The cytoskeleton-associated protein SCHIP1 is involved in axon guidance, and is required for piriform cortex and anterior commissure development

During nervous system development, axon outgrowth and guidance are key processes for the proper formation of neuronal circuits. Schwannomin-interacting protein 1 (SCHIP1) is a cytoplasmic protein identified as a partner of the tumor suppressor protein schwannomin [also known as merlin or neurofibromatosis 2 (NF2)] (Goutebroze et al., 2000). The IQCJ-SCHIP1 isoform is a component of axon initial segments and nodes of Ranvier, two regions highly enriched in Na_v channels important for the initiation and the saltatory conduction of action potentials, respectively (Martin et al., 2008). In these regions, Na_v channels sit in multimolecular complexes comprising KCNQ channels and cell adhesion molecules (CAMs), NRCAM and neurofascin, anchored to the actin cytoskeleton through their interaction with the ankyrin_G/βIV spectrin-based cortical cytoskeleton (Buttermore et al., 2013). In this complex, SCHIP1 interacts directly with ankyrin_G (also known as ankyrin 3) (Martin et al., 2008). SCHIP1 is also able to interact with ankyrin_B (also known as ankyrin 2), which binds adhesion molecules involved in axon formation (Bennett and Lorenzo, 2013). Ankyrin_B regulates both axon elongation and navigation (Colavita and Culotti, 1998; Gil et al., 2003; Kunimoto and Suzuki, 1995; Nishimura et al., 2003), although its precise in vivo contribution is poorly characterized (Scotland et al., 1998).

The Schip1 gene is expressed in the brain during stages of active axonogenesis (http://www.genepaint.org/cgi-bin/mgrqcgi94) suggesting that it could function in axon outgrowth and/or navigation. Using a gene-trap strategy, previous studies revealed the consequences of SCHIP1 loss of function on embryonic development, but did not investigate specific morphological brain abnormalities in mutant mice (Chen et al., 2004; Schmahl et al., 2007,, 2008). It is noteworthy that the gene-trap sequence was introduced into the Schip1 gene at a position where it was expected to alter the expression of two isoforms (SCHIP1a and IQCJ-SCHIP1) only.

In the present study, we investigated the role of SCHIP1 in brain development by generating Schip1Δ10 mutant mice, which are impaired in the expression of all Schip1 isoforms. We show that Schip1Δ10 mice display a partial agenesis of the anterior commissure (AC). The piriform cortex, characterized by its three-layered organization, is one of the main sources of AC axons (de Castro, 2009) and plays key roles in odor discrimination, association and learning (Bekkers and Suzuki, 2013). Multiple guidance signals are known to be expressed in the piriform cortex and surrounding regions and to control the development of the AC, including ephrin, netrin and semaphorin family members (Lindwall et al., 2007).

Using a combination of in vivo and in vitro approaches, we show that Schip1Δ10 mice display decreased thickness of the piriform cortex, which may result from cell death, and early AC axon developmental defects, which are likely to be associated with impaired axon elongation and guidance. Our results further suggest that the Schip1Δ10 mutation changes axon sensitivity to the repulsive guidance cues semaphorin 3F (SEMA3F) and Eph receptor B2 (EPHB2), which are known to participate in AC formation (Henkemeyer et al., 1996; Sahay et al., 2003).

The Schip1 gene encodes six isoforms expressed in the mouse brain (supplementary material Fig. S1A). All isoforms differ in their N-terminus and share a C-terminal domain encoded by exons 9 to 14. This domain includes a ∼40 residue leucine zipper that is predicted to adopt a coiled-coil conformation and is required for ankyrin and schwannomin binding (Goutebroze et al., 2000; Martin et al., 2008). Mutant mice were generated by deletion of Schip1 exon 10, which is predicted to result in a frameshift that creates a STOP codon in exon 11 (supplementary material Fig. S1B,C). We checked whether this led to the expression of truncated proteins lacking the conserved C-terminal domain common to all isoforms (supplementary material Fig. S1D-F). Immunoprecipitation experiments followed by immunoblotting using different combinations of SCHIP1 antibodies did not reveal any truncated proteins in brain extracts (supplementary material Fig. S1F and Fig. S2). Homozygous mutant mice (termed Δ10) born from heterozygous crossings were fertile and survived as long as wild-type (WT) mice (>2 years), although they suffered from a mild growth delay from birth to adulthood (data not shown).

Although the global anatomy of adult mutant brains did not differ from that of WT littermates, histological analyses revealed white matter defects in the AC and to a lesser extent in the corpus callosum (CC) (Fig. 1). The AC is formed by two main branches. The anterior branch (ACa) is composed of axons from the anterior olfactory nucleus (AON) and the anterior piriform cortex, whereas the posterior branch (ACp) is composed of axons from the posterior piriform cortex and the amygdala (Cummings et al., 1997; Pires-Neto and Lent, 1993) (Fig. 1A). Horizontal sections showed a severe decrease in ACa thickness and an absence of ACp in all mutants (Fig. 1B). Serial coronal sections labeled with neurofilament antibodies confirmed these observations, showing a reduced ACa bundle area in mutants compared with WT sections. This decrease was more severe in the ACa caudal region (Fig. 1Ci,ii,D). The phenotype was even more striking for the ACp, which was missing in all mutants (Fig. 1Ciii). In caudal regions, mutant mice also displayed a thinner CC at the midline compared with WT mice (Fig. 1Ciii,E). Mid-sagittal brain sections further showed a decreased CC length along the rostrocaudal axis (Fig. 1F,G). These observations suggested a reduction of the number of axons in the CC and the ACa, and an absence of ACp axons in Δ10 mutants.

To identify the origin of axonal tract defects in mutants, we studied the development of the AC, which was more severely affected than the CC. The AC starts to develop between E13.5 and E14.5, with pioneer L1CAM-immunopositive axons detected at the midline at E14.5 (Schneider et al., 2011). We therefore immunolabeled serial coronal sections of E13.5 and E14.5 brains with L1CAM antibodies (Fig. 2A,B). At E13.5, the first ACa pioneer axons were visible in WT brains, whereas they were not detectable in mutants, whatever the section plane (Fig. 2A). Moreover, at E14.5 no AC axons were detected at the midline in mutants (Fig. 2B). These results suggested that SCHIP1 is important for early AC development.

To further characterize the AC defects, we immunolabeled serial horizontal sections from E16.5 brains with L1CAM antibodies. In WT embryos, both AC branches were fully developed as previously shown by Schneider et al. (2011). By contrast, E16.5 mutant embryos displayed thinner ACa tracts (Fig. 2C) with axons that never reached the midline. Selecting specific mutant horizontal sections in which the ACa axons were closest to the midline, we determined that ACa axons terminated 432±142 µm from the midline (mean±s.e.m., n=4 Δ10 embryos). The ACp tract was absent in all mutant embryos. Very few axons arose from the posterior piriform cortex and seemed stunted, often misprojecting to rostral regions (Fig. 2C).

We further performed tracing experiments with DiI (as sketched in Fig. 2Da) to determine whether AC axons of mutant embryos misprojected towards other brain regions. In serial horizontal slices of WT DiI-injected brains, both ACa and ACp axons crossed the midline and projected towards the contralateral hemisphere. By contrast, mutant ACa labeled axons appeared less numerous and stopped before crossing the midline without aberrant projections to other regions (Fig. 2Db). Labeled axons from the posterior piriform cortex were very rare in mutant embryos and stopped growing close to the rostral ipsilateral cortex instead of projecting to the midline (Fig. 2Dc). Altogether, these observations (as summarized in Fig. 2E) suggested that SCHIP1 could contribute to AC formation by playing a role in both axon outgrowth and guidance of projecting neurons.

ACp axons arise mainly from the piriform cortex and the amygdala (Huang et al., 2014). This prompted us to ask whether these structures were affected in Δ10 mutants. The areas of the lateral and the basolateral nuclei of the amygdala were similar in adult Δ10 and WT mice (supplementary material Fig. S3), suggesting that the amygdala was not affected in mutants. We also analyzed the overall thickness of the piriform cortex from E14.5 to adulthood in Cresyl Violet-stained coronal sections. We defined as anterior the piriform cortex at the level of the ACa along the rostrocaudal axis (Fig. 3A, ANT) and defined as posterior the piriform cortex at the level of the ACp (Fig. 3A, POST). The thickness of the mutant anterior piriform cortex was not significantly affected during development (Fig. 3B). By contrast, from E16.5 on, the posterior piriform cortex was thinner in the mutant than in WT brains (Fig. 3B). This was not the case for the neocortex (supplementary material Table S1). Since AC axons start to grow at E13.5 (Schneider et al., 2011), and as the AC is fully established by E16.5 in mice, these results showed that the defect in posterior piriform cortex thickness coincided with the completion of AC development.

We further characterized the three-layered organization of the piriform cortex and analyzed the thickness of each layer. In WT and mutant embryos, the layers were not distinguishable at E14.5. From E16.5 on, the piriform cortex started to organize into three distinct layers with increased cell density in layer 2, even though the precise layer delimitations remained poorly defined (data not shown). At P0, the three layers were clearly distinct (Fig. 3C). Both layer 2 and layer 3 were thinner in the mutant than in the WT posterior piriform cortex (Fig. 3D). Decreased thickness was restricted to layer 2 in the anterior piriform cortex (Fig. 3D), which may account for the undetectable overall thickness reduction (Fig. 3B).

We then attempted to identify the origins of layer 2 defects. Confocal imaging of serial coronal sections stained with DAPI showed no difference in cell density between mutant and WT newborn mice (supplementary material Table S2). Hence, the reduced thickness of the layers was likely to result from an overall decrease in cell number. Layer 2 contains pyramidal cells, which are mostly generated at E12.5 (Sarma et al., 2011). BrdU injection in pregnant females at E12.5 showed no difference in the density of BrdU-immunopositive cells between WT and mutant piriform cortices at E14.5 (supplementary material Table S3), suggesting that the early migration of mutant neurons was not affected and that neurons whose axons initially form the AC were present in normal numbers in mutant piriform cortex at this stage. Also, no difference in BrdU-immunopositive cell density could be detected in layer 2 between WT and mutant at E18.5 (supplementary material Table S3). However, since layer 2 could clearly be distinguished and was already thinner in mutants at this stage (supplementary material Table S3), this indicated that a population of projecting cells disappeared between E14.5 and E18.5. Immunolabeling for cleaved caspase 3 on serial sections showed increased cell death at E14.5 in mutant compared with WT piriform cortices, which was not the case in the neocortex (Fig. 3E,F). The increased cell death in mutants was more pronounced in the posterior than in the anterior piriform cortex (Fig. 3E,F), consistent with the more severely decreased thickness observed in the posterior piriform cortex at later embryonic stages.

Altogether, our results suggested that SCHIP1 could play specific roles in the development of piriform cortex neurons in vivo. In situ hybridization of WT coronal brain sections indeed confirmed that Schip1 is expressed in the piriform cortex during brain development at E13.5, E14.5 and P0 (supplementary material Fig. S4A). Interestingly, at P0, Schip1-expressing cells were mainly localized within layer 2, the most affected layer in Δ10 mutant mice (supplementary material Fig. S4A, high magnification panel). RT-PCR also revealed the expression of Schip1 isoforms (mainly SCHIP1b, SCHIP1c and IQCJs-SCHIP1) in WT E14.5 piriform cortex (supplementary material Fig. S4C). Schip1 expression appeared to be higher in the anterior piriform cortex than in the posterior piriform cortex (supplementary material Fig. S4B,D).

The early expression of Schip1 prompted us to further characterize SCHIP1 functions during axon formation in piriform cortex neurons using in vitro approaches. Cultures were performed from E14.5 embryos, a stage of active axon growth in WT embryos and at which the thickness of the cortex is unaffected in mutants (Fig. 3B). We first estimated axon growth by quantifying the surface covered by the axons of anterior and posterior piriform cortex explants immunolabeled for TUJ1 (TUBB3) after 2 days in culture (Fig. 4A). Axon growth areas of mutant explants were significantly reduced compared with those of WT explants (Fig. 4B). Defects were similar for anterior and posterior piriform cortex explants.

To decipher which step of axon development was impaired by the Δ10 mutation, we cultured dissociated piriform cortex neurons. Characterization of these cultures indicated that neurons (TUJ1-immunopositive cells), which represented more than 80% of the total number of cells, started to develop by generating lamellipodia and later minor processes, and then formed one longer process (axon) within 2 days in vitro (DIV). In addition, 80% of neurons expressed the cell adhesion molecule NRCAM, which is known to be important during piriform cortex development (Lustig et al., 2001) (data not shown). RT-PCR confirmed that WT dissociated neurons expressed the same Schip1 isoforms as those expressed in E14.5 piriform cortex in vivo (data not shown). No differences in neuronal density or percentage of cleaved caspase 3-immunopositive neurons were observed between WT and mutant cultures up to DIV4 or DIV6, respectively (supplementary material Table S4), demonstrating that, unlike in vivo, there was no obvious increase in cell death in cultures from mutant embryos.

We next analyzed neuronal morphology in these cultures. The immature neuritogenesis stage was over-represented in mutant cultures at DIV1 and DIV2 (Fig. 4C), suggesting a role for SCHIP1 in axon initiation. Measurements performed on neurons that exhibited one longer process and several shorter neurites (Fig. 4D) showed decreased axon length in mutant neurons at DIV1, DIV2 (Fig. 4D,E) and DIV4 (data not shown). This defect was rescued by expression of the Flag-tagged SCHIP1b isoform in mutant neurons (Fig. 4F,G). These data indicated a role for SCHIP1 in axon elongation, which was consistent with the distribution of SCHIP1 within growing neurons. Flag-tagged SCHIP1b indeed displayed a punctate localization along the neurites and in the growth cones, where it was found in the central region but also in the peripheral F-actin-rich domains that drive axon elongation (lamellipodia and filopodia) (Fig. 4F). The same localization was observed for SCHIP1c and IQCJs-SCHIP1 isoforms (data not shown).

Our results suggested that AC axons from mutant embryos might be misguided. Therefore, we asked how they respond to guidance cues from surrounding tissues. The lateral striatum is the first structure encountered by piriform cortex axons when they approach the midline and is known to express several guidance cues that control AC formation (Falk et al., 2005; Ho et al., 2009; Sahay et al., 2003). We analyzed the response of axons from E14.5 anterior and posterior piriform cortex explants to molecules secreted by lateral striatum explants in co-culture assays. Cortical explants, but not striatal explants, developed long axons within 24 h (Fig. 5A). We blindly determined a guidance index for cortical explants, as described previously (Castellani et al., 2000; Falk et al., 2005), which takes into account the density and the length of axons growing towards and away from the striatal explant (Fig. 5A). We analyzed the response of piriform cortex axons to striatal explants from embryos of either the same genotype or the ‘opposite' genotype. In WT co-cultures, axons of anterior cortical explants were repelled by the striatum, whereas the axons of posterior cortical explants appeared unaffected, revealing specific and differential intrinsic guidance properties of the anterior and the posterior piriform cortices (Fig. 5B). The differential response of WT anterior and posterior piriform cortex axons was similar when co-cultured with Δ10 mutant striatum (Fig. 5B). The repulsive response of mutant anterior piriform cortex axons to the striatum was unaffected (Fig. 5B). By contrast, mutant posterior piriform cortex axons displayed significant repulsive responses when co-cultured with either WT or mutant striatum (Fig. 5B). Thus, the guidance properties of WT and mutant striatum appeared to be similar, suggesting that SCHIP1 does not play a key role in guidance molecule expression/secretion by the striatum. Accordingly, the expression patterns of the striatum-expressed guidance cue SEMA3F (Falk et al., 2005; Sahay et al., 2003) were similar in mutant and WT embryos (supplementary material Fig. S5). Overall, this suggested that SCHIP1 has an intrinsic function in the guidance mechanisms of ACp axons, allowing them to project toward the striatum.

SEMA3F plays a key role in AC formation through repulsive activity mediated by its interaction with the receptor neuropilin 2 (NPN2, or NRP2) expressed in the piriform cortex (Chen et al., 2000; Giger et al., 2000; Sahay et al., 2003). Mice mutant for Sema3f or Npn2 display agenesis of the ACp, resembling the phenotype of Δ10 mice (Giger et al., 2000; Sahay et al., 2003). The expression pattern of Npn2 was similar in E14.5 WT and Δ10 embryos (supplementary material Fig. S5). We therefore asked whether SCHIP1 could be directly implicated in the response of piriform cortex axons to SEMA3F. We performed collapse assays on dissociated anterior and posterior piriform cortex neurons separately by adding Fc-coupled SEMA3F (SEMA3F-Fc) or the Fc fragment alone to the culture medium. In control conditions, the proportion of collapsed growth cones immunolabeled for NPN2 (Fig. 5C) was low and similar in WT and mutant cultures, showing that the Δ10 mutation did not affect the basal rate of collapse (Fig. 5D). SEMA3F induced a collapse response of growth cones from both anterior and posterior piriform cortex neurons (Fig. 5C,D). Interestingly, the percentage of collapsed growth cones was 15.8% higher in anterior compared with posterior piriform cortex cultures (Fig. 5D). This indicated that, at E14.5, anterior piriform cortex neurons were more sensitive to SEMA3F than posterior piriform cortex neurons. The response of mutant anterior piriform cortex neurons to SEMA3F was similar to that of WT neurons (Fig. 5D). Intriguingly, mutant posterior piriform cortex cultures showed a 20.5% increase in the collapse response to SEMA3F as compared with WT cultures (Fig. 5D), indicating that SCHIP1 influences the response of posterior piriform cortex to SEMA3F. These results underscore the increased repulsion observed for mutant piriform cortex explants co-cultured with striatum (Fig. 5B), although axon guidance in explants undoubtedly results from a combination of responses to several guidance cues.

Non-secreted guidance molecules also control AC development, including EPHB2 and EPHA4 (Henkemeyer et al., 1996; Ho et al., 2009). Mice mutant for Ephb2 present severe ACp defects, which resemble those of Δ10 mice. ACp defects are different in Epha4 mutant mice, since ∼50% of ACp axons are misdirected into the ACa tract while the remainder exit the midline and resume projection to the contralateral hemisphere. We therefore chose to further address the potential role of SCHIP1 in response to the EPHB2 guidance cue. The interaction of EPHB2 with ephrin B2 expressed by AC axons is necessary for the proper guidance of ACp axons, notably in overriding repulsion mechanisms (Cowan et al., 2004). The extracellular domain of EPHB2 is sufficient to induce the collapse of growth cones expressing ephrin B (Mann et al., 2003).

In situ hybridizations showed that the expression patterns of EPHB2 were similar in E14.5 WT and Δ10 embryos (supplementary material Fig. S5). This prompted us to evaluate whether SCHIP1 could also be implicated in the response of axons to EPHB2 by adding the Fc-coupled extracellular domain of EPHB2 (EPHB2-Fc) to the culture medium. Immunolabeling for human Fc after cell fixation confirmed that EPHB2-Fc bound to piriform cortex neurons (Fig. 6A). EPHB2-Fc induced the collapse of growth cones after 10 min in WT and mutant cultures (Fig. 6B,C). However, after 30 min, the number of collapsed growth cones was reduced in WT, whereas it remained high in mutant cultures (Fig. 6C). This effect was also observed in anterior and posterior piriform cortex cultures performed separately (data not shown). It was abolished by expression of the Flag-tagged SCHIP1b isoform in mutant neurons (Fig. 6D). In addition, the number of membrane-associated EPHB2-Fc puncta was higher in mutant than in WT collapsed growth cones after 30 min of EPHB2-Fc incubation, whereas it was similar after 10 min of incubation (Fig. 6E). These observations suggested that SCHIP1 influences growth cone recovery after EPHB2-induced collapse, possibly by controlling ephrin B levels at the membrane.

We further investigated this phenomenon by performing time-lapse imaging of growth cones (Fig. 6F,G). The onset of growth cone collapse was comparable in WT and mutant neurons (Fig. 6F). At 10 min of EPHB2-Fc incubation, the proportion of collapsed growth cones was similar in WT and mutant cultures. This proportion decreased in WT at 15 min of EPHB2-Fc incubation, whereas it remained unchanged in mutant cultures (Fig. 6G). These observations confirmed that SCHIP1 also controls the response of growth cones to EPHB2, allowing them to quickly recover after collapse.

In the present study, we demonstrate that the cytoskeleton-associated protein SCHIP1 plays an important role during brain development. We generated Δ10 mutant mice deficient for expression of all Schip1 isoforms and showed that they present a thinner piriform cortex and severe AC defects, which are likely to result from impaired axon growth and navigation. In vitro assays indicated that SCHIP1 modulates axon outgrowth and responses of growth cones to repulsive guidance cues in piriform cortex neurons. Remarkably, Schip1^{Gt(ROSA)77Sor} mice, which lack expression of the SCHIP1a and IQCJ-SCHIP1 isoforms only (Chen et al., 2004), do not present AC defects (supplementary material Fig. S6). This supports the importance of SCHIP1b, SCHIP1c and IQCJs-SCHIP1, which are the most highly expressed isoforms during piriform cortex development, in AC formation.

In Δ10 mice, ACa and ACp display different morphological abnormalities: the ACa is present but thinner than in WT mice, whereas the ACp is absent. Developmental studies suggest that both ACa and ACp abnormalities result from a combination of altered axon outgrowth and guidance errors. In vitro studies of piriform cortex neurons support this hypothesis. Neurons from anterior and posterior mutant piriform cortex display impaired axon elongation without cell death and an impaired response to guidance cues required for AC formation. Expression analyses indicate that the different phenotypic severities of AC branches are not directly linked to the expression levels of Schip1 in piriform cortex, but might be due to specific and differential intrinsic properties of the anterior and the posterior piriform cortex neurons. Consistently, we show that WT anterior piriform cortex axons are repelled by the striatum, whereas posterior piriform cortex axons are not. In addition, anterior piriform cortex neurons from WT E14.5 embryos display a higher sensitivity to SEMA3F than posterior piriform cortex neurons. Interestingly, the Δ10 mutation triggers the repulsion of posterior piriform cortex axons by the striatum, and enhances the response of posterior piriform cortex growth cones to SEMA3F. These results highlight specific guidance properties for posterior piriform cortex axons. Growth cone recovery after EPHB2-induced collapse also appears delayed in mutant axons. This indicates that SCHIP1 is required for the posterior piriform cortex axon response to at least two guidance cues, and further suggests that ACp agenesis in Δ10 mutants might result from impaired responses to these molecules, in combination with axon outgrowth defects.

Of note is the fact that Δ10 mutants represent the first mouse model with ACp agenesis associated with an increased axon sensitivity to repulsive guidance cues. So far, ACp agenesis was observed in mutant mice lacking guidance molecules, including SEMA3F, EPHB2, and their receptors NPN2 and ephrin B2, respectively. Thus, the similarity of ACp phenotypes in these mice and Δ10 mutants might appear counterintuitive. However, the developmental origins of the ACp agenesis in mice mutant for guidance molecules are likely to differ from that in Δ10 mutants. Indeed, in the absence of EPHB2, for instance, the ACp axons project aberrantly to the ventral forebrain (Henkemeyer et al., 1996; Ho et al., 2009), whereas this is not the case in Δ10 mutants. In mice deficient for axon guidance molecules, ACp agenesis may result from mistargeting of posterior piriform cortex axons to other brain regions followed by pruning. Changes in axon response to repulsive guidance cues in Δ10 mice could contribute to axon defects in different ways. The development of the axons requires adaptation mechanisms that allow their elongation toward a guidance cue gradient or the adjustment of their sensitivity to the environment (Ming et al., 2002; Piper et al., 2005). Growth cone recovery after collapse, which was observed by others (Campbell and Holt, 2001; Mann et al., 2003), could be one of these mechanisms, and the decrease of growth cone recovery seen in Δ10 mice could impair axon ability to navigate in a guidance cue gradient. In addition, the increased sensitivity of axons could induce axon degeneration and the death of the neurons, which normally project their axon through the AC. Indeed, both semaphorins and ephrins have been shown to promote cell death (Park et al., 2013; Vanderhaeghen and Cheng, 2010), and we observed cell death in E14.5 mutant piriform cortex, which is likely to depend on extrinsic factors since mutant neurons do not exhibit increased death in vitro. Alternatively, or in addition, mutant neurons impaired in axon growth might not reach anti-apoptotic factors, such as neurotrophic factors expressed at the midline (Barnes et al., 2007; Huang and Reichardt, 2001).

The exact mechanisms by which SCHIP1 controls axon development are unclear. Spontaneous neuronal activity was proposed to stimulate axon growth (Mire et al., 2012). The IQCJ-SCHIP1 isoform is a late component of the axon initial segment and could therefore play a role in controlling the excitability of mature neurons (Martin et al., 2008). However, this is unlikely to be the case in our in vitro experiments since we analyzed axon outgrowth at stages at which the axon initial segments have not yet formed (data not shown), hence suggesting earlier functions of SCHIP1 during axon development. Indeed, as an ankyrin-binding protein, SCHIP1 could participate in transmembrane protein localization and stabilization during early axon development. Axon outgrowth and guidance rely on a cascade of cellular events, including signal integration by CAMs and guidance receptors at the membrane and dynamic cytoskeletal reorganizations at the growth cones. L1-CAMs, which associate directly with ankyrins, have been implicated in multiple aspects of these processes (Gil et al., 2003; Herron et al., 2009; Kiryushko et al., 2004; Nishimura et al., 2003). Thus, SCHIP1 could mediate axon outgrowth by regulating ankyrin interactions and the responsiveness of L1-CAMs, such as L1CAM and NRCAM, expressed by piriform cortex neurons.

A role for SCHIP1 as an adaptor-associated protein within the cortical cytoskeleton would also be consistent with its involvement in the response of growth cones to repulsive cues. In the context of AC axons, the SEMA3F-activated signaling pathway involves the interaction of its receptor NPN2 with NRCAM (Falk et al., 2005). The interaction of NPN1 (NRP1) with L1CAM and TAG-1 (CNTN2) in response to SEMA3A induces endocytosis of the complex, which enhances the signaling pathway leading to repulsion (Castellani et al., 2000,, 2004; Schmid et al., 2000). Similar mechanisms involving NRCAM could operate in AC axons in response to SEMA3F. Ankyrins have been suggested to control NRCAM dynamics at the membrane and L1CAM endocytosis (Falk et al., 2004; Needham et al., 2001). SCHIP1 could therefore participate in the SEMA3F-induced response by regulating ankyrin interactions with NRCAM and the stability of the NPN2-NRCAM complex at the membrane. In addition, ephrin B signal transduction upon EPHB-ephrin B interaction involves molecular pathways leading to repulsion (Xu and Henkemeyer, 2012). This process needs termination mechanisms, notably clathrin-mediated endocytosis of EPHB-ephrin B complexes, to lower receptor levels at the membrane and allow growth cone desensitization and adaptation to guidance cues (Georgakopoulos et al., 2006; Marston et al., 2003; Parker et al., 2004; Tomita et al., 2006; Zimmer et al., 2003). The persistence of high numbers of EPHB2-Fc clusters on the growth cone surface after long-lasting EPHB2-Fc incubation in mutant cultures suggests that SCHIP1 could contribute to the regulation of ephrin B internalization. It has been shown that L1CAM/CHL1 interacts with ephrin A and is required in repulsive responses (Demyanenko et al., 2011). Thus, SCHIP1 could be a component of ephrin-associated protein complexes containing L1-CAMs and ankyrins.

Beside ankyrins, SCHIP1 is able to associate with schwannomin, an actin-binding protein that links membrane proteins to the cortical cytoskeleton (McClatchey and Fehon, 2009). Schwannomin has been shown to play roles in axon growth and guidance (Lavado et al., 2014; Schulz et al., 2010; Yamauchi et al., 2008). However, whether schwannomin contributes to SCHIP1-mediated axon outgrowth and guidance is uncertain since SCHIP1 and schwannomin expression patterns are different during brain development (Huynh et al., 1996). In addition, Nf2^F/F;Emx1-cre mice do not present AC defects but instead CC agenesis, which does not result from a lack of schwannomin in callosal neurons or their progenitors but in midline neural progenitor cells (Lavado et al., 2014). Schwannomin appeared to regulate guidance cue expression, whereas SCHIP1 does not and is directly involved in axon development.

In conclusion, our study reveals a role for a cytoskeleton-associated protein in AC and, to a lesser extent, in CC axon development. It highlights that intracellular mechanisms, directly downstream of surface molecules, also need to be considered in axonal tract development. Schip1Δ10 mice potentially represent a unique model to better understand these mechanisms. In addition, these mice can be helpful in dissecting piriform cortex functions in olfaction since they display behavioral and cognitive deficits that might be related to piriform cortex dysfunction (our unpublished results).

The description and characterization of SCHIP1 antibodies are presented in supplementary material Methods and Fig. S2; other antibodies are listed in supplementary material Table S5.

The identification of mouse SCHIP1 isoforms, generation and characterization of Schip1 mutant mice, and animal procedures (including ethics statement) are described in supplementary material Methods, Fig. S1 (Schip1Δ10 mice) and Fig. S6 [Schip1^{Gt(ROSA)77Sor} mice].

Primary cultures of dissociated neurons were performed as described in supplementary material Methods. To express SCHIP1 isoforms in neurons, Schip1 cDNAs were fused by PCR 3′ to nucleotides encoding the tag Flag and cloned into a pCIG plasmid (Megason and McMahon, 2002) that allows co-expression of EGFP in nuclei. Constructs were electroporated in neurons before plating using the Neon Transfection System (Invitrogen) according to the manufacturer's protocol (two 20-ms pulses, 1350 V). For collapse assays, SEMA3F-Fc (R&D Systems, #3237-S3-025), EPHB2-Fc (R&D Systems, #467-B2-200) and control Fc fragment (Rockland, #009-0103) were added to the culture medium 2 days after plating, at the lowest concentration at which a collapse response was observed (250 ng/ml). For explant cultures, pieces of tissue comprising the anterior piriform cortex and the AON or more posteriorly located piriform cortex were dissected from E14.5 embryos and manually cut into ∼500 µm diameter explants. For co-culture experiments, the lateral striatum lining the piriform cortex was dissected and cut into ∼700 µm diameter explants. Explant cultures were performed in three-dimensional plasma clots as described previously (Castellani et al., 2000) and fixed in 4% paraformaldehyde/10% sucrose/PBS after 24 h in culture.

Tissue processing, immunostaining, immunoprecipitations from brain lysates, acetylcholinesterase and X-gal staining, Nissl coloration, BrdU injection and detection, axon visualization by fluorescent lipophilic tracers and in situ hybridization were performed using standard protocols as described in supplementary material Methods.

Images were acquired as described in supplementary material Methods and analyses were performed using ImageJ software (NIH). The thickness of the neocortex, the piriform cortex and its three layers was measured on coronal sections throughout the anteroposterior axis. Three sections were quantified and averaged per animal. Three equidistant lines perpendicular to the surface of the layers were drawn on each image as described previously (Sarma et al., 2011). The thickness of each layer was measured along these three lines and averaged per section. For morphological studies of dissociated neurons, the length of axons was measured using the NeuronJ plug-in. For explant axon growth experiments, the axon growth area was calculated by subtracting the explant area from the total area occupied by the explant and the axons. To estimate axon guidance we used a qualitative guidance index as described previously (Castellani et al., 2000; Falk et al., 2005). The global influence of the lateral striatum on the piriform cortex axon trajectories was scored blindly from −2 (when most axons grew away from the lateral striatum) to 2 (when most axons grew toward the lateral striatum).

To quantify the proportions of collapsed growth cones, we defined as control growth cones those that exhibited complex profiles with multiple filopodia and broad lamellipodia, and as collapsed growth cones those that typically lacked lamellipodia and possessed only one or two major F-actin-positive filopodia. For time-lapse experiments, phase-contrast imaging revealed very small protrusions (<2 µm) in addition to major filopodia in certain collapsing growth cones, which were not detectable by phalloidin staining on fixed neurons and that were assumed to correspond to retracting filopodia undergoing actin depolymerization. They were not considered as ‘major’ filopodia. For quantification, a value of 1 was assigned to collapsed growth cones and a value of 0 to non-collapsed growth cones. The proportion was defined as the number of values 1 compared with the total number of values. Statistical analyses were performed as described in supplementary material Methods.

We are grateful to R. Boukhari, A. Rousseau, S. Thomasseau and M. Savariradjane (Institut du Fer à Moulin) for animal care, genotyping and assistance with microscopy; D. Godefroy (Institut de la Vision imaging platform) for technical support with nanozoomer acquisitions; A. Bellon (Development and Neuroscience Department, Cambridge University) for advice on piriform cortex cultures; and M. Nosten-Bertrand, F. Francis, P. Gaspar and R. Belvindrah (Institut du Fer à Moulin) for help with statistical analyses and critically reading the manuscript.