Hedgehog-dependent proliferation drives modular growth during morphogenesis of a dermal bone

During formation of the skeleton, bones and cartilages with unique morphologies are produced to provide structural support and framework for the body and protection to vital organs. Understanding the regulation of bone morphogenesis, the acquisition of form through shape and/or size changes, is of particular interest as alterations to these controls can result in skeletal defects in human inherited disorders and might drive morphological change during vertebrate evolution (Zelzer and Olsen, 2003; Kimmel et al., 2007).

The shape of dermal bones (i.e. those that develop from the direct differentiation of mesenchymal cells into osteoblasts) is especially dependent upon two important developmental events: condensation and osteoblast recruitment. This dependence is highlighted by the fact that the shape and size of mesenchymal condensations, groups of previously unorganized cells prior to overt differentiation into bone-forming osteoblasts, is predictive of early bone morphology (Hanken and Hall, 1993; Hall and Miyake, 2000; Eames et al., 2003). After the condensation has formed, bone shaping is thought to be controlled through the recruitment of new osteoblasts to the growing bone in a space- and time-dependent manner (Kimmel et al., 2010). Nevertheless, the molecular determinants that regulate dermal bone shape throughout development remain obscure, in part because dermal bone morphogenesis is complex and involves multiple developmental origins and processes over long periods of time.

To understand such complex bone morphogenesis, researchers have long used the rodent mandible as a model (Atchley and Hall, 1991; Hanken and Hall, 1993). The dermal bone component of the mandible grows appositionally: differentiating osteoblasts surrounded by their proliferating progenitors line the pre-existing bone surfaces (Atchley and Hall, 1991; Ramaesh and Bard, 2003). Mandibular morphogenesis occurs over a long period of development, and interactions of the osteogenic tissues with teeth and muscle are important for its later form and functional integrations (Atchley and Hall, 1991). Furthermore, it is hypothesized that the mandibular dermal bone comprises two distinct morphogenetic components, the anterior alveolar region and the posterior ascending ramus (Atchley and Hall, 1991; Hanken and Hall, 1993; Klingenberg et al., 2003; Klingenberg, 2009). These independent components, or modules, are characterized as having strong internal integration, but only sharing weak interactions with other modules (Raff, 1996; von Dassow and Munro, 1999; Wagner et al., 2007; Klingenberg, 2009). Developmental modularity might allow different signaling pathways to regulate the rates of growth independently within each module at specific times, allowing for precise regulation of bone shape. However, even though many genes are involved in mandibular formation (Bronner et al., 2010), no functional evidence exists that demonstrates how specific molecular activities can impose such regulation.

Here, we use the zebrafish opercle (Op) as a model system for understanding at cellular resolution the genetic regulation of growth and shaping of bone. The Op, a neural crest-derived dermal bone in the second pharyngeal arch, undergoes a series of characterized shape changes during larval development (Kimmel et al., 2010). Initial ossification of the Op occurs through the formation of a linear spur of bone matrix surrounded by osteoblasts that projects ventrally away from the joint it forms with the hyosymplectic cartilage (Cubbage and Mabee, 1996; Kimmel et al., 2010). Further morphogenesis of the Op throughout larval development occurs allometrically, with shape and size changes accompanied by new osteoblast arrangements and localized deposition of mineralized matrix along the pre-existing surfaces of the bone (Kimmel et al., 2010). Mutational analyses have identified genes in the endothelin 1 signaling pathway, particularly edn1 and its effectors, Dlx genes and mef2ca, to be crucial determinants of Op shape (Kimmel et al., 2003; Walker et al., 2006; Miller et al., 2007). The specific timing of Op growth and shape changes implies that Op morphogenesis is modular (Kimmel et al., 2010), and we take a genetic approach to test this hypothesis and address how molecular determinants fine tune bone shape.

We identify indian hedgehog a (ihha), encoding a Hedgehog (Hh) family ligand, and patched 1 (ptch1; formerly ptc2 – Zebrafish Information Network), encoding a Hh receptor, as key regulators of Op morphogenesis. Our data provide genetic evidence that the bone-shaping process is regulated in a modular fashion, as ihha and ptch1 mutants display spatially localized defects in Op growth. Through time-lapse analysis of Op development at cellular resolution, we show that Hh signaling is required regionally for the proper rate of nascent osteoblast addition to the growing bone. Furthermore, we find that levels of Hh signaling activity positively correlate with amounts of cell proliferation along the ventral Op edge. Finally, fusions of the Op and adjacent dermal bones in ihha mutants indicate that the Hh-dependent module may extend locally to immediately adjacent regions of separate bones.

Zebrafish were reared according to standard protocols (Westerfield, 2007) and staged as previously described (Kimmel et al., 1995; Parichy et al., 2009). All experiments conducted were approved by the University of Oregon Institutional Animal Care and Use Committee (IACUC). Zebrafish lines were as described: ihha^hu2131 (Parkin et al., 2009), ptch1^tj222 (Koudijs et al., 2005), Tg(sp7:EGFP)b1212 (DeLaurier et al., 2010) and trps1^J1271aGt (Talbot et al., 2010). ihha mutants were identified by PCR genotyping with the primers ihhaf6 (5′-CTGTGCCACCGTACCACTC-3′) and ihhar5 (5′-GCTACATTTGGACTAAACTGCAT-3′), and wild-type allele cut with NspI. ptch1 mutants were genotyped with the primers lepf6 (5′-TGGAAACCTGGCTACTTTTTG-3′) and lepr5 (5′-AAAGCGGCGGTCCTCTCTTCG-3′), and wild-type allele cut with TaqαI.

Skeletal staining was performed as previously described (Kimmel et al., 2010). For single-labeling experiments, larvae of all ages were stained overnight in the dark with 50 μg/ml Alizarin Red in E2 embryo medium. For double labeling experiments, larvae were incubated in 50 μg/ml Alizarin Red and 10 mM HEPES in E2 for 2 hours at 5 dpf and 50 μg/ml calcein at 7 dpf before being imaged at 8 dpf.

Larvae were treated with cyclopamine (Toronto Research Chemicals) as previously described (Winata et al., 2009). Larvae were exposed to 100 μM cyclopamine/1% ethanol in E2 from 4-6 dpf. Controls were treated with 1% ethanol in E2.

Larvae (4 dpf) were incubated from 4-5 dpf in 1 mM EdU (5-ethynyl-2′-deoxyuridine) in E2. Larvae were then anesthetized and fixed overnight at 4°C in 4% PFA/1×PBS. Following fixation, fish were washed with PBSTx (1% Triton X-100/1× PBS) and incubated with anti-GFP Alexa Fluor 488 (Invitrogen) 1:500 in PBSTx overnight at 4°C. After several washes in PBSTx, the Click-iT (Invitrogen) EdU reaction mixture was made according to manufacturer’s instructions and larvae were incubated in the reaction mixture for 1 hour. Apoptotic cells were labeled by TUNEL using the in situ cell death detection kit with TMR red (Roche) as previously described (Wu et al., 2006). For cell labeling, larvae were fixed overnight in 4% PFA/1×PBS, permeabilized with PBSTx and incubated with SYTO-59 (Invitrogen) at a 1:1000 dilution in PBSTx for 1 hour.

Zebrafish were reared in 15 mg/l PTU (1-phenyl-2-thiourea). Whole-mount labeling was performed essentially as described (Talbot et al., 2010). Larvae (5 dpf) were permeabilized with 10 μg/ml Proteinase K in PBST at room temperature for 45 minutes. Probes used have been described previously: ihha and col10a1 (Avaron et al., 2006), ihhb (Currie and Ingham, 1996), ptch1 (Lewis et al., 1999), ptch2 (Concordet et al., 1996), sp7 (DeLaurier et al., 2010), gli1 (Karlstrom et al., 2003), gli2a (Karlstrom et al., 1999), gli3 (Tyurina et al., 2005), and runx2a and runx2b (Flores et al., 2004).

Imaging was conducted using either a Zeiss LSM 5 Pascal confocal or Leica SD6000 spinning disk confocal with Borealis illumination technology. For time-course analysis, fish from a single clutch were only used for imaging at one stage to prevent stress-induced developmental delay. Opercle length measurements were obtained using ImageJ (National Institute of Health). For time-lapse imaging, larvae were anesthetized in 80 mg/l clove oil (Hilltech), mounted in 0.4% agarose in glass-bottomed Petri dishes, covered with E2 containing clove oil, and imaged at 30 minute intervals. Larvae imaged in this manner appeared similar to staged, non-imaged controls. Movies were constructed from z-projections using Metamorph imaging software (Molecular Devices). Quantitation of osteoblast addition was conducted manually across individual sections from time-lapse movies. The anterior and posterior halves of the Op are defined by the major axis of a best fit ellipse to each individual bone.

For time-lapse alignments, stacks of z-projections were aligned at 3 dpf by the joint and angle of the linear Op using ImageJ. The resultant channels were then thresholded individually at each stage. The area within the thresholded Op was filled using the fill hole command in ImageJ. Four regions of interest were constructed on the aligned Op based on divisions created using the draw ellipse macro in ImageJ. Percent overlap between images for each region of interest was calculated on the threshold, aligned images using the colocalization threshold plug-in.

Morphometric analysis was performed on a series of 14 digitized and Procrustes aligned landmarks on the Op shown in Fig. 1I, essentially as previously described (Kimmel et al., 2010). The 11 landmarks between the three apices along the bone were treated as sliding semi-landmarks. Visualizations of the landmark configurations as wire-frame diagrams, Principal component analysis and discriminant function analysis were implemented in MorphoJ version 1.02f (Klingenberg, 2011). For the presentation in Fig. 1I, the shape configurations were taken out of Procrustes alignment by scaling to their correct (non-normalized) sizes and manually aligning to the joint region, where bone growth initiates (Kimmel et al., 2010).

The opercle (Op) of zebrafish larvae homozygous for the putative null ihha^hu2131 allele (Parkin et al., 2009) is malformed in shape and reduced in size compared with wild type (Fig. 1A,B). To investigate the developmental basis of the morphological defects, we performed a systematic analysis of early larval Op development in ihha mutants carrying the sp7:EGFP transgene, which labels osteoblasts (DeLaurier et al., 2010). The Op forms a linear spur surrounded by sp7:EGFP-expressing osteoblasts at 3 days post fertilization (dpf) in wild type (Kimmel et al., 2010) and is unchanged in ihha mutants (Fig. 1C,D). Expansion of the ventral edge from 3-5 dpf forms a triangular Op, bounded by edges jp, vj and vp (Fig. 1E) (Kimmel et al., 2010). At 5 dpf, Op morphology is altered in ihha mutants: there are fewer osteoblasts near the v apex (Fig. 1F). Growth at this stage results in an overall size increase of the Op by 8 dpf (Fig. 1G), but ihha mutants display an apparent reduction in both osteoblasts and bone near the v apex (Fig. 1H). The shaping of other craniofacial dermal bones appears unaltered at this time, suggesting that ihha uniquely regulates Op morphology (data not shown).

Geometric morphometrics with landmarks placed along the Op edges (Kimmel et al., 2010) allows us to assess bone size and shape deformations quantitatively in ihha mutants. Mean centroid size of the ihha mutant Op is significantly smaller than that of the wild-type Op (Fig. 1I), and the mean shape difference appears as a local reduction of the ventral region of the bone near the v apex (Fig. 1I), matching our interpretations described above. Principal component analysis shows a clear separation in Op shapes of wild types and ihha mutants along the principal axis of shape variation (supplementary material Fig. S1). Furthermore, Procrustes distance, a univariate measure of shape disparity between the two groups, is substantial and highly significant. These results demonstrate that ihha is a key regulator of Op size and shape.

We used confocal time-lapse imaging of larvae with transgenically labeled osteoblasts from 3-4 dpf to examine the function of ihha at the cellular level in Op morphogenesis. In both wild-type and ihha mutant larvae, early Op morphogenesis occurs mainly through the addition of nascent (i.e. newly expressing) sp7:EGFP-expressing cells to the ventral edge of the linear spur, rather than by division of pre-existing sp7:EGFP-expressing cells (supplementary material Movies 1-3). Quantitation of the rate of nascent sp7:EGFP-expressing osteoblast addition to the Op reveals a spatially localized deficit in cells added to the anteroventral Op edge in ihha mutants (Fig. 2A). Direct comparison by alignment of time-lapse movies reaffirms that at first (3 dpf) the ihha mutant Op has wild-type morphology (Fig. 2B; supplementary material Movies 1, 2). The wild-type Op enters a second phase of morphogenesis between 3 and 3.5 dpf, indicated by the onset of osteoblast addition to both the anterior and posterior sides of the ventral tip, resulting in a fan-shaped array of osteoblasts (supplementary material Movies 1, 2). By 3.5 dpf and increasingly evident at 4 dpf, ihha mutants have fewer cells along the anterior of the ventral edge in comparison with wild type (Fig. 2C,D, arrows; supplementary material Movies 1, 2). This variation is not observed between alignments of two wild-type Ops (supplementary material Movie 3).

To verify our observation that defects in osteoblast addition in ihha mutants result in a region-specific Op shape change, we used aligned time-lapse images to calculate the fraction of wild-type EGFP-derived fluorescence overlapping with that of ihha mutant larvae in quadrants defined by the major and minor axes of a best fit ellipse to the Op (WT:ihha). As a control for variation in wild-type development, we also compared overlapping fluorescence between pairs of aligned wild-type larvae (WT:WT). The WT:ihha percent overlap at 3 dpf is not significantly different from WT:WT in any quadrant (Fig. 2E). ihha mutant Op shape is distinct from wild type at 3.5 and 4 dpf, specifically in the anteroventral quadrant, matching our observations above (Fig. 2F,G). At no time points are any other quadrants affected (Fig. 2E-G). Thus, the effects of ihha loss are specific in space and time.

The major modes of growth during larval Op morphogenesis are extensions in length along the vj, jp and vp edges at precise times (Fig. 1) (Kimmel et al., 2010). We thus tested whether the early defects in cell addition along the vp edge in ihha mutants cause a specific shortening of this edge. Quantitation of Op edge lengths supports this idea, revealing a significant reduction in vp edge length in ihha mutants, but no change in vj or jp (Fig. 3A-E). Furthermore, double labeling of the Op bone matrix at 5 dpf with Alizarin Red and then at 8 dpf with another calcified bone label, calcein, demonstrates that the reduction in ventral outgrowth is localized to the vp edge (Fig. 3F,G). Posterior outgrowth of the vp edge near the posterior (p) apex is relatively unaltered (Fig. 3F,G). Overall, these data support a model in which ihha coordinates morphogenesis within a specific ventral region of the Op by promoting local cell addition and bone extension exclusively along the vp edge.

As Hh signaling has been shown to promote proliferation of chondrocytes as well as of a variety of other cell types (St-Jacques et al., 1999; Long et al., 2001; Agathocleous et al., 2007; Hammond and Schulte-Merker, 2009), we asked whether Hh signaling controls cell addition and vp edge length by regulating the rate of cell proliferation. In wild-type larvae at 5 dpf, many proliferative cells just ventral to and including the sp7:EGFP-expressing vp edge osteoblasts are EdU positive following a 24 hour incubation (Fig. 4A). ihha mutants show a significant decrease in EdU-labeled cells in both sp7:EGFP-expressing vp edge osteoblasts and in sp7:EGFP-negative cells within two cell diameters of the vp edge (Fig. 4B,C). The total number of cells in this region is decreased by ∼30% in ihha mutants at 5 dpf; however, the number of proliferating cells is reduced by 60-70%, and therefore the decrease in cells cannot completely account for reduced proliferation (Fig. 4C; data not shown). Interestingly, the reduction in total cells correlates exactly with the 30% reduction in vp edge length apparent at this time (Fig. 3, Fig. 4C). Notably, ihha mutants show no significant change in the number of EdU-positive cells dorsal to the Op in a region that is likely to be occupied by Op-associated muscles (data not shown), indicating that ihha mutants do not have decreased proliferation throughout the entire animal. We saw no significant change in cell death by a TUNEL assay (data not shown). These data provide evidence that reduced osteoblast addition to the ihha mutant Op is due to locally decreased cell proliferation.

The Hh receptor ptch1 (formerly ptc2) is a negative regulator of Hh signaling, and fish homozygous for the presumably null ptch1^tj222 allele display elevated levels of Hh signaling activity and increased proliferation in pharyngeal chondrocytes (Koudijs et al., 2005; Hammond and Schulte-Merker, 2009). We therefore hypothesized that ptch1 mutants might display proliferation and Op morphogenesis defects that are opposite to those observed in ihha mutants. In 6 dpf ptch1 mutants, the Op appears slightly malformed with the vp edge seemingly extended relative to the other two edge lengths (Fig. 5A). As ptch1 mutants are smaller than wild type (Koudijs et al., 2005) and Op size scales directly with standard length (Kimmel et al., 2010), we analyzed ratios of Op edge lengths to normalize for body size differences. At all time points measured, vp/jp and vp/vj edge length ratios are significantly increased in ptch1 mutants, whereas the vj/jp edge length ratio remains unaltered (Fig. 5B). The elongation of the vp edge relative to the other edges in ptch1 mutants is accompanied by increased cell proliferation in both sp7:EGFP-positive and -negative cells within two cell diameters of the vp edge (Fig. 5C,D). Together, these data indicate that active Hh signaling is sufficient to drive cell proliferation and induce growth along the vp edge.

That both ihha and ptch1 mutants show localized defects in bone morphogenesis and proliferation of osteoblast precursors suggested molecular components of the Hh pathway might also be regionally restricted, so we examined this by in situ hybridization. High levels of sp7 expression span the wild-type vp edge, matching transgenic sp7:EGFP expression in osteoblasts (Fig. 6A; supplementary material Fig. S2A,B). Wild-type expression of ihha colocalizes with sp7 within these same cells (Fig. 6A). Lower levels of sp7 are apparent throughout the entire Op; however, ihha expression appears exclusively in the more highly sp7-expressing osteoblasts along the vp edge (supplementary material Fig. S2C,D). Expression levels of sp7 appear unchanged in ihha and ptch1 mutants (Fig. 6B; supplementary material Fig. S3). The ihha transcript in ihha mutants remains colocalized with sp7 along the vp edge, albeit at seemingly decreased levels compared with wild type (Fig. 6B). We note that the ihha^hu2131 allele contains a nonsense mutation; hence, the apparent decrease in ihha expression might be the consequence of nonsense mediated mRNA degradation (Chang et al., 2007).

To address which cells receive the Ihha signal, we investigated the expression patterns of the genes encoding the Hh signaling receptors/transcriptional targets ptch1 and ptch2 (Concordet et al., 1996; Lewis et al., 1999), as well as Gli genes, which are Hh signaling effectors (Ruiz i Altaba et al., 2007). At 5 dpf, wild-type expression of ptch1 colocalizes with ihha in osteoblasts along the vp edge and is also apparent in cells just ventral to the ihha/sp7-expressing cells (Fig. 6C). ptch1 expression is reduced within both of these regions in ihha mutants, indicating that Ihha might signal in both autocrine and paracrine manners (Fig. 6D). ptch2 expression appears similar to ptch1, is reduced in ihha mutants and is increased in ptch1 mutants specifically along and ventral to the vp edge (Fig. 6E,F; supplementary material Fig. S3A,B). At 5 dpf, expression of gli1 and gli3 in wild type colocalizes with ihha along the vp edge and is reduced in this region of ihha mutants (Fig. 6G-J). A local increase in gli1 expression is apparent along the vp edge in ptch1 mutants (supplementary material Fig. S3C,D). By contrast, the majority of gli2a expression is apparent in a broad region just ventral to the sp7-positive vp edge cells (Fig. 6K), and no notable changes in gli2a expression are visible in ihha or ptch1 mutants (Fig. 6L; supplementary material Fig. S3E,F). The localized expression of ihha and downstream Hh signaling markers specifically in the vp edge region matches exactly where morphogenesis is altered in ihha and ptch1 mutants, and supports the model that this edge of the Op is a Hh-dependent module (see Discussion).

Development of the bone lineage involves the sequential transition of cells through discrete stages (Bruder and Caplan, 1989; Bruder and Caplan, 1990; Li et al., 2009), and Ihh function has previously been associated with regulating osteoblast differentiation during dermal bone development (Abzhanov et al., 2007; Lenton et al., 2011). As a change to this ordered progression could result in the observed phenotypes, we looked at bone lineage markers at 3 dpf and 5 dpf. sp7, a marker of early osteoblasts (Li et al., 2009), is expressed at high levels in the ventral Op at 3 dpf. runx2a and runx2b, genes that encode transcription factors that directly activate and are required for sp7 transcription (Nakashima et al., 2002; Nishio et al., 2006; Li et al., 2009), are expressed in the ventral-most sp7-expressing cells of the wild-type Op, as well as in pre-osteoblasts that surround the sp7-expressing cells (Fig. 7A,B). col10a1, an early osteoblast marker that persists in later-stage osteoblasts (Li et al., 2009) is co-expressed with dorsal-most sp7 expression at 3 dpf, and also is expressed at high levels in the dorsal and presumably more mature cells of the wild-type Op (Fig. 7C). In ihha mutants, expression of these markers appears unchanged at 3 dpf (Fig. 7A-C, insets). The nested expression pattern of these markers at 5 dpf, when the ihha mutant Op phenotype is clearly apparent, is similar to the patterns observed at 3 dpf and is unchanged in ihha mutants (Fig. 7D-F). These data suggest that ihha is not required for progression through stages of osteoblast differentiation during Op morphogenesis. Notably, at both 3 and 5 dpf, the runx2a/b expression domain corresponds to where the Hh receptor ptch1 is expressed and also where proliferative defects are observed in ihha and ptch1 mutants, suggesting that Ihha signals to runx2a/b-expressing pre-osteoblasts.

As ihha is expressed all along the vp edge, yet only appears to control the morphology of the most anterior region, we examined whether its co-ortholog, ihhb, may function redundantly with ihha. Expression of ihhb is not apparent in any bone-forming cells associated with the Op of either wild-type or ihha mutant larvae (supplementary material Fig. S4). However, we note that ihhb expression is visible in a domain just dorsal to the jp edge of the Op that corresponds to the location of muscles that elevate the Op (supplementary material Fig. S4). Thus, Ihhb could possibly signal to the posterior portion of the vp edge and compensate locally for the loss of ihha. Therefore, we used cyclopamine (CyA) to inhibit all Hh signaling from 4-6 dpf and test whether other Hh ligands than Ihha contribute to Op morphogenesis. We find that addition of CyA either to wild-type or ihha mutant larvae yields a quantitatively equivalent Op phenotype to untreated ihha mutants (Fig. 8). That CyA treatment matches and does not exacerbate the ihha mutant phenotype suggests that Ihha is the only Hh ligand required for early Op patterning and, furthermore, that the requirement for Hh signaling is during the 4-6 dpf time window, corresponding to the period of early Op morphogenesis when we first observe the ihha mutant phenotype.

To assess whether morphogenetic problems persist in ihha mutants during later larval stages, we analyzed Op development in larvae stained vitally with Alizarin Red from 10-21 dpf. At 10 dpf, wild-type larvae display a large fanned Op that is located dorsal and lateral to the emerging subopercle (Sop) (Fig. 9A). Continual outgrowth of the vp edge causes the Op to overlap the Sop by 14 dpf (Fig. 9B). Then, the Op grows proportionally larger by 17 dpf (Fig. 9C) and eventually assumes a trapezoidal shape at 21 dpf while becoming overlapped by the interopercle (Iop) (Fig. 9D). At all of these time points, ventral, but not posterior, outgrowth of the Op is reduced in ihha mutants, matching our earlier observations (supplementary material Fig. S5). In addition to growth defects, ihha mutants display fusions between opercular dermal bones that normally form functional articulations (Hulsey et al., 2005). Approximately 64% of ihha mutants at 10 dpf exhibit fusions of the anterior vp edge to the adjacent dorsal edge of the Sop (Fig. 9E,I). Extension of this fusion results in nearly complete suturing of the two bones at 14 dpf (Fig. 9F). Additionally, an ectopic site of ossification of unknown origin is often apparent between the branchiostegal rays in ihha mutants (Fig. 9F, arrow). At 17 dpf, ∼22% of ihha mutants have fusions between the Sop and Iop (Fig. 9G, arrow), whereas ∼55% persistently exhibit Op-Sop fusions (Fig. 9I; data not shown). By this time, fusions between all three of these bones are also seen in roughly 23% of ihha mutants, and at 21 dpf these phenotypes continue to occur at nearly equivalent frequencies (Fig. 9H,I; data not shown).

To examine whether joint identity might be affected by the loss of ihha in regions of dermal bone fusions, we used a transgenic line, trps1^J1271aGt (Talbot et al., 2010) (abbreviated here to trps1:EGFP), which marks many skeletal joints. At 17 dpf, high trps1:EGFP expression is apparent ventral to the vp edge in a region that overlaps the dorsal part of the Sop (Fig. 9J, arrow). In the corresponding region of ihha mutants, the Op is fused to the Sop, and trps1:EGFP expression appears reduced (Fig. 9K). In total, these analyses demonstrate that the morphogenetic domain of ihha function may include the adjacent edges of dermal bones and, furthermore, suggest a function for Hh signaling in maintaining joint identity.

Our study of opercle (Op) morphogenesis in ihha and ptch1 mutant zebrafish larvae reveals that Hh signaling can regulate bone morphogenesis by control of pre-osteoblast proliferation. As determined by EdU incorporation, proliferation of cells along the ventral edge of the Op is decreased in ihha mutants and increased in ptch1 mutants compared with wild-type siblings. Because we observe runx2a/b and ptch1/2 expression in exactly the region of EdU incorporation, the cells receiving the Hh signal are likely to be pre-osteoblasts and incorporate EdU prior to transitioning to sp7:EGFP-expressing osteoblasts. This supposition is supported by our observation that early Op growth occurs largely through the recruitment, rather than division, of sp7:EGFP-expressing osteoblasts. Therefore, the amount of pre-osteoblast proliferation would positively correlate with Hh signaling.

Hh signaling has been linked to cell cycle control in a variety of cell types through its activation of various proto-oncogenes such as cyclins (Pasca di Magliano and Hebrok, 2003; Ehlen et al., 2006). Interaction between Ptch1 and Cyclin B1 (Barnes et al., 2001; Jenkins, 2009), as well as transcriptional regulation of various cyclin genes by Gli proteins (Kenney and Rowitch, 2000; Long et al., 2001; Yoon et al., 2002; Mill et al., 2003) are important for cell cycle progression. Because we observe the presence of Gli gene expression as well as Hh-dependent ptch1/2 expression in cells where proliferation is altered in the ihha and ptch1 mutants, it is reasonable to suppose that Ihha signaling directly stimulates pre-osteoblast proliferation through activation of cyclins. The bone fusion phenotype that we observe in later ihha mutant larvae might also somehow be a secondary effect of the early proliferation decrease, but this possibility requires further study.

Given the localized nature of the mutant phenotypes, our study provides genetic support for the hypothesis that control of Op morphogenesis is modular (Kimmel et al., 2010). Owing to its specific dependence upon Hh signaling, vp edge morphogenesis is regulated independently from other facets of Op shaping/growth. Spatially localized control of cell proliferation by Hh signaling along the ventral Op controls two parameters of morphogenesis: vp edge length and overall ventral outgrowth. The ability of a local signal to alter specifically a local region of morphology of a given bone is consistent with the concept of dissociability, which proposes that functional or developmental units, normally integrated, might be experimentally separated from one another (Needham, 1933). Genetically based modular organization (Raff and Raff, 2000), as we observe in the Op, allows regional rates of bone growth to be individually regulated so as not to have morphogenetic consequences in other regions of the bone. Our results thus help to explain the molecular basis of allometric growth seen in many dermal bones, including the Op (Kimmel et al., 2010) (Kimmel et al., 2012). In addition, that ihha mutants display fusions between dermal bones that would normally articulate might mean that these adjacent regions of functionally related bones are all included in a single Hh-dependent module that extends along the lengths of the neighboring edges.

Previous work has revealed conflicting mechanisms by which Hh signaling may regulate dermal bone development (Abzhanov et al., 2007; Lenton et al., 2011). In one study, Ihh gain- and loss-of-function experiments in chick and mouse suggest that Hh signaling blocks the transition of pre-osteoblasts to osteoblasts (Abzhanov et al., 2007). An opposing study reported an overall decrease in markers of osteoblast differentiation in the calvaria of Ihh-null mice, suggesting that Hh signaling is pro-osteogenic (Lenton et al., 2011). As we did not detect any changes in osteoblast stage markers in ihha mutants, our data more closely fit the pro-osteogenic model put forth by Lenton et al., as their observed reduction in osteoblast markers might reflect differences in the rate of progenitor proliferation.

In the case of the Op, we show that Ihha is secreted exclusively from sp7-expressing osteoblasts along the vp edge and is likely to signal to pre-osteoblasts to maintain high levels of proliferation and consequently growth along the length of the vp edge as well as ventral outgrowth (Fig. 10). Some of these proliferating pre-osteoblasts are recruited to the vp edge and subsequently transition into sp7-expressing osteoblasts, causing the Op to grow in a manner similar to that of other dermal bones (Wilkie, 1997; Opperman, 2000; Ramaesh and Bard, 2003). The loss of Hh signaling leads to a decrease in pre-osteoblast proliferation, which ultimately results in a reduction in vp edge length as there are fewer cells available for recruitment to the vp edge (Fig. 10). We observe the converse proliferative and morphogenetic defects in ptch1 mutants and expect mutants for other Hh pathway repressors to exhibit similar phenotypes. Therefore, the cellular function of Hh signaling in Op morphogenesis resembles its role in cartilage morphogenesis, during which Ihh, which is secreted from pre-hypertrophic chondrocytes that have just left the proliferative pool, feeds back to maintain high levels of chondrocyte proliferation (Vortkamp et al., 1996; St-Jacques et al., 1999; Karp et al., 2000; Kronenberg, 2003). Our findings are also consistent with the pro-osteogenic role of Hh signaling in endochondral ossification (St-Jacques et al., 1999; Hammond and Schulte-Merker, 2009). Thus, Hh signaling is likely to be part of a genetic toolkit that is reused to produce similar outputs in multiple skeletogenic contexts. Importantly, our data reveal that a molecularly defined module can be restricted to a very precise region of a single bone to provide exquisite genetic control of the bone shaping process. Experimental dissociation of apparently unitary structures indicates that multiple processes must be involved in the formation of that particular structure (Raff, 1996). Accordingly, our model predicts the existence of additional modules within the Op controlled by other gene networks that regulate growth along the other edges.

We thank April DeLaurier and Mark Sasaki for dedicating their expert assistance in experimental design; Phil Ingham and Christiane Nüsslein-Volhard for providing fish strains; John Dowd and the University of Oregon fish facility for care and maintenance of animals; members of the Kimmel Lab for thoughtful conversations; Yasuko Honjo for gli1, gli2a and gli3 probes; and Rolf O. Karlstrom for discussion of data.