Craniofacial development is the result of carefully regulated interactions of numerous tissues, including neural crest cells (NCCs), paraxial mesoderm, and pharyngeal endoderm (for review, see Yelick and Schilling, 2002). Discrete NCC populations contribute to cartilage, pigment, neuronal, and glial cell fates, whereas paraxial mesoderm contributes to muscle and endothelia. The neurocranium is derived from both cranial neural crest (CNC) and mesoderm, whereas the pharyngeal skeleton, including the jaw and branchial arches (Fig. 1), is derived solely from cranial NCCs. NCCs exhibit segmental patterning based on their anterior–posterior (AP) location, and migrate in three discrete streams from dorsal positions along the neural/non-neural border, into the ventrally located pharyngeal pouches. The first and second migrating NCC streams contribute to first arch ventral Meckel's and dorsal palatoquadrate cartilages and second arch ventral ceratohyal and dorsal hyosymplectic cartilages. The third migrating NCC stream splits to form posterior arch ceratobranchial 1–5 (cb1–5) cartilages and pharyngeal teeth present on the cb5 arch. Signals emanating from the pharyngeal endoderm lining each pharyngeal pouch direct the differentiation, growth, and patterning of adjacent pharyngeal arch cartilage elements (Crump et al., 2004).
The transforming growth factor-beta (TGFβ) super family is an extensive group of structurally related, secreted growth factors that exhibit essential roles in a broad range of developmental processes, including AP and dorsoventral (DV) axis patterning, tissue differentiation, cell proliferation, and apoptosis (Massagué, 1998; Waite and Eng, 2003). TGFβ family members are further classified into activin, bone morphogenetic protein (Bmp), TGFβ, and nodal related subfamilies (for review, see Massagué, 1998). TGFβ family member signaling complexes consist of ligand and type I and II receptor dimers (Wrana et al., 1994), and their signaling pathway partners exhibit extensive conservation between invertebrates and vertebrates, suggesting highly conserved evolutionary functions for these growth factors (Liu, 2003).
Characterizations of activin signaling in developmental model systems have revealed functions both in early embryonic patterning and in later tissue-specific differentiation events. In the chick, activin (acv) and activin receptor (acvr) genes are first expressed along the primitive streak, consistent with functions in early mesoderm induction (Stern et al., 1995). At later stages, acvr2a expression is restricted to the notochord and myotome, whereas acvr2b mRNAs are restricted to the dorsal neural tube, proximal–anterior part of the limb bud, sensory placodes, and specific regions of the forebrain and midbrain (Stern et al., 1995). Activins can induce both mesodermal and neural tissues in animal cap explants from Xenopus embryos and have been shown to regulate both DV and AP patterning of the amphibian embryo (Sokol and Melton, 1991; Green et al., 1994, 1997; Nagaso et al., 1999).
Activin type I and II receptors exhibit fairly promiscuous binding properties, similar to other TGFβ family member receptors, and have been demonstrated to bind in vitro to activins (Donaldson et al., 1992; Matthews and Vale, 1993; ten Dyke et al., 1996), Bmp7 (Greenwald et al., 2003), and to proteins such as Crypto that modulate activin receptor signaling (Gray et al., 2003). The binding partners of zebrafish Acvr2a and Acvr2b have not been identified to date, although craniofacial expressed TGFβ family member ligands, including Contact (Bruneau et al., 1997), are potential Acvr2a and Acvr2b binding partners. Activin signaling has also been implicated in apoptosis, where extensive in vitro studies support roles for TGFβ family members as inducers of apoptosis (Chen et al., 2002, and references therein; Schuster and Krieglstein, 2002), through activation of bax (Fukuchi et al., 2001) and bcl-Xs (Koseki et al., 1998). The bax and bcl-X gene products normally repress the caspase inhibitor bcl2 (Koseki et al., 1998), resulting in increased caspase activity and subsequent DNA fragmentation (Fukuchi et al., 2001). Overexpression of activin decreases tumorigenicity in human prostate cancer LNCaP cells (Zhang et al., 1997), underscoring the importance of activins in cancer research.
Roles for activin signaling in craniofacial development were demonstrated using targeted mutagenesis approaches in mice. These studies revealed that mice lacking activinβA (acvβA) and activinβB (acvβB) exhibited discrete craniofacial defects, including the absence of whiskers and lower incisors and clefting of the secondary palate (Matzuk et al., 1995a, b). Mice lacking the acvr2a gene exhibited several defects observed in activin-deficient mice, including cleft palate and the absence of incisors, and also exhibited unique mandibular defects not present in acv knockout mice. These results are consistent with the observed expression of acvr1 and acvr2 during craniofacial intramembranous and endochondral bone formation in rats (Shuto et al., 1997) and also suggest that Acvr2 can serve as a receptor for ligands other than activins.
Activin is expressed in the dental mesenchyme of early bud stage teeth (Ferguson et al., 1998), and phenotypes of acvβA knock out mice reveal spatial and temporally restricted roles for acvβA in tooth development (Ferguson et al., 1998). AcvβA mutant mice exhibit incisors and mandibular molar teeth that fail to develop beyond the bud stage, whereas maxillary molars develop normally, suggesting that their development is activin-independent. Tissue recombination experiments demonstrated that acvβA mutant dental epithelium can induce tooth development, and treatment of acvβA null molar tooth primordia with exogenous ActivinβA completely rescued tooth development in E11.5 but not in E12.5 stage mouse embryos, suggesting that activin signaling is required before bud formation. Similar tooth phenotypes were observed in acvr2 and Smad2 mouse mutants, consistent with the defined roles of these genes in activin signaling pathways (Ferguson et al., 2001).
Here, we have explored the roles of acvr2 genes in zebrafish craniofacial development. The expression of acvr2a and acvr2b suggests distinct roles in AP patterning of the craniofacial complex. Targeted protein depletion studies also support roles for each acvr2 gene in AP patterning. In all, our results reveal previously uncharacterized and distinct developmental roles for acvr2a and acvr2b in neurepithelial and NCC patterning, cranial NCC migration, apoptosis, pharyngeal arch cartilage articulation and joint formation, skeletogenesis, and pharyngeal tooth development.
Characterization of Zebrafish acvr2a and acvr2b
Degenerate reverse transcriptase-polymerase chain reaction (RT-PCR) and cDNA library screening (Sagerstrom et al., 2001) was used to identify zebrafish acvr2a and acvr2b cDNAs: 4 overlapping cDNAs for acvr2a (1 full-length and 3 partial) and 11 overlapping clones for acvr2b (4 full-length and 7 partial) were isolated. Zebrafish acvr2a exhibited 70 % and 71% nucleotide, and 79% and 79% amino acid sequence identities to human and mouse acvr2a genes, respectively, whereas zebrafish acvr2b exhibited nucleotide sequence identities of 83% and 81%, and amino acid sequence identities of 72% and 72%, to human and mouse acvr2b genes, respectively. Based on these identities, the isolated zebrafish cDNAs were called zebrafish acvr2a and acvr2b.
Developmental Expression of acvr2a and acvr2b
The developmental expression of zebrafish acvr2a and acvr2b was examined by whole-mount in situ hybridization (WISH) analyses (Fig. 2a), which revealed that both genes are ubiquitously expressed as maternal mRNAs (Fig. 2a, A and G). Ubiquitous expression was observed until approximately 24 hours postfertilization (hpf) when both genes became predominantly restricted to developing neural tissues: acvr2a is expressed throughout neural tissues (Fig. 2a, C), while acvr2b is expressed specifically in the midbrain–hindbrain boundary (mhb), and hindbrain tissues (Fig. 2a, I). At 48 hpf, acvr2a expression remains ubiquitous in the head (Fig. 2a, D), while acvr2b exhibited discrete expression in the hindbrain (Fig. 2a, I). By 72 hpf, expression of both genes is down-regulated in the developing head (data not shown). At 96 hpf, transcripts for both receptors localized to the maxillary processes (Fig. 2a, E and K), oral epithelium (Fig. 2a, F and L), developing gut (not shown), and tail bud (not shown).
RT-PCR was used to compare the developmental expression of zebrafish acvr2a and acvr2b with that of previously identified receptor genes alk8 and tgfβr2 (Fig. 2b; Yelick and Stashenko, 1996; Yelick et al., 1998; de Caestecker et al., 2002). RT-PCR products were quantitated by their integrated density value (IDV), the absolute sum of pixilated area of each RT-PCR product, as generated from digitally imaged ethidium bromide (EtBr) -stained RT-PCR product normalized to actin controls. Zebrafish acvr2a and acvr2b expression levels closely mirrored one another, increasing slightly until shield stage and remaining at fairly constant levels until 5 days postfertilization (dpf). In contrast, tgfbr2 expression decreased from maternal to tail bud stage, slightly increased to the 15-somite stage, and was maintained at that level until 5 dpf. The expression of alk8 closely paralleled that of tgfbr2, consistent with previously demonstrated interaction of these two receptors (de Caestecker et al., 2002).
Genetic Mapping of the Zebrafish acvr2a and acvr2b Genes
We previously mapped zebrafish acvr2a and acvr2b to linkage groups (LG) 9 and 24, respectively (Postlethwait et al., 1998). We further examined the orthologous relationships of acvr2 genes by comparing their genomic contexts in zebrafish and human (Fig. 3). Our results demonstrated that acvr2a is embedded in a very large conserved synteny previously noted in zebrafish, where LG9 is an ortholog of the long arm of human chromosome 2 (Hsa2q; Postlethwait et al., 1998, 2000; Gates et al., 1999; Barbazuk et al., 2000; Woods et al., 2000). Zebrafish acvr2a is on ctg9401, which contains several other loci from Hsa2q, and from Hsa13q. A sequence called DKFZP566F2124 in human is three genes distal to ACVR2A in Hsa2q, and an ortholog of this is adjacent to acvr2a in ctg9401 (Fig. 3a). A chromosome rearrangement near the zebrafish ortholog of DKFZP566F2124 appears to have disrupted conserved syntenies between human and zebrafish. Conserved syntenies are not as dramatic for acvr2b. Besides acvr2b, LG24 contains an ortholog of NKTR, which in human is quite close to ACVR2B on Hsa3p (Fig. 3b). The acvr2b gene is present on the zebrafish genomic fragment NA7275, with no other identified loci. Altogether, these conserved synteny results are consistent with the assigned orthologies of the zebrafish acvr2 genes and suggest that the chromosome segments they inhabit were inherited from the last common ancestor of zebrafish and human 450 million years ago.
Morpholino-Based Targeted Knockdown of acvr2a and acvr2b
Targeted depletion of Acvr2a and Acvr2b results in discrete embryonic patterning defects.
The zebrafish craniofacial complex consists of distinct cartilage, skeletal, and pharyngeal tooth elements (Figs. 1, 4a). The functions of Acvr2a and Acvr2b in the development of craniofacial structures were examined using a targeted protein depletion strategy (Nasevicius and Ekker, 2000). Antisense morpholino oligomers (MOs) targeted to the ATG start sites of acvr2a and acvr2b mRNAs were used to generate embryos depleted of Acvr2a and Acvr2b. Injection of MOacvr2a at 4 mg/ml resulted in an early lethal phenotype at approximately 4 dpf (not shown), whereas injections at 2 mg/ml resulted in embryos that survived until approximately 7 dpf and exhibited distinct defects including a shortened AP body axis, eye swelling, and mediolaterally displaced eyes, relative to mock-injected control embryos (Fig. 4a, A vs. K). Abnormal fusion of maxillary processes was also observed in acvr2a morphants (Fig. 4a, A, arrow), consistent with the fact that, when the normal expression of acvr2a within the maxillary processes (Fig. 2a, E) is disrupted, aberrant development of the maxilla ensues. Similarly, MOacvr2b injections at 2 mg/ml resulted in embryos that appeared shortened along the AP axis, and which exhibited eye swelling and mediolaterally expanded eyes and pharyngeal arches (Fig. 4a, F). The acvr2b morphants also exhibited abnormal maxillary fusion (Fig. 4a, F, arrow), which is again consistent with the normally observed expression of acvr2b within the maxillary processes (Fig. 2a, K). In general, the phenotypes of acvr2b morphants appeared less severe than those of acvr2a morphants, regardless of the concentration of injected antisense MO.
The specificity of each morphant phenotype was confirmed by coinjection of wild-type acvr2a and acvr2b mRNAs (Fig. 4b). Coinjection of wild-type acvr2a mRNAs at 200 and 300 ng/μl (1- to 3-nl injection volume) resulted in rescue of 16% and 67%, respectively. Coinjection of wild-type acvr2b mRNAs at 200 ng/μl resulted in the rescue of 71% of the acvr2b MO phenotype. Injection of wild-type acvr2b mRNAs at 300 ng/μl resulted in an acvr2b overexpression phenotype in greater than 50% of the injected embryos (data not shown). The detection of Acvr2a and Acvr2b protein in wild-type and morphant embryos is not possible at this time due to the absence of suitable antibodies.
Pharyngeal arch cartilage defects.
Alcian blue staining was used to reveal pharyngeal arch cartilage patterning defects in Acvr2a- and Acvr2b-depleted embryos (Fig. 4a, B, G, compared with L). Embryos injected with 4 mg/ml of antisense acvr2a MO exhibited the complete absence of all cranial NCC-derived cartilages, and early lethal phenotypes, by 4–5 dpf (data not shown). In contrast, embryos injected with 2 mg/ml of antisense acvr2a MO lived to approximately 7 dpf, and exhibited distinct anterior and posterior pharyngeal arch cartilage defects (Fig. 4a, B). The first and second arch cartilages of acvr2a morphants appeared disorganized and either did not extend to or were fused at articulation points. Fusions between dorsal and ventral first arch cartilages (palatoquadrate and Meckel's) and between dorsal cartilages of the first and second arches (the palatoquadrate and hyosymplectic, respectively) were commonly observed (Fig. 4a, B and E). Another common feature of acvr2a morphants is the absence of ceratobranchial (cb) arch cartilages 2, 3, and 4 (Fig. 4a, B and E, red bracket), with only the first and fifth cb cartilages commonly present albeit often in very truncated form. Together, these analyses revealed roles for acvr2a in the patterning and segmentation of both anterior and posterior pharyngeal arch cartilages.
Pharyngeal arch cartilage defects were also evident in acvr2b morphants (Fig. 4a, G and J). Whereas the anterior first and second arch cartilages exhibited fairly normal patterning, posterior arch cartilages were either reduced or missing in acvr2b morphants, in a graded manner from least severely affected cb1 to most severely affected cb4 arch cartilages (Fig. 4a, G and J). Cb5 arch cartilages appeared truncated in acvr2b morphants but not as severely as those of acvr2a morphants. Embryos injected with high MOacvr2b concentrations (4 mg/ml) still possessed well-formed and well-articulated first and second arch cartilages, whereas the posterior arch cb1–5 exhibited severe defects (not shown), consistent with prominent roles for acvr2b in the proper development of posterior but not anterior pharyngeal arch cartilages.
Bone and teeth defects in acvr2a and acvr2b morphants.
By 7 dpf, wild-type zebrafish possess five dermal and five endochondral bones, and exhibit three well-formed pharyngeal teeth on each cb5 cartilage (Fig. 4a, M). In contrast, 7 dpf acvr2a morphants possessed only four dermal bones, had no endochondral bone, and only 0–1 teeth (Fig. 4a, C, D, E). The dermal bones present in acvr2a morphants, the dentary, maxilla, opercle, and the posterior branchiostegal, appeared small and misshapen, and the anterior branchiostegal was completely missing (Fig. 4a, D and E). Of 25 MOacvr2a-injected embryos, 11 exhibited the first pharyngeal tooth, 4V1, which was usually malformed, and 14 morphants had no teeth at all. These observations suggest that acvr2a is required for dermal bone growth, for the initiation of endochondral bone development, and for normal tooth development.
Acvr2b depleted larvae exhibited bone phenotypes quite similar in severity to those of acvr2a morphants (Fig. 4a, I and J). Acvr2b morphants possessed four dermal bones—the dentary, maxilla, opercle, and posterior branchiostegal—and were missing the anterior branchiostegal, as was observed in acvr2a morphants. Also similar to acvr2a morphants, no cartilage replacement bones were present, suggesting roles for both acvr2 genes in endochondral bone development. A notable difference was that Acvr2b-deficient larvae possessed more teeth than acvr2a morphants. Of 28 acvr2b morphants examined at 7 dpf, 13 exhibited only one tooth (4V1), whereas 15 possessed three, albeit slightly malformed, pharyngeal teeth (4V1, 3V1, and 5V1), as normally present in wild-type embryos (Fig. 4a, H and J vs. M and O). Neither acvr2 receptor mRNA was detectable at any stage of tooth development, suggesting the observed tooth defects are secondary.
Molecular characterization of early embryonic defects of acvr2 morphants.
To identify early patterning defects that may contribute to the later observed developmental defects, WISH analyses were performed on acvr2a and acvr2b morphants and wild-type controls to examine pax2a, krox20, and dlx2 expression (Fig. 5). Analysis of pax2a expression revealed AP compression and medial–lateral expansion of neural tissues in acvr2a morphants, relative to wild-type controls (Fig. 5B vs. A). Pax2a was almost completely absent from the putative optic nerve and spinal cord neurons of acvr2a morphants and was severely reduced in the mhb, suggesting a requirement for acvr2a in the development of these neural tissues. Pax2a expression appeared robust in the otic vesicle and pronephric duct, although pronephric duct tissues appeared shortened along the AP axis. WISH analysis of krox20 revealed poorly defined and mediolaterally expanded hindbrain rhombomeres 3 and 5 (r3, r5) in acvr2a morphants relative to control embryos (Fig. 5E vs. D). The mediolateral expansion of hindbrain suggests that both acvr2a and acvr2b MO embryos exhibit dorsalized phenotypes. Furthermore, krox20-positive CNC streams were not present at r5, as observed in wild-type embryos. WISH analysis of dlx2 revealed aberrant CNC cell development in acvr2a morphants compared with age-matched wild-type siblings (Fig. 5H vs. G), characterized by fused and reduced CNC masses within arches 1, 2, and 3, 4. These defects are consistent with hindbrain segmentation defects in acvr2a morphants.
Whereas targeted depletion of Acvr2a affected both anterior and posterior arch structures, Acvr2b morphants exhibited defects largely restricted to posterior arch tissues. MOacvr2b-injected embryos exhibited only slightly reduced pax2a expression in anterior optic stalk, mhb, and otic vesicle tissues, which otherwise appeared fairly well-patterned (Fig. 5C vs. A). In contrast, WISH analysis of krox20 revealed hindbrain defects including mediolateral expansion, AP shortening, and abnormal chevron-shaped segmentation of acvr2b morphant hindbrain compared with wild-type controls (Fig. 5F vs. D). Analysis of dlx2 expression revealed well-defined anterior first and second CNC segments, whereas posterior segments 3 and 4 appeared disorganized in acvr2b morphants (Fig. 5I vs. G).
Embryos depleted of Acvr2a and Acvr2b exhibit increased apoptosis.
Because activins and their receptors have been demonstrated to play significant roles in apoptosis (reviewed in Chen et al., 2002), we used the terminal transferase-mediated dUT nick end-labeling (TUNEL) method (Yager et al., 1997) to reveal critical roles for both acvr2 genes in cell survival (Fig. 6). Both 24 and 48 (not shown) hpf acvr2a morphants exhibited increased apoptosis throughout neural tissues (Fig. 6D, H), while acvr2b morphants exhibited apoptotic cell populations largely restricted to dorsal neural fold tissue (Fig. 6C, G). Control morpholino-injected embryos exhibited apoptotic cell populations equivalent to age-matched wild-type control embryos (Fig. 6B, F vs. A, E), demonstrating that morpholino oligomer injection alone did not result in increased apoptosis. To quantitate apoptosis in acvr2a and acvr2b morphants, TUNEL-positive cells were counted in discrete regions of the midbrain and hindbrain (white boxed regions in Fig. 6A). Both acvr2a and acvr2b morphant embryos exhibited increased apoptosis in the midbrain and hindbrain, compared with age-matched control injected embryos (Fig. 6I, J). However, patterns of apoptosis were distinct between acvr2a and acvr2b morphants. The acvr2a morphants exhibited a ubiquitous punctate pattern of apoptosis throughout the midbrain and hindbrain, whereas acvr2b morphants exhibited apoptotic cell populations predominantly in the most dorsal aspects of hindbrain neural folds (Fig. 6C vs. A, B, and D). There was no obvious difference in the level of apoptosis in the anterior versus posterior hindbrain of acvr2a or acvr2b morphants. The observed patterns of apoptosis closely correlated with the expression patterns of acvr2a and acvr2b mRNAs (Fig. 2a), consistent with roles for these genes in cell survival, may account for reduced CNC cell masses observed in embryos depleted of Acvr2a and Acvr2b.
Recent reports have characterized the roles of the growth factor receptor genes including alk8 (Yelick et al., 1998; Payne et al., 2001), Bmpr1a and -1b (Ashique et al., 2002), et(a)r (Clouthier et al., 1998, 2000; Kempf et al., 1998), fgfr1 (Trokovic et al., 2003), and ryk (Halford et al., 2000) in vertebrate craniofacial development. Several of these have been implicated in human craniofacial disorders, including DiGeorge/velocardiofacial syndrome (Clouthier et al., 1998, 2000) and various craniosynostosis and chondrodysplasia syndromes (Bachler and Neubuser, 2001; Britto et al., 2001). Characterizations of receptor functions in early embryonic development can prove particularly difficult to interpret, due to the fact that they are often ubiquitously and transiently expressed, have multiple ligands, and can exhibit global and early lethal defects when disrupted. Nonetheless, these studies are critically important, because receptor mutants do not always phenocopy putative cognate ligand mutations and, therefore, can reveal previously unknown functions. In this report, we reveal distinct roles for zebrafish acvr2a and acvr2b in patterning and development of the neurectoderm, CNC, cartilage, bone, and teeth. We also describe extensive synteny of zebrafish and human acvr2 genes, which suggests conserved roles for these genes in vertebrate craniofacial patterning.
Acvr2a and Acvr2b Exhibit Discrete Functions in Craniofacial Patterning
Zebrafish acvr2a and acvr2b mRNAs exhibit similar but distinct developmental expression patterns, and acvr2a and acvr2b MOs exhibit similar but distinct phenotypes. Consistent with broader expression domains, Acvr2a MOs exhibit more severe defects than Acvr2b depleted embryos, including the absence of, or severe reduction in, the CNC and CNC-derived pharyngeal arch cartilages. In contrast, acvr2b MOs exhibit defects largely restricted to posterior arch cartilages (Fig. 4). Similar phenotypes were observed in targeted mutagenesis of acvr2 genes in the mouse, suggesting evolutionarily conserved roles in craniofacial development (Ferguson et al., 2001). The dorsalized phenotypes of acvr2a and acvr2b MOs, as revealed by mediolaterally expanded krox20 expression in rhombomeres 3 and 5, resembles that of dorsalized DN alk8 mRNA-injected embryos (Payne et al., 2003) and is consistent with known roles for acvr2 genes in dorsoventral patterning (Sokol and Melton, 1991; Green et al., 1994, 1997). We found that acvr2a mRNAs are expressed throughout neurectodermal tissues, and are required to maintain cell viability, and proper patterning of CNC and CNC derived pharyngeal arch cartilages, in both anterior and posterior arches. In contrast, acvr2b mRNAs are restricted to the hindbrain, and are required to inhibit apoptosis in discrete cell populations of hindbrain ventricles, and for posterior CNC and pharyngeal arch cartilage patterning and development.
Of interest, these studies suggest possible roles for Acvr2a in joint formation. Over half of all MOacvr2a-injected embryos (16 of 25) exhibited skeletal cartilage elements that did not articulate, or the fusion of at least one pair of first or second arch cartilages. The acvr2a MO-fused joint phenotype resembles that of the zebrafish et-1/suc mutant (Miller et al., 2000), with notable differences. Whereas et1/suc mutant defects were restricted to ventral arch cartilages (Miller et al., 2000; Kimmel et al., 2003), acvr2a MO defects affected both ventral and dorsal arch cartilage elements. These results suggest that Acvr2a may mediate signaling by joint expressed genes, including gdf5 (Storm and Kingsley, 1996), nog (Brunet et al., 1998), and/or bapx1 (Miller et al., 2003). Acvr2a-mediated articulation defects might also result from early pharyngeal arch patterning defects. For example, AP compressed and fused cartilages observed in MOacvr2a-injected embryos were consistent with early patterning defects, including AP compacted rhombomeric segments and fused CNC segments. Because acvr2a mRNAs were ubiquitously expressed in the developing head until 60 hpf, and are not specifically expressed in the pharyngeal endoderm, CNC, or developing joints, it is difficult to assess the developmental origin of these defects at this time. Further investigations with joint- and arch-expressed genes are required.
It is notable that the fused pharyngeal cartilage defects were largely asymmetric—one side was typically more severely affected than the other. This phenotype resembles Treacher Collins syndrome and craniosynostosis in humans, where physical constraints imposed by aberrant or premature fusion of craniofacial structures on one side result in compensatory malformations on the opposing side (Francis-West et al., 2003). This asymmetric pattern is also consistent with the well-characterized role for type II activin receptors in left–right patterning of the vertebrate embryo (Oh et al., 2002). Elucidation of the molecular signals regulating asymmetric craniofacial growth and development will be extremely important for the development of clinically relevant therapies for these types of human craniofacial abnormalities.
Zebrafish Acvr2a and Acvr2b Are Required for Endochondral Bone Development
Several reports suggest roles for activins and activin receptors in bone development and repair. ActivinβA has been demonstrated to stimulate osteoblast proliferation (Hashimoto et al., 1992), enhance synthesis of bone matrix proteins (Oue et al., 1994), and induce ectopic bone formation (Ogawa et al., 1992). Both acvβA and acvr2a null mice exhibit craniofacial defects, including cleft palate and mandibular hypoplasia (Matzuk et al., 1995a, b). Our studies reveal similar roles for acvr2 genes in zebrafish endochondral bone development. By 7dpf, wild-type zebrafish exhibit a complex bony skeleton consisting of both intramembranous and endochondral bones (Fig. 1), with ossification limited to the mandibular and hyoid arches, and fifth arch cleithrum and tooth bearing cb5 (Cubbage and Mabee, 1996). Dermal cranial bones present in 7dpf zebrafish include the dentary, maxilla, opercle, two branchiostegal rays, and the cleithrum, the dermal progenitor of the pectoral girdle. Endochondral bones include the quadrate, retroarticular, hyosymplectic, ceratohyal, and cb5.
While acvr2a and acvr2b MOs exhibited distinct cartilage defects, acvr2 MOs exhibit similar defects in bone formation, suggesting overlapping requirements for the acvr2 genes in craniofacial skeletogenesis. Both acvr2 MOs formed dermal bones, but they appeared small and misshapen. In contrast, neither MO developed endochondral bone. Because dermal bone formation slightly precedes that of cartilage replacement bone (Cubbage and Mabee, 1996), it is possible that the absence of endochondral bone formation reflects a developmental delay rather than a direct effect on endochondral bone formation. We find this unlikely, however, because no endochondral bone formation was observed, even in the less-affected first and second arches at 7–8 dpf. Moreover, endochondral ossification of the cb5 arch was rarely observed, even when the cb5 cartilage and pharyngeal teeth were present, and when it did occur, as observed (in 12 of 53 cases, ∼23%), it was limited to the site of tooth attachment. These data are consistent with reports of type II activin receptor expression in osteoblasts initiating endochondral bone formation in rats (Shuto et al., 1997).
Acvr2a and Acvr2b Depletion Results in Maxillary Defects
By 96 hpf, both acvr2a and acvr2b mRNAs exhibited discrete expression within the maxillary processes (Fig. 2a, E and K). When Acvr2a or Acvr2b function is abrogated, the maxillary processes do not fuse (Fig. 4a, A and F). This defect is similar to that observed in activinβA knockout mice, in which 33% exhibited a cleft secondary palate (Matzuk et al., 1995a). In addition, skeletal preparations of the activinβA knockout mice not exhibiting a cleft palate revealed that most either lacked or exhibited incomplete development of the hard palate. Similarly, acvr2 knockout mice exhibited cleft palates (Matzuk et al., 1995b). In mammals, the primary constituent of the secondary palate is the palatine bone, and a secondary palate cleft occurs when the right and left palatine bones do not fuse. Although there are no palatine bones in teleost fishes, a robust argument could be made that the pterygoid process of the palatoquadrate is an analogous structure. Like the palatine in mammals, the pterygoid process develops laterally and grows medially to articulate with the ethmoid plate. Acvr2a- and Acvr2b-depleted zebrafish tended to have short pterygoid processes that did not articulate with the ethmoid. Unfortunately, the premaxilla does not form by 8dpf in zebrafish, precluding detailed analyses of primary palate defects in acvr2a and acvr2b morphants. Taken together though, these data argue that both acvr2 genes are required for the proper development of midline structures in the zebrafish head.
Zebrafish acvr2 Genes and Pharyngeal Tooth Development
Our studies demonstrate that Acvr2a-deficient larvae exhibit a more severe tooth phenotype than those lacking Acvr2b. Only one tooth (4V1) is present in 7 dpf acvr2a MOs, and it appears abnormal (Fig. 4A,C), whereas three teeth (3V1, 4V1, 5V1) are present in age-matched wild-type siblings (Fig. 4a, M). In contrast, 7 dpf acvr2b MOs exhibited up to three teeth, with 4V1 appearing fairly normal, whereas 3V1 and 5V1 appeared malformed (Fig. 4a, H). WISH analyses demonstrated that acvr2a and acvr2b mRNAs themselves are not detectable in developing pharyngeal tooth tissues. These results suggest that downstream acvr2a and acvr2b signaling partners may play significant roles in the observed pharyngeal tooth defects. Further molecular characterization of tooth development in wild-type and acvr2a and acvr2b MOs, and the identification of zebrafish acvr2a and acvr2b mutants, will likely facilitate our understanding of the roles for these receptors in the initiation and development of both primary and adult replacement teeth.
Our analyses of zebrafish acvr2a and acvr2b MOs suggest critical and distinct roles for each acvr2 gene in various aspects of craniofacial development, including neurepithelial and hindbrain patterning, CNC specification and segmentation, and in the regulation of apoptosis. Our studies reveal that regionally restricted early developmental defects are subsequently reflected in later craniofacial phenotypes. We conclude that acvr2a and acvr2b are critical regulators of craniofacial patterning and development. The acvr2a gene acts more globally to pattern both anterior and posterior arch structures, while acvr2b acts in a more restricted manner to pattern posterior arch cartilages. Ongoing characterization of acvr2 signaling, including the identification of type I receptor, ligand, and downstream signaling partners, will provide a more detailed understanding of the roles of acvr2 genes in craniofacial development. The extensive genomic synteny at the acvr2a locus suggests that regulatory elements also may have been conserved, making studies in zebrafish particularly relevant to human craniofacial development.
Fish Husbandry and Collection of Embryos
Zebrafish husbandry and embryo production were performed as described (Westerfield, 1995).
Isolation of Full-Length Zebrafish acvr2a and acvr2b cDNA Clones
Full-length cDNAs were isolated, as previously described (Yelick et al., 1998), by RT-PCR using degenerate primer pairs to conserved serine/threonine kinase domains of type I and II TGFβ family member receptor genes. RT-PCR products exhibiting close nucleotide sequence identity to identified acvr2a and acvr2b genes were used to screen a zebrafish gastrula staged cDNA library. One full-length and three partial acvr2a clones, and four full-length and seven partial acvr2b cDNA clones were identified.
The acvr2a and acvr2b genes were mapped by single-strand conformation polymorphism on the HSP doubled-haploid mapping panel as described (Kelly et al., 2000; Postlethwait et al., 2000; Woods et al., 2000). PCR primers were designed to amplify 3′-untranslated regions (UTRs) of the two genes: for acvr2a, F-CCA GAG GAA ATA GTC ACC and R-GCT TTC AGA GGA AAC ACG; and for acvr2b, F-GGA GCG CAT CTC TCA GAT and R-GTT CCT GAA CAC CCG TTA G. Orthologies were determined by the reciprocal best blast method (Altschul et al., 1997; Hirsh and Fraser, 2001). Human gene locations were obtained from LocusLink (http://www.ncbi.nlm.nih.gov/LocusLink/). Contigs containing acvr2 genes were identified in the Sanger Center zebrafish genome sequencing database (http://www.sanger.ac.uk/Projects/D_rerio/).
MO Design and Injections
Antisense MOs (Gene Tools, LLC, Philomath, OR) targeted to the translational start site of acvr2a and acvr2b were as follows: MOacvr2a, 5′-GCAGGTCCCATTTTTTCACTCTTCT-3′; and MOacvr2b, 5′-GCAGAGAAGCGAACATATTCCTTT-3′; the antisense ATG start site is underlined. The MO standard control was used as a negative control for the injection procedure (Gene Tools, LLC Philomath, OR). Nacre embryos (Lister et al., 1999) were injected (approximately 1- to 3-nl injection volumes) at the yolk/blastoderm interface of one- to two-cell stage embryos at concentrations of 0.5, 1.0, 2.0, and 4.0 mg/ml in morpholino buffer (120 mM KCl, 20 mM HEPES-NaOH pH 7.5, 0.25% phenol red). Subtle but consistent phenotypes were observed using, in embryos injected at 2.0 mg/ml, the concentration subsequently used to characterize cartilage, bone, and tooth phenotypes in 5–8 days postfertilization (dpf) embryos. More severe phenotypes were observed in embryos injected at 4.0 mg/ml, the concentration used to characterize early patterning defects up to 48 hpf. Coinjection of acvr2a and acvr2b MOs resulted in additive effects, which are not presented here. The specificity of each MO phenotype was confirmed by rescue with coinjected wild-type acvr2a and acvr2b mRNAs, respectively.
WISH analysis was performed as described (Thisse et al., 1993), using digoxigenin-labeled antisense riboprobes synthesized from the following cDNAs: pax2a (Brand et al., 1996), krox20 (Oxtoby and Jowett, 1993), and dlx2 (Akimenko et al., 1994).
Apoptotic cells were identified in acvr2a and acvr2b morphant embryos, and in control MO-injected embryos, using the Apoptag Kit (Chemicon Corp, Temecula, CA), per manufacturer's instructions.
Cartilages of larval fish aged 4–7 dpf were stained with Alcian blue using a protocol adapted from Potthoff (1983), kindly provided by Dr. Jackie Webb. Skeletal bone was visualized by using the vital stain Quercetin (Sigma), which stains calcium, as follows. Larval fish aged 7–8 dpf were placed in Quercetin (0.1 mg/ml) for 1–3 hr, and mineralized bone and tooth tissues visualized under ultraviolet (UV) were using a Zeiss M2-Bio microscope, photographed using a Zeiss Axiocam digital imaging system, and processed in Adobe Photoshop.
Developmental RT-PCR Analysis of Zebrafish Type I and II TGFβ Family Member Receptor Genes
RT-PCR was performed using the Superscript First-Strand Synthesis RT-PCR Kit (Invitrogen, Carlsbad, CA), per manufacturer's suggested protocol. PCR amplification of alk8, tgfbr2, acr2a, and acr2b (Yelick and Stashenko, 1996; Postlethwait et al., 1998; Yelick et al., 1998; de Caestecker et al., 2002) was performed with unique primers chosen from the 3′-UTRs of each cDNA, including: alk8 forward and reverse primers, 5′-TCATCATCCTGTTCCTGC-3′ and 5′-TTAGACGCGATAAGCCCA-3′; acvr2a forward and reverse primers, 5′-GGTGTCCTCACAACATTG-3′ and 5′-TCACCGGTCACTCGACAC-3′; acvr2b forward and reverse primers, 5′-CAAACCAGCCATCGCACA-3′ and 5′-TCACACCAGTCTACGACC-3′; tgfbr2 forward and reverse primers, 5′-GGCGACGTAGAAGAATAC-3′ and 5′-ACACCAATTCATGCACGCA-3′; and actin forward and reverse primers, 5′-TCAGCCATGGATGATGAAAT-3′ and 5′-GGTCAGGATCTTCATGAGGT-3′; under optimized conditions (PCR Optimizer Kit, Invitrogen, Carlsbad, CA). RT-PCR products were size fractionated by electrophoresis in 1.8% agarose containing EtBr, and imaged under UV illumination. The IDV PCR product densities were determined using the Spot density program of the Alpha Innotech Fluorchem imaging system, and normalized to actin control products.
The authors thank the members of the Yelick Laboratory, past and present, for helpful discussions and contribution to these studies. In particular, we thank John Hubbard, Kevin Tong, Loic Fabricant, and Seija Cope for expert zebrafish husbandry.