brother of cdo (umleitung) is cell-autonomously required for Hedgehog-mediated ventral CNS patterning in the zebrafish

Hedgehog (Hh) signaling is required for numerous developmental processes in vertebrates, including neurogenesis, axon guidance, myogenesis, organogenesis and limb formation (Fuccillo et al., 2006; Hughes et al., 1998; Panman and Zeller, 2003; Reidy and Rosenblum, 2009; Ruiz i Altaba et al., 2002; Sanchez-Camacho and Bovolenta, 2009). Sonic Hedgehog (Shh) regulates several distinct cellular behaviors in the embryo, including cell differentiation, proliferation and survival (Jacob and Briscoe, 2003; Tannahill et al., 2005). Shh functions in the central nervous system (CNS) as a morphogen, being secreted by the notochord and floor plate of the spinal cord to elicit concentration-dependent transcriptional responses in neural precursors that then determine neural fates along the dorsoventral axis (Ericson et al., 1997). Shh from the floor plate also serves as a chemoattractant for ventrally migrating axonal growth cones in the neural tube (Charron et al., 2003).

Hh signaling gradients in developing embryos are influenced by numerous positive and negative regulators, some of which directly interact with the Shh protein. The Hh receptor Patched and Hedgehog-interacting protein (Hhip) can bind Shh and help shape the Shh gradient (Briscoe et al., 2001; Dessaud et al., 2007; Ochi et al., 2006). Both of the genes that encode these proteins are Class II Hh target genes (transcriptionally activated by Shh signaling) and both are expressed in cells that require Shh for proper differentiation (Chuang and McMahon, 1999; Goodrich et al., 1996).

Shh binding proteins that are transcriptionally repressed by Hh signaling (Class I Hh target genes) include Cell adhesion molecule-related/downregulated by oncogenes (Cdo) and Brother of Cdo (Boc) (Tenzen et al., 2006). These proteins belong to a group of immunoglobulin (Ig) and fibronectin type III (FNIII) domain-containing integral membrane proteins (Kang et al., 1997; Kang et al., 2002). Boc acts as a receptor for Shh and mediates commissural axon guidance in the spinal cord (Okada et al., 2006). Boc is also required to repel retinal ganglion cell (RGC) growth cones from regions expressing Shh in the mouse forebrain (Fabre et al., 2010) and for the proper formation of ventrally projecting axon tracts in the zebrafish brain (Connor et al., 2005).

Roles for Boc and Cdo in Hh-mediated CNS patterning have begun to be elucidated, but the details of how these proteins affect the cellular Hh response in vivo is poorly understood. Cdo normally promotes Hh signaling, and Cdo mutant mice display phenotypes consistent with a loss of Hh signaling (Tenzen et al., 2006; Zhang et al., 2006). Although in vitro studies indicate that Cdo and Boc can both bind Shh to positively modulate Hh signaling, Boc mutant mice are reported to have normal ventral CNS patterning (McLellan et al., 2008; Okada et al., 2006). Intriguingly, both Boc and Cdo are primarily expressed in dorsal CNS tissues that do not respond to Hh signals, suggesting a negative role in Hh signal regulation in these regions (Mulieri et al., 2002; Mulieri et al., 2000; Tenzen et al., 2006). Consistently, overexpression of Boc and Cdo in CNS explants results in non-cell-autonomous inhibition of the Shh response (Tenzen et al., 2006). These Hh binding proteins might also function to amplify Shh signals at the dorsalmost limits of the Shh gradient, and/or might help to refine Shh gradient formation (Dessaud et al., 2008).

The Drosophila homologs of Cdo and Boc, interference Hedgehog (iHog) and brother of iHog (boi), respectively, were recently shown to be required for Hh signaling in the developing wing disc. boi and iHog mutants have defects in Hh-mediated embryonic patterning and have reduced Hh target gene expression that is enhanced in double mutants (Yao et al., 2006). Both of these proteins work at the level of Ptc and can bind directly to Hh via an extracellular FN domain. iHog overexpression enhances the localization of the Ptc receptor at the cell surface of Hh-receiving cells (Zheng et al., 2010). It appears that a high level of Hh signaling might involve interactions between Hh-bound protein complexes with a number of cell-surface receptors (Wilson and Chuang, 2006).

A large-scale forward genetic screen in zebrafish identified mutations in many components of the Hh signaling pathway (Haffter et al., 1996). These mutations were isolated because of defects in body axis formation, slow muscle differentiation and/or retinal axon guidance (Baier et al., 1996; Brand et al., 1996; Karlstrom et al., 1996; Trowe et al., 1996; van Eeden et al., 1996). Known components in the Hh pathway that were identified in this screen include Disp1 (Nakano et al., 2004), Shh (Shha – Zebrafish Information Network) (Schauerte et al., 1998), the transmembrane proteins Smo (Varga et al., 2001), Hhip, Ptch2 and Ptc2 (Ptch1 – Zebrafish Information Network) (Koudijs et al., 2008; Koudijs et al., 2005; Lee et al., 2008), the cytoplasmic regulator Sufu (Koudijs et al., 2005), and the Hh responsive transcription factors Gli1 (Karlstrom et al., 2003) and Gli2 (Gli2a – Zebrafish Information Network) (Karlstrom et al., 1999). Novel components identified include the extracellular protein Scube2 (Kawakami et al., 2005; Woods and Talbot, 2005) and the intracellular protein Dzip1 (Sekimizu et al., 2004; Wolff et al., 2004).

Here, we present the molecular characterization of umleitung (uml), the last of the ipsilateral class of zebrafish mutants from this screen (Karlstrom et al., 1996) to reveal itself. We show that uml encodes Boc and that Boc is cell-autonomously required for Hh signaling in the ventral CNS. These studies uncover new roles for Boc in vertebrate development.

uml^ty54 was generated in the Tü strain and crossed to the TL, AB and Brian's wild-type strains for genetic mapping (Talbot and Schier, 1999) using simple sequence repeat markers (Knapik et al., 1996). Other fish lines used were yot^ty119(gli2) (Karlstrom et al., 1999), dtr^ts269(gli1) (Karlstrom et al., 2003) and Tg(nkx2.2a:megfp)^vu17 (Kirby et al., 2006).

cDNA from wild-type or uml mutant individuals was PCR amplified using four sets of overlapping primers (Connor et al., 2005): cdsBOC5′UTR.fw, ATGCAAGAAGTTGGGGAGCCG; EXON5.rv, ATTCCTCAAGTCCAGGCCGTC; EXON4.fw, TGCTGCTTACAACCCCGTCAC; EXON10.rv, CGTCACATAAGCGGAGGTCTCGGT; EXON9.fw, GAAGAGGGGGACCAATAGGAC; EXON15.rv, AGCGGGACCATGGTATACTGG; EXON14.fw, CCTGGAGATCTGCCTTACCTC; EXON18.rv, TCCATCTTCTGGATCATGTGT. DNA fragments were gel purified (Qiagen Gel Extraction Kit) and sequenced commercially (GeneWiz).

For genotyping, fin clip DNA was PCR amplified using primers designed to discriminate between the mutant and wild-type sequence (umlEXON5.MUT.FW, GAAGCGGCTCGTATCATCTAA; UMLEXON5.RV, TCAGCGAGACACACGTAAGTG) resulting in a 207 bp PCR product only in mutant carriers. The wild type control primer was: umlEXON5.WT.FW, GAAGCGGCTCGTATCATCTAT. Differences in the wild type and mutant primers are underlined.

Whole-mount in situ hybridization (ISH), immunohistochemistry, cryostat sectioning and imaging were performed as described previously (Karlstrom et al., 1999; Guner and Karlstrom, 2007; Barresi et al., 2005) using published probes. Bone and cartilage were labeled as described by Walker and Kimmel (Walker and Kimmel, 2007). DiI/DiO axon labeling was performed as described by Karlstrom et al. (Karlstrom et al., 1996). ISH probes for cdo and foxa2 were made by PCR using gene specific primers containing the T7 or SP6 RNA polymerase binding sites (indicated in capitals): CDO.fw.Sp6, ATTTAGGTGACACTATAGAAtacgatctggagtggagagc; CDO.rv.T7, TAATACGACTCACTATAGGGAaagagatgtgaggcccagtg; Foxa2.fw1.Sp6, ATTTAGGTGACACTATAGAAacagcgttaagagccagcag; Foxa2.rv1.T7, TAATACGACTCACTATAGGGAtgcccgtgttgacataggac.

shh and boc mRNA were synthesized (Message Machine, Ambion) using the shh/T7TS (Ekker et al., 1995) and boc/PCS2⁺ (Connor et al., 2005) plasmids as templates. The translation-blocking boc morpholino (MO1-boc: 5′-AATCCAATTCAACGTCCCAGACATC-3′) has been described previously (Connor et al., 2005). Embryos were injected at the two-cell stage as described previously (Westerfield, 2000).

Cell transplantation was performed as described previously (Carmany-Rampey and Moens, 2006) with donor embryos injected with or without the boc morpholino. Approximately 20 donor cells were transplanted into the presumptive medial spinal cord region at the gastrula stage (Woo and Fraser, 1995). Paired donor/host embryos were grown in 12-well plates and fixed at 24 hours postfertilization (hpf). Cyclopamine (Toronto Chemicals) treatments were carried out in 12-well plates with 40 embryos per well as described previously (Guner and Karlstrom, 2007).

The umleitung (uml) (meaning `detour' in German) mutant was identified in a large-scale screen of N-ethyl-N-nitrosourea (ENU)-generated mutations based on defects in retinal axon projections to the tectum (Karlstrom et al., 1996). In 5 days postfertilization (dpf) uml mutants, some retinal ganglion cell (RGC) axons failed to cross the ventral midline of the forebrain and instead projected ipsilaterally (Fig. 1A,B). Despite these midline crossing errors, RGC axons reached their correct topographic targets in the optic tectum (Fig. 1B). At 2 dpf, uml mutant RGC axons approached, but did not cross, the midline (Fig. 1C,D). The post optic commissure (POC) failed to form in the diencephalon of uml mutants, whereas the anterior commissure (AC) in the telencephalon formed normally (Fig. 1C-F). These axon phenotypes are similar to, but less severe than, those seen in other ipsilateral class mutants such as yot(gli2) and dtr(gli1) (Barresi et al., 2005; Culverwell and Karlstrom, 2002; Karlstrom et al., 1996).

To better understand these forebrain axon guidance defects, we assayed axon guidance cues in the region. Specialized glial fibrillary acidic protein (Gfap)-expressing glial cells span the midline prior to axon crossing (Barresi et al., 2005). In uml mutants, the glial bridge in the post optic area (POA) was disrupted, whereas the glial bridge in the AC region appeared to form normally (Fig. 1E,F insets). Slit guidance molecules are thought to act primarily as growth cone repellents, with bands of slit expression helping to position correctly the optic nerves and POC in the forebrain (Fricke et al., 2001). slit2 and slit3 expression domains were expanded into the POA in uml mutants (Fig. 1G,H; data not shown). Interestingly, uml mutants also had reduced slit1a and sema3d expression in the ventral forebrain (Fig. 1I,J; data not shown). These defects in axon guidance molecule expression were very similar to those seen in yot(gli2) mutants (Fig. 1H,J) (Barresi et al., 2005) suggesting that axons failed to cross the midline in uml mutants because of glial patterning defects and/or the expansion of slit repellent molecules into the POA (Barresi et al., 2005).

Next, we determined whether earlier forebrain patterning events are disrupted in uml mutants. The homeodomain transcription factor dlx2 (dlx2a – Zebrafish Information Network) serves as a good marker for patterning within the telencephalon and diencephalon (Akimenko et al., 1994). dlx2a expression was absent in the POA of uml mutants, whereas dlx2a expression appeared normal in the telencephalon and the more posterior diencephalon (Fig. 1K,L), indicating regional patterning defects in the forebrain. This misexpression was nearly identical to that seen in the Hh signaling pathway mutants yot(gli2) (Fig. 1L, inset) and dtr(gli1) (data not shown). Together, these data suggest that the uml mutation might affect Hh-mediated forebrain patterning.

Given these phenotypic similarities with known Hh pathway mutants, we examined the expression of several Class I and Class II Hh target genes in uml mutant embryos. ptch2 (previously Ptc1, now Ptch2 – Zebrafish Information Network) expression is a well-established read-out of Hh signaling (Concordet et al., 1996). In uml mutants, ptch2 expression was reduced but not absent throughout the CNS (Fig. 2A,B) and somites (not shown), suggesting an overall reduction in Hh signaling levels. Consistently, expression of the floor plate marker foxa2 was reduced in the brain and absent in lateral floor plate cells but remained in the medial floor plate (Fig. 2C,D), as is the case in the Shh mutant syu(shh) (Schauerte et al., 1998).

The Class II genes nkx2.9 and nkx2.2a are expressed in the ventral brain and lateral floor plate (Guner and Karlstrom, 2007). nkx2.9 expression was completely absent in uml mutants (Fig. 2F), whereas nkx2.2a expression was regionally absent in the brain and completely absent in the spinal cord (Fig. 2H). This is similar to the pattern seen in yot(gli2) and dtr(gli1) mutants (Guner and Karlstrom, 2007). In wild type, the Class II genes olig2 and nkx6.1 were regionally expressed in the brain and were expressed slightly more dorsally than nkx2.2a in the ventral spinal cord (Fig. 2I,K). In uml mutants, expression of olig2 and nkx6.1 was reduced in the brain and spinal cord (Fig. 2J,L). In the spinal cord, the expression domains of both genes were shifted ventrally by a few cell diameters (Fig. 2J′,L′). By contrast, expression of the Class I Hh target genes pax3a and pax7a (Seo et al., 1998) appeared to be unaffected in uml mutants (Fig. 2M,N), similar to the situation in yot(gli2) and dtr(gli1) (Guner and Karlstrom, 2007). Together, these results indicate that mutations at the uml locus lead to reduced Hh signaling levels, similar to other zebrafish Hh pathway mutants.

To determine the molecular cause of these Hh signaling defects, we next identified the gene encoded at the uml locus. Using simple sequence repeat markers (Knapik et al., 1996) and a mapping panel encompassing over 3200 meiotic events, we mapped uml to a 2.5 cM genetic interval on zebrafish chromosome 24. Given the lack of genomic sequence information available in this region, we identified a syntenic region on Tetraodon nigroviridis chromosome 6 and identified genes likely to be located near the uml locus. By combined synteny and recombinant analyses we further refined the uml genetic interval to a 0.3 cM region containing approximately nine genes, including brother of cdo (boc) (Fig. 3A).

Co-segregation analysis showed that boc was tightly linked to the uml locus (0 recombinants/3242 meioses). Sequencing of the boc coding region in uml mutants revealed a single point mutation in exon 5 (T to A) that is predicted to lead to a severely truncated protein (Tyr238 to a STOP codon) (Fig. 3B) containing only the signal sequence and two immunoglobulin (Ig) domains (Fig. 3D). Sequence alignments indicate that zebrafish Boc is closely related to mammalian Boc, but is also similar to the closely related Cdo protein. Both of these proteins are conserved across multiple species from Drosophila to humans (Fig. 3C).

To further confirm that the uml(boc) mutation leads to a loss of Boc function, we injected translation blocking boc morpholinos (Connor et al., 2005) into wild-type and uml mutant zebrafish. Injecting 5 ng of boc morpholino into wild-type embryos produced early defects indistinguishable from those observed in the uml mutants, including curved body axes (Fig. 4A-C) and regional loss of the Class II Hh target gene nkx2.2a (Fig. 4D-F). We also reproduced the RGC axon path-finding defects seen in 48 hpf uml mutants, although not as robustly as the earlier phenotypes (Fig. 4G-I). Injection of boc morpholinos (MOs) into uml mutants did not enhance any of these phenotypes (data not shown). Together, these results strongly suggest that uml^ty54 represents a boc loss-of-function mutation that results from a severe truncation in the Boc extracellular domain.

To further investigate the role of Boc in Hh signaling, we overexpressed boc and shh in uml and wild-type embryos. boc mRNA injections (250 pg) partially rescued the loss of nkx2.9 and nkx2.2a (Fig. 5E,K) in 100% of uml mutants. Interestingly, whereas boc mRNA injection restored expression of nkx2.9 and nkx2.2a in the normal ventral domain of uml mutants, little ectopic Hh target gene expression was seen either in uml mutants or wild-type embryos (Fig. 5B,H). In fact, ectopic expression of nkx2.9 was not detectable by in situ hybridization (Fig. 5B,E), and ectopic expression of nkx2.2a was only detected in the Tg(nkx2.2a:megfp) transgenic line (Ng et al., 2005), which allows for amplification of GFP fluorescence in nkx2.2a-expressing cells (Fig. 5H). By contrast, injecting 100 pg of shh mRNA into wild-type individuals led to a massive expansion of nkx2.9 and nkx2.2a expression (Fig. 5C,I). Injecting this same amount of shh mRNA into uml mutants led to a much less robust expansion of nkx2.9 and nkx2.2a expression but did rescue nkx2.9 and nkx2.2a expression in ventral regions (Fig. 5F,L). Unlike the case in yot(gli2) and dtr(gli1) mutants, cells throughout the CNS of uml mutants responded to increased levels of shh, but only weakly. Together, these results suggest that (1) both Boc and Shh overexpression can compensate for the loss of Boc function in the ventral CNS and (2) cells can respond to Shh in the absence of Boc function but this response is much weaker than in the presence of Boc.

To begin to assess how Boc influences Hh-mediated embryonic patterning, we examined boc expression in relation to Hh-mediated embryonic patterning events (see Fig. S1 in the supplementary material). Expression of boc was first seen in the ventral neural plate and spinal cord at 8-10 hpf, just prior to the first expression of the Hh target gene nkx2.2a (see Fig. S1A-C in the supplementary material). By 12 hpf, boc was predominantly expressed in the dorsal spinal cord, although weak labeling was still present ventrally (see Fig. S1D in the supplementary material). boc expression in the dorsal spinal cord generally overlapped with expression of the closely related cdo gene (see Fig. S1E,I in the supplementary material). Expression of boc was maintained dorsally in the brain and spinal cord and was expressed regionally in the eye throughout embryonic and larval stages (see Fig. S1F-Q in the supplementary material). boc was weakly expressed in cells that give rise to the pituitary gland (see Fig. S1G in the supplementary material) and was more strongly expressed in the somites (see Fig. S1D,E,J in the supplementary material) and brachial arches (see Fig. S1P,Q in the supplementary material). The restriction of boc expression to the dorsal CNS was reported previously (Connor et al., 2005) and is consistent with boc being a class I Hh target gene. This was confirmed by the loss of boc expression following injection of shh mRNA (see Fig. S1M in the supplementary material) and by the ventral expansion of boc expression seen in smoothened mutants (see Fig. S1N in the supplementary material). Interestingly, no ventral expansion of boc or cdo expression was seen in the spinal cord of uml(boc) mutants (see Fig. S1H,O in the supplementary material), consistent with the normal expression of the class I Hh target genes pax3a and pax7a in the dorsal CNS of uml(boc) mutants (Fig. 2). boc gene expression itself was slightly reduced in the dorsal CNS of uml(boc) mutants (see Fig. S1H in the supplementary material).

Given the expression of boc in tissues outside the CNS and the requirement for Hh/Gli signaling during zebrafish pituitary, jaw and slow muscle fiber differentiation (Du et al., 1997; Eberhart et al., 2006; Sbrogna et al., 2003), we examined these tissues more carefully for defects in uml(boc) mutants. Similar to dtr(gli1) mutants, expression of the Hh target gene nkx2.2a was reduced in the anterior adenohypophysis in uml(boc) mutants, whereas posterior pax7a expression was not affected (see Fig. S2A,B, in the supplementary material; data not shown). Our previous work showed that Hh dosage regulates endocrine cell differentiation, with Prolactin- (Prl), Adrenocorticotropic hormone- (Acth; Pomca – Zebrafish Information Network) and Growth hormone- (Gh; Gh1 – Zebrafish Information Network) secreting cells being the most sensitive to reduced Hh signaling levels (Guner et al., 2008). Consistent with a mild loss of Hh signaling in uml(boc) mutants, gh-expressing cells were absent, acth- and prl-expressing cells were greatly reduced, whereas more posterior endocrine cells were largely unaffected (see Fig. S2C-G in the supplementary material). pomc-expressing cells in the ventral hypothalamus were disorganized, as in other Hh pathway mutants (see Fig. S2F in the supplementary material), consistent with Hh-related forebrain patterning defects.

Reduced Hh signaling leads to fused upper jaw cartilages and reduced lower jaw structures (Brand et al., 1996; Schwend and Ahlgren, 2009; Teraoka et al., 2006). In uml(boc) mutants, neurocranial elements, including the trabeculae, were fused, consistent with a loss of Hh signaling in the upper jaw (see Fig. S2 in the supplementary material). Surprisingly, the lower jaw was expanded in uml(boc) rather than reduced or absent. As Hh signaling is needed for proliferation and chondrogenesis during jaw growth (Schwend and Ahlgren, 2009; Teraoka et al., 2006), this phenotype might point to a repressor role for Boc in Hh-mediated craniofacial development.

Although somite borders appeared morphologically normal in uml(boc) mutants (Fig. 4), the fact that boc was expressed in developing adaxial cells and somites (see Fig. S1 in the supplementary material) led us to examine muscle differentiation in uml(boc) mutants more carefully. Expression of myoD (myod1 – Zebrafish Information Network) in early slow muscle fiber precursors (adaxial cells) was reduced but not absent in uml (see Fig. S2R,S in the supplementary material), similar to yot(gli2) heterozygotes (data not shown). Expression of ptch2 was also generally reduced in the somites (data not shown), suggesting a reduction in Hh signal levels in this tissue. The number of slow muscle fibers, which require Hh signaling for differentiation (Barresi et al., 2000), was slightly reduced in uml mutants (see Fig. S2T,U in the supplementary material). These data suggest that Boc plays a minor role in slow muscle fiber differentiation, and might act as an activator of the Hh response during adaxial cell specification.

The ventral shift in olig2 and nkx6.1 expression in uml mutants (Fig. 2) suggested that Hh signal levels are generally reduced in the ventral CNS. To further examine how loss of Boc function affects overall levels of Hh signaling, we treated progeny from uml(boc) heterozygous parents with varying concentrations of the dose-dependent Hh signaling inhibitor cyclopamine (CyA) (Chen et al., 2002). In the absence of CyA, ∼17% of embryos from a uml clutch had reduced olig2 expression, indicating that this phenotype is not completely penetrant (see Fig. S3 in the supplementary material). Low doses of CyA (20 μM) had almost no effect on embryos from wild-type clutches, but increased the number of affected embryos in uml(boc) clutches from 17% to 33%. Furthermore, 13% of the embryos from these clutches had a more severe loss of olig2 expression. Slightly higher doses of CyA (40 μM) increased the severity of defects, with 45% of the embryos in uml clutches having a severe loss of olig2 expression. These CyA doses caused severe olig2 expression defects in less than 5% of embryos from wild-type clutches. Both wild-type and uml mutant embryos had more severe olig2 expression defects when Hh signaling was blocked further (see Fig. S3C,D in the supplementary material). These data suggest that uml homozygous and heterozygous embryos are more sensitive to the loss of Hh signaling than wild-type embryos and points to a general attenuation of Hh signaling in the absence of Boc function.

To determine the site of action for Boc we generated genetic mosaic embryos by cell transplantation. To simplify the assay for the Hh transcriptional response, we performed these experiments using Tg(nkx2.2a:megfp) transgenic embryos (Ng et al., 2005). When targeted to the lateral floor plate region of uml mutant (or boc MO-injected) host embryos, isolated wild-type cells were able to respond to Hh signals, as indicated by expression of GFP from the nkx2.2a:megfp transgene (Fig. 6C). In the opposite experiment, boc mutant or morphant cells carrying the nkx2.2a:megfp transgene were never found to express GFP, even when located in the lateral floor plate region where nkx2.2a is normally expressed (Fig. 6D). These results show that Boc function is required cell-autonomously for cells to respond to maximal Hh levels in the ventral CNS (Fig. 6E,F). Importantly, normal Boc function in adjacent cells did not compensate for the lack of Boc function in transplanted cells.

Our analysis of the zebrafish uml(boc) mutant shows for the first time that Boc is required for the transcriptional response to Hh in the ventral vertebrate CNS. Boc appears to be required for the transcriptional activation of Class II Hh target genes that require maximal Hh signal levels, as transcription of the most ventrally expressed Hh-regulated genes fails to be activated in the absence of Boc function (Fig. 2). Increasing Hh signals artificially in uml(boc) mutants rescued gene expression in the normal expression domains (Fig. 5), suggesting that Boc acts to amplify Hh signals that are otherwise subthreshold for the activation of these genes (Fig. 6E). The ventral shift in the olig2 and nkx6.1 expression domains suggests that the overall Hh signal gradient is also shifted ventrally in uml(boc) mutants (Fig. 6F). Consistent with this idea, Boc dosage seems to affect cellular sensitivity to Hh, as indicated by the increased sensitivity to the Hh-inhibiting drug CyA seen in heterozygous uml(boc) mutants (see Fig. S3 in the supplementary material).

Consistent with an activator function for Boc in Hh signaling, we show that overexpression of boc or shh can rescue Hh target gene expression in the lateral floor plate in uml(boc) mutants (Fig. 5). The fact that this rescue is not uniform could indicate that the dose of Boc, which is variable in mRNA injection experiments, is crucial for a proper Hh response. Shh injections into uml(boc) mutants resulted in minimal ectopic Hh target gene activation, suggesting that Boc can affect the Hh response throughout the CNS. Given the fact that Boc can directly bind to Shh (Tenzen et al., 2006) and our data that Boc is required cell-autonomously for maximal Hh signaling (Fig. 6), we suggest that Boc amplifies early Hh signaling by increasing local Hh concentrations and/or the duration of Hh binding to responding cells (Fig. 6E). Intriguingly, boc expression quickly becomes downregulated in the ventral CNS, suggesting that this amplification serves as a transient kick-start needed for maximal Hh signaling in the ventral CNS.

Studies using neural explants indicated that Boc can function cell-autonomously to activate Hh signaling (Tenzen et al., 2006); however, loss of Boc function in mice did not appear to grossly affect Hh-mediated CNS patterning (Okada et al., 2006), hindering a closer examination of this issue. Recent work in Drosophila uncovered a subtle requirement for the Boc homolog (boi) in the induction of Hh target genes, and this phenotype was enhanced when the closely related iHog protein was also removed (Zheng et al., 2010), suggesting overlapping functions for these two molecules. Given our results and the Drosophila data, the lack of obvious ventral CNS phenotypes in Boc mutant mice might be explained by a functional overlap with the closely related molecule Cdo. Although Cdo is expressed in the floor plate in both mouse (Tenzen et al., 2006) and fish (see Fig. S1E in the supplementary material), the timing of expression appears to be slightly later in fish, which might prevent Cdo from compensating for the loss of Boc function. If true, we would expect that roles for these proteins in activating Hh signaling would be revealed in mouse Boc-Cdo double mutants.

Transcription of boc is clearly repressed by Hh in zebrafish and mouse (see Fig. S1 in the supplementary material), consistent with negative regulators of the Hh response such as cdo and gas1 (Tenzen et al., 2006). Gas1 and Cdo can bind Shh and might act to shape the Hh response gradient by limiting diffusion or activity of Hh proteins. Indeed, explant studies show that Boc can act in a non-cell-autonomous manner to negatively regulate Hh signaling in cells adjacent to Boc-expressing cells (Tenzen et al., 2006). We did not see changes in the expression of dorsal Hh-regulated genes in the spinal cord of uml(boc) mutants. One possibility is that Cdo, Gas1 and/or other dorsally expressed Hh-binding proteins could mask a role for Boc as a negative regulator of Hh signals.

The fact that the lower jaw is present in uml(boc) mutants clearly indicates that Boc is not required as an activator of Hh-mediated cell specification and proliferation in this tissue (Brand et al., 1996; Schwend and Ahlgren, 2009; Teraoka et al., 2006). In fact, the expansion of jaw tissue in uml mutants suggests that Boc could be acting as a negative regulator of Hh signaling in this tissue. Given the direct role for Hh at later stages of jaw development (Schwend and Ahlgren, 2009) and the high level of boc expression in the jaw at this time (see Fig. S1 in the supplementary material), we propose that high levels of Boc in the jaw could bind Shh and limit the ability of Hh to activate chondrogenesis and proliferation in this tissue.

Forebrain patterning defects in uml(boc) mutants are strikingly similar to those seen in the Hh pathway mutants yot(gli2) and dtr(gli1) (Figs 1, 2) (Karlstrom et al., 1996; Karlstrom et al., 2003), consistent with a role for Boc in Hh-mediated forebrain patterning. This is supported further by the rescue of forebrain gene expression following overexpression of boc or shh in uml(boc) mutants (Fig. 5). Although a role for Boc in mammalian forebrain development has not been reported, loss of the closely related Cdo protein leads to reduced Hh target gene expression in the forebrain and defects in midline facial and brain structures associated with holoprosencephaly (HPE) that are enhanced by reductions in Hh signaling (Zhang et al., 2006). Ventral boc expression in the early zebrafish brain (see Fig. S1A in the supplementary material) is consistent with an early cell-autonomous role for Boc in Hh-mediated forebrain patterning, as we show for the spinal cord (Fig. 6). This suggests that boc might be a candidate gene for human HPE (Ming and Muenke, 1998).

Despite the fact that Boc is known to play a direct role in axon guidance of mouse commissural neurons (Okada et al., 2006) and retinal ganglion cells (Fabre et al., 2010), retinal axon guidance defects in uml(boc) mutants (Fig. 1) are probably indirect. In fact, we saw no evidence of boc expression in zebrafish RGCs up to 5 dpf, suggesting that Boc does not play a direct role in RGC axon guidance. Our work in the Hh pathway mutants yot(gli2) and dtr(gli1) showed that expanded expression of Slit-repellent cues accounts for the ipsilateral RGC axon projections seen in these mutants (Barresi et al., 2005). Similar to these mutants, slit expression is expanded across the POA in uml(boc). Midline-spanning glia are reduced in uml(boc), as they are in yot(gli2) and dtr(gli1), suggesting that POC and RGC axon guidance defects are likely to be caused by earlier defects in the forebrain growth substrate.

The reduced expression of nkx2.2a in the developing pituitary placode, combined with reductions in the pituitary endocrine cell types that require the highest levels of Hh signaling (see Fig. S2 in the supplementary material) (Guner et al., 2008; Devine et al., 2009), suggest a novel role for Boc in Hh-mediated pituitary patterning. Low-level boc expression was detected in the placode region prior to 24 hpf, however boc expression was not seen in the placode by 36 hpf. Thus, Boc might be transiently required for the maximal response to Hh in the early pituitary placode, similar to the situation in the ventral spinal cord.

Although no muscle phenotypes have been reported for Boc mutant mice, Cdo mouse mutants have delayed development of all somitic musculature (Cole et al., 2004). In cultured cells, Boc was shown to bind Cdo at its extracellular domain to form a myogenesis-promoting complex (Kang et al., 2002). The number of slow muscle fibers is slightly reduced in zebrafish uml(boc) mutants (see Fig. S3 in the supplementary material), confirming a role for Boc in muscle differentiation. The reduction in myoD expression in adaxial cells (see Fig. S3 in the supplementary material) and of ptch2 expression throughout developing somites (data not shown) in uml(boc) mutants is similar to that seen in yot(gli2) heterozygotes (Karlstrom et al., 2003), and is consistent with a reduction in Hh signal levels in these cells. Starting at 20 hpf, boc is expressed in superficial muscle fibers that are distant from the midline source of Shh (see Fig. S1J in the supplementary material) suggesting that Boc could continue to play a role in myogenesis that is independent of Hh signaling, as was recently shown for Cdo (Lu and Krauss, 2010).

The first large-scale zebrafish genetic screen uncovered 13 mutations in the Shh pathway (to date), including six mutations that affect retinotectal axon guidance. Here, we show that the last of the original ipsilateral mutants (Karlstrom et al., 1996) affects a poorly understood component of the Hh signaling pathway, Boc. Ironically, despite a known role for Boc in Hh-mediated axon guidance in the mouse spinal cord (Okada et al., 2006), our studies indicate that the retinal axon guidance errors seen in uml mutants are likely to be an indirect outcome of earlier Hh-mediated forebrain patterning defects, with Boc acting as an early activator of Hh signaling. A later role for Boc as a repressor of Hh signals is consistent with boc expression in the dorsal CNS and with the increased lower jaw structure seen in uml mutants. The uml(boc) mutant thus provides a new tool for dissecting the overlapping roles of multiple Hh repressor proteins in development. Given the key role Hh signaling plays in regulating neural and cancer stem cell proliferation, a detailed knowledge of Hh inhibitory proteins might be particularly important for the treatment of human cancers as well as for stem cell therapies.

Thanks to Brian Key for the boc MO and PCS2+boc plasmid, Bruce Appel for the Tg(nkx2.2a:megfp) line and, as always, the zebrafish community for providing in situ hybridization probes. Special thanks to Meng-Chieh Shen for technical assistance, Judy Bennett for fish care and the Karlstrom lab for help proofing the manuscript. This work was supported by NIH grants NS039994 and HD044929 (R.O.K.) and training fellowship T32 MH020051 (S.A.B.). Deposited in PMC for release after 12 months.