Cadherin-7 enhances Sonic Hedgehog signalling by preventing Gli3 repressor formation during neural tube patterning

Sonic Hedgehog (Shh) is a ventrally enriched morphogen controlling dorsoventral patterning of the neural tube. In the dorsal spinal cord, Gli3 protein bound to suppressor-of-fused (Sufu) is converted into Gli3 repressor (Gli3R), which inhibits Shh-target genes. Activation of Shh signalling prevents Gli3R formation, promoting neural tube ventralization. We show that cadherin-7 (Cdh7) expression in the intermediate spinal cord region is required to delimit the boundary between the ventral and the dorsal spinal cord. We demonstrate that Cdh7 functions as a receptor for Shh and enhances Shh signalling. Binding of Shh to Cdh7 promotes its aggregation on the cell membrane and association of Cdh7 with Gli3 and Sufu. These interactions prevent Gli3R formation and cause Gli3 protein degradation. We propose that Shh can act through Cdh7 to limit intracellular movement of Gli3 protein and production of Gli3R, thus eliciting more efficient activation of Gli-dependent signalling.


Introduction
The vertebrate neural tube is patterned along the dorsoventral (DV) axis to form different progenitor domains, identified by unique patterns of transcription factor expression and distinct neuronal fates [1]. This DV patterning results from a ventral-to-dorsal gradient of Sonic Hedgehog (Shh), which is complemented by bone morphogenetic proteins (BMPs) and Wnts acting in the dorsal neural tube. These morphogens provide positional information by promoting specific transcription factor profiles in each progenitor domain [2,3]. Although progenitor domains are sharply delimited along the DV axis, the mechanisms underlying well-defined boundaries of gene expression in the neural tube are still unclear.
Shh is secreted by the notochord and the neural tube floor plate, generating a dynamic ventral-to-dorsal gradient of concentration that results in spatiotemporally graded Shh pathway activation and ventral neural tube patterning [2,4,5]. Shh peptide is proteolytically cleaved to generate an N-terminal biologically active fragment (N-Shh) [6], which binds the transmembrane protein Patched1 (Ptch1) [7]. In the absence of N-Shh, Ptch1 represses the movement of another transmembrane protein, Smoothened (Smo), to the primary cilium [8]. When Ptch1 is bound to N-Shh, Smo can accumulate in the primary cilium and transduce Shh signalling intracellularly, modulating the activity of Gli1-3 transcription factors [6].
Gli2 and Gli3 exist as either full-length transcriptional activators (Gli2FL and Gli3FL) or truncated N-terminal fragments generated by partial proteolysis of the carboxyl terminus (Gli2R and Gli3R), acting as transcriptional repressors [9]. Suppressor-of-fused (Sufu) functions as an inhibitor of Shh signalling by binding Gli2/3FL proteins in the cytoplasm and preventing them from reaching the nucleus [10]. In the absence of Shh, Gli2/3FL proteins associated with Sufu are thought to be phosphorylated by PKA, GSK3b and CKI at the base of the primary cilium, causing their conversion into Gli2/3R [10][11][12][13]. Gli2R is completely degraded, but Gli3R translocates to the nucleus and represses transcription of Shh-target genes [14]. Activation of Shh signalling is believed to cause increased localization of unphosphorylated Gli2/3FL within the primary cilium, causing their transport to the nucleus, where they can activate Shh-target genes [10]. While Gli2FL is a strong transcriptional activator, Gli3FL has a low transcription-promoting activity and it rapidly undergoes proteosomal degradation when Shh signalling is active [11,12,15]. Therefore, GliR function can be mainly ascribed to Gli3R, while Shh-dependent transcriptional activation is mostly mediated by Gli2FL [10]. Despite major headway in dissecting the regulation of Shh signal transduction, there is still limited knowledge of the molecular mechanisms controlling trafficking and processing of Gli proteins.
Cdh7 is a member of the classical cadherin family, which is expressed in the intermediate neural tube region at the lower end of the Shh morphogen gradient [6]. Here, we show that Cdh7 is required to ventrally delimit the expression domain of dorsal spinal cord genes, thus allowing the specification of ventral territories. We provide evidence that Cdh7 exerts this function by positively regulating Shh signalling. The extracellular region of Cdh7 is bound by N-Shh, which alters Cdh7 distribution within the cell membrane. This promotes association of the intracellular domain of Cdh7 with Sufu and Gli3FL. As a result of these interactions, Gli3R production and nuclear accumulation are inhibited, and Gli3 protein is instead targeted for degradation, thus enhancing Shh signalling levels. These results reveal a new mechanism of Shh signalling regulation and help to understand how a highly dynamic spatio-temporal Shh gradient can result in well-defined boundaries of gene expression.  [17 -20]. Confirming previous observations [18],  1ab,ac). Therefore, these results suggest that Cdh7 expression occurs within a specific range of Shh signalling activation. This was confirmed by analysing Cdh7 expression, along with that of the ventral spinal cord marker Isl1/2, the intermediate marker Lim1/2 and the dorsal marker Pax7, in explants of presumptive intermediate neural tube that were cultured for 20 h with different concentrations of recombinant N-Shh (rN-Shh) (figure 1ad,ae). Isl1/2 expression was observed at 4 nM rN-Shh and Lim1/2 expression at 2 nM N-Shh, while Pax7 expression was repressed by rN-Shh. Cdh7-positive cells were detectable only in the presence of 2 nM rN-Shh, indicating that Cdh7 expression is specifically promoted by low/moderate doses of Shh signalling (figure 1ad,ae).
Similar results were obtained using SAG (3- To gain more insight into Cdh7 function in the DV patterning of the neural tube, we electroporated a plasmid coexpressing green fluorescent protein (GFP) and Cdh7 into HH st. 10 chick spinal cord, and analysed its effects on Pax7 expression at HH st. 15. Control embryos were electroporated with a plasmid expressing GFP only. In control embryos, GFPexpressing cells located in the dorsal half of the spinal cord were mostly Pax7-positive (Pax7 þ ) (figure 2a-c). By contrast, a significant fraction of the cells overexpressing Cdh7 were Pax7-negative (Pax7 2 ) (figure 2d-g; yellow-labelled and blue-labelled cells or arrows in figure 2c,f,g correspond to Pax7 2 and Pax7 þ -electroporated cells, respectively). When Pax7 þ cells were counted in the intermediate neural tube close to the ventral border of Pax7 expression domain, a clear reduction in the fraction of Pax7 þ cells was observed in Cdh7-expressing cells in comparison with cells expressing GFP only (figure 2h; electronic supplementary material, figure S1a-n). Dorsally localized Cdh7-overexpressing cells, however, retained Pax7 expression (figure 2g; electronic supplementary material, figure S1g-n). By staining electroporated embryos with an antibody against Dbx1, a marker of the p0 progenitor domain [23], we found ectopic Dbx1 þ cells close to the dorsal border of the Dbx1 expression domain upon Cdh7 overexpression, but not in embryos expressing GFP only (figure 2i-m; electronic supplementary material, figure S1o-ad). Thus, Cdh7-dependent signalling can repress Pax7 expression specifically in the intermediated spinal cord region, leading to the acquisition of more ventral fates.
To test whether Cdh7 is required for proper spinal cord patterning, we electroporated HH st. 10 embryos with a plasmid expressing a Cdh7-targeting siRNA (siRNA-Cdh7), which could effectively knock down Cdh7 expression (figure 2n,o,y-ac), or with a control siRNA plasmid (siRNA-control). Electroporation of siRNA-Cdh7, but not of siRNA-control, caused ventral expansion of the Pax7 þ domain (figure 2p-x).    Dbx1 Cdh7 Cdh7 Dbx1  Notably, Cdh7 overexpression significantly enhanced Shh signalling in the presence of 2 nM rN-Shh, but not in the presence of 4 nM rN-Shh or in the absence of rN-Shh (figure 3a; electronic supplementary material, table S1). The specificity of this effect was confirmed by co-electroporation of plasmids expressing Cdh7 and siRNA-Cdh7 (figure 2n,o), which abolished the increase in reporter expression observed in Cdh7expressing explants (figure 3b; electronic supplementary material, table S1). Furthermore, reporter expression was not enhanced upon overexpression of truncated versions of Cdh7 lacking the intracellular or the extracellular domains, respectively (figure 3c,d; electronic supplementary material, table S1). N-cadherin (Ncdh) and cadherin-20 (Cdh20) were unable to enhance Shh signalling in this assay (figure 3e; electronic supplementary material, table S1). These observations suggest that Cdh7 is a specific binding partner of Shh and a positive regulator of Shh signalling in the presence of low/moderate levels of Shh ligand. We investigated the mechanisms of Cdh7-dependent regulation of Shh signalling in more detail. We found that 25 nM cyclopamine-KAAD, a specific Smo inhibitor [21], abrogated Cdh7-dependent upregulation of GBS-luc reporter expression (figure 3f; electronic supplementary material, table S1), indicating that Cdh7 requires Smo function to promote Shh signalling. In agreement with these results, Cdh7 was able to reinforce GBSluc reporter expression when overexpressed together with a constitutively active form of Smo (SmoM2) [21], especially in the presence of 2 nM rN-Shh (figure 3g; electronic supplementary material, table S1). We then analysed whether Cdh7 could exert similar effects following co-overexpression with the Smo effectors Gli1, Gli2 and Gli3. Notably, Cdh7 could enhance reporter expression in combination with 2 nM rN-Shh and Gli3FL (figure 3h; electronic supplementary material, table S1), but not Gli1, Gli2 or Gli3R (electronic supplementary material, figure S3; all p-values for the experiments shown in figure S3 are reported in electronic supplementary material, table S2). Similar results were obtained when performing GBS-luc reporter assays in vitro using NIH3T3 cells, which bear a functional Shh signalling pathway [24]. Confirming assays performed with chick embryo explants, GBS-luc reporter activity was significantly increased by 4 nM rN-Shh, but not by 2 nM rN-Shh. Furthermore, Cdh7 expression in NIH3T3 cells significantly enhanced Shh signalling together with 2 nM rN-Shh, but not with 4 nM rN-Shh or without rN-Shh. The effects of Cdh7 could not be mimicked by Cdh20 or deletion mutants of Cdh7, and they were sentitive to the presence of siRNA-Cdh7 and KAAD-cyclopamine. Moreover, in the presence of 2 nM rN-Shh, Cdh7 could strengthen Shh signalling in collaboration with SmoM2 or Gli3FL, but not with Gli1, Gli2 and Gli3R (electronic supplementary material, figure S4; all p-values for the experiments shown in figure S4 are reported in electronic supplementary material, table S3). These results suggest that Cdh7 modulates Shh signalling by acting at the level of Gli3FL, but not through Gli1, Gli2 or Gli3R. They also suggest that Smo-dependent signal transduction is necessary for Cdh7 action and facilitates it.
Endogenously produced N-Shh peptide undergoes posttranslational modifications through covalent binding of a cholesterol moiety at its C-terminus and a palmitate moiety at its N-terminus [10]. To confirm whether Cdh7 could functionally interact with fully lipidated N-Shh, we performed GBS-luc reporter assays with explants from chick embryos co-electroporated with plasmids encoding for Cdh7 and Figure 2. (Opposite.) Shh-dependent Cdh7 expression controls specification of the Pax7 þ /Pax7 2 neural tube boundary. (a-c) Immunostaining analysis with anti-Pax7 (red staining) and anti-GFP (green staining) mAbs on HH st. 15 chick neural tube, following unilateral electroporation of a control plasmid driving expression of a chimeric GFP protein fused to a nuclear localization signal; (c) shows a higher magnification of boxed regions in (a) and (b). Yellow and blue circles indicate Pax7negative and Pax7-positive electroporated cells, respectively, within the intermediate spinal cord region (bracket). Within this region, most of the GFP-expressing cells are Pax7-positive. f, floor plate. Scale bars, 50 mm. (d-g) Immunostaining with anti-Pax7 (red staining) and anti-GFP (green staining) antibodies on HH st. 15 neural tube, following unilateral electroporation at HH st. 10 of a plasmid driving expression of both Cdh7 and a chimeric GFP protein fused to a nuclear localization signal; (f ) shows a higher magnification of (d,e) at the level of the dorsal spinal cord, while (g) shows further enlargement of the electroporated intermediate spinal cord region. Cells overexpressing Cdh7 (green cells in (e), (f ) and (g)) within the intermediate or the dorsal spinal cord are mostly negative or positive for Pax7, respectively. Brackets indicate the dorsal half of the spinal cord in (d ), and the intermediate (bottom bracket) or the dorsal (top bracket) spinal cord regions in (f ). In (f,g), yellow arrows point to electroporated cells that are Pax7 2 , whereas blue arrows point to electroporated Pax7 þ cells. f, floor plate; so, somite. Scale bar, 100 mm for (e); 50 mm for f; 25 mm for (g). (h) Quantification of the percentage of the electroporated Pax7 2 or Pax7 þ cells within the intermediate spinal cord region, following electroporation of expression plasmids (EP) encoding for both Cdh7 and GFP (blue bars) or GFP only (orange bars). At least 13 sections from three different embryos were used for these analyses. Error bars represent s.e.m.; *p , 0.005 according to Student's t-test. (i-l) Immunostaining with anti-Dbx1 (red staining) and anti-GFP (green staining) antibodies on HH st. 15 neural tube, following unilateral electroporation at HH st. 10 of a plasmid driving expression of both Cdh7 and a chimeric GFP protein fused to a nuclear localization signal; (l) shows a higher magnification of the boxed area (i,j,k). Blue arrows point to electroporated

Cdh7 binds Shh protein and enhances Shh signalling
As shown above, Cdh7 can repress Pax7 expression in the intermediate spinal cord region, but not in more dorsal areas. We speculated that this spatially limited activity might be due to a molecular interaction between Cdh7 and Shh, which is not present in the dorsal spinal cord, and explored this possibility by means of co-immunoprecipitation assays. As described in the Material and methods (see electronic supplementary material), lysates of NIH3T3 cells transiently expressing chick Cdh7 were incubated with rN-Shh, followed by immunoprecipitation using an anti-Shh monoclonal antibody (mAb) or an anti-Cdh7 monoclonal antibody (mAb). Cdh7 (106 kDa) co-precipitated with N-Shh and vice versa (figure 4a), indicating direct binding between Cdh7 and N-Shh. N-Shh was not precipitated in pulldown experiments with anti-Cdh7 mAb using control NIH3T3 cells, which do not express Cdh7 (data not shown). Co-immunoprecipitation of endogenous N-Shh and Cdh7 was also observed using extracts from HH st. 17 spinal cord tissue (figure 4b). By contrast, rN-Shh did not co-immunoprecipitate with Ncdh or Cdh20 in assays with NIH3T3 cells transiently    [25,26], and Cdh20 is the closest cadherin family member to Cdh7 [27], indicating that the binding between N-Shh and Cdh7 is highly specific.
We performed saturation binding experiments using COS-7 cells expressing Cdh7, which were treated with conditioned medium from 293 cells expressing N-Shh tagged with alkaline phosphatase (AP) (figure 4e; electronic (e) . By performing immunoprecipitation assays with different Cdh7 deletion mutants, we found that cadherin repeat 1 (CR1) and cadherin repeat 2 (CR2) domains are the major binding sites for N-Shh ( figure 4g,h). Binding between N-Shh and Cdh7 was not prevented by anti-Shh mAb (clone 5E1), which inhibits N-Shh binding to Ptch1 (electronic supplementary material, figure S7) [28], suggesting that Shh interacts with Cdh7 and Ptch1 through different domains. Sufu plays a pivotal role in controlling Gli3 processing to Gli3R [9,10]. Remarkably, co-immunoprecipitation assays using anti-Cdh7 mAb and Cdh7-expressing NIH3T3 cells showed stronger association of Cdh7 with Gli3FL and Sufu in cells cultured in the presence of rN-Shh, in comparison with control cells. In particular, treatments with 2 nM rN-Shh roughly doubled the amount of Gli3FL and Sufu bound to Cdh7 in comparison with untreated cells (figure 4i-k; western blot quantifications are reported in electronic supplementary material, table S5). By contrast, Cdh7 did not interact with Gli3R (figure 4l; electronic supplementary material, table S5) or GSK-3b neither in the absence nor in the presence of rN-Shh (figure 4m; electronic supplementary material, table S5). Co-immunoprecipitation assays with anti-Cdh7 or anti-Ncdh mAbs using HH st. 17 chick embryos confirmed that Cdh7, but not Ncdh, interacts with Gli3FL in vivo ( figure 4n-p). These results suggest that Cdh7 can bind Gli3FL and Sufu at the intracellular level and that these interactions are strengthened when the extracellular region of Cdh7 is bound to Shh.

Shh promotes Cdh7 aggregation within the cell membrane
As Shh enhances the association of Cdh7 with Gli3FL and Sufu, we speculated that Shh might alter Cdh7 distribution within the cell membrane. To address this question, we performed immunocytochemical analysis of NIH3T3 cells transiently

Cdh7 prevents Gli3R production and promotes degradation of Gli3 protein in the presence of Shh
As Shh is able to alter Cdh7 distribution at the cell membrane, we investigated whether Cdh7 aggregates forming in In the absence of exogenous rN-Shh, Cdh7 was present across the entire cell membrane and Sufu was weakly detectable in the cytosol (figure 6a-c). Following rN-Shh treatments, Cdh7 aggregates were associated with Sufu at the cytoplasmic side of the cell membrane (figure 6d-s), while no association was detectable between Sufu and Ncdh, Cdh20 or a truncated Cdh7 version lacking the intracellular domain (electronic supplementary material, figure S9 and figure S10a-i). Artificially cross-linking Cdh7 molecules on the cell surface using anti-Cdh7 mAb resulted in association with Sufu, but could not activate GBS-luc reporter expression, indicating that Cdh7 aggregation per se is not sufficient to promote Shh signalling (electronic supplementary material, figure S10j-m). These results suggest that Cdh7 positively modulates Shh signalling by interacting with Sufu at the intracellular level and that this interaction involves Shh-dependent regulation of Cdh7 distribution within the cell membrane.
To investigate whether Cdh7 interaction with Gli3FL and Sufu can modulate Gli3R production, we generated a construct encoding for a fusion protein of mouse Gli3FL with an N-terminal GFP tag and a C-terminal V5 tag (GFP-Gli3FL-V5). This allowed to detect nuclear accumulation of Gli3R as positive GFP nuclear staining together with negative V5 staining, while co-localization of GFP and V5 staining or the absence of GFP staining would indicate decreased production of Gli3R. We transfected GFP-Gli3FL-V5 along with either control or Cdh7-expressing vectors into NIH3T3 cells, followed by cell culture in the presence of 0, 2 or 4 nM rN-Shh and immunocytochemical localization of GFP-tagged and/or V5-tagged peptides.
No GFP staining was detectable after treatments with 4 nM rN-Shh (figure 6z-ab), indicating that elevated levels of Shh signalling can prevent Gli3R formation. Nuclear V5 staining was detectable at these doses of rN-Shh, suggesting that they cause Gli3 protein degradation to C-terminal fragments [12], which translocate into the nucleus. These degradation products are likely to coexist with Gli3R at lower rN-Shh doses (2 nM), while they predominate over Gli3R at higher rN-Shh doses (4 nM). In the absence of rN-Shh, Cdh7 expression had no impact on Gli3R formation (figure 6ag-ai).
No GFP staining, however, was detectable after treatments of Cdh7-expressing cells with 2 nM rN-Shh (figure 6aj,ap-as), suggesting that Cdh7 efficiently blocks Gli3R production in the presence of these lower doses of rN-Shh. V5 staining was largely excluded from the nucleus in Cdh7-expressing cells treated with rN-Shh (figure 6ak,al,an,ao,ap-as), suggesting that C-terminal fragments resulting from Gli3 degradation in the presence of both rN-Shh and Cdh7 do not undergo nuclear translocation. V5 staining and Cdh7 aggregate staining, however, did not co-localize (electronic supplementary  rsob.royalsocietypublishing.org Open Biol. 7: 170225 material, figure S11), indicating that Gli3 C-terminal fragments forming in these conditions do not associate with Cdh7 aggregates. Immunoblotting analysis using nuclear extracts led to results consistent with immunostaining data. In this case, cells were transfected with a construct encoding for Gli3FL-V5 fusion protein, with or without Cdh7-expressing vector, and an antibody against the Gli3 N-terminal region was used to detect Gli3R or Gli3FL. Cells were co-transfected with a control GFP vector to account for transfection efficiency. In agreement with immunostaining results, a nuclear 80 kDa Gli3R band was detectable with anti-Gli3 N-terminus antibody both in the absence of rN-Shh and in the presence of 2 nM rN-Shh, but it was reduced to approximately half of the levels detected in untreated cells after treatments with 4 nM rN-Shh (figure 6at,aw; western blot quantifications are reported in electronic supplementary material, table S6). By contrast, the levels of nuclear 190 kDa Gli3FL band, as detected with anti-V5 antibody, were roughly doubled in rN-Shh-treated cells in comparison with untreated cells (figure 6au,av; electronic supplementary material, table S6). These results confirm the idea that Gli3R formation can take place at lower doses of rN-Shh (2 nM) but it is efficiently inhibited at higher rN-Shh doses (4 nM). They further suggest that rN-Shh treatments lead to Gli3 protein degradation to C-terminal fragments along with limited nuclear accumulation of Gli3FL (which is detectable by immunoblotting, but not by immunostaining). Notably, in Cdh7-expressing cells treated with 2 nM rN-Shh, the nuclear Gli3R band was decreased to levels comparable to those present in cells treated with 4 nM rN-Shh and roughly corresponding to half of the levels detected in the absence of rN-Shh (figure 6aw, ay; electronic supplementary material, table S6). These results confirm that Cdh7 can inhibit Gli3R formation in the presence of low/moderate doses of rN-Shh. No V5-positive 190 kDa Gli3FL band was detectable by immunoblotting in nuclear extracts of Cdh7-expressing cells (figure 6ax, ay; electronic supplementary material, table S6), indicating that, in the presence of rN-Shh, Cdh7 causes efficient Gli3 protein degradation to Gli3 C-terminal fragments.
We confirmed that, in immunostaining assays, detection of V5 signal without GFP signal was due to C-terminal Gli3 peptides forming as a result of Gli3 degradation by means of immunoblotting assays with cytoplasmic extracts of cells transfected with Gli3FL-V5 using antibodies against the Gli3 N-terminal region or V5 tag. While no 80 kDa Gli3R or 190 kDa Gli3FL bands were detectable, we observed several low molecular weight bands of various sizes (20 -70 kDa) by immunoblotting with anti-V5 antibodies (electronic supplementary material, figure S12). This suggests that Gli3FL is unstable in the cytoplasm and undergoes partial proteolysis generating C-terminal fragments. Cotransfection of plasmids expressing Cdh7 and siRNA-Cdh7 prevented the effects of Cdh7 on both Sufu localization (electronic supplementary material, figure S13a-i) and Gli3 processing (electronic supplementary material, figure S13j-r), indicating their specificity.
Taken together, these results indicate that, by associating with Sufu and Gli3FL, Cdh7 can effectively collaborate with Shh to prevent formation of Gli3R, leading to Gli3FL cytoplasmic degradation and enhanced activation of Shh signalling.

Discussion
A gradient of Shh protein, released from the notochord and floor plate, plays a crucial role in ventral neural tube patterning [2,5,6,29]. Analysis of a variety of mutant mice, lacking different components of the Shh signalling pathway, shows that Shh-dependent patterning extends up to a sharp boundary between Pax7-negative and Pax7-positive domains in the ventral and dorsal neural tube, respectively [2,5,6,29]. As a result of Shh graded action, the Pax7 2 region becomes subdivided into the floor plate and five sharply delimited neural progenitor domains ( p3, pMN, p2 -p0), each of them expressing specific transcription factors [2,5,6,29]. A crucial question that remains only partially addressed is how an apparently continuous gradient of Shh protein results in distinct, well-defined boundaries of gene expression at different levels of the spinal cord DV axis. A crucial element in the interpretation of the Shh gradient appears to be the regulatory architecture of the transcriptional network activated by Shh signalling, which involves cross-repressive interactions between different transcription factors [30][31][32]. This mechanism ensures both activation of distinct genes at different thresholds of Shh signalling, and the establishment of mutually exclusive gene expression domains delimited by sharp borders [30 -32]. Although much progress has been made to clarify how floor plate, p3, pMN and p2 domains are specified in the ventral spinal cord, the molecular mechanisms establishing the boundary between the Pax7 2 and Pax7 þ regions in the intermediate spinal cord remain poorly investigated.
Cdh7 is expressed on the ventral side of the Pax7 þ /Pax7 2 boundary in the intermediate spinal cord region overlapping with the dorsal tail of the Shh gradient, and its dorsal limit of expression is confined to the ventral border of the Pax7 þ region (figure 7a). Confirming previous results [18], we found that Cdh7 expression in this intermediate spinal cord domain is dependent on low/moderate Shh levels. We show for the first time that Cdh7 overexpression represses Pax7 expression, while Cdh7 knock-down causes ventral expansion of Pax7 þ domain. In these assays, Cdh7 could inhibit Pax7 expression only in the proximity of the Pax7 þ /Pax7 2 boundary, but not in the most dorsal spinal cord region, which is devoid of Shh. Therefore, Shh promotes Cdh7 expression in the intermediate spinal cord and also collaborates with Cdh7 to negatively regulate Pax7 expression. Supporting this idea, we found that Cdh7 interacts with Shh protein and cooperates with Shh to prevent Gli3R formation. This interaction boosts Shh signalling at low/moderate doses of the Shh ligand. We speculate that this Cdh7-dependent mechanism of Shh pathway enhancement leads to a sharp drop in the levels of Shh signalling on the dorsal side of the Cdh7 2 /Cdh7 þ boundary with respect to cells on the ventral side of the border. This, in turn, facilitates establishment of Pax7 2 and Pax7 þ domains on each side of this border.
Based on these observations, we propose the model shown in figure 7a. According to this model, ventral to the Cdh7 2 /Cdh7 þ boundary, Shh-dependent Cdh7 expression reinforces Shh signalling by inhibiting formation of Gli3R, thus preventing Pax7 expression and specification of dorsal identities. Dorsal to the Cdh7 2 /Cdh7 þ boundary, however, lower levels of Shh in the absence of Cdh7 are not sufficient to prevent Gli3R formation and Pax7 expression. As previously described [33 -36], BMP and Wnt signals play a major role in promoting Pax7 expression in the dorsal spinal cord either directly or indirectly via positive regulation of Gli3 expression, while Pax7-dependent repression of Cdh7 expression [18,37] probably contributes to sharpening the boundary between the Cdh7 þ /Pax7 2 and Cdh7 2 /Pax7 þ regions. One possible caveat to this model comes from previous reports showing that the Pax7 þ domain in the dorsal spinal cord is not affected in Gli3 mutant mice [38], suggesting that Cdh7 may also modulate Pax7 expression via Gli3-independent mechanisms. It should be noted, however, that the Pax7 þ domain spreads ventrally in either Shh or Smo mutant mice, whereas it is correctly confined to the dorsal spinal cord in Shh/Gli3 or Smo/Gli3 double mutant mice [38 -40]. Therefore, inhibition of Gli3R formation in the intermediate spinal cord is both necessary and sufficient to ventrally delimit the Pax7 þ domain, even though Pax7 expression in the dorsal spinal cord can be maintained independently of Gli3. Although Cdh7-dependent homophilic interactions may also have a role in refining the Pax7 þ / Pax7 2 boundary and delimiting coherent progenitor domains, Cdh7-dependent regulation of Shh signalling clearly plays a major role in this process, because Cdh7 overexpression in the dorsal neural tube, outside of the Shh range, was unable to affect Pax7 expression and alter DV patterning. Further work will be needed to elucidate possible roles of Gli3-independent and/or cell adhesion-dependent mechanisms of Cdh7 function in spinal cord patterning.
In spite of continuous progress in the field of Shh signal transduction, the biochemical machinery controlling processing of Gli proteins in the absence or in the presence of Shh remains only partially understood. According to recent interpretations [10,41,42], in the absence of Shh, Gli2/3 proteins bind Sufu in the cytoplasm and transit through the primary cilium. When crossing the base of the primary cilium, Gli2/3FL undergo phosphorylation by PKA, GSK3b and CKI, leading to their conversion into Gli2/3R. While Gli2R is degraded, Gli3R accumulates in the nucleus and represses transcription of Shh-target genes. Activation of Shh signalling is thought to increase the residence time of Gli2/3FL -Sufu complexes within the primary cilium, thus protecting Gli2/3FL from PKA/GSK3b/CKI-dependent phosphorylation and preventing formation of Gli2/3R. This leads to nuclear accumulation of Gli2FL, causing activation of Shh targets, while unphosphorylated GLI3FL appears to be quickly degraded by the proteasome [10 -12,41].
We found that the extracellular region of Cdh7 binds Shh and this interaction promotes the association of Cdh7 with Sufu on the intracellular side of the cell membrane. We also found that Shh stimulates Cdh7 aggregation, suggesting that Cdh7 movement within the cell membrane is reduced in the presence of Shh and that a mechanism involving Shh-dependent Cdh7 translocation to the primary cilium is unlikely. Preliminary evidences suggest that Cdh7 is excluded from the primary cilium either in the presence or in the absence of Shh (data not shown). Based on these observations, we propose that, following binding to Shh, Cdh7 associates with Sufu and retains the Sufu -Gli3FL complex in the cytoplasm, thus preventing its translocation at the base of the primary cilium and Gli3FL processing to Gli3R (figure 7b,c). As we observed extensive degradation of the Gli3 protein in the presence of both Cdh7 and Shh, we speculate that inhibiting Sufu -Gli3FL movement to the base of the primary cilium and Gli3FL phosphorylation by PKA/ rsob.royalsocietypublishing.org Open Biol. 7: 170225 GSK3b/CKI eventually leads to Gli3FL dissociation from Sufu and degradation, irrespective of whether this involves increased residence of the Sufu -Gli3FL complex within the primary cilium (in the presence of high Shh levels without Cdh7) or within the cytoplasm (in the presence of both Cdh7 and low/moderate Shh levels). It should be noted, however, that Cdh7-dependent activation of Shh signalling is sensitive to Smo inhibitors and is enhanced by an active form of Smo, indicating that Smo-mediated signal transduction, possibly involving the primary cilium, is required for Shh signalling activation even in the presence of Cdh7. Cdh7 ability to enhance Shh signalling was limited to low/ moderate doses of Shh. No Cdh7-dependent increase in Shh signalling was detectable at higher Shh levels, suggesting that the cross-talk between Cdh7 and Shh is likely to involve complex, dose-dependent mechanisms. Future investigations will need to shed light on how Cdh7-dependent and Smo-dependent mechanisms converge on the regulation of Shh-target genes, and on how the interactions between Cdh7 and the Shh signalling pathway are influenced by the levels of the Shh ligand.
In conclusion, Cdh7 expressed in the intermediate region of the vertebrate spinal cord is required for proper neural tube patterning. This role is due to the ability of Cdh7 to enhance Shh signalling in the presence of low/moderate levels of Shh, by causing more efficient inhibition of Gli3R formation than would be achieved by these levels of Shh on their own. We propose that this mechanism leads to a marked difference in the levels of Shh signalling between Cdh7 þ and Cdh7 2 cells exposed to limited amounts of the Shh ligand. This, along with opposite regulation of Cdh7 expression by Shh and Pax7, ensures establishment of a sharp boundary between the Pax7 þ and Pax7 2 neural tube regions. By focusing mainly on the role of transcription factors, previous landmark studies have shown that the interpretation of the Shh gradient depends on the regulatory logic of downstream molecular networks [30 -32,43], and this work excitingly shows how embedding transmembrane proteins such as Cdh7 in Shh-dependent molecular networks can set the stage for morphogen-dependent tissue patterning.

Material and methods
Detailed description of experimental procedures is available in the electronic supplementary material.
Data accessibility. Additional data are available in the electronic supplementary material.