2001b). Expression of eng2 in moz mutants at 28 hpf is not seen homeotically duplicated in the second arch(Fig. 6A,B). myodexpression marks all pharyngeal arch myogenic condensations(Schilling and Kimmel, 1997). We examined myod expression in moz mutants at 44 hpf, a stage soon after myod expression labels the first and second arch myogenic condensations (Fig. 6C,D). The arrangement of myod-expressing cores in moz mutants at this stage appears grossly indistinguishable from the wild-type pattern (Fig. 6C,D). Slightly later in development at 54 hpf, subtle defects are observed in moz mutant myod-expressing myogenic condensations. Ectopic patches of myod expression are seen in the intermediate second arch of moz mutants (Fig. 6E,F).
Despite the lack of dramatic early muscle phenotypes, the larval musculature at 4 days is radically transformed in moz mutants(Fig. 6G,H). In wild types, a large jaw-closing adductor mandibulae muscle connects the dorsal and ventral cartilages in the first arch, whereas no prominent muscles connect the dorsal and ventral cartilages in the second arch(Fig. 6G). In mozmutants, the first arch muscle pattern appears unaffected, except for late subdivision and patterning of CD, which normally inserts upon the HM cartilage. In most (61%, 14/23) moz mutants, a muscle of variable size in the second arch connects the dorsal and ventral cartilages(Fig. 6H). These late muscle phenotypes in moz mutants resemble a slightly stronger version of those reported for hox2-MO injected zebrafish(Hunter and Prince, 2002). In moz mutants in which a muscle did not connect the dorsal and ventral hyoid cartilages, hyoid musculature was variably disorganized. Without specific markers for individual hyoid muscles, we were unable to assign identity to these. The third arch ventral muscle (transversus ventralis)(Schilling and Kimmel, 1997)was slightly enlarged in moz mutants, suggesting, like the cartilage phenotype, a mild anterior homeotic transformation. Together these results indicate that moz mutants display late, but not early, anterior homeotic transformations of second and third arch musculature.
Defects in pharyngeal epithelia are not detected in mozmutants
Pharyngeal endoderm is required for many aspects of CNC patterning(Piotrowski and Nüsslein-Volhard,2000; Piotrowski et al.,2003; Couly et al.,2002) and chondrification of CNC requires contact with pharyngeal endoderm (Epperlein, 1974). Given that the hyomandibular (HM) region of the dorsal second arch cartilage almost never chondrifies in moz mutants(Table 2), we wondered whether missing, mispositioned or mis-specified pharyngeal pouches contribute to the moz mutant phenotype.
We directly assayed developing pharyngeal pouch morphology and specification by following the expression of the FGF target gene pea3(Roehl and Nüsslein-Volhard,2001) at 24, 34 and 54 hpf. Each pharyngeal pouch, which separates the arch primordia, consists of an epithelial bilayer with an AP polarity:expression of pea3 and the secreted ligand edn1 are both expressed in posterior, but not anterior, pharyngeal endodermal epithelia(Fig. 7A,B)(Miller et al., 2000). The first pharyngeal pouch, which abuts dorsal second arch CNC, is present in moz mutants and appears similarly patterned as its wild-type counterpart at all three stages examined(Fig. 7A,B; data not shown). The same was true for more posterior pharyngeal pouches(Fig. 7A,B; data not shown). We wondered if the mirror-image duplication of the first arch pattern in the moz mutant second arch could be due to defects or possibly even reversals in pharyngeal pouch polarity. However, expression of pea3(Fig. 7A,B) and edn1(data not shown) in moz mutants revealed no defects in pouch polarity.
Surrounding ectodermal epithelia also contribute to CNC patterning(Tyler and Hall, 1977). shh is specifically expressed in a thin stripe of second arch surface ectoderm at the posterior margin of the second arch resembling an identified shh expression domain in chicks and mice (posterior ectodermal margin or PEM of the second arch) (Wall and Hogan, 1999). Given the second arch specificity of this pattern and the defect in second arch identity in moz mutants, we wondered whether moz mutants might lack this expression domain of shh. Instead, expression of shh in moz mutants is still present in the PEM of the second arch, although expression is perhaps slightly reduced ventrally(Fig. 7C,D). Thus, this epithelial aspect of hyoid segmental identity is retained in mozmutants.
Between the inverted ectopic jaw and the enlarged third arch in moz mutants, large bilateral openings in the pharynx resemble ectopic mouths (see Fig. 2B). To determine if gene expression data support this interpretation, we examined expression of a stomadeal marker, pitx2c(Essner et al., 2000; Schweickert et al., 2001), in moz mutants. pitx2c expression in wild-type embryos at 54 hpf strongly labels the mouth, presumably the ectodermal derivatives of the stomodeum (Fig. 7E,F). Expression of pitx2c in moz mutants was not detected ectopically in these enlarged pharyngeal openings, providing no evidence for stomadeal identity.
In chicks, the thymus forms largely from the third and fourth endodermal pharyngeal pouches, which attract blood-borne lymphocyte precursors(LeDouarin and Jotereau,1975). In mice, hox3 genes regulate thymus formation(Manley and Capecchi, 1995; Manley and Capecchi, 1998). In zebrafish, rag1 expression in lymphocytes marks the early thymus(Willett et al., 1997). In pbx4(lzr) mutants, which have reduced hox3expression, the thymus fails to form as assayed by rag1 expression(Popperl et al., 2000). Thus,we similarly asked whether moz mutants form a thymus by examining rag1 expression at 4 days. rag1 expression in the thymus is present in moz mutants, although reduced(Fig. 7G,H). Taken together,these results reveal relatively minor defects in arch epithelial tissues in moz mutants and are consistent with the idea that many aspects of the arch environment are set up independent of the CNC(Veitch et al., 1999; Gavalas et al., 2001).
Early CNC generation appears normal in moz mutants but mispositioned and misshapen condensations form
Finding evidence suggesting that early patterning of non-CNC arch tissues is normal in moz mutants, we next analyzed the CNC. In mice, Hox genes not only control segmental identity, but also control the generation of CNC (Gavalas et al., 2001). To determine whether the broad Hox expression defects result in a defect in CNC generation, we examined early pharyngeal arch primordia in living embryos with the fluorescent dye BODIPY ceramide. This vital labeling offers nice histological resolution of all major differentiated cell types(Kimmel et al., 2001b; Yan et al., 2002). Examining arch primordia in labeled embryos from clutches of moz mutants revealed mutants to be morphologically indistinguishable from their wild-type siblings around 28 hpf when postmigratory CNC has populated the arch and surrounded the mesodermal cores. No deficit in hyoid CNC was apparent(Fig. 8A,B). Consistent with this, expression of the broadly expressed CNC marker dlx2 at 28 hpf appears unaffected in moz mutants (data not shown). Slightly later,around a 34 hpf stage, the moz mutant hyoid arch appears slightly hypoplastic (Fig8C,D), although gross arch morphology appears relatively normal. These same optical sections confirm our in situ results that no gross changes are apparent in the early pharyngeal pouches in moz mutants (see above).
In jellyfish/sox9a mutant zebrafish, which lack nearly all cartilage, condensations form but subsequently fail to differentiate(Yan et al., 2002). We wondered if the moz mutant hyomandibular cartilage, which almost never forms (Fig. 2A,B; Table 2), forms a condensation but fails to differentiate. Examining BODIPY-labeled animals at later timepoints revealed the condensation pattern in moz mutants to be extremely aberrant around 48 hpf and predictive of disruption of the larval cartilage pattern at 3.5 days (Fig. 8E-H). In the moz mutant dorsal second arch, no trace of an HM condensation is seen. Instead, the dorsal second arch condensation is shifted ventrally (Fig. 8E,F). Likewise, the lateral end of the ventral second arch condensation is displaced anteriorly, abutting the lateral end of the lower jaw cartilage condensation(Fig. 8E-H). Thus, the positioning and shaping processes that moz controls in the larval cartilage pattern begin prior to condensation formation.
moz and hox2 genes repress early second arch expression of bapx1, which is required for aspects of moz-mediated homeosis
Reduced hox2 expression can at least in part account for the anterior transformation homeotic skeletal and muscular phenotypes observed in the second arch of moz mutants. To investigate the molecular consequences of hox2 downregulation in the early second arch primordium, we analyzed embryonic expression of a known hox2 target gene, bapx1 (Pasqualetti et al.,2000), in moz mutants and in embryos injected with morpholinos to reduce function of hoxa2b and hoxb2a.
In embryos and larvae, bapx1 is expressed in a patch of intermediate first arch, but not second arch, mesenchyme(Fig. 9A,D)(Miller et al., 2003). In moz mutants at 33 hpf, first arch bapx1 expression is present while an ectopic bapx1 domain is seen in the second arch(Fig. 9B,E), providing molecular confirmation of an anterior transformation in second arch CNC of moz mutants. Ectopic second arch bapx1 expression is also observed in embryos injected with hoxa2b and hoxb2amorpholinos (95%, 37/39; Fig. 9C), demonstrating that hox2 dysfunction is sufficient to result in these homeotic molecular changes. Ectopic bapx1 expression is present in moz mutants and hox2-MO injected animals when bapx1 expression first initiates arch expression, around 30 hpf (data not shown). This molecular homeosis is stable, as bapx1 expression is maintained in the second arch at 54 hpf(Fig. 9F,G) and 4 days(Fig. 9H,I).
We next asked if this ectopic bapx1 expression domain is functional in moz mutants. Reducing bapx1 function in a wild-type background specifically eliminates the jaw joint and does not affect the second arch joint (Miller et al.,2003) (Fig. 9J). Reducing bapx1 function in a moz mutant background now results in loss of joints in both arch one and two(Fig. 9K; Table 4). Thus, the moz mutant second arch requires bapx1 function for formation of the dorsal/ventral joint, providing functional genetic evidence that the moz mutant second arch phenotype is homeotic.
moz and hox2 genes regulate the mediolateral pattern of goosecoid expression in the second arch
We extended our analyses of early CNC patterning in moz mutants by studying expression of a second known hox2 target gene, goosecoid (gsc)(Grammatopoulos et al., 2000; Pasqualetti et al., 2000; Hunter and Prince, 2002). Misexpression of hoxa2 in the chick and Xenopus induces gsc expression (Grammatopoulos et al., 2000; Pasqualetti et al.,2000). In zebrafish, hoxa2b and hoxb2a were reported to positively regulate early second arch gsc expression,although only lateral views were reported(Hunter and Prince, 2002).
Viewing early embryonic gsc expression from a ventral aspect reveals gsc expression to consist of a thin medial crescent in the ventral first arch and a broad lateral crescent in the ventral second arch(Fig. 10A). Examination of gsc expression in moz mutants from this ventral aspect reveals a startling patterning change. In moz mutants, ventral second arch expression of gsc appears as a thin medial crescent, mirroring the first arch pattern (Fig. 10B). Like the bapx1 expression change, this shifting of gsc expression is also observed in embryos injected with hoxa2b and hoxb2a morpholinos (92%, 23/25; Fig. 10C), demonstrating that hox2 dysfunction is also sufficient to result in this homeotic molecular change of gsc expression. At 41 hpf, dorsal arch two expression of gsc is reduced(Fig. 10D,F). However, lateral views do not reveal maintenance of the patterning change that ventral views do: the shifting of lateral gsc expression to the medial second arch(Fig. 10E,G). The pattern at 41 hpf is slightly different than the 33 hpf pattern(Fig. 10H), suggesting gsc expression is either dynamic, and/or that movements of gsc-expressing cells occur. Together our results suggest moznot only controls maintenance of an early pattern, but also specification of subsequent dynamic changes in patterning in the second arch CNC well before differentiation begins.
An essential embryonic role for the oncogene moz
In humans, translocation breakpoints within MOZ result in leukemia(Borrow et al., 1996; Aguiar et al., 1997; Carapeti et al., 1998; Carapeti et al., 1999; Liang et al., 1998; Chaffanet et al., 2000; Kitabayashi et al., 2001b). Other regulators of Hox genes, including the mammalian ortholog of Drosophila trithorax, MLL also mutate to cause human leukemia(reviewed by Ernst, 2002). As Hox genes are normally expressed in hematopoietic lineages, it has been proposed that deregulated Hox expression contributes to many forms of leukemia (reviewed by Look, 1997). In support of this, deregulated Hox expression is associated with leukemia, and misexpression of individual Hox genes is sufficient to induce leukemia in mice(Armstrong et al., 2002; Yeoh et al., 2002; Rozovskaia, 2001; Kroon et al., 1998; Magli et al., 1997). Our findings that zebrafish moz regulates Hox gene expression raises the question of whether deregulated Hox gene expression underlies MOZ-mediated leukemias.
Although no other in vivo functional data have been reported for moz in other vertebrates, biochemistry on human MOZ has revealed multiple functional domains. These domains include a founding HAT domain of the MYST family, an N-terminal transcriptional repression domain, and a C-terminal transactivation domain(Champagne et al., 2001; Kitabayashi et al., 2001a). MOZ additionally contains two C4HC3 zinc fingers and a C2HC nucleosome recognition motif (Borrow et al.,1996). This composite structure suggests MOZ can bind other proteins and chromatin, acetylate histones, and modulate transcription. Our moz alleles are predicted to truncate the C terminus, causing loss of a transcriptional activation (TA) domain, but leaving the HAT domain intact. However, the severely reduced moz mRNA levels detected by in situ hybridization in mozb719 mutants makes it likely that the overall activity of the remaining protein would be greatly reduced. This reduced expression of moz in mozb719 mutants suggests that mozb719 mutant transcripts are unstable or that moz directly or indirectly regulates its own transcription.
Both moz lesions we present are loss of function lesions, as mutant phenotypes are phenocopied by morpholino injections. However, we cannot rule out the possibility that both alleles and morpholino phenotypes are all hypomorphic until deletion alleles are found. Although we have no evidence of a fish-specific duplication of moz, vertebrates do have a closely related gene, Morf (monocytic leukemia zinc finger protein related factor; Myst4 - Mouse Genome Informatics), which is also mutated in human leukemias (Champagne et al.,1999; Panagopoulos et al.,2001). An embryonic function has been reported for Morf(named Querkopf) in mice (Thomas et al.,2000). An insertion in the 5′UTR of mouse MORF causes skull and forebrain defects, but hindbrain or homeotic pharyngeal arch defects were not reported (Thomas et al.,2000). Whether MORF and MOZ have overlapping functions is unknown.
Biochemical analyses of human MOZ have also revealed multiple MOZ-interacting partners. MOZ physically interacts with RUNX1 (AML1) and RUNX2(AML3 or CBFA1) (Kitabayashi et al.,2001b; Pelletier et al.,2002). The RUNX2 interaction is particularly interesting, as this is the osteogenic gene shown in mice to be repressed by Hoxa2(Kanzler et al., 1998). Thus,in our analyses of segmental identity in the pharyngeal arches of mozmutants, we examined pharyngeal bones expecting to see ectopic bone formation. However, the hyoid bone pattern (Kimmel et al., 2003) was undetectable and duplicated mandibular bones were not seen in moz mutants (C.M., unpublished). Although we did not examine bone patterning in hox2-MO injected fish, one possibility is that MOZ functionally interacts with RUNX2 during pharyngeal development in zebrafish.
Whether mammalian MOZ regulates Hox gene expression, as we predict, awaits generation of moz mutant mice. The regulation of Hoxa2 by the transcription factor AP2 is conserved from mammals to fish(Maconochie et al., 1999; Knight et al., 2003). However,no reduction in AP2 expression was observed in moz mutants at 28 hpf(C.M., unpublished results). Thus, the regulation of hox2 genes by moz appears to act through another mechanism, possibly by directly transactivating Hox genes.
moz regulates segmental identity of pharyngeal cartilages
Our results extend the understanding that Hox genes specify segmental identity in the vertebrate pharynx, as the transformed pharyngeal segments in moz mutants correlate with reduced Hox expression in pharyngeal arch primordia. A mirror-image duplicated jaw replaces the hyoid cartilages in moz mutants, resembling the phenotype seen in animals co-injected with hoxa2b and hoxb2a morpholinos (C. T. Miller, PhD Thesis, University of Oregon, 2001)(Hunter and Prince, 2002). However, the moz mutant transformation is more complete, as dorsal transformations are more severe and dorsal fusions more common than in hox2-MO injected animals. It is likely that morpholinos cause incomplete loss-of-function at later developmental timepoints when the injected morpholino is significantly diluted. Alternatively, moz might regulate other genes that are also expressed in hyoid CNC and contribute to segmental identity.
Despite this stronger phenotype, the homeotic transformation in the hyoid arch of moz mutants is still not complete in that the pterygoid process of the palatoquadrate (PTP) is not seen duplicated. Perhaps in moz mutants, as has been proposed for mouse Hoxa2 mutants,only certain axial levels of CNC are transformed, i.e. perhaps PTP is derived from midbrain crest whose derivatives are not seen duplicated in the second arch of Hoxa2 mutants(Köntges and Lumsden,1996). de Beer (de Beer,1937) proposed that PTP was a premandibular element.
The pharyngeal arches are more sensitive to partial reduction of Hoxa2 function in the mouse(Ohnemus et al., 2001). The partial transformations observed in these hypomorphic mouse mutants led these authors to propose that homeosis was not an `all-or-nothing' phenomenon, as the second arch did not act as a developmental unit as a whole. Our results,in which a hypomorphic allele and low-level injections of Moz-MOs also separate particular homeotic phenotypes from others, strongly support this conclusion.
Experiments in Xenopus with an inducible Hoxa2 construct revealed that the time of Hoxa2 overexpression affected the resultant phenotype: overexpressing Hoxa2 early during CNC migration resulted in `segmentation' phenotypes, where arch derivatives were fused, while overexpressing late in postmigratory CNC resulted in `homeotic' phenotypes,where cartilage shapes were altered(Pasqualetti et al., 2000). As interarch fusions could also be interpreted as homeosis (i.e. loss of individual arch identity), this distinction is debatable. However, we note that the moz mutant phenotype contains more frequent shape changes(`homeotic') than fusions (`segmentation'), both dorsally and ventrally for both mutant alleles and for all three morpholinos at two different doses each. pbx4(lzr) mutants might display the converse phenotype, i.e. segmentation appears more affected than homeosis. Interarch fusions seen in moz mutants resemble the pbx4(lzr) mutant phenotype, although fusions are more severe in pbx4(lzr)mutants (Pöpperl et al.,2000). The more severe phenotype of pbx4(lzr)mutants might reflect differences in the set of affected target genes and/or temporal differences of target gene regulation (e.g. initiation versus maintenance).
moz mutants also present with mild anterior homeotic transformations of pharyngeal arches three and four (branchial or gill-bearing arches one and two). Arch three and four cartilages in moz mutants are slightly thicker, and typically contain an enlarged process on their lateral end, resembling the retroarticular process of Meckel's (the lower jaw)cartilage. Especially in the third arch of moz mutants at 5 days, an ectopic dorsal cartilage is also frequently seen. These transformations in the moz mutant anterior branchial arches are not seen in hox2-MO injected animals (Hunter and Prince,2002). These phenotypes probably result from additional Hox genes(e.g. Hox3 and Hox4 genes) that moz regulates (see above). Once genetic alleles of these zebrafish Hox genes are isolated, their function in specifying pharyngeal segmental identity can be assessed.
This third arch cartilage phenotype in moz mutants somewhat resembles the phenotype of valentino (val) mutants(Moens et al., 1998) (reviewed by Kimmel et al., 2001a),which was interpreted to be an ectopic interhyal cartilage based on ectopic hoxb2a expression in the third arch of val mutants. As neither hoxa2 nor hoxb2 is expressed in the mozmutant third arch, we propose that the moz mutant third arch has adopted mandibular fate. We did not detect ectopic bapx1 expression in third or fourth arch CNC of moz mutants. However, given the subtle nature of the skeletal change, the causative gene expression changes would probably be subtle as well.
Moz is required for Hox maintenance and behaves like a trithorax group factor
Consistent with the homeotic pharyngeal arch phenotype, we observe defects in hox1-4 gene expression in moz mutants. Three pieces of evidence suggest that Moz functions similarly to trithrorax (trxG) factors in regulating maintenance of Hox gene expression, as we discuss below.
First, we do not detect changes in initiation of Hox gene expression in the hindbrain of moz mutants. By later stages, graded reduction of most hox1-4 expression domains in the CNS domains is apparent. Defects in Hox gene maintenance, but not initiation, are hallmarks of trxmutants in Drosophila (Breen and Harte, 1993) and mouse (Yu et al., 1998). Loss of Hox group 1 or 2 gene function in zebrafish or in mice can cause severe homeotic neuronal transformations in the hindbrain and motor axon pathfinding defects(McClintock et al., 2002; Cooper et al., 2003; Studer et al., 1996; Gavalas et al., 1997; Gavalas et al., 1998; Gavalas et al., 2003; Rossel and Capecchi, 1999; Gendron-Maguire et al., 1993; Rijli et al., 1993). Supporting a role for moz in maintenance of Hox expression in the hindbrain, we find that early neuronal specification in the hindbrain and axonal trajectories in the head periphery are approximately normal in moz mutants. The only consistent neuronal defect we are able to detect in moz mutants is the disruption of facial motor neuron migration. This phenotype may be consistent with the defect in maintenance of hoxb1a expression in moz mutants, as loss of hoxb1ain zebrafish or Hoxb1 in mice causes a similar defect(McClintock et al., 2002; Studer et al., 1996). In the mouse Mll (trx) mutant, cranial ganglia are condensed and fail to innervate the pharyngeal arches(Yu et al., 1998), but more specific neuronal defects have not been reported and MLL mutant zebrafish have not been described.
Second, Moz has a HAT domain, which for human MOZ has been demonstrated to have HAT activity (Champagne et al.,2001), and HAT activity has been associated with trxG factors(Petruk et al., 2001; Milne et al., 2002). Furthermore, HAT activity is required for Moz function, as treatment with a histone deacetylase inhibitor rescues many aspects of the moz mutant phenotype.
Third, we find that moz mutant homeosis and Hox maintenance defects are rescued by TSA, and TSA has been shown to rescue defects caused by trxG mutations in Drosophila(Sollars et al., 2003) and human cells (Milne et al.,2002). Why does inhibition of histone deacetylase activity rescue a putative decrease of histone acetylation? It seems likely that the transcriptional on or off state of Hox genes is maintained through a balance of chromatin modification activities, including histone acetylation by trxG factors and histone deacetylation by Polycomb group(PcG) factors (Milne et al.,2002) (reviewed by Francis and Kingston, 2001; Simon and Tamkun, 2002). In support of this idea, trxG and PcG factors are antagonistic for proper Hox expression. For example,homeotic axial transformations and Hox expression defects of Mll-deficient mice and Bmi1-deficient mice are rescued when function of both genes is removed (Hanson et al., 1999; Yu et al.,1995; Yu et al.,1998). We do not yet know whether Moz directly acetylates histones associated with Hox regulatory regions. Western analyses show no detectable decrease of acetylated histone H4 levels in moz mutants compared with wild-type siblings (L.M., unpublished). We might expect to see rhombomere- or arch-specific defects in acetylated histone H4 levels at specific Hox genes,but at the present time this is very difficult to test.
Taken together, these findings implicate Moz as a trxG factor. trxG genes have been genetically defined as suppressors of PcG mutant phenotypes (reviewed by Kennison, 1995). Further studies demonstrating genetic interactions between moz and PcG genes would provide firm support for moz as a trxG gene.
One interesting aspect of Hox regulation that has emerged from our analysis of moz mutants is that in general, there appears to be a gradient effect of Moz activity within a Hox complex. We find that the hox1-4requirement for Moz activity ranges from strong for group 1 to weak for group 4, while group 5 and 6 Hox genes show no Moz requirement. We see this gradient effect on similar paralogs, where they have segmental domains in hindbrain and CNC, but we do not see a moz requirement for hoxa1aexpression, even though moz appears to be expressed ubiquitously throughout the head. Our findings suggest that Moz activity plays a global role in Hox locus regulation, possibly through HAT activity. MYST family HAT activity has been shown to have a chromosomal gradient of transcription control in yeast (Kimura et al.,2002). Whether a Moz-mediated gradient of histone acetylation exists across group 1-5 genes in Hox clusters remains to be determined.
moz is required for late but not early patterning of head musculature
Our data suggest that at times when severe hox2, bapx1 and gsc expression defects are present in postmigratory hyoid CNC of moz mutants, head mesodermal patterning appears unaffected. eng2, with eng3 the only segmentally restricted head mesodermal marker that we know of (Ekker et al., 1992; Hatta et al.,1990), appears appropriately confined to the first arch dorsal muscle core (constrictor dorsalis) of moz mutants. Likewise, the early myod expression pattern, which labels all proposed arch myogenic cores (see Kimmel et al.,2001b), appears normal in moz mutants. The absence of eng2 duplication or myod pattern disruption could be due to residual hox2 activity in moz mutants. Alternatively, moz and hox2 genes could play no role in restricting eng expression to the first arch or setting up the pattern of myod-expressing myogenic cores. The homeotic late muscle pattern seen in moz mutants perhaps results from transformed CNC-derived connective tissue, which has been shown to pattern paraxial-mesodermally derived and somitic-derived myocytes(Noden, 1983a; Noden, 1983b; Noden, 1986).
Dramatic changes in CNC expression of bapx1 and gsc prefigure the moz mutant phenotype
In stark contrast to the apparently normal early patterning of mozmutant mesoderm, endoderm and surface ectoderm, expression of two known hox2 target genes, bapx1 and goosecoid(gsc), is radically perturbed in postmigratory CNC of mozmutant second arch primordia.
Within postmigratory CNC, bapx1 expression is confined to a patch of intermediate first arch mesenchyme which appears to prefigure the jaw joint(Miller et al., 2003). We previously identified edn1 and hand2 (dHAND) as positive and negative regulators, respectively, of bapx1 expression. bapx1 expression spreads ventrally in hand2 mutants(Miller et al., 2003). We report that moz and hox2 genes also contribute to positioning bapx1 to the jaw joint, although these genes prevent bapx1 from being expressed in an intermediate domain of the hyoid arch. Thus, bapx1 integrates positional information from both the DV(edn1, hand2) and AP axes (moz, hox2) to achieve its jaw-joint-restricted expression. Furthermore, as an aspect of the mozmutant homeotic pattern (the jaw joint) requires bapx1, these results identify bapx1 as a crucial downstream effector contributing to the homeotic transformation.
Microarray comparisons of gene expression in the second arches of wild-type and Hoxa2 mutant mice revealed Pitx1 to be upregulated in the Hoxa2 mutant second pharyngeal arch primordial(Bobola et al., 2003), similar to what we report here for bapx1. These authors report finding no confirmed gene that is downregulated in Hoxa2 mutant second arches,and suggest that Hoxa2-mediated segmental identity in the second arch might largely involve repression of the first arch program. Although microarray analyses promise to provide a global view of overall changes in gene expression in Hoxa2 mutant arches, our demonstration of spatially shifted gsc expression highlights the need to also analyze potential spatial reorganization of affected genes.
In the mouse, gsc expression is spatially restricted within first and second arch CNC (Gaunt et al.,1993). Although gsc is required for specific aspects of mouse craniofacial development, defects in first, but not second, arch derivatives were reported (Rivera-Perez et al., 1995; Yamada et al.,1995). In both fruitflies and vertebrates, gsc functions as a transcriptional repressor (Danilov et al., 1998; Ferreiro et al.,1998; Mailhos et al.,1998; Latinkic and Smith,1999; Yao and Kessler,2001), although precedent exists for gsc positively regulating target genes (frzb)(Yasuo and Lemaire, 2001). The identity of gsc target genes and the nature of their regulation in the pharyngeal arches remains to be determined. bapx1 and gsc are expressed in strikingly complementary patterns in the first two arches (compare Fig. 9A-Cwith Fig. 10A-C), suggesting one might repress expression of the other. Our previous report that in hand2 mutant zebrafish ventral first arch expression of gscis lost, while bapx1 expression expands ectopically into this domain(Miller et al., 2003), is consistent with gsc repressing bapx1 arch expression.
The inverted gsc expression domain in the early second arch primordia of moz mutants suggests reorganization of the fate map at an early prechondrogenic stage has occurred. This model is consistent with our finding that moz mutant prechondrogenic condensations are mispositioned and the finding that in the mouse Hoxa2 mutant,chondrogenesis is induced in different regions of the arch than in wild types(Kanzler et al., 1998). This latter study additionally showed that transgenically driving Sox9expression in the Hoxa2 domain partially phenocopied the Hoxa2 mutant phenotype. Furthermore, transgenically driving Hoxa2 with an Msx2 promoter resulted in loss of cranial bones, suggesting Hoxa2 represses both cartilage and bone formation(Kanzler et al., 1998). Our bapx1 and gsc expression data suggests that moz and hox2 affect patterning within the second arch primordia long before cartilage or bone differentiation occurs.
Perhaps gsc expression labels chondrogenic cells and these cells shift medial in the early CNC cylinder. Alternatively, gsc expression could label non-chondrogenic cells. Although the first model is more consistent with the demonstrated cell-autonomous function of gsc in mice (Rivera-Perez et al.,1999), the latter is more consistent with the moz mutant phenotype, in which the lateral end of the duplicated lower jaw cartilage fuses laterally with the lower jaw near the jaw joint. In mice, although gsc was found to be cell autonomous, gsc-null cells in the presence of wild-type cells could contribute to the condensation of the tympanic bone, a bone that never forms in gsc mutants(Rivera-Perez et al., 1995; Rivera-Perez et al., 1999; Yamada et al., 1995). However, these gsc-null cells were not maintained(Rivera-Perez et al., 1999). Exogenous gsc can induce neighboring cells to form a secondary axis,suggesting in some contexts, Gsc can have non cell-autonomous functions(Cho et al., 1991; Niehrs et al., 1993).
A central mystery remaining is why the duplication in moz or hox2-deficient animals is mirror image. At the time the gscexpression defect appears, hoxa2 expression appears to mark all hyoid postmigratory CNC. Thus, the spatially complex gsc defect in hox2-injected animals is hard to reconcile with a model in which hox2genes simply positively regulate gsc. The gsc expression defect is also hard to reconcile with a model in which hox2 genes modify responsiveness of second arch CNC to a single cue emanating from the arch 1/2 boundary (Rijli et al.,1993). We propose a modified version of the model of Rijli et al.,in which hox2 modifies the responsiveness of hyoid CNC to multiple environmental signals. The mediolateral inversion of gsc could be explained if hox2 genes conferred responsiveness of second arch CNC to a lateral surface ectodermal signal cue to activate gsc while repressing responsiveness to a medial endodermal cue that normally repressed gsc expression. Continued forward genetic screens in zebrafish could reveal components of these putative signaling pathways that underlie segmental identity in the pharyngeal arches.
We thank Angel Amores, Bruce Draper, Andrew Waskiewicz and Cecilia Moens for providing PAC DNA pools; Shin-ichi Higashijima and Hitoshi Okamoto for providing the Islet1:GFP strain; Vicky Prince and Michael Hunter for sharing data prior to publication; Vicky Prince for Hox probes; Tom Schilling and David Stock for insightful suggestions; the UO zebrafish community,particularly those involved in the mutagenesis screens; past members of the Kimmel laboratory, particularly Bruce Draper, Cecilia Moens and Sharon Amacher; Charlene Walker for help in the beginning phases, and Macie Walker and Gage Crump for help in later phases of the Alcian screen; Eric Selker and Bea Darimont for advice; and Cecilia Moens for comments on an earlier version of the manuscript. C.T.M. was supported by an NSF graduate research fellowship, L.M. was supported by a Damon Runyon Cancer Research Fund postdoctoral fellowship, and this research was supported by NIH DE13834 and HD22486 to C.B.K.