Regulation of neurogenesis by Fgf8a requires Cdc42 signaling and a novel Cdc42 effector protein.

Fibroblast growth factor (FGF) signaling is required for numerous aspects of neural development, including neural induction, CNS patterning and neurogenesis. The ability of FGFs to activate Ras/MAPK signaling is thought to be critical for these functions. However, it is unlikely that MAPK signaling can fully explain the diversity of responses to FGFs. We have characterized a Cdc42-dependent signaling pathway operating downstream of the Fgf8a splice isoform. We show that a Cdc42 effector 4-like protein (Cdc42ep4-l or Cep4l) has robust neuronal-inducing activity in Xenopus embryos. Furthermore, we find that Cep4l and Cdc42 itself are necessary and sufficient for sensory neurogenesis in vivo. Furthermore, both proteins are involved in Fgf8a-induced neuronal induction, and Cdc42/Cep4l association is promoted specifically by the Fgf8a isoform of Fgf8, but not by Fgf8b, which lacks neuronal inducing activity. Overall, these data suggest a novel role for Cdc42 in an Fgf8a-specific signaling pathway essential for vertebrate neuronal development.


INTRODUCTION
Neuronal differentiation within the neural plate is controlled by the inhibition of Notch signaling (Chitnis et al., 1995), resulting in the expression of neurogenic bHLH transcription factors including neurog2 and neuroD (Lee et al., 1995;Turner and Weintraub, 1994). The specific cell fates of primary neurons (i.e., larval-specific) are regulated differently along the mediolateral axis of the embryo, which is later transposed into dorsoventral patterning of the neural tube following neurulation. In amphibians, primary neurons appear in three stripes on either side of the dorsal midline within the neural plate, corresponding to lateral sensory neurons, intermediate interneurons and medial motorneurons. Primary neuronal differentiation takes place mainly in the deep (sensorial) layer of the neural plate, the cells of which are not polarized, and is excluded from the apically polarized cells of the superficial epithelium (SE; Hartenstein, 1989;Chalmers et al., 2002). Cell cycle exit and neuronal differentiation occur in a spatiotemporal wave from the lateral to medial neural plate (Hartenstein, 1989). The lateral primary sensory (Rohon-Beard) neurons are the first to terminally differentiate, with birthdating experiments suggesting this occurs mostly during the mid-gastrula stages in both frogs and fish (Lamborghinr 1980;Metcalfe et al., 1990).
Different neuronal subtypes are specified by graded amounts of specific molecular signals. Sonic hedgehog signaling from the notochord and floor plate is largely responsible for the specification of the ventral motorneurons and interneurons (Briscoe and Ericson, 2001). By contrast, the dorsal primary sensory neurons are specified independently from Shh and are regulated by the same signals that also induce non-neuronal fates in cells derived from the neural plate border region. These cell fates include hatching gland, preplacode, and neural crest, all of which also derive from the deep layer of the ectoderm (Park et al., 2012;Rossi et al., 2008;Rossi et al., 2009). Dorsal neuron and neural plate border inducing signals include BMPs and Wnts from the non-neural ectoderm as well as FGFs. In particular, graded Fgf8a induces preplacode and neural crest fates at low doses, whereas high doses induce sensory neurons and inhibit neural plate border fates (Fletcher et al., 2006;Hardcastle et al., 2000;Park et al., 2012).
In Xenopus, FGF signaling through Fgf4 and multiple Fgf8 splice isoforms is critical for many aspects of both mesoderm and neural development. Interestingly, Fgf8 splice isoforms, Fgf8a and Fgf8b, differ by 11 amino acids but elicit different responses in vitro and in vivo (Fletcher et al., 2006;Lee et al., 1997;Liu et al., 1999;MacArthur et al., 1995;Olsen et al., 2006;Sato et al., 2001;Sato and Nakamura, 2004). In Xenopus, Fgf8a induces ectopic neurogenesis throughout the epidermis whereas Fgf8b induces mesoderm (Fletcher et al., 2006;Hardcastle et al., 2000). The basis for these differences is not known, but may result from weaker binding of Fgf8a to the set of Fgf8-binding receptors (Olsen et al., 2006). In the chick, Fgf8a is thought to act as a weak Fgf8 ligand, since electroporation of low levels of Fgf8b resembled the Fgf8a effect (Sato and Nakamura, 2004). However, in frog embryos, high amounts of fgf8a did not mimic fgf8b, suggesting a qualitative difference (Fletcher et al., 2006). Prior to the work presented here, specific signaling pathways downstream of Fgf8a have not been identified.
Activation of FGFR signaling and tyrosine phosphorylation generates a host of wellcharacterized signal transduction responses (Dailey et al., 2005). The Ras/MAPK kinase cascade is a commonly studied outcome of FGFR signaling, but Rho GTPases (Rho,Rac Cdc24,et al.) can also be activated (Dailey et al., 2005). The specific functions of Rho GTPases are mediated by the binding of different effector proteins. Individual Guanine nucleotide exchange factors (GEFs) are thought to help recruit specific effectors to the monomeric GTPases they activate, thereby assembling compartmentalized signaling complexes (Bos et al., 2007;Sinha and Yang, 2008).
We have identified a role for one such effector protein in neuronal development, a member of the Cdc42 effector protein family (a.k.a., Borgs or CEPs; (Hirsch et al., 2001;Joberty et al., 1999)), Cdc42ep4-like/Cep4l. Cep4l and Cdc42ep4 proteins have well-characterized Nterminal CRIB (Cdc42/Rac Interactive Binding) domain that mediates binding to activated Cdc42 and Rac, but notto Rho (Burbelo et al., 1995). Cdc42ep family proteins have been characterized using in vitro, cell line and yeast-two hybrid assays, and bind only to activated Cdc42 and TC10/RhoQ, but not to activated Rac or Rho (Hirsch et al., 2001;Joberty et al., 1999). Cdc42ep4 has been shown to induce pseudopodia formation when transfected into fibroblasts (Hirsch et al., 2001), butthe overall functions of the Cdc42ep family are unknown.
In this study, we show for the first time that a Cdc42 and Cdc42 effector-regulated signaling pathway is necessary and sufficient for neurogenesis in Xenopus. We show that Cep4l and Cdc42 overexpression in Xenopus embryos induces primary sensory neurons throughout the deep layer of the epidermis, and antisense knockdown of these molecules reduces this neuronal population. These effects resembled the induction of neurons in the epidermis by the Fgf8a splice isoform and we further show that Fgf8a stimulates Cep4l association with Cdc42. We also find that these proteins are required for robust neuronal induction by Fgf8a. Overall, these data identify a specific Cdc42-dependent molecular mechanism for neuronal specification likely acting downstream of Fgf8a isoform signaling.

Xenopus Embryos and Explants
Xenopus adult females were induced to lay eggs by injection ofhuman chorionic gonadotropin (hCG). Eggs were fertilized in vitro using a sperm suspension, cultured in 0.1X MMR (1X MMR: 0.1 M NaCl, 1.8 mM KCl, 2.0 mM CaCl 2 , 1.0 mM MgCl 2 , 15.0 mM HEPES, pH 7.6) and dejellied in cysteine (2% in 0.1X MMR). Animal caps were dissected at stage 9 (Nieuwkoop and Faber, 1994) and cultured in 1X MMR to the indicated stage. For dp-ERK assays, caps were cultured for 60 minutes to avoid wound-induced ERK activation.

Plasmids and mRNA Synthesis
Full-length cDNAs for cep4l and cdc42 in pCMV-SPORT6 were obtained commercially (Open Biosystems). The cep4l and cdc42 coding regions (CDS), or cep4l CDS along with the 5'UTR, were amplified by PCR and cloned into pCR8/GW/TOPO (Invitrogen). Clones were sequenced and inserted into pCS2+/GW, pCS2-HA/GW, or pCS2-FLAG/GW custom Gateway converted vectors via LR recombination (Invitrogen). For pCS2+/GW, reading frame A cassette (Invitrogen vector conversion kit) was inserted into the StuI site of pCS2+. For the HA-tag vector, reading frame B cassette was cloned into ClaI digested/Klenow filled pCS2+HA to generate a 2x HA fusion C-terminally. For the FLAG-tag vector, reading frame A was cloned into the StuI site of pCS2+FLAG to generate a 2x FLAG fusion Nterminally. The cassettes were chosen so that constructs with complete triplet codons cloned into pCR8/GW/TOPO and recombined into the vectors will be in frame with the epitope tags. The parent vectors were gifts from J. Yang. A CRIB domain mutant (cep4lH31A,H34A-pCS2) was produced using site-directed mutagenesis (Quickchange II kit; Stratagene). Template DNA was linearized for sense transcription of mRNAs from cep4l-pCS2, cep4lH31A,H34A-pCS2, CDC42EP4-pCS2, cep4lCDS-pCS2-HA, cep4lUTR-pCS2-HA, cdc42-pCS2, cdc42G12V-pCS2, and cdc42-pCS2-FLAG with NotI or cdc42-RN3-GFP with SfiI. Fgf8a-pCS107 (a gift from R. Harland) was linearized with AscI. Betagalactosidase-pING4 (β-gal) was linearized with BamHI. Capped mRNAs were synthesized using the T3 (cdc42-GFP) or SP6 mMESSAGE mMACHINE kit (Ambion). PCR primers for cloning and mutagenesis are shown in Supplementary Information (Table S1).
Quantification of in situ hybridization stained neurons was performed by counting individual neurons in control and unilaterally-injected embryos. Significance was determined using two-tailed paired t-tests in Excel.

Coimmunoprecipitation and immunoblotting
In vitro translation was performed using TnT SP6 Coupled rabbit reticulocyte lysate kit (Promega). Capped mRNA synthesized as above was used as the template for protein translation according to manufacturer instructions. Half-volume reactions were incubated at 30°C for 90 minutes.
Coimmunoprecipitation was performed as described in Schneider et al., (2012) except that crosslinkers were not used. Coimmunoprecipitation of in vitro translated proteins with active His-tagged Cdc42 and Rac1 proteins (Cytoskeleton) was performed using nickel-chelated agarose beads (Pierce). Beads and protein pellets were resuspended in SDS-PAGE buffer, boiled and subjected to electrophoresis.

Immunostaining
Embryos for immunostaining were fixed in MEMFA, washed in MeOH and bleached in 70% MeOH/10% H 2 O 2 . Following rehydration, embryos were incubated in PBT (1% Triton in 1X PBS) plus 10% goat serum three to four hours at room temperature and washed 30 minutes in PGT (PBT plus 1% goat serum). Mouse zn-12 hybridoma supernatant (DSHB) was diluted 1:5, and rabbit anti-diphospho-p44/42 MAPK (dp-Erk, Cell Signaling) was diluted 1:1000 in PGT. Embryos were incubated overnight at 4°C and washed five times for one hour in PGT. Peroxidase-conjugated anti-rabbit or anti-mouse IgG secondary antibodies (Jackson ImmunoResearch) were diluted 1:500 in PGT and incubated overnight at 4°C. Embryos were washed five times one hour in PGT and 30 minutes in PBS. DAB detection was performed using 0.3% DAB and 0.01% hydrogen peroxide diluted in PBS and stopped after 5-10 minutes bywashing in PBS/0.1% Tween-20. TrueBlue substrate (KPL) detection was carried out for 15 minutes and reaction was stopped with water.

Gene Expression Analysis by RT-PCR
Total RNA isolation was performed as described (Houston and Wylie, 2005) using proteinase K and RNase-free DNase I. Negative controls lacked reverse transcriptase during cDNA synthesis to show absence of genomic DNA contamination. Both semi-quantitative and quantitative PCR using the Lightcycler480 (Roche) were carried out using ornithine decarboxylase (odc) as a loading control or normalization for quantitative PCR. PCR primers for RT-PCR are shown in Supplementary Information (Table S1).

Cep4l is a vegetally localized RNA in Xenopus oocytes
Cuykendall and Houston (2010) identified a number of novel mRNAs localized to the vegetal cortex. One of these mRNAs, cep4l, was verified by RT-PCR and showed an expression pattern during oogenesis similar to that of vg1 and vegt (Cuykendall and Houston, 2010). Preliminary overexpression of cep4l caused morphogenetic defects (see below), suggesting it has functions in early development. We therefore cholfc this cDNA for further characterization.
We obtained a full-length EST for Xenopus laevis cep4l from a commercial source. This EST encodes a predicted protein of 306 amino acids with a domain structure classifying the protein as a homologue of Cdc42 Effector Protein 4 (Cdc42ep4) (Fig. S1). Cdc42ep family proteins are conserved throughoutthe chordate lineage, with Cdc42ep4 showing independent duplications in Xenopus and in Teleost fish. We therefore designated this gene as Cdc42 effector protein 4-like (Cdc42ep4-like or Cep4l). Cdc42ep4 proteins contain three conserved domains, a Cdc42/Rac interactive binding (CRIB) domain and two other conserved regions, the CI and CII domains, located C-terminally to the CRIB domain (Fig. 1A). The CI/II regions and the acidic-rich C-termini of Cdc42ep4 homologues have no known function.
Cep4l mRNA is present maternally and expression continues throughout early development (Fig. 1B). Spatial analysis of RNA expression during oogenesis using in situ hybridization shows that cep4l mRNA is expressed uniformly in previtellogenic oocytes (data not shown). In full-grown stage VI oocytes, cep4l mRNA is localized in a broad expression domain in the cortex of the vegetal hemisphere (Fig. 1C). During embryogenesis, cep4l is expressed diffusely in the blastula and becomes enriched more in prospective dorsal mesoderm and ectoderm around the blastopore during gastrulation. In later gastrula stages, expression becomes enriched on the dorsal side of the blastopore ( Fig. 1D-E). In early neurulae, cep4l is first expressed in patches adjacent to the anterior neural plate. This expression pattern persists, and by late neurulae cep4l is present in the migrating cranial neural crest ( Fig. 1F-G,I-J). In tailbud stage embryos, cep4l is found in the branchial arches populated by migrating cranial neural crest as well as in the somites at stage 25, and in migrating trunk neural crest at stage 31 ( Fig. 1H,K-L).

Overexpression of Cep4l induces ectopic neurogenesis
To identify functions of Cep4l in development, we injected cep4l mRNA into 2-4 cell stage embryos. The effects of cep4l overexpression were first seen in blastula stage embryos as a disruption in the maternal animal cap pigment ( Fig. 2A, D). Cep4l-injected embryos developed a perinuclear accumulation of pigment in affected blastomeres (data not shown), causing asymmetrical distribution of pigment as cell division proceeded. This pigment disruption was evident at stage 7, prior to the activation of zygotic transcription at MBT, indicating a direct effect of cep4l (data not shown). Other morphological effects of cep4l whole embryo overexpression include delayed closure of the blastopore, convergent extension defects and shortened anteroposterior axis, and defects in anterior structures, such as reduced forebrain and eyes ( Fig. 2A-F).
In identifying potential effects of Cep4l expression on the nervous system, we examined ntubulin/tubb2b expression in primary neurons and noticed that cep4l-injected embryos displayed robust ectopic expression of tubb2b ( Fig. 2G-H). Xenopus primary neurons are stereotypically organized in three mediolateral stripes. The medial and intermediate stripes (motor and interneurons) were unaffected in cep4l-injected embryos, whereas the lateral, sensory neurons were markedly increased and expanded throughout the lateral and ventral epidermis (102/145 embryos, 70%; over many experiments) (Fig. 2G, H). We also examined the function of a mammalian Cdc42ep4 homologue, which is a distinct paralogue of Cep4l. Interestingly, the neuronal-inducing activity was conserved in human CDC42EP4 mRNA (14/29 embryos, 48%) ( Fig. 2I). Sectioning confirmed that ectopic neurons induced by Cep4l formed in the deep layer of the ectoderm, as do endogenous primary neurons ( Fig.  2J-K). To provide evidence that Cep4l was inducing differentiated neurons, we stained tailbud embryos using a monoclonal antibody against HNK-1 (mAb zn-12), a marker ofbranching neurons (Metcalfe et al., 1990). As with tubb2b, embryos overexpressing cep4l demonstrated ectopic neurons expressing HNK-1 well outside of the neural tube and in ventral regions (Fig. 2L, M).
We next injected cep4l unilaterally at the 2-cell stage and counted lateral domain neurons to determine whether the apparent increase in tubb2b was indeed owing to an increased number of neurons and not to a dispersal of existing neurons (Fig. 3). The average number of neurons on each side in control embryos was similar ( Table 1; mean ~64 on each side), however cep4l-injected embryos showed a significant increase in neurons on the injected side (Table 1; >100 on injected side, Fig. 3A, E). Because the effect of cep4l overexpression appeared confined to the lateral neuronal cells (sensory), we confirmed their identity using more specific markers. Consistent with an induction of primary sensory neurons, cells expressing runxl, which is expressed by a subset of Rohon-Beard neurons (Park et al., 2012), were increased in cep4l-injected embryos (77/100 embryos, 77%; Fig.  3B, F). Quantification of neurons in unilaterally injected embryos showed a consistent statistically significant increase in mean numbers on the cep4l-injected side (Table 1; Fig,  3I).
Counting islet-expressing cells in these embryos again indicated a significant increase in mean numbers of sensory neurons on the injected side, while control embryos showed no change (Table 1; Fig. 3I). For the interneuron population, counting of pax2 neurons showed no change in mean numbers of interneurons on the injected side compared to the uninjected side (Table 1; Fig. 3I).
Overall, these data show that misexpression of cep4l results in induction of primary neurons in the epidermal region of the embryo. Based on the expression of different neuronal markers (islet1, runx1), these ectopic neurons are likely to be subclasses of primary sensory neurons.

Cep4l is required for primary sensory neurogenesis
We next turned to loss-of-function experiments to determine the extent that Cep4l is required for endogenous neurogenesis. Cep4l is expressed maternally and maternal depletion of cep4l led to gastrulation lethality (data not shown), the etiology of which is under investigation. Additionally, we wished to specifically inhibit late Cep4l function in embryos, however the cep4l locus lacks introns (data not shown), precluding a splice-blocking oligonucleotide approach. We therefore designed a morpholino oligo (MO) that hybridizes to the 5' UTR upstream of the starting methionine to block translation and production of the Cep4l protein (Fig. 4A). We tested the efficacy of the cep4l MO by immunoblotting against C-terminally HA-tagged constructs with or withoutthe 5' UTR MO target sequence. Addition of the HA tag did not hinder Cep4l function in overexpression assays (data not shown). Two-cell embryos were coinjected with cep4lCDS-HA or cep4lUTR-HA mRNAs and increasing doses of cep4l MO. Western blotting showed that Cep4lCDS-HA was expressed at similar levels in the presence or absence of the cep4l MO, whereas Cep4lUTR-HA was greatly reduced at all doses of MO (Fig. 4B).
To test the requirement for Cep4l in neurogenesis, we first injected both blastomeres at the two-cell stage with 40 ng of cep4l MO. Whole embryo depletion of Cep4l caused numerous morphological effects, including delay or failure ofblastopore closure, shortened anteroposterior axis, and reduced anterior structures ( Fig. 4C-F), possibly owing to interference with Cep4l function in the vegetal hemisphere. Since we were unable to assess any direct roles of Cep4l in these morphologically abnormal embryos, we more specifically tested the requirement of Cep4l by performing targeted blastomere injections. Eight-cell embryos were injected with 8 ng cep4l MO into either one dorsal animal (DA) or one ventral animal (VA) blastomere on the future right side of the embryo, and stained for neurons using in situ hybridization for tubb2b and islet1 ( Fig. 4G-N). These injections are expected to target predominantly either the brain or the posterior CNS, respectively (Moody, 1989), and should avoid targeting maternally-derived vegetal cep4l transcripts. Cep4l MO injection into DA blastomeres caused a delay in blastopore and neural tube closure on the injected side, resulting in abnormal head morphogenesis. There was also a complete loss of tubb2b staining in the anterior neural tube and the trigeminal placode (16/20 embryos, 80%; Fig. 4K). Posterior tubb2b staining was unaffected. Unfortunately, we were unable to rescue this anterior morphogenetic defect using a wide range of cep4l mRNA doses (not shown), making interpretation of these effects problematic.
In contrast to dorsal animal (DA) injections, injection of cep4l MO into ventral animal (VA) derivatives strongly and reproducibly reduced spinal sensory neurons on the injected side and produced minimal morphogenetic defects. Expression of tubb2 and islet1 in the sensory neuron domain was reduced on the Cep4l-depleted side (44/56 embryos, 79%; Fig. 4L-N). Expression of these markers in the remainder of the neural tube, in particular the motor neuron expression of islet1 (Fig. 4N), was unaffected on the injected side. VA-targeting of cep4l MO did not affect overall expression or patterning of a neural marker (pax6; Fig.  S2A,C), suggesting that overall CNS formation is not dependent on Cep4l. Cep4l is also not required for neural crest fate specification, since a neural crest marker (snail/snail, Fig.  S2B,D) was expressed normally. Spatially disrupted snail expression in later tailbud stages indicated that Cep4l could have an additional role in crest migration, although we did not pursue this further since Cep4l also affected gastrulation and neurulation, which could confound effects on the neural crest.
We confirmed the specificity of neuronal differentiation deficiencies in Cep4l-depleted embryos by performing rescue experiments (Fig. 4O-V). Embryos at the four-cell stage were injected on one side with low dose (125 pg) cep4l and β-gal as a lineage tracer (or βgal alone), and subsequently injected with cep4l MO into a VA blastomere on same side at the eight-cell stage. Cep4l mRNA (lacking the MO-binding site) co-expression in VA blastomeres at this sub-phenotypic dose was able to restore sensory neurons on the injected side in the majority of cases in at least three separate experiments (25/35 embryos; 71%; Fisher's exact test p < 0.001, Fig. 4Q, U), indicating that the defects in neurogenesis were specifically the result ofloss of Cep4l protein.
Taken together with our overexpression results, these data demonstrate that Cep4l activity is sufficient to induce primary sensory neurons in epidermal tissue, and is also necessary for endogenous sensory neuron formation in Xenopus.

Cdc42 activity is required for primary neurogenesis
Since Cep4l is a predicted Cdc42-effector protein, we further explored the functional relevance of its interaction with Cdc42. Prior studies on Cdc42ep proteins (Hirsch et al., 1999;Joberty et al., 1999) as well as high-throughput protein-protein interactome mapping (Rual et al., 2005), demonstrated that Cdc42ep4 binds specifically to activated Cdc42 and Rhoq, but not Rac or Rho proteins. We performed in vitro binding studies using commercially supplied active Cdc42 or Rac1 His-tagged proteins and in vitro translated Cep4l-HA. Consistent with these data, we found that purified active Cdc42 but not active Rac1 protein was able to interact with full-length Cep4l-HA (Fig. 5A). We did see a shorter product pulled down by active Rac1, which could represent use of an alternate downstream starting methionine. Altering the length can change the specificity of CRIB domains (Thompson et al., 1998), suggesting that a shorter Cep4l CRIB domain may weakly bind activated Rac1. We found no evidence that this shorter form is expressed from injected mRNA in vivo. Additionally, we confirmed that Cep4l-HA interacts with FLAG-Cdc42 in immunoprecipitations of embryo lysates (see below). Finally, we did not detect any expression of RhoQ in early embryos (data not shown). These data suggested that Cep4l would act primarily as an effector for activated Cdc42 during Xenopus embryogenesis.
Since members of the Cdc42ep family require interaction with Cdc42 for their function, we hypothesized that a Cep4l mutant CRIB domain deficient in Cdc42 binding would be inactive for neuronal inducing activity. We mutated well-characterized histidine residues in the Cep4l CRIB domain essential for binding to activated Cdc42 (Ash et al., 2003;Hirsch et al., 2001), cep4lH3lA,H34A and compared expression of full-length and mutant proteins in embryo lysates. No differences in protein expression levels were found by immunoblotting against Cep4l using a guinea pig antisera raised against Xenopus Cep4l (Fig. 5B). The mutant constructwas inactive in neuronal induction (Fig. 5D-F), and in pigment disruption (data not shown), indicating a requirement for activated Cdc42 binding. We confirmed that Cep4l-H31A,H34A failed to interactwith Cdc42 using co-immunoprecipitation of embryo lysates (Fig. 5C) We next tested the idea that Cdc42 activation might be sufficient to induce neuronal differentiation. We first overexpressed wildtype Cdc42 (untagged) or a constitutively active mutant (Cdc42G12V) in embryos and assayed for ectopic tubb2b expression. Although ectopic neurons could be generated in over half the cases (29/45 embryos, 64%), associated blastopore or neural tube closure defects made it unclear whether these neurons were secondary to these other defects (data not shown). We therefore asked whether Cdc42 and Cep4l could synergize in neuronal induction. We co-injected sub-phenotypic doses of cep4l and either wildtype or constitutively active mutant Cdc42 constructs unilaterally into twocell embryos. Injection of low doses (< 100 pg) of cdc42 or cdc42G12V did not affect tubb2b expression, nor did injection of the low dose of cep4l, alone or in combination with cdc42 ( Fig. 5G-K). Coinjection of the cdc42G12V constitutively active mutant with cep4l, however, induced ectopic neurons and expanded the region of lateral neurons in the neural tube on the injected side (9/17 embryos, 53%; Fisher's exact test p < 0.001; Fig. 5L).
Since these results suggested that Cep4l/Cdc42 interactions were sufficient to induce neurogenesis, we next determined the effect of Cdc42 loss-of-function on endogenous neuronal differentiation. We depleted Cdc42 in embryos using a translation-blocking morpholino that was used previously in X. tropicalis, and that recognizes the identical sequence in X. laevis (Rana et al., 2006). We confirmed that this oligo was effective in reducing expression of exogenous GFP-tagged Cdc42 (Fig. 6A). Injection of 60 ng of MO into one or two blastomeres of two-cell embryos did not visibly disrupt normal early development, although we did observe that tubb2b expressing neurons were reduced in a majority of cdc42 MO-injected embryos (44/57 embryos, 77%; Fig. 6B-F). In comparison, control morpholino injection at the same dose did produce a minor reduction in sensory neurons in a minority of embryos (22/65 embryos, 34%). The difference between the control and cdc42-MO groups remains statistically however (Fisher's exact test, p < 0.01). These data suggest that Cdc42 plays a critical role in sensory neurogenesis, but also that these cells may be sensitive to high oligo doses. In unilateral injections, numbers of sensory domain tubb2b-positive neurons were moderately but significantly decreased on the Cdc42-depleted side compared to control uninjected embryos (Table 1, Fig. 6D-F). Overall, the synergy of neuronal induction in coinjection experiments, and the comparable effects of depletion of Cep4l and Cdc42, suggest that these proteins act together in endogenous sensory neuronal differentiation.

Cep4l is likely not a BMP antagonist or neural plate inducer
We next performed additional experiments to determine potential mechanisms by which Cdc42/Cep4l signaling might generate ectopic neurons. First, we treated cep4l-injected embryos with a combination of hydroxyurea and aphidicolin (HUA) to block the cell cycle in S phase (Harris and Hartenstein, 1991). Ectopic neurons were still induced by Cep4l (Fig.  S3A-F), ruling out that ectopic neurons simply result from stimulation of cell division by Cep4l. Second, we found that ventral-animal overexpression of cep4l neither induced nor inhibited markers of neural plate (sox2), neural crest (sox9), or epidermis (xk8lal/epidermal keratin; Fig. 7A-E). Third, we analyzed Cep4l activity in tissue explants and found that Cep4l does not dorsalize ventral marginal zone explants (data not shown) or induce neural fate in animal cap (Fig. 7F-I; Fig. S3G). These negative data suggest that Cep4l does not have strong pro-or anti-BMP activity or obvious pro-or anti-Wnt activity. Next, we looked at possible effects on Notch signaling, which is known to inhibit neuronal differentiation (Chitnis et al., 1995), but can also inhibit multiciliated cell (MCC) and ionocyteWell (IC) fates (Stubbs et al., 2006). These cell types are derived from deep-layer ectoderm in the epidermal regions (Hayes et al., 2007). Surprisingly, however, overexpression of cep4l greatly suppressed MCC and IC fates (marked by ciliary-rootlet coiled coil (crcc) and foxi1, respectively; Hayes et al., 2007) in a majority of cases (crcc 10/20, 50%; foxi1 24/34, 71%) while inducing neurons in those areas (Fig. 7N-O). Since Notch activation or Notch inhibition affects the differentiation of neurons, MCCs and ICs in the same manner, we conclude from these experiments that Cep4l is not directly involved in Notch signaling.
Finally, we compared the activity of Cep4l to Fgf8a, which induces primary sensory neurons throughout the lateral epidermis similarly to Cep4l (Hardcastle et al., 2000;Fletcher et al., 2006). We found that Fgf8a also suppressed MCC and IC fates while inducing neurons (Fig.  7P). The extent of the Fgf8a effect was greater in these assays owing to the fact that Fgf8a is secreted and easily diffusible and can act over a greater area. These roughly comparable effects of Fgf8a and Cep4l therefore suggested that Cep4l might functionally interact with Fgf8a signaling in regulating neuronal cell fate.

Cep4l is required downstream ofFgf8a in neuronal induction
In Xenopus, Fgf8a is required for both for posterior neural patterning and for neuronal differentiation (Fletcher et al., 2006). It is unclear however whether these activities involve similar signaling mechanisms, since posterior neural patterning requires Ras/MAPK signaling and neuronal differentiation does not (Ribisi et al., 2000). Since our initial experiments suggested that gain or loss of Cep4l function did not directly affect neural patterning, we more specifically examined the extent of Cep4l and Fgf8a functional interaction during neurogenesis using molecular epistasis experiments.
We first used minimally effective low doses of cep4l mRNA and human FGF8a protein (injected into the blastocoel) either singly or in combination. Individual treatment alone caused minimal ectopic tubb2b expression, whereas co-injection oflow-dose FGF8a and cep4l resulted in many ectopic tubb2b-expressing cells throughout the epidermis (19/19 embryos; Fig. 8A-D). We additionally performed double-depletion experiments. Inhibition of Fgf8a protein expression using a splice-blocking MO specific for the Fgf8a isoform has been shown to reduce neuronal differentiation (Fletcher et al., 2006 ; Fig. S4A-F). We injected fgf8a MO and cep4l MO singly or in combination and assayed tubb2b expression. To minimize confounding morphological effects and provide an internal control, we injected fgf8a MO bilaterally into the ventral blastomeres at the four-cell stage and cep4l MO unilaterally into one ventral animal blastomere at the eight-cell stage. Consistent with Fgf8a/ Cep4l interaction in overexpression experiments, the reduction of neuronal differentiation by Fgf8a knockdown was further enhanced on the side injected with cep4l MO (48/76 embryos, 63%; Fisher's exact test p <0.001; Fig. 8E-H). Fgf8a-depleted embryos had ~55 neurons on each side, and this was reduced to an average of six per side when both Cep4l and Fgf8a were depleted (Table 1).
Next, we analyzed the extent that Cep4l and Cdc42 were necessary for the induction of ectopic neurons by Fgf8a. We injected controls and Cep4l or Cdc42-depleted embryos (VA injection) with fgf8a mRNA (unilateral at two-cell) and examined tubb2b or runx1 expression. Fgf8a induction of neurons was robust on both sides of control embryos (91/91, 100%). Importantly, this activity was unilaterally reduced on the sides of embryos also injected with cep4l MO (47/57, 82%; Fig. 8I-L, Fig. S5) or with cdc42 MO (29/53, 55%; Fig. S5). It should be noted that Cep4l or Cdc42 depletion did not affect the cell nonautonomous induction of neurons on the contralateral side by Fgf8a. This observation suggests that Cep4l/Cdc42 are required in the receiving cell and not for aspects of Fgf8 processing or secretion.
Lastly, to more directly address a possible role for Cep4l in posterior neural patterning by Fgf8a, we depleted Cep4l in the context of Fgf8a overexpression and examined the effects on anteroposterior neural pattern and neural crest formation. Unilateral injection fgf8a caused an expansion of sox2, krox20 and hoxb9 throughout the embryo and repressed snai1 expression on the injected side whilst inducing it on the contralateral side (Fig. S6). These observations indicate that Cep4l is not required for these effects of Fgf8a on the nervous system. Overall, these data demonstrate that primary sensory neurogenesis in Xenopus requires Fgf8a signaling involving activated Cdc42 and the Cep4l effector protein.

Fgf8a signaling is independent of MAPK signaling and specifically promotes the association of Cep4l with Cdc42
The molecular basis for the differential signaling activities of Fgf8 isoforms in chicken and frog embryos is not well understood. It has been suggested that Fgf8a/b isoforms differentially activate ERK during chicken midbrain-hindbrain development (Sato and Nakamura, 2004). We therefore tested the extent that this differential activation might occur in Xenopus. We injected embryos with equal doses of fgf8a or fgf8b mRNAs (500 pg) and dissected animal caps at the late blastula stage. These were cultured and assayed for the diphosphorylated, active form of MAPK (dp-ERK) by immunoblotting (Fig. 9A). Caps expressing fgf8b showed robust activation of MAPK, whereas control caps and fgf8a injected caps lacked MAPK activation (Fig. 9A). Similar to fgf8a, cep4l (500 pg) also failed to induce MAPK in animal caps (Fig. 9A). Since Cep4l and Fgf8a do not induce neurogenesis in animal caps, but do so in whole embryos, we considered that caps might lack an additional component needed for MAPK signaling and neurogenesis. We therefore confirmed these results using whole-mount immunostaining against dp-ERK in intact gastrula stage embryos (Fig. 9B-E). These data are consistent with the results shown in the chick and suggest that in the frog, Cep4l and Fgf8a regulate neurogenesis independently of immediate MAPK pathway activation.
The functional interaction data suggested that Fgf8a might stimulate the activation Cdc42 and increase recruitment of Cep4l. For these experiments, late blastula embryos expressing tagged Cep4l and Cdc42 constructs were treated with commercially prepared, purified FGF8 proteins injected into the blastocoel (Pera et al., 2003). We then performed immunoprecipitations 60 minutes later and analyzed the resulting immune complexes. In control embryos co-injected with FLAG-cdc42 and cep4l-HA, anti-FLAG (Cdc42) immunoprecipitation was able to bring down detectable amounts of Cep4l-HA (Fig. 9F). Interestingly, in the presence of exogenous human FGF8a protein injected into the blastocoel, the amount of Cep4l pulled down by FLAG-Cdc42 was increased (Fig. 9F). Whereas FGF8a significantly stimulated Cdc42/Cep4l complex formation, FGF8b did not, suggesting that regulation of Cdc42/Cep4l interaction is a specific outcome of Fgf8a signaling. The results from three independent experiments were quantitated, and indicated an average ~2 fold increase of Cep4l/Cdc42 interaction following Fgf8a treatment. Fgf8b treatment produced a small but reproducible reduction in Cep4l/Cdc42 interaction.
Fgf8a has been proposed to act as a weak Fgf8 ligand, since a one hundred-fold reduction in Fgf8b dose caused Fgf8a-like effects (Sato and Nakamura, 2004;Olsen et al., 2006). We therefore performed experiments using a fifty-fold dilution ofFGF8b. We found that this dose was similarly unable to induce neurons (Fig. S7). These results suggest that Fgf8 ligands exert primarily qualitative effects on downstream signaling mechanisms. To further test this, we also examined Fgf17 activity in Xenopus. Fgf17 (and Fgf18) is predicted to have a receptor affinity in between Fgf8a and Fgf8b and can produce an Fgf8a-like effect in chick neural tubes (Liu et al., 2003;Olsen et al., 2006). Consistent with this idea, we found that human FGF17b also greatly induced ectopic neurons in embryos (17/19 cases; 89%; three experiments; Fig. S7).
Since we observed a high level of Cdc42/Cep4l interaction in control embryos, we suspected this was the result of endogenous Fgf8a signaling. To determine whether this was the case, we performed immunoprecipitations to analyze Cep4l/Cdc42 interactions in embryos depleted of Fgf8a. Consistent with this hypothesis, the amount of Cep4l-HA immunoprecipitated by FLAG-Cdc42 was reduced in embryos co-injected with Fgf8a MO (Fig. 9H). Thus, Fgf8a is both sufficient and necessary to regulate Cep4l/Cdc42 interaction in the late blastula/early gastrula stages.

Cdc42 effector protein 4-like (Cep4l) regulates neurogenesis
As part of an ongoing effort to study the functions of maternally-localized transcripts in the Xenopus oocyte, we identified and characterized a Xenopus homologue of Cdc42 effector protein 4 (Cdc42ep4). Cep4l belongs to a little-studied class of Cdc42 effector proteins (Cdc42ep/Borg/CEP) implicated in actin filament regulation and cell shape (Hirsch et al., 2001). Through initial gain-of-function experiments in Xenopus embryos, we showed that Cep4l dramatically induced neurogenesis throughout the deep layer of the epidermis. The induced neurons were derived from the lateral domain of sensory neuron precursors present in the early neurula and express markers of primary sensory/Rohon-Beard neurons including pan-neuronal tubulin (tubb2b), islet1, runx1 and the zn-12 antigen. Markers of the motor and interneuronal subtypes were unaffected, suggesting a specific induction of the sensory population. Counting of neurons indicated that sensory neurons were increased in number in a cell/tissue autonomous fashion, suggesting that Cep4l does indeed induce neuronal fate as opposed to causing a dispersal of existing neuronal precursors. Interestingly, human CDC42EP4 had the same activity, indicating that this role in neuronal fate may be conserved. Nothing is known about the role of the Cdc42ep family in mammalian development however.
Loss-of-function experiments using antisense MOs support the conclusion that Cep4l is required for primary sensory neuron formation in vivo in Xenopus. Depletion of Cep4l in regions fated to form sensory neurons leads to a reduction in these neurons that can be rescued by MO-resistant cep4l mRNA. Importantly, this rescue could be achieved at doses that do not normally induce neurons and thus demonstrate the specificity of the MO injection. Primary spinal sensory/Rohon Beard neurons become post-mitotic primarily during the late gastrula stages, with 80% being born at stage 11-12 (Lamborghini, 1980). Cep4l is expressed maternally in the vegetal hemisphere and is zygotically expressed around the blastopore and to a lesser extent throughout the animal cap by the midgastrula stage. Cep4l then becomes restricted to the dorsal half of the blastopore by late gastrula. Spinal cord and Rohon-Beard cells originate from the lateral blastoporal regions of the gastrula (Keller et al., 1992;Rossi et al., 2008), which would overlap with cep4l expression at these stages. Taken together with the gain-of-function data, we conclude that Cdc42ep4 proteins play a critical role in the specification and differentiation of primary sensory neurons in Xenopus.
Cep4l is likely to have other functions as well, since we saw effects on pigment dispersal as well as morphogenetic defects at high doses. Depletion of Cep4l in the whole embryo or in dorsal injections also results in gross morphogenetic defects, although we were unable to rescue this aspect. It is possible that precise dosage and temporal/spatial regulation of Cep4l expression would be needed to restore normal morphogenesis in depleted embryos. Cdc42ep proteins have been implicated in cytoskeletal regulation in cultured cells, however it is unclear to what extent Cep4l acts similarly in Xenopus. Cdc42ep4 and Cep4l are notable in their lack of an N-terminal extension of the conserved Cdc42ep CII motif, termed the Borg homology domain 3 (BD3), which has been characterized as a Septin-binding, cytoskeletal regulatory motif (Joberty et al., 2001). Also, our preliminary experiments ruled out a direct role for Cep4l in actin assembly (data not shown), it is therefore possible that Cep4l overexpression affects morphogenesis indirectly by affecting intracellular transport. A more detailed analysis of Cep41 function at the cellular level would be instrumental in addressing these questions.

Role of Cdc42 in amphibian primary neurogenesis
One surprising finding from this work was the novel observation that Cdc42 is required for primary sensory neuron formation. The presence of a CRIB domain Cep4l and the characterization of related Cdc42ep proteins suggested that Cep4l would interact with activated Cdc42 and we verified this using in vitro and in vivo assays. Additionally, we made a CRIB domain mutant Cep4l thatwas unable to bind Cdc42 and correspondingly had no gain-of-function effect when expressed in embryos. This observation suggested that Cdc42 itself might function in sensory neuron specification. The ability of a constitutively active Cdc42 construct along with co-injected Cep4l to induce neurons supports this idea. Additionally, depletion of endogenous Cdc42 by MO injection greatly reduced the numbers of sensory neurons, demonstrating a requirement for Cdc42-mediated signaling in Xenopus neurogenesis. It was somewhat curious to note that injection of high doses of Cdc42 MO did not cause morphogenetic or other visible defects. Also, we did not observe the high rate of lethality seen when this MO was used in X. tropicalis (Rana et al., 2006). Cdc42 is likely to be an abundant protein and therefore MOs may be less effective than dominant-negative constructs which might act as sinks for a variety of Rho GTPase regulators.
It is unclear if Cep4l/Cdc42 signaling would have a conserved role in mammalian neurogenesis, since Cdc42ep4 gene targeted mice have not been reported. In mice Cdc42 is required to activate apical aPKC signaling to maintain ventricular zone neuronal progenitors and deletion of Cdc42 results in expansion of subventricular zone progenitors and expanded neurogenesis (Cappello et al., 2006). It is unlikely that Cdc42 is acting similarly at the level of maintaining apical identity in the context of primary neurogenesis in Xenopus. We find that gain of Cdc42 activity results in ectopic neurogenesis whereas deletion has this effect in mice (Capello et al., 2006). Additionally, primary neurogenesis occurs in unpolarized inner layer cells in the amphibian neural plate and is inhibited in the apically polarized superficial layer cells (Chalmers et al., 2002;Hartenstein, 1989), likely through aPKC activation (Ossipova et al., 2007;Sabherwal et al., 2009). These differences in the role of aPKC and Cdc42 could reflect bona fide species differences in function or could indicate differences in primary larval neurogenesis versus neurogenesis of adult neurons.

Regulation of Cep4l by Fgf8a isoform activity
A second major finding of this work is the involvement of Cep4l and Cdc42 in the regulation of neurogenesis by Fgf8a signaling. We show that Cep4l and Cdc42 are each needed for the neuron-inducing activity of Fgf8a. Furthermore, we find that Fgf8a is necessary and sufficient for the interaction of Cep4l with Cdc42. Fgf8a has been previously implicated in regulating neuronal fate, but the exact mechanisms and downstream signaling events have not been comprehensively identified. Many aspects of FGF signaling are mediated by MAPK activation, including anterior-posterior neural patterning, however the regulation of neurogenesis appears to occur independently of this pathway (Ribisi et al, 2000). We have shown that Fgf8a and Cep4l fail to activate MAPK in animal caps, whereas Fgf8b is strongly active in MAPK stimulation. These data are consistent with the ability of Fgf8b to activate brachyury (xbra) expression in Xenopus (Fletcher et al., 2006) and MAPK in the chick neural tube (Sato and Nakamura, 2004). Although previous Fgf8a gain-offunction experiments showed MAPK activation in a manner restricted to the diencephalon in the chick, this effect was likely indirect since it was enhanced rather than blocked by a dominant-negative Ras construct (Sato and Nakamura, 2004). Fgf8a is required for anteriorposterior neural patterning in Xenopus (Fletcher et al., 2006) and may thus also regulate MAPK indirectly in neural tissue. Since Cep4l gain-or loss-of-function does not appear to specifically regulate neural patterning, this aspect of Fgf8a signaling is likely to be independent of Cep4l. Our data are the first to suggest a distinct, direct outcome of Fgf8a signaling in the context of neurogenesis.
It is interesting to speculate on whether Fgf8a/Cdc42ep signaling might function in other aspects of vertebrate development. Fgf8 isoforms (of which Fgf8b is the major form) are coexpressed many tissues in the embryo, including the gastrula, the mesoderm, the midbrainhindbrain boundary and the limb bud apical ectodermal ridge (Crossley et al., 1995). Interestingly, elimination of the Fgf8a-containing spliceforms (a, c, e, g) in mouse, through gene targeted alteration of the relevant splice acceptor site, failed to cause embryonic defects in Fgf8-expressing tissues (Guo et al., 2010), in contrast to deletion of the b-containing forms or complete Fgf8 nulls. Additionally, compound mutants of Fgf8a and Fgf17, which likely has Fgf8a-like activity, did not result in additional defects (Guo et al., 2010). Residual Fgf8a-like activity could be supplied by Fgf18 in these animals, but triple mutants have not been described. Interestingly, Fgf17 and 18 orthologues and Fgf8 splice variants other than a/b are absent in Xenopus (Lea et al., 2009), which could explain the more severe effect of Fgf8a depletion on neural patterning in the frog compared to the mouse.
The mechanisms by which different Fgf8 family members and sliceforms could activate distinct signaling pathways are unclear. In vitro, Fgf8a and Fgf8b bind the same complement of FGFRs, namely the c-isoforms of FGFRs1-3 and FGFR4 with similar affinities, although Fgf8a binds them with about 10-fold weaker affinity (Olsen et al., 2006). Analysis of the Fgf8b-FGFR2c 3D structure suggested that Fgf8b binding creates a close apposition of FGFR tyrosine kinase domains, through binding-induced rotation of the linker region of the receptor (Olsen et al., 2006). It is possible that this tighter conformation is critical for MAPK activation, whereas weaker Fgf8a binding results in a looser arrangement allowing for activation of Cdc42 or other signaling cascades. Interestingly, we have shown that FGF17b regulates an Fgf8a-like response (Fig. S7), even though this isoform contains the same 11 amino acid region required for the high affinity interaction of Fgf8b, including a critical phenylalanine residue (F32, Olsen et al., 2006). It will thus be important to investigate additional determinants of signaling specificity of Fgf8 family ligands.
Following Cdc42 activation, binding of Cep4l or other effectors can occur. Cep4l and other Cdc42ep family members are hypothesized to be scaffold proteins (Hirsch et al., 2001), and thus Cep4l is likely to facilitate the formation of protein complexes to mediate neurogenic signaling. These signaling pathways remain unknown, but the identification of Cdc42/ Cep4l-interacting proteins would likely provide clues to downstream pathways activated by Fgf8a. Surprisingly little is known about the specificity and usage of different Cdc42effector proteins. The binding of activated Cdc42 to CRIB domains is considered rather promiscuous, and therefore compartmentalization of function is likely critical for the diversity of Cdc42 functions in the cell. The particular GEFs involved in initial activation are thought to provide much of the specificity in small GTPase effector recruitment (Bos et al., 2007;Sinha and Yang, 2008). We would therefore expect that specific Cdc42-activating GEFs activated by Fgf8a would recruit Cep4l and function in the context of sensory neuron fate in Xenopus.

Developmental regulation of sensory neuron fate
It is becoming increasingly appreciated that primary sensory neuron fates are regulated similarly to other neural plate border derivatives, including placode and neural crest derivatives (Hong et al., 2008;Park and Saint-Jeannet, 2008;Rossi et al., 2008;Cornell and Eisen, 2000). Notably, both neural and non-neural ectoderm can contribute to sensory neuron and neural crest lineages (Rossi et al., 2008;Mancilla and Mayor, 1996;Moury and Jacobson, 1990;Selleck and Bronner-Fraser, 1995). Graded amounts of Fgf8a and intermediate BMP levels are thought to specify different neural plate border lineages, with lower doses inducing preplacode and neural crest fates and higher doses inducing sensory neuron fates and inhibiting neural plate border fates (Hong et al., 2007;Hardcastle et al., 2000;Fletcher et al., 2006;Park et al., 2012). Our data showthat Cep4l/Cdc42 signaling regulates sensory neurogenesis but not neural or neural plate border fates, suggesting that the Cep4l pathway is activated only at high levels of Fgf8a. Alternatively, Cep4l could allow integration with BMP and Wnt gradients, which are also involved in sensory neuron fate (Hong et al., 2008). Interestingly, we have found that Cep4l and Fgf8a overexpression induce sensory neurons but also inhibit the specification of intercalating deep layer epidermal cell types, multiciliated cells and ionocytes. Notch signaling is known to control the ultimate terminal differentiation of these cells, regardless of neural or epidermal fate. Fgf8a/Cep4l/Cdc42 signaling might therefore act to promote sensory neuron fate in multipotential deep layer ectodermal progenitors in both neural and non-neural ectoderm.

Supplementary Material
Refer to Web version on PubMed Central for supplementary material.
• We examine the role of a Cdc42 effector protein (Cep4l) in Xenopus development.
• Cep4l binds Cdc42 and induces primary sensory neurogenesis in epidermal regions.
• Fgf8a signaling is necessary for and specifically promotes Cdc42/Cep4l binding.
• Both Cep4l and Cdc42 are required for Fgf8a-induced primary neurogenesis.