Modulation of dorsal root ganglion development by ErbB signaling and the scaffold protein Sorbs3

Organisms must coordinate the processes of differentiation and migration, both to effect normal embryonic development and to prevent diseases such as metastatic cancer. A particularly interesting system in which to study cell fate specification in the context of migration is the neural crest (NC). The NC arises at the conjunction of neural and non-neural ectoderm; cells then migrate and differentiate into a variety of tissue types (Le Douarin and Dupin, 2003). Signals for specification/induction, epithelial-to-mesenchymal transition, migration and differentiation all converge in this lineage. Consequently, it can be thought of as a microcosm of developmental signaling in the embryo as a whole and as a useful model for studying the coordination of these processes.

In the trunk, NC cells that migrate medially, between the neural tube and somite, give rise to sensory neurons and glia of the segmentally arranged bilateral dorsal root ganglia (DRG) near the ventralmost aspect of the neural tube. In mammals, DRG precursor migration proceeds in waves of tens to hundreds of cells, followed by significant apoptosis to produce ganglia that contain the full complement of neurons in neonates. In zebrafish, however, the initial migratory population of NC is composed of only 10-12 cells per somite, of which only one or two cells assume a sensory neuronal fate (Raible et al., 1992; Raible and Eisen, 1994; McGraw et al., 2008; McGraw et al., 2012). This relatively simple developmental pattern makes the zebrafish particularly amenable to the study of the earliest processes of DRG sensory neurogenesis.

Some degree of cell fate specification occurs prior to NC migration, but environmental cues also influence the fates of cells in different migratory paths (reviewed by Marmigère and Ernfors, 2007; Pavan and Raible, 2012). The exact mechanisms required for the integration of migration and differentiation signals in the developing NC are not known. Although a common neuro-glial progenitor population appears to give rise to neurons and glia, the timing of migration may influence whether these cells are associated with DRG (Le Douarin, 1986; Serbedzija et al., 1989; Raible and Eisen, 1996; Wright et al., 2010). Upregulation of neuron-specific transcription factors such as neurogenin1 (neurog1) and neurog2 are some of the first events in DRG neurogenesis (Greenwood et al., 1999; Ma et al., 1999; Perez et al., 1999; McGraw et al., 2008). Mechanisms that drive only some cells to activate these transcription programs while others assume glial fates instead are not well understood, although lateral inhibition via Notch/Delta signal (Hu et al., 2011; Mead and Yutzey, 2012), Hedgehog signaling (Ungos et al., 2003) and Wnt signaling (Lee et al., 2004) have been implicated.

The ErbB family of receptor tyrosine kinases plays diverse roles during development, and the paralogs erbb2 and erbb3 influence DRG formation in both mammals and zebrafish (Britsch et al., 1998; Riethmacher et al., 1997; Honjo et al., 2008; Morris et al., 1999). ErbB3 is required for several early processes crucial to DRG neuron development, including NC migration (Budi et al., 2008; Honjo et al., 2008) and glial development (Riethmacher et al., 1997; Erickson et al., 1997; Garratt et al., 2000; Lyons et al., 2005). The requirements for ErbB receptors in multiple stages of DRG development complicate our understanding of their distinct roles. ErbB receptors are also implicated in both the initiation and progression of several cancers (reviewed by Feigin and Muthuswamy, 2009), and thus unraveling methods of integration during differentiation and migration during development could have clinical implications.

We have isolated a zebrafish mutant, ouchless, in which most DRG fail to develop whereas other NC-derived tissues appear unaffected. NC migration and condensation at sites of DRG appear unchanged, yet most DRG progenitors fail to undergo neurogenesis. The mutant affects the expression of the scaffold protein gene sorbs3, also known as vinexin (Kioka et al., 1999), which is known to interact with the ErbB pathway in cell culture (Akamatsu et al., 1999; Suwa et al., 2002; Mitsushima et al., 2004; Mitsushima et al., 2006; Mitsushima et al., 2007; Mizutani et al., 2007). We show that sorbs3 is both necessary for DRG neurogenesis and sufficient to rescue the ouchless DRG phenotype. We further show that ouchless is required cell-autonomously with respect to the NC for DRG neuron differentiation, and we present evidence suggesting that sorbs3 acts in the same pathway as erbb3. We propose that sorbs3 acts to modulate a subset of ErbB signaling required specifically for DRG neurogenesis, and integrates this signal with others to promote expression of the proneural transcription factor neurog1.

AB, ouchless (sorbs3^w35), picasso (erbb3b^wpr2e2) (Budi et al., 2008), neurogenin1 (neuroD3^hi1089) (Golling et al., 2002), Tg(sox10:nlsEos) and reck (reck^w12) (Prendergast et al., 2012) embryos were obtained from natural spawning or in vitro fertilization. They were raised under standard conditions in EM (Westerfield, 1994) and staged according to Kimmel et al. (Kimmel et al., 1995). Zebrafish care followed standard procedures approved by the University of Washington Institutional Care and Use Committee.

ENU mutagenesis and early pressure screens were carried out as described previously (Lister et al., 1999; Owens et al., 2008). Using bulk segregant analysis, the ouchless lesion was linked to chromosome 8, then fine resolution mapping of a total of 1304 individuals narrowed the region to between markers z53446 and z25210. We designed primers to amplify polymorphic CA-repeats scattered throughout the region, and further narrowing it to 0.34 Mb with ca37 and ca48 as flanking markers. These simple sequence length polymorphisms (SSLP) are amplified with the following primers (5′-3′): ca-48 fwd, TGGCACCTTAAACTGATACTC; ca48 rev, GCCTTCAAATTCACCATAAA; ca37fwd, TGAGTGAAGTGAAGTAAGCCT; ca37 rev, GTTGTGGCAAGTTAGTTGG.

Embryos were fixed in 4% paraformaldehyde for 2 hours at room temperature or overnight at 4°C, then immunostained as described previously (Ungos et al., 2003). Antibodies were used with 20% goat serum at the following concentrations: mouse αElavl (mAB 16A11, Invitrogen), 1:500; rabbit αGFP (Invitrogen), 1:1000; mouse αGFP (Invitrogen), 1:100; rabbit αSox10 (Park et al., 2005; a gift from B. Appel), 1:1000; rabbit αMBP (Lyons et al., 2005; a gift from W. Talbot), 1:50. Embryos were imaged as described previously (McGraw et al., 2008).

RNA in situ hybridization was performed as described previously (Andermann et al., 2002). Digoxigenin-labeled probes for sox10 (Dutton et al., 2001), neurog1 (Korzh et al., 1998), crestin (Rubinstein et al., 2000) and neuroD (Blader et al., 1997) were generated as described. sorbs3 riboprobes were generated by amplifying fragments from cDNA clones, inserting them into the pCRII-TOPO vector (Clontech), linearizing the vector and transcribing with RNA polymerase using the following oligos: sorbs3F, AACCCCTGATAAAGCATGTAC; sorbs3R, CCCGATCCTCAGGCTATG; pdlim2F, GCGATGAGAACCAGAACTAC; pdlim2R, GTTTGTGTGCGTGTATATCTT.

Four hundred to 450 dechorionated embryos were dissociated for 20-30 minutes in 5.5 ml of digestion solution [0.25× trypsin; HyClone), 5 mM EDTA, 10 μl DNaseI (Invitrogen) in 1× PBS] with trituration through a micropipette tip. Kill solution (1.4 ml) [50% FBS (Gibco), 5 mM CaCl₂ in 1× PBS] was added to stop digestion. Cells were centrifuged for 3 minutes at 1300 g, rinsed with 1× PBS, centrifuged for 3 minutes at 1300 g and resuspended in 1 ml of suspension solution [900 μl of 1× Leibovitz’s L15 media (Gibco), 10% FBS and 0.8 mM CaCl₂]. Eos+ cells were sorted with a FACSAria flow cytometer. RNA was isolated using RNeasy (Qiagen); cDNA was amplified from DNase-treated RNA using Superscript III (Invitrogen); RT-PCR was performed using Ssofast supermix (Bio-Rad) on a CFX connect cycler and analyzed using CFX manager 3.0 software (Bio-Rad). Primers used for amplification from cDNA were bactinF (CCCAAGGCCAACAGGGAAAA), bactinR (GGTGCCCATCTCCTGCTCAA), sox10F (ATCCCGAGTACAAGTACCAGCCAC), sox10R (TGCAGGCTCTGTAATGCGATTGG), sorbs3F (TCTTCCACATCCCAGTCTCC), sorbs3R (CACATGCTTGGTGAAGAACG).

A MO predicted to block splicing of exons 5 and 6 of Sorbs3 with the sequence TTCCGACAGGGAAAGCACATACC was obtained from Gene Tools. Two to 6 ng MO in water + 0.2% Phenol Red was injected into a one-cell embryos carrying the TgBAC(neurog1:dsred) transgene (Drerup and Nechiporuk, 2013). DsRed-positive DRG were counted at 2-4 days. Fish were anesthetized in MS-222, then fixed for 2 hours at room temperature or overnight at 4°C in 4% paraformaldehyde.

BAC DNA was prepared from clones zK179C10 and zH17L17 (ImaGenes GmbH) using the Nucleobond Extra-Midi Plus kit (Machery Nagel). One to 2 μl DNA, 0.1 M KCl and 0.2% Phenol Red was injected into ouchless; Tg(neurog1:GFP) embryos at the one-cell stage. Embryos were fixed at 3 days postfertilization (dpf) and stained with mouse α-Elavl and rabbit α-GFP antibodies. sorbs3-coding regions were inserted into pCS2MT, and mRNA was synthesized from linearized DNA using the mMessage mMachine in vitro transcription kit (Ambion). mRNA was diluted in water and 0.2% Phenol Red and injected into ouchless; Tg(neurog1:EGFP) embryos. GFP⁺ DRG were counted on one side of the embryos at 48 hpf, 72 hpf and 96 hpf using a Nikon SMZ1500 fluorescent dissecting microscope.

Mosaic embryos were generated by cell transplantation at early gastrula stages as described previously (Carmany-Rampey and Moens, 2006). Donor embryos were injected at the one-cell stage with 1 nl 0.2% rhodamine-dextran (Sigma) and cells from sphere stage donors were transplanted into shield stage host embryos. Melanophore-deficient nacre hosts (Lister et al., 1999) were used to identify mosaic embryos in which transplanted cells had contributed to NC.

ErbB inhibitors PD158780 (Tocris Bioscience) and AG1478 (Tocris Bioscience) and MEK inhibitor PD0325901 (Stemgent) were dissolved in DMSO. Embryos were treated with drug and a final concentration of 1% DMSO at 18 hpf, and replenished with fresh drug and media at 42 hpf. At 70-72 hpf, embryos were fixed and stained with mouse α-Elavl and rabbit α-GFP antibodies. Embryos were genotyped as described previously for neurog1 (Golling et al., 2002) or for reck (Prendergast et al., 2012).

Contingency tables were analyzed using Statistics To Use (http://www.physics.csbsju.edu/stats/). All other statistics were generated using GraphPad Prism for MacOSX version 5.0 (GraphPad, La Jolla, CA).

To discover genes with essential functions in sensory neuron development, we conducted a screen for mutants with abnormal DRG. Fish were examined at 3 dpf using antibody against Elavl, which marks differentiated neurons. The ouchless mutant, though viable and morphologically indistinguishable from wild-type siblings (Fig. 1A,B), lacks most DRG at 3 dpf. DRG development was examined using the Tg(neurog1:EGFP) line (McGraw et al., 2008), where GFP is expressed in newly specified neurons and subsequently downregulated in mature Elavl⁺ cells. We found severe reductions in DRG in ouchless mutants (Fig. 1C,D). The few DRG that develop are most frequently found in the first eight segments. ouchless DRG contain fewer neurons than those in wild type (supplementary material Fig. S1C,D). Spatial arrangement of each remaining ganglion, as well as orientation of axonal projections, appears normal in mutants (supplementary material Fig. S1A,B). More neurons differentiate in wild-type DRG as animals age (McGraw et al., 2012); by contrast, ouchless fish show reduced neuron addition in the few DRG that form (supplementary material Fig. S1C-D). ouchless mutants do not add DRG after 3 dpf, suggesting the defect does not result in a simple developmental delay. Although ouchless mutants eventually grow to adult stages, they never gain normal numbers of DRG (not shown). We were not able to find any changes in Elavl expression in or morphology of sympathetic ganglia (SG), enteric neurons or cranial ganglia in mutants (supplementary material Fig. S2). These results indicate a specific requirement for ouchless in formation of DRG, but not other NC-derived neurons.

DRG development is the result of several coordinated events, including NC specification, precursor migration and differentiation of neurons and glia. As other NC-derived cells develop normally in ouchless mutants, we reasoned that NC specification and migration were unlikely to be perturbed. Indeed, sox10 and crestin expression by in situ hybridization showed no differences in the pattern of ventrally migrating cells at 24 hours post-fertilization (hpf) between wild type and ouchless mutants (Fig. 1G,H; data not shown).

The earliest marker for newly formed DRG sensory neurons is neurog1, which can be detected as early as 24 hpf. Segmented clusters of cells ventral to the spinal cord express neurog1 in wild-type embryos, but only occasional neurog1⁺ cells could be detected in ouchless embryos (Fig. 1I,J). neuroD, a marker of maturing neurons (Blader et al., 1997), was expressed in a similar sporadic pattern in ouchless at 36 hpf (Fig. 1K,L), suggesting that the few cells that initiate the neuronal program continue their differentiation. These results suggest that lack of DRG neurons in mutants originates from impaired specification of the correct number of neurog1⁺ precursors. Other peripheral neurons that require neurog1 function, such as Rohon-Beard cells and cranial ganglia (Andermann et al., 2002), develop normally (supplementary material Fig. S2), suggesting that defects in the initiation of neurog1 expression are limited to migrating trunk NC.

We next sought to determine the gene affected in ouchless embryos. Bulk segregation analysis localized the mutation to chromosome 8, confirmed by analysis of 1304 individual embryos to a region between marker z53446 and z25210. Analysis of de novo generated SSLP markers narrow the region to the 342 kb flanked by markers ca37 and ca48 (Fig. 2A). This interval contained two annotated genes, pdlim2 and sorbs3. We sequenced the entire coding sequence of both genes from wild type and ouchless but were unable to identify a lesion, suggesting that mutation in a regulatory region may be responsible.

We used a rescue approach to further limit the region containing the ouchless lesion. We injected the bacterial artificial chromosome (BAC) clone zK179C10, which includes the gene sorbs3 and two other genes outside of the recombination interval but not pdlim2, into ouchless embryos (Fig. 2A,B). This resulted in a dose-dependent rescue of the number of DRG at 3 dpf. We confirmed this result using a second BAC, zH17L17, which contains only the sorbs3 gene. The other two genes contained in zK179C10 were sequenced and no lesions found. Furthermore, analysis of their mRNA expression patterns by in situ hybridization did not reveal expression consistent with roles in DRG development (not shown). Taken together with recombination mapping data, this suggests that the ouchless lesion affects the sorbs3 gene.

The mouse Sorbs3 gene, which encodes several alternatively spliced isoforms of the Sorbs3/vinexin protein, consists of 19 exons. A long isoform, Sorbs3α, contains a sorbin homology domain in the N-terminus and 3 C-terminal Src homology 3 (SH3) domains. An intronic promoter controls transcription of a shorter Sorbs3β isoform that lacks 5′ sequences, including the sorbin domain (Kioka et al., 1999). The zebrafish sorbs3 locus contains 21 exons and is predicted to encode a protein with the same domain arrangement as the mammalian form (Fig. 2A). We have isolated transcripts containing message complementary to exons 14-21, which probably encodes Sorbs3β, and longer transcripts that encode isoforms corresponding to mammalian Sorbsα. Sorbs3 mRNAs are appropriately spliced in ouchless; we confirmed that splice donor and acceptor sites for each exon are unaffected in mutants with genomic DNA sequencing.

To test whether re-expression of sorbs3 was sufficient to rescue mutants, we generated mRNA expression constructs in which the sorbs3-coding sequences were fused to an N-terminal Myc tag. mRNA was injected into one-cell ouchless; Tg(neurog1:EGFP) embryos. DRG were counted each day from 2 to 3 dpf, and at 3 dpf animals were fixed and stained with EGFP and Elavl antibodies to label differentiated neurons. mRNA (50 pg) encoding a long isoform of sorbs3 was able to partially rescue the DRG defect (Fig. 2C). Higher concentrations of mRNA were deleterious to embryos. We observed no changes in DRG development with overexpression of sorbs3 in wild-type embryos. mRNA for the shorter sorbs3b isoform was also able to rescue DRG defects (Fig. 2C). This suggests that the N-terminal region of Sorbs3, including the sorbin domain, is dispensable for DRG development and that the crucial region is contained in the C-terminal SH3 domains.

To validate that sorbs3 function affects DRG development, we designed a MO to block splicing of the fifth and sixth exons of sorbs3, injected it into one-cell stage Tg(neurog1:EGFP) embryos, fixed embryos at 3 dpf and stained with GFP and Elavl antibodies. We observed a dose-dependent loss of DRG in injected embryos; wild-type embryos injected with 6 ng MO developed significantly fewer DRG than uninjected embryos (Fig. 3A). Fish heterozygous for ouchless developed significantly fewer DRG than their wild-type counterparts when injected with as little as 2 ng MO. This sensitization to MO injection provides additional support to the hypothesis that the ouchless lesion affects sorbs3. Mis-splicing of cDNA after MO injection was confirmed using RT-PCR (Fig. 3B). MO injection at these concentrations causes few other phenotypes, suggesting that the decrease in DRG formation is specific (Fig. 3C).

We designed an in situ hybridization probe that recognizes mRNA corresponding to exons 4-21 of sorbs3. At 19 hpf, sorbs3 is expressed in the anterior neural tube, with expression decreasing posteriorly (Fig. 4A). This pattern continues through 24 hpf, but expression is downregulated by 30 hpf (Fig. 4E). Strong sorbs3 expression is found in the pronephros and more moderate expression in forebrain and skin. Although ouchless embryos show similar sorbs3 expression patterns to wild-type embryos in head and pronephros, they exhibit lower levels in the neural tube (Fig. 4B,F). In transverse sections at the level of the yolk extension of wild-type embryos, sorbs3 expression is concentrated to the ventral neural tube, but there is also weaker expression in cells resembling NC ventrolateral to the spinal cord (Fig. 4C, arrows). These could not be found in ouchless mutants (Fig. 4D). By RT-PCR analysis, expression of sorbs3 mRNA in ouchless embryos is approximately twofold less than wild-type levels at 28-30 hpf (Table 1). Using fluorescence-activated cell sorting, we confirmed that both NC and non-NC cells express sorbs3 mRNA in wild-type embryos at 28-30 hpf (Table 1).

As sorbs3 is expressed both in NC and in surrounding cells, we constructed mosaic embryos to elucidate its tissue-specific requirement for DRG development. We transplanted donor cells from 3 hpf embryos into the prospective NC domain at 6 hpf. We used nacre hosts, which lack NC-derived melanophores due to a mutation in the mitfa gene (Lister et al., 1999) to confirm that NC was successfully targeted by transplantation. We reasoned that if sorbs3 is required cell-autonomously in the NC, then wild-type cells transplanted into ouchless hosts should be able to generate sensory neurons. If, on the other hand, sorbs3 acts non-cell autonomously, wild-type cells should not be able to give rise to DRG more frequently than ouchless mutant cells.

Chimeras of rhodamine labeled Tg(neurog1:gfp) donor cells transplanted into nacre hosts were analyzed at 3 dpf for pigment, Tg(neurog1:gfp) and Elavl expression, and rhodamine distribution. Of 44 transplants, 31 (70%) contained pigmented cells, demonstrating transfer of NC precursors, and of those 48% had DRG derived from donors (Fig. 5A,B, Table 2). In complementary experiments, transplanting ouchless cells (neurog1:gfp+, rhodamine labeled) into nacre embryos, resulted in a comparable percentage of pigmented larvae (66%, 37/56) but ouchless cells never gave rise to DRG (Fig. 5D-F). In both sets of experiments, the nacre host cells gave rise to normal Elavl⁺ DRG neurons (Fig. 5C,F). These results showed that sorbs3 is required cell-autonomously for DRG development, but did not exclude a second, non-cell autonomous function for sorbs3. We addressed this question by transplanting wild-type cells into ouchless mutants and found that wild-type cells were able to generate sensory neurons, indicating that DRG precursors do not require extrinsic sorbs3 for their development (Fig. 5G,H). Taken together, these data suggest a cell-autonomous role for ouchless within DRG precursors.

Despite their severe DRG defects, a proportion of homozygous ouchless mutants survive until adulthood and are able to reproduce. Though grossly indistinguishable from wild-type and heterozygous siblings as larvae (Fig. 1A,B), adult mutants can be identified by their slow growth. Mutants largely fail to produce the metamorphic pigment cells that begin to appear at 3-4 weeks and do not fully form the adult pigment pattern of wild-type zebrafish (Fig. 6A,B). This eventually results in broken melanophore stripes that consist of few and dispersed melanophores surrounded by iridophores and xanthophores in adults (Fig. 6C,D). The mutant pattern phenotype is strikingly similar, though less severe than that of picasso mutants, which have been shown to possess a mutation in the erbb3b gene (Budi et al., 2008) (Fig. 6E,F). erbb3b mutants also have a DRG phenotype resembling that of ouchless mutants (Honjo et al., 2008).

The phenotypic similarities of ouchless and erbb mutants led us to hypothesize that Sorbs3 might be a downstream modulator of ErbB signaling during the development of DRG. In cultured cells, the third SH3 domain of Sorbs3 binds to Son-of-sevenless (Sos), and this binding activity is known to modulate ErbB receptor tyrosine kinase signaling (Akamatsu et al., 1999). In the absence of erbb3, neither DRG sensory neurons nor glia develop in zebrafish embryos, because NC cells, instead of pausing to condense and form ganglia, continue to migrate ventrally (Honjo et al., 2008; Lyons et al., 2005; Britsch et al., 1998). To examine this possibility, we visualized DRG progenitor condensation at 24 and 36 hpf in wild-type, ouchless and erbb3b embryos. Sox10-positive NC cells, migrating ventrally in a comb-like pattern over the trunk at 24 hpf, were arranged in distinct clusters adjacent to the ventral neural tube in both wild-type and ouchless embryos by 36 hpf (Fig. 6G,H,J,K). Consistent with previous results (Honjo et al., 2008), erbb3b mutants did not form any Sox10-expressing aggregates at 36 hpf, even though initial NC migration was normal at 24 hpf (Fig. 6I,L). Thus, although erbb3 mutants have dramatic defects in DRG condensation, this process appears to be independent of the changes in sorbs3 found in ouchless mutants.

ErbB signaling, besides being required for DRG progenitor condensation, has a separate function required for the development of glial cells in the DRG and other tissues (Riethmacher et al., 1997; Britsch et al., 1998; Lyons et al., 2005; Chen et al., 2003; Sharghi-Namini et al., 2006; Pogoda et al., 2006) (reviewed by Britsch, 2007). We sought to examine the development of peripheral glia in ouchless mutants to determine whether this ErbB-dependent process is also dependent on Sorbs3. The dearth of markers of DRG satellite glia prevented us from directly examining the development of these cell types, so we examined the development of other peripheral glia cell types known to be affected in erbb mutants. We stained fish at 5 dpf for myelin basic protein (MBP), a marker for myelinating glia (Brösamle and Halpern, 2002) and found MBP+ cells surrounding both the posterior lateral line nerve (PLL) in both wild type and ouchless mutants (Fig. 6M,N). The PLL of erbb3b mutants lacked MBP staining (Fig. 6O), as described previously (Lyons et al., 2005). Patterning of the lateral line system, which is dependent on peripheral glial cells (Grant et al., 2005), is also normal in ouchless mutants but defective in erbb3 mutants (data not shown). To exclude the possibility that the lack of a glial phenotype may be due to potential residual sorbs3 function in ouchless mutants, we injected ouchless embryos carrying Tg(sox10:nlsEos) with 6 ng sorbs3 MO. We examined the peripheral glia of the DRG at 30 hpf and of the lateral line at 5 dpf, and found no differences compared with wild-type fish (Fig. 6P-S). Thus, the development of peripheral glia, although requiring ErbB function, does not seem to require sorbs3.

To further assess the idea that Sorbs3 and ErbB functions intersect, we tested for genetic interactions. Fish heterozygous for either ouchless or erbb3b were indistinguishable from wild type (Table 3). Those heterozygous for both ouchless and erbb3b showed significantly fewer DRG (P>0.05) than either wild-type or ouchless/+ fish, but did not differ significantly from erbb3b/+ larvae. To further characterize this subtle difference, we scored whether DRG were missing in the trunk or in the tail. Fish of all genotypes occasionally lacked DRG in late-developing posterior tail segments, but we found that the number of fish lacking DRG in the trunk region of double heterozygous fish was significantly larger than all other genotypes (Table 3). Although this genetic interaction is subtle, it suggests that sorbs3 and erbb3 may function in the same biochemical pathway.

We reasoned that if Sorbs3 acts in the same pathway as ErbB3, heterozygous ouchless larvae might be sensitized to further disruptions in that pathway. Lyons et al. (Lyons et al., 2005) have previously shown that treatment of erbb3 heterozygotes with pharmacological ErbB inhibitors shows a dose-dependent effect on myelin gene expression. We tested whether similar interactions might occur to regulate DRG development using two inhibitors: AG1478 (Lyons et al., 2005; Levitzki and Mishani, 2006) and PD158780 (Fry et al., 1997; Rewcastle et al., 1998; Frohnert et al., 2003). Treating fish from 18-72 hpf with either inhibitor resulted in a specific and dose-dependent loss of DRG (Fig. 7A, data not shown). PD158780 showed less toxicity than AG1478, so we continued using this inhibitor.

Activation of ErbB signaling is known to initiate several signaling cascades, including MAP kinase activation. We treated fish with the MEK (MAP kinase kinase) inhibitor PD0325901 over a range of concentrations from 18-72 hpf, then counted the number of neurog1:gfp-positive DRG. PD0325901 showed dose-dependent effects on DRG development (Fig. 7B), similar to that for ErbB inhibitors. These data suggest that the MAP kinase pathway is required for the development of DRG.

We next sought to determine whether heterozygous mutant embryos were sensitized to ErbB inhibition. When treated at a suboptimal concentration of 0.5 μM PD158780, both ouchless/+ and erbb3b/+ fish showed sensitization to inhibition when compared with wild-type fish (Fig. 7C). neurog1 appears to be downstream of both ErbB3 and Sorbs3, as its expression is rarely initiated in picasso or ouchless DRG. We therefore tested whether neurog1 heterozygotes were also sensitized to ErbB inhibition. When neurog1/+ embryos (Golling et al., 2002) were treated with ErbB3 inhibitors, they developed significantly fewer DRG than their wild-type siblings (Fig. 7C). Embryos heterozygous for reck loss of function, another gene necessary for DRG development (Prendergast et al., 2012), are not sensitized to ErbB inhibition, (Fig. 7C), suggesting that ErbB inhibitor treatment is specifically affecting a developmental pathway that includes a novel link to sorbs3 and neurog1.

In this study, we have identified and characterized a zebrafish mutant, ouchless, which develops only a small subset of the normal complement of DRG, fails to add neurons to the DRG that do develop at a normal rate and fails to develop normal numbers of adult melanophores. All of these phenotypes are reminiscent of a subset of those seen in mutants for the EGFR receptor tyrosine kinases ErbB2 and ErbB3 (Honjo et al., 2008; Budi et al., 2008; Honjo et al., 2011). Zebrafish adult pigment cells form from progenitor cells that are sometimes localized to peripheral nerves and DRG, and specification of these cells is dependent on erbb3b (Budi et al., 2011). Together, these observations support the idea that ErbB and Sorbs3 are required in the same population of progenitor for the differentiation of all of these cell types.

The ouchless mutation affects the expression of sorbs3, a scaffold protein known to interact with components of focal adhesions, as well as with effectors of ErbB signaling, including Sos, Raf and ERK (Kioka et al., 1999; Akamatsu et al., 1999; Mitsushima et al., 2004; Matsuyama et al., 2005; Mitsushima et al., 2007; Gehmlich et al., 2007). Our experiments support the hypothesis that Sorbs3 and ErbB2/3 are part of the same genetic pathway. ErbB2/3 and Sorbs3 are both required for neurog1 expression in DRG precursors, ouchless and the erbb3b allele picasso interact genetically in DRG neurogenesis, and neurog1/+ embryos are also sensitized to the DRG-depleting effects of ErbB signaling. Taken together, these results suggest a crucial role for Sorbs3 in DRG development and shed light on the processes of neuron differentiation through ErbB signaling. Although we favor a model where Sorbs3 directly modulates ErbB signaling, we cannot rule out that interactions may be indirect through parallel pathways.

Sorbs3 has been previously implicated in neuronal function but not initial specification or differentiation. Sorbs3 is expressed in the developing rat and mouse brains (Kawauchi et al., 2001), and is localized to developing growth cones and filopodia, as well as to synapses (Ito et al., 2007; Ito et al., 2008). However, both isoform-specific and complete knockouts of Sorbs3 in mice do not have any apparent DRG or pigment phenotypes (Matsuyama et al., 2005; Kioka et al., 2010). There are several possible explanations for the differences between phenotypes of ouchless mutants in zebrafish and mouse Sorbs3 knockouts. One possibility is that mammalian paralogs of Sorbs3, CAP/Ponsin/Sorbs1 and ArgBP2/Sorbs2 (Kioka et al., 2002) may compensate for the loss of Sorbs3 in most developmental processes, including the development of DRG. Transcriptome analysis of mouse embryos suggests that at least one of these paralogs is expressed in developing DRG (Diez-Roux et al., 2011). Alternatively, these phenotypic differences, like those observed between the mouse and fish after ErbB loss of function, may be a result of distinct regulatory pathways for DRG development within these organisms.

We propose a model in which Sorbs3 acts as a scaffold to bind components of ErbB signaling complexes to promote the expression of neurog1 and subsequent differentiation of DRG neurons. This function of Sorbs3 would likely occur within the precursors of DRG neurons, as Sorbs3 is required cell-autonomously. Although Sorbs3 has not been directly implicated to modulate ErbB2/3 signaling, there is extensive evidence for its interaction with EGFR signaling. Sorbs3 binds to the adapter protein Sos (Akamatsu et al., 1999), and this binding is required for the activation of the MAPKs ERK and JNK by the ErbB ligand EGF (Akamatsu et al., 1999; Suwa et al., 2002). Sorbs3 binds to ERK1/2, as well as to c-Raf (Mitsushima et al., 2004; Matsuyama et al., 2005), and prevents the dephosphorylation, thus promoting the kinase activities of both EGFR and ERK1/2 (Mitsushima et al., 2006; Mitsushima et al., 2007). At the same time, ERK itself phosphorylates Sorbs3, which inhibits cell migration (Mitsushima et al., 2004; Mizutani et al., 2007). Expression of Sorbs3 in v-src transformed cells also suppresses their enhanced migration phenotype (Umemoto et al., 2009). EGFR and ErbB2/3 signaling share many of these signaling components, including Sos, Raf and ERK (Yarden and Sliwkowski, 2001), suggesting that similar interactions may underlie Sorbs3 and ErbB2/3 functions during DRG sensory neuron development.

Although MAPK signaling has been implicated in regulating several aspects of mammalian neuron development, the roles of specific signaling components in DRG specification remain unclear. ERK5, which is activated downstream of ErbB2/3 (Esparís-Ogando et al., 2002), regulates neurog1 during cortical neurogenesis (Cundiff et al., 2009), suggesting it might play a similar role during DRG development. Activated ERK2 is found in developing NC and peripheral nervous system (Corson et al., 2003), its activity is required for some aspects of NC development (Newbern et al., 2008), and its phosphorylation can activate neurogenesis programs (Kim et al., 2004). However, although genetic inactivation of ERK1/2 results in significant effects on sensory neuron axon outgrowth and survival, loss of ERK1/2 or ERK5 results in little gross changes in initial DRG formation (Newbern et al., 2011). The specific components of the MAPK signaling pathway that are involved in zebrafish sensory neuron development remain to be identified.

Disruption of ErbB2/3 signaling has a greater phenotypic effect than that resulting from interference with Sorbs3. In particular, glial development appears normal in ouchless mutants, in contrast to disruptions caused by loss of ErbB signaling. These observations suggest that while ErbB2/3 plays multiple roles in NC development, the outcomes of signaling are refined by Sorbs3 interaction through integration of other signals. A candidate for an additional signaling pathway modulated by Sorbs3 is focal adhesion signaling. Sorbs3 is often localized to, promotes the assembly of and binds to several components of focal adhesions (Kioka et al., 1999; Gehmlich et al., 2007; Thompson et al., 2010). Several integrins, which are extracellular components of focal adhesions, have been implicated in NC migration, and loss of these proteins cause defects in NC development (reviewed by Perris and Perissinotto, 2000). Focal adhesions are signaling centers, and are often characterized by high concentrations of ErbB2/3 dimers (reviewed by Pinon and Wehrle-Haller, 2011), poising ErbBs and other focal adhesion components to physically interact. The coordination of ErbB and focal adhesion signaling in the DRG is plausible, given their synergism in other systems. ErbB2 promotes the formation and stabilization of focal adhesions in breast carcinoma (Zaoui et al., 2010; Marone et al., 2004; Zaoui et al., 2008) and cultured adult neurons (Grimm et al., 2009; Grimm et al., 2010). Scaffolding of adhesion signals has already been demonstrated in the NC. Nedd9, a member of the p130Cas family of scaffold proteins is required for the integration of integrin-based adhesions and NC migration (Aquino et al., 2009). Interestingly, p130Cas family proteins are known to coordinate signals from adhesions and ErbB2 to promote aggressiveness and metastases in several cancer models (reviewed by Cabodi et al., 2010).

A crucial step in DRG development regulated by Sorbs3 and by ErbB2/3 may be NC cell migration. Sorbs3 has been implicated in keratinocyte cell migration and wound healing, events also regulated by EGFR signaling (Kioka et al., 2010). In addition, Sorbs3 overexpression inhibits cell migration in vitro, and phosphorylated forms localize to the leading edge of migrating cells (Mitsushima et al., 2004; Mizutani et al., 2007; Umemoto et al., 2009). There is a large body of evidence that ErbB receptors regulate cell migration (reviewed by Feigin and Muthuswamy, 2009), and zebrafish NC cell migration is altered after loss of erbb2/3 (Honjo et al., 2008). Although we have found no gross deficits in ouchless mutants or after sorbs3 MO injection, whether there are subtle effects on migration of specific DRG precursors within the migrating NC population will require more detailed analysis.

ErbB signaling is already a well-established player in several cancers, including those of the NC. The Sorbs3 paralog ArgBP2 has been implicated in anti-oncogenic processes (Roignot and Soubeyran, 2009), raising the possibility that Sorbs3 could also be an important modulator of ErbB signaling in cancers. As many developmental signaling processes are recapitulated in error to cause cancer, further study of the role of the Sorbs3 protein in developmental processes may provide new insight into cancer biology.

We thank the Parichy lab for picasso mutants, and Will Talbot and Bruce Appel for antibodies. We thank Dave White and staff of the University of Washington (UW) fish facilities, and are grateful to the UW fish and developmental biology communities for advice and support.