Sonic Hedgehog switches on Wnt/planar cell polarity signaling in commissural axon growth cones by reducing levels of Shisa2

Commissural axons switch on responsiveness to Wnt attraction during midline crossing and turn anteriorly only after exiting the floor plate. We report here that Sonic Hedgehog (Shh)-Smoothened signaling downregulates Shisa2, which inhibits the glycosylation and cell surface presentation of Frizzled3 in rodent commissural axon growth cones. Constitutive Shisa2 expression causes randomized turning of post-crossing commissural axons along the anterior–posterior (A–P) axis. Loss of Shisa2 led to precocious anterior turning of commissural axons before or during midline crossing. Post-crossing commissural axon turning is completely randomized along the A–P axis when Wntless, which is essential for Wnt secretion, is conditionally knocked out in the floor plate. This regulatory link between Shh and planar cell polarity (PCP) signaling may also occur in other developmental processes.


Introduction
The complex neuronal network is in part assembled through the stepwise events of axon pathfinding. Many axonal growth cones follow the instruction of a set of guidance cues until they reach their intermediate targets; they then switch their responsiveness to follow new guidance cues which allow them to navigate to the next intermediate targets (or the final targets). The mechanisms of these important switches are not well understood. In the developing spinal cord, commissural axons arise from the dorsal margin, project along the dorsal-ventral axis and then turn anteriorly after they have reached and crossed the ventral midline (Bovolenta and Dodd, 1990). Wnts are expressed in a decreasing anterior-to-posterior gradient in the floor plate and direct the post-crossing commissural axons to turn anteriorly after midline crossing (Lyuksyutova et al., 2003). Previous work showed that components of the planar cell polarity (PCP) signaling pathway mediates Wnt attraction and anterior turning of commissural axons (Lyuksyutova et al., 2003;Shafer et al., 2011;Zou, 2012;Onishi et al., 2013;Onishi et al., 2014). We report here that Shh-Smoothened (Smo) signaling activates PCP signaling by inhibiting the expression of Shisa2, an inhibitor of Wnt signaling that blocks Frizzled3 (Fzd3) glycosylation and cell-surface presentation.

Increased Shisa2 expression in dorsal commissural neurons in Smoothened (Smo) conditional knockout mice
As Shh is a major signaling molecule in the ventral midline, we hypothesized that Shh signaling may regulate the expression of genes relevant to the switch of responsiveness of commissural neurons during midline crossing. We, therefore, performed transcriptome analyses in the dorsal spinal cord, where the cell bodies of commissural neurons reside, from control Smo fl/fl and Smo conditional KO (cKO) E11.5 embryos crossed with Wnt1-Cre. Consistent with a previous study, we observed compromised anterior turning of commissural axons in Smo cKO (Figure 1-figure supplement 1) (Yam et al., 2012). Dorsal margins of the spinal cord were dissected (50-100 mm from the dorsal edge) and lysed to extract total RNA. Stranded-mRNAs were prepared and sequenced by singleend RNA-sequencing ( Figure 1A). Among a total of 15,737 identified genes, 363 genes showed significant differences in mRNA levels ( Figure 1B). The expression of several functional groups related to neuronal differentiation, axon morphogenesis and synaptic function was found to be changed significantly ( Figure 1C). We confirmed that Smo mRNA was indeed significantly reduced in Smo cKO ( Figure 1F). We then compared the expression levels of all core PCP genes, such as Fzd3, Celsr3 and Vangl2, that are required for A-P guidance of post-crossing commissural axons ( Figure 1D) (Shafer et al., 2011;Onishi et al., 2013). There was no significant difference in mRNA levels between control Smo fl/fl and Smo cKO. Interestingly, a homolog of a Wnt inhibitor, Shisa2, showed a 3.68-fold increase in Smo cKO ( Figure 1E) (Yamamoto et al., 2005).
We next tested whether Shisa2 mRNA level is increased in the dorsal spinal cord commissural neurons in Smo cKO using in situ hybridization (Figure 2A,B). To localize commissural neuron cell bodies, we performed combined fluorescent in situ hybridization and immunohistochemistry with a dI1 neuronal marker, Lhx2. Lhx2 is expressed in dI1 neurons derived from Atoh1-positive pdI1 progenitors. The dI1 neurons, which express high levels of Lhx2 (Lhx2 high ) at E11.5, project axons contralaterally (dI1c), whereas the Lhx2 low (and Lhx9 high ) dI1 neurons, which are located ventral to the Lhx2 high neurons at E11.5, project axons ipsilaterally (dI1i) (Figure 2A) (Ding et al., 2012). The axons of Lhx2 high dI1 neurons reach and cross the floor plate at E11.5, whereas the Lhx2 low dI1 neurons are a later population and extend axons ventrally at E12.5. We found that Shisa2 mRNA level was low in the control Smo fl/fl dorsal spinal cord but increased in the Lhx2 high dI1 neurons in Smo cKO ( Figure 2B,C). These results suggest that the Shisa2 expression in dI1c commissural neurons is regulated by Shh-Smo signaling during midline crossing. We noticed that, in Smo cKO, a few Lhx2 low neurons dorsal to the Lhx2 high dI1 neurons also showed upregulated levels of Shisa2 mRNA. These may be later-born Lhx2 high dI1 neurons, whose expression of Lhx2 is still usually not yet fully activated.

Shisa2 inhibits Fzd3 glycosylation and cell surface presentation
Shisa2 is a member of a family of transmembrane proteins with a single transmembrane domain. Xenopus Shisa, xShisa, interacts with xFrizzled8 and retains xFrizzled8 in the endoplasmic reticulum (ER), thus inhibiting canonical Wnt signaling (Yamamoto et al., 2005). We first tested whether mouse Shisa2 interacts with mouse Fzd3 by co-immunoprecipitation and found that Fzd3 and Shisa2 interact with each other ( Figure 3A,B). As a control, we also tested interaction between Shisa2 and Insulin receptor b (IRb), a cell-surface receptor unrelated to Fzd3, and found that IRb was not coimmunoprecipitated with Shisa2 (Figure 3-figure supplement 1). We previously reported that mouse Fzd3 is glycosylated (Onishi et al., 2013). When expressed in HEK293 cells, Fzd3 shows three major bands, which correspond to 'phosphorylated and glycosylated ', 'glycosylated' and 'unphosphorylated and unglycosylated (unmodified)' proteins ( Figure 3B). We found that Shisa2 reduced the amount of 'phosphorylated and glycosylated' and 'glycosylated' Fzd3 ( Figure 3A,B); the major form of Fzd3 that co-immunoprecipitated with Shisa2 is the 'unmodified' form ( Figure 3B). We then analyzed the level of Fzd3 at the cell surface using a surface biotinylation assay and found that Shisa2 expression robustly decreased the amount of Fzd3 on the cell surface ( Figure 3C,D); the cell surface level of IRb is meanwhile unaffected ( Figure 3C,D). GAPDH is a cytoplasmic marker and IRb is a cell surface marker, which allow assessment of fractionation. IRb was detected both in cell surface fraction and in total cell extract, whereas GAPDH was only found in total extract ( Figure 3C,D). Finally, we performed immunocytochemistry to characterize Fzd3 localization. In the absence of Shisa2, Fzd3-EGFP is localized diffusely in the membrane compartments of the entire cytoplasmic area and on the plasma membrane ( Figure 3E). In Shisa2-myc-expressing cells, Fzd3-EGFP and Shisa2-myc are colocalized with Calnexin, an ER marker, suggesting that Shisa2 blocks Fzd3-EGFP transport from ER to Golgi and cell surface ( Figure 3E,F). Therefore, our results suggest that Shisa2 may inhibit Wnt/PCP signaling by inhibiting Fzd3 glycosylation and trafficking to the cell surface.

Fzd3 glycosylation is required for its cell surface expression
To further test the importance of the glycosylation of Fzd3 in its trafficking, we set out to identify the glycosylation site(s) of Fzd3. A recent co-crystal structure of xWnt8 and xFz8 revealed a glycosylation site in the cysteine rich domain (CRD) region of xFz8, which is conserved among all vertebrate Frizzle proteins and which may be important for binding to Wnt ligands (Janda et al., 2012). In mouse Fzd3, this site is N42 ( Figure 4A). Fzd3 and Fzd6 have an additional putative glycosylation site (N356 in mouse Fzd3; Figure 4A) in their extracellular loop II. To test whether these putative sites are glycosylated, we made two point mutants, N42Q and N356Q, and the double mutant (2NQ [or N42Q + N356Q]). These Fzd3 mutants, N42Q, N356Q and 2NQ, showed faster mobility ( Figure 4B,C), suggesting that N42 and N356 are indeed glycosylation sites. The cell-surface level of Fzd3 (N42Q) is significantly reduced, whereas Fzd3 (N356) is completely undetectable ( Figure 4D), suggesting that glycosylation at N356 is more crucial for Fzd3 trafficking than that at N42Q. To better characterize Fzd3 glycosylation, we enriched glycosylated proteins using the Glycoprotein Isolation Kit, WGA (Thermo Scientific) ( Figure 4E). The wheat germ agglutinin (WGA) preferentially interacts with N-acetyl glucosamine (GluNAC), terminal GluNAC, and sialic acid structures, so that WGA-immobilized agarose allows us to enrich glycoproteins with modified GluNAC or sialic acid. We found that wildtype Fzd3 is detected in the glycoprotein fraction, whereas mutant Fzd3 (2NQ) is absent from the glycoprotein fraction, further confirming that N42 and N356 are glycosylation sites in Fzd3.
To ask whether Shh-Smo signaling regulates Fzd3 glycosylation in vivo, we examined the glycosylation level of endogenous Fzd3 in the dorsal spinal cord in Smo cKO. We generated polyclonal anti-Fzd3 antibodies and performed western blot. We detected two bands corresponding to 'glycosylated' and 'unmodified' bands, which are both absent from Fzd3 -/lysate (Figure 4-figure supplement 1). However, we noted a non-specific band that runs at the same position as the 'phosphorylation and glycosylation' band (marked with an asterisk). Therefore, this antibody can be utilized to detect 'glycosylation only' and unmodified Fzd3, but not the 'phosphorylation and glycosylation' band. We isolated glycoprotein using the Glycoprotein Isolation Kit, WGA (Thermo Scientific) and performed western blot to detect how much Fzd3 is glycosylated. We found that Fzd3 glycosylation ('glycosylation only' band) was robustly reduced in Smo cKO ( Figure 4F), suggesting that Shh-Smo signaling does indeed regulate Fzd3 glycosylation in vivo.

Forced expression of Shisa2 lead to A-P guidance defects of postcrossing commissural axons
To test whether Shisa2 downregulation is necessary for proper A-P guidance of commissural axons, we forced expression of Shisa2 using a heterologous CAG promoter in commissural neurons. We coelectroporated Shisa2-expressing plasmids with tdTomato-expressing plasmids into the dorsal spinal cord commissural neurons and analyzed their axon trajectory revealed by tdTomato in an 'openbook' culture ( Figure 5A) (Lyuksyutova et al., 2003;Wolf et al., 2008;Shafer et al., 2011;Onishi et al., 2013). Axons expressing control plasmid extended to the midline, crossed, and turned normally anteriorly ( Figure 5B,C; 85.9 ± 6.79%). By contrast, we found that axons of Shisa2-expressing neurons turned both anteriorly and posteriorly randomly after midline crossing ( Figure 5B,C; 54.3 ± 7.89%). This result suggests that the downregulation of Shisa2 is necessary for proper A-P guidance of commissural axons. As xShisa also binds to FGFR and as our RNA-seq data suggest that FGFR1, FGFR2 and FGFR3 are expressed in dorsal spinal cord (Figure 5-figure supplement 1A) (Yamamoto et al., 2005), we tested whether FGFRs are required for A-P guidance of commissural axons by blocking FGFR activity using pharmacological inhibitor LY2874455 (inhibits FGFR1, FGFR2, FGFR3, FGFR four and VEGFR2; Figure 5-figure supplement 1B). First, we tested whether LY2874455 can abolish FGFR autophosphorylation in our 'open-book' explants and found that LY2874455 suppressed phosphor-Y653/654 significantly (Figure 5-figure supplement 1C; Vehicle (DMSO), 1 (normalized); LY2874455, 0.43 ± 0.05). We then found that LY2874455 did not affect A-P guidance of commissural axons ( Figure 5-figure supplement 1D,E), suggesting that FGFR activity is not required for anterior turning of post-crossing commissural axons.

Knocking down Shisa2 caused precocious anterior turning of precrossing axons through activation of Wnt-PCP signaling
To investigate whether Shisa2 prevents pre-crossing commissural axons from responding to Wnts, we developed shRNA constructs against Shisa2 ( Figure 6A,B). As a control, we utilized scrambled shRNA (Shafer et al., 2011;Onishi et al., 2013). We then electroporated control and Shisa2 shRNAs into dorsal spinal cord. We found that the majority of control-shRNA-expressing commissural axons turned anteriorly after crossing the floor plate ( Figure 6C,D; 84.3 ± 2.3%), while a small subset of axons turned before crossing (12.6 ± 2.3%) or in the floor plate (3.0 ± 1.0%). This is consistent with our previous finding using this 'open-book' explant culture system (Wolf et al., 2008). By contrast, the expressino of both Shisa2 shRNAs significantly elevated the proportion of commissural axons that turned before crossing and in the floor plate ( Figure 6C,D) and decreased the proportion of those turning after crossing. To control for off-target effects of these shRNAs, we also performed rescue experiments. We introduced four silent mutations in the shRNA#1 targeting site, and two in the #2 site, and confirmed that the expression level of this mutant Shisa2 (which still encodes WT Shisa2) is not suppressed by shRNAs ( Figure 6A,B). We then electroporated Shisa2 shRNAs together with the Shisa2 rescue construct and found that it rescues the phenotype of Shisa2 shRNAs ( Figure 6C,D).
To test whether the precocious turning in Shisa2 knockdown mutants is caused by the premature activation of Wnt-PCP signaling, we performed double knockdown of Shisa2 and Vangl2, an essential component of Wnt-PCP signaling. We previously reported that Vangl2 is required for Wnt-stimulated axon outgrowth and essential for anterior turning of dorsal commissural axons (Shafer et al., 2011). We utilized the same Vangl2 shRNA and confirmed that Vangl2 knockdown led to A-P guidance defects (Figure 6-figure supplement 1). Then, we co-electroporated Shisa2 shRNA and Vangl2 shRNA into the dorsal spinal cord, and found that precocious anterior turning does not occur ( Figure 6C,D), supporting our notion that the precocious turning produced by Shisa2 knockdown is due to the premature activation of Wnt-PCP signaling.
To confirm that Shisa2 regulates Fzd3 trafficking in neurons, we directly tested whether Shisa2 knockdown can change cell surface levels of Fzd3 protein in the growth cone of commissural neurons. We utilized a construct that we developed in a previous study, in which Fzd3 was tagged with tdTomato to the carboxyl domain and the FLAG epitope tag was engineered to the extracellular N-terminus ( Figure 6-figure supplement 2A) (Onishi et al., 2013). We can label the cell surface Fzd3 using anti-Flag antibodies, while the total Fzd3 protein can be detected by the tdTomato signal. In neurons expressing control shRNA, only a small proportion of the Fzd3 protein was on the cell surface (  Although we have obtained genetic evidence for the essential roles of several PCP components in the A-P guidance of commissural axons, the genetic evidence for a role for Wnts in this guidance is still lacking. This is in part due to the fact that at least five Wnts are expressed in a decreasing anterior-to-posterior gradient (Lyuksyutova et al., 2003;Agalliu et al., 2009) our unpublished results).
As Wntless (Wls) is essential for the secretion of all Wnt proteins ( Figure 7A) (Bänziger et al., 2006;Fu et al., 2009), we knocked out Wls specifically in the floor plate by crossing the Wls floxed allele with the Shh-CreER T2 line; in the resulting line, Cre is only expressed in the floor plate and can be activated by intraperitoneal injection of tamoxifen into pregnant females (Harfe et al., 2004). First, we confirmed the specificity of Shh-CreER T2 by crossing with R26R-LSL-tdTom mice. After administration of tamoxifen at E8.5, embryos were collected at E11.5 and transverse sections were stained with floor plate and ventral spinal cord markers ( Figure 7B). We observed that tdTomato expression is specifically activated in the notochord and floor plate, where endogenous Shh is expressed ( Figure 7C). We also confirmed the knockout of the Wls gene from the floor plate by in situ hybridization ( Figure 7D). Second, we tested whether conditional knockout of Wls in the floor plate may cause patterning and cell fate defects. At a much later stage, at E13.5, the number of Lhx3 + ; Islet1/2 + Median Motor Column (MMC) neurons was found to be reduced, as was that of Lhx3 -. Islet1/2 + Hypaxial Motor Column (HMC) neurons were found to be increased in Wnt4/5 KO (Agalliu et al., 2009). We analyzed whether the cell numbers of different motor neuron pools are affected in Wls cKO spinal cord at E11.5 (Figure 7-figure supplement 1A). We did not observe any differences in the total numbers of motor neurons, marked by Islet1/2, between control and Wls cKO mice or the numbers of MMC (Lhx3 + ; Islet1/2 + ) and (HMC + PGC) (Lhx3 -; Islet1/2 + ) at this early stage (Figure 7-figure supplement 1B-D; p>0.05). In addition, the numbers of motor neuron progenitors (pMNs, marked by Oligo2) and V3 interneuron progenitors (p3, marked by Nkx2.2) were also unchanged (Figure 7-figure supplement 1B-D; p>0.05).
We then analyzed the midline pathfinding of commissural axons in the Wls cKO mice. The isolated spinal cords were prepared as 'open-book', and commissural axons were visualized using iontophoretic injection of DiI at a quarter of the way into the dorsal margin of the spinal cord. In control Wls fl/fl mice (without Shh-CreER T2 ), commissural axons turn anteriorly after floor plate crossing ( Figure 7E,F; 85.6 ± 15.7%). On the other hand, in Wls cKO mice spinal cords (with Shh-CreER T2 ), commissural axons turned randomly along the A-P axis after midline crossing ( Figure 7E,F; 38.7 ± 17.8%), providing genetic evidence that the floor-plate-secreted Wnts are essential A-P guidance cues for post-crossing commissural axons. Due to the intrinsic features of the DiI labeling technique, DiI sometimes over diffuses to label other classes of neurons. Other normal axons tend to have a clear growth pattern and look different from misguided axons in mutants, which tend to show inconsistent wandering patterns. In Figure 7E, some neurons, probably located more ventral to the dl1 neurons, were also labeled. Some of them turned before crossing. Other labeled neurons appear to run along the A-P axis close the injection sites. This occurs in both WT control and mutants. These  keeping Wnt/PCP signaling inactive. Consistent with this, Shisa2 expression using a heterologous promoter led to randomized turning of commissural axons along the A-P axis after midline crossing. Furthermore, Shisa2 knockdown induced Vangl2-dependent precocious anterior turning before midline crossing. At this stage of development, Shh is expressed only in the floor plate and other Hedgehogs are not expressed in or near the spinal cord (Zhang et al., 2001). Secreted Shh has two different lipid modifications at its N-terminus and C-terminus (fatty-acid and cholesterol, respectively) that prevent it from diffusing over a long distance (Sloan et al., 2015). Therefore, we propose that Shh in the ventral midline is detected by commissural axons and signals retrogradely to inhibit Shisa2 expression in the cell body. This inhibition allows Fzd3 to be trafficked to the cell surface, resulting in the activation of Wnt/PCP signaling in commissural axon growth cones ( Figure 7G). We also showed here that Wls, an essential component of Wnt secretion, is required for A-P guidance of commissural axons, providing genetic evidence for the role of Wnts in A-P guidance of postcrossing commissural axons. This regulatory link between Shh and PCP signaling pathways may also be important in other developmental processes. The mechanisms for axon responsiveness switches are not fully understood and have attracted increasing attention in recent years. The midline is an excellent model system for addressing this fundamental question. Spinal cord commissural neurons are first guided by Netrin1 to grow from the dorsal spinal cord to the ventral midline, where they cross to the contralateral side Serafini et al., 1994;Serafini et al., 1996;Podjaski et al., 2015;Yung et al., 2015). After midline crossing, they lose responsiveness to Netrin1 and become responsive to repellents in the floor plate and the ventral spinal cord, such as Semaphorins and Slits, which turn their trajectory from the dorsal-ventral axis to the anterior-posterior axis (Zou et al., 2000). Recent studies suggest that Netrin1 in the ventricular zone is sufficient to guide commissural axons to grow into the floor plate and that floor plate Netrin1 is not required, indicating that Netrin1 functions at a much shorter-range than previously thought (Dominici et al., 2017;Varadarajan et al., 2017). After midline crossing, commissural axons become responsive to the Wnts, which direct them to turn anteriorly (Lyuksyutova et al., 2003). Here, we show that Shh is a midline switch for Wnt/PCP signaling that operates by regulating the mRNA levels of Shisa2, which in turn regulates Fzd3 trafficking. We previously observed that it takes approximately 8 hr for commissural axons to cross the midline. This should be sufficient time to downregulate Shisa2 mRNA and protein levels and to allow Fzd3 to be glycosylated and trafficked to the cell surface. This switch mechanism may be optimal for the activation of PCP signaling after midline crossing, given the time it takes to cross the midline. We showed previously that Shh switches on repulsive responses to Semaphorin after midline crossing by inhibiting protein kinase A (PKA) (Parra and Zou, 2010). Therefore, Shh is a switch for growth cone responsiveness to both Wnts and Semaphorins. In addition to Shh, other switches in the midline have been proposed. At the Drosophila midline, Frazzled/DCC, a Netrin receptor, transcriptionally activates the expression of Commissureless, allowing attraction to be coupled to the downregulation of repulsion in precrossing commissural axons (Neuhaus-Follini and Bashaw, 2015). In the vertebrate midline, NrCAM and Gdnf activates Plexin-A1 to switch on responsiveness to Sema3B (Nawabi et al., 2010;Charoy et al., 2012). Floor-plate-derived neuropilin-2, probably functioning as a Semaphorin sink, was also proposed to be a switch mechanism for Semaphorin responsiveness (Yang et al., 2009;Hernandez-Enriquez et al., 2015). In rodents, cell-intrinsic signaling proteins, namely the 14-3-3 proteins, were proposed to switch Shh-mediated attraction to repulsion during midline crossing (Yam et al., 2012). In chick, Shh binds Glypican 1 to induce the expression of Hhip, Tamoxifen (0.1 mg/g of mother) was injected intraperitoneally at E8.5. Embryos were harvested and dissected at E11.5 for immunohistochemistry (IHC), in situ hybridization (ISH) and DiI tracing experiments. (C) Cre recombination (tdTomato signal) was detected specifically in the notochord and floor plate. After intraperitoneal injection of tamoxifen (0.1 mg/g mother) at E8.5, Figure 7 continued on next page which was proposed to mediate repulsion by Shh (Bourikas et al., 2005;Wilson and Stoeckli, 2013). Our study here showed that Shh switches on Wnt/PCP signaling by removing an inhibitory mechanism involving Shisa2. Taken together, there may exist highly sophisticated and tightly regulated switch mechanisms to ensure proper changes of responsiveness for axons at intermediate targets. These switch mechanisms can involve gene expression, signal modulation and ligand distribution, which may function together to ensure fidelity.
Although a gene expression regulatory mechanism linking Shh, Tbx2/Tbx3, Shisa3 and canonical Wnt signaling has been observed recently (Lüdtke et al., 2016), our study identified a link from Shh to Shisa2 and to non-canonical Wnt/PCP signaling, which regulates cellular and tissue morphogenesis through cytoplasmic signaling. Therefore, Shh may determine the time and space of cell polarity signaling through PCP signaling, as well as by gene expression through the canonical Wnt pathway. Currently, we do not know whether Shisa2 is a direct target of Shh-Smo signaling. It is also possible that the Shisa2 mRNA level is regulated by a posttranscriptional mechanism.
In this study, we identified two glycosylation sites of Fzd3, one is in the CRD region (N42) and another is in extracellular loop II (N356). It is possible that glycosylation at N42 might be important for the binding affinity with Wnts. Interestingly, the lowest band of the N42Q mutant showed faster mobility than WT ( Figure 4B), suggesting the possibility that all Fzd3 protein is glycosylated at N42. We show that glycosylation at N356 is more crucial for cell surface expression of Fzd3, this is a unique glycosylation site for Fzd3 and Fzd6, which may also be important for regulating interactions between Fzd3 and other PCP components.
This study also presents the missing genetic evidence for Wnts as axon guidance molecules in vertebrates. Owing to the high redundancy of Wnt family members (there are 19 Wnts in mammals), it has been hard to demonstrate that mutations of individual Wnts lead to axon guidance defects in vertebrates. Here, we conditionally knocked out Wls from the floor plate and thus perturbed the secretion of all Wnts from the floor plate. We found that post-crossing commissural axons completely lost directionality along the A-P axis, suggesting that Wnts are the key A-P guidance cues for commissural axons in the ventral midline.  Signaling, #9740, 1/2000, RRID:AB_11178519); and anti-Flag (SIGMA, M2, 1/2000, RRID:AB_439685) were purchased from the indicated venders. Anti-Lhx3 and Islet1/2 antibodies were kind gifts from Dr Pfaff of the Salk Institute. Alexa-conjugated secondary antibodies for mouse/rabbit/rat IgG and mouse IgM were purchased from Molecular Probes (1/500 dilution). HRP-conjugated secondary antibodies for mouse/rabbit/goat IgG were purchased from Jackson ImmunoResearch Laboratories (1/ 10000 dilution). LY2874455 was purchased from Selleckchem. Anti-Fzd3 rabbit polyclonal antibodies (1/2000 dilution) were generated by the Zou lab. The C-terminal cytoplasmic domain (505-666 a.a.) of mouse Fzd3 fused with glutathione S-transferase (GST) was generated using pGEX4T-1. Recombinant GST-Fzd3cyto protein was expressed in BL21 Escherichia coli and purified with glutathione sepharose 4B (GSH beads, GE Healthcare). Anti-Fzd3 polyclonal antibody was generated in rabbits by injection with recombinant GST-Fzd3cyto protein (Covance). Specificity of the anti-Fzd3 antibodies was validated using lysates from Fzd3 knockout dorsal spinal cord (Figure 4-figure supplement  1).

Cell lines and transfection
COS-7 (RRID:CVCL_0224) cells and HEK293T cells (RRID:CVCL_0063) were purchased from ATCC and maintained in Dulbecco's modified Eagle's medium (DMEM) containing 10% fetal bovine serum. Transfection of both COS-7 and HEK293T cells was carried out using 1 mg/ml Polyethyleneimine MAX (Polyscience). Mycoplasma contamination was monitored by DAPI staining. These cell lines are not on the lists of Cross-contaminated or Misidentified Cell Lines from the International Cell Line Authentication Committee. The dorsal margins where cell bodies of dorsal commissural neurons are located were dissected by L-shaped tungsten needle (50-100 mm from the dorsal edge) and collected into microcentrifuge tubes. Total RNA was extracted using RNeasy Plus Micro Kit (QIAGEN). We dissected a total of 14 embryos (seven for each genotype). Total RNA quality was assessed using an Agilent Tapestation. Samples had an RNA Integrity Number (RIN) of greater than 8.0. RNA libraries were generated using Illumina's TruSeq Stranded mRNA Sample Prep Kit using 300 ng of total RNA. RNA libraries were multiplexed and sequenced with 100 base pair (bp) paired single end reads (SR100) to an average depth of approximately 35 million reads per sample on an Illumina HiSeq2500 using V4 chemistry. Sequence results were analyzed using BaseSpace Cloud Computing System (Illumina). RNA reads were aligned using TopHat Alighment App, followed by Cufflinks Assembly and DE App to determine mRNA expression level. RNA levels were quantified using FPKM (Fragment per kilobase of exon per million fragments mapped). Gene Ontology analysis was performed using GO term mapper (http://go.princeton.edu/cgi-bin/ GOTermMapper).

Dorsal spinal cord extract
Mouse E11.5 embryos from control Smo fl/fl or Smo cKO were dissected and isolated spinal cords were prepared as 'open-book'. Dorsal margin of the spinal cord was collected into microcentrifuge tubes and then lysed with RIPA buffer (20 mM Tris HCl [pH 7.5], 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 5 mM NaF, 10 mM b-glycerophosphate, 1 mM Na 3 VO 4 , 1 mM DTT and protease inhibitor cocktail (SIGMA), 1% TX-100% and 0.1% SDS). Lysates were subject to immunoblotting with anti-Fzd3 and anti-GAPDH antibodies. The band intensity was quantified with ImageJ. Statistical analysis was performed with Prism7 (GraphPad Software).

Glycoprotein isolation
Isolation of glycoprotein from tissue/cell extracts was performed using a Glycoprotein Isolation Kit, WGA (ThermoScientific). For quantification of Fzd3 glycosylation, we prepared lysates from three control Smo fl/fl and three Smo cKO embryos. The band intensity was quantified with ImageJ. Statistical analysis was performed using Prism7 (GraphPad Software).

Surface biotinylation assay
The surface biotinylation and NeutrAvidin pull down methods have been described previously (Shafer et al., 2011;Onishi et al., 2013). Briefly, 48 hr after transfection with indicated plasmids, HEK293T cells (seeded on 20 mg/ml PDL-coated six-well plate) cells were washed with ice-cold PBS (pH 8.0) three times and incubated with 1 mg/ml Sulfo-NHS-LC-Biotin (ThermoFisher Scientific)/PBS for 2 min at room temperature to initiate the reaction, followed by incubation on ice for 1 hr. After quenching active biotin by washing with ice-cold 100 mM Glycine/PBS twice followed by normal icecold PBS, the cell lysates were incubated with NeutrAvidin agarose for 2 hr and then precipitated. For quantification, three independent experiments were performed and the band intensity was quantified with ImageJ. Statistical analysis was performed with Prism7 (GraphPad Software). of Shisa2 overexpression on commissural axon turning, 116 axons were counted for control, and 98 axons were counted for Shisa2 overexpression from eight different spinal cords (from four different litters). To quantify the effect of Shisa2 knockdown, 336 axons were counted for control (11 embryos), 427 axons were counted for shRNA#1 (eight embryos), 320 axons were counted for shRNA#2 (eight embryos), 92 axons were counted for Vangl2 shRNA (five embryos), 101 axons were counted for shRNA#1 + rescue (four embryos), 92 axons were counted for shRNA#2 + rescue (four embryos), 102 axons were counted for shRNA#1 + Vangl2 shRNA (four embryos) and 82 axons were counted for shRNA#2 + Vangl2 shRNA (four embryos). Immunohistochemistry E11.5 mouse embryos were fixed in 4% PFA for 2 hr on ice. After equilibration with 30% (w/v) sucrose in PBS, the fixed embryos were embedded in OCT compound (SAKURA) and frozen. Transverse sections were prepared with a cryostat (CM3050S, Leica) at a thickness of 20 mm and mounted on glass slides (Superfrost Plus, Fisher Scientific). Slides were washed in PBST (PBS + 0.1% TritonX-100) and incubated in 3% BSA in PBST (blocking solution) for 1 hr at room temperature. Slides were further incubated with primary antibodies diluted in blocking solution overnight at 4˚C, washed three times for 10 min each time in PBST and then incubated for 2 hr with secondary antibodies diluted in blocking solution at room temperature. The slides were washed again and mounted using Fluoromount G (Southern Biotech). All fluorescence images were taken using a Leica SP5 confocal microscopy. All quantification data were obtained using ImageJ software.

Statistical analysis
Statistical analysis of multiple comparisons was performed using one-way ANOVA followed by a Tukey-Kramer post-hoc test for multiple comparisons using Prism7 (Figure 1-figure supplement 1C, Figure 4C,D, Figure 6B,D, Figure 6-figure supplement 2). To compare two groups, Student's t test was performed (two-tailed distribution) using Prism7 ( Figure 3D, Figure 4F, Figure 5C In the RNA-seq analysis, q-values (modified p-values) were used to estimate significance (consider q < 0.05 as significant).