N-cadherin SPRY motifs bind unconventionally-secreted Fbxo45 and regulate multipolar neuron migration

During development of the mammalian neocortex, the orientation of migrating multipolar projection neurons is controlled by Reelin, a secreted glycoprotein, which increases cell-surface expression of N-cadherin. Although N-cadherin regulates cell-cell adhesion, recent results suggest that its adhesive function is not required to orient multipolar neuron migration. To understand N-cadherin function in multipolar migration, we performed two independent screens for embryonic brain proteins that bind the N-cadherin extracellular domain. Both screens detected MycBP2 and SPRY-domain protein Fbxo45, two components of an intracellular E3 ubiquitin ligase. We found that Fbxo45 is secreted by a non-classical mechanism, not involving a signal peptide and not requiring endoplasmic reticulum to Golgi transport. Secreted Fbxo45 stimulated neurite branching in culture. A SPRY-motif N-cadherin mutant did not bind Fbxo45 and failed to rescue neuron migration even though it still formed trans-homophilic adhesions. The results suggest that secreted Fbxo45 may regulate neurite branching and bind N-cadherin to orient multipolar neuron migration.


INTRODUCTION
The complex architecture of the mammalian neocortex arises through the generation, specification, migration and connection of different types of neurons (1)(2)(3). Projection neurons, born and specified in the pallial ventricular zone (VZ), migrate outwards to form discrete layers before they establish connections. Their migrations pass through discrete stages. They first migrate radially from the VZ to the subventricular zone (sVZ)/intermediate zone (IZ), where they become multipolar and migrate randomly (4). When they arrive in the upper IZ, they assume a bipolar morphology and migrate radially outwards, passing by the earlier-born neurons in the cortical plate (CP) until they stop at the marginal zone (MZ) and undergo terminal somal translocation. Genetic disruption of pathways that regulate neuron migration is associated with neurodevelopmental disorders including lissencephaly, epilepsy and schizophrenia.
Because cadherin trans-interactions are critical for cell-cell adhesion and recognition (5), we hypothesized that homophilic interactions between NCad on migrating neurons and NCad on other neurons, axons, or radial glia may orient multipolar migration. Indeed, neurons in culture will polarize towards an external source of NCad (23). However, our recent work revealed that a NCad W161A mutant, which cannot form "strand-swap" trans homodimers (24)(25)(26)(27), stimulates multipolar migration (28). We further found that NCad binds and activates fibroblast growth factor receptors (FGFRs) in cis (on the same cell) and this interaction is needed for multipolar migration (28). ECad is not able to replace NCad for this function even though binds FGFR. Mechanistically, NCad but not ECad protected FGFR from being degraded, and the first two of five extracellular calcium-binding domains on NCad (EC1 and EC2) were critical. These results leave open the possibility that additional proteins binding NCad EC1 or EC2 may regulate neuron migration.
Fbxo45 (F box/SPRY domain-containing protein 1) is a little-studied protein that is highly expressed in the nervous system and required for cortical lamination, axonal outgrowth and synaptic connectivity (29-31). Most F-box proteins bind Skp1, Cul1 and Rbx1 to form a SCF (Skp1-Cul1-F-box) E3 ubiquitin ligase complex. Fbxo45 is atypical in that it does not bind Cul1 or Rbx1 and instead associates with MycBP2/PAM (Myc binding protein 2/protein associated with Myc), forming a Fbxo45-Skp1-MycBP2 complex that has E3 ligase activity in vitro (30). The SPRY domain of Fbxo45 potentially interacts with substrates. Curiously, NCad was detected in a Fbxo45 interaction screen (32). Furthermore, knockdown of Fbxo45 decreased NCad expression and impaired differentiation of neuronal stem cells (32), suggesting that Fbxo45 interaction with NCad is involved in brain development.
Here we set out to identify secreted proteins that interact with the ectodomain of NCad and may regulate radial migration. Two different unbiased proteomics approaches detected Fbxo45 and MycBP2 as major binding partners for the extracellular domain of NCad. We found that the SPRY domain of Fbxo45 binds a SPRY domain consensus in the EC1 region of NCad but not ECad. Furthermore, Fbxo45 is secreted by an unconventional mechanism from neurons and other cells and regulates dendritic arborization in vitro. Cell autonomous knockdown of Fbxo45 interfered with neurogenesis, which precluded analysis of neuron migration. Therefore, we tested the potential role of Fbxo45 in neuron migration by an alternative approach. We generated a NCad mutant which does not bind to Fbxo45 but still forms calcium-regulated transhomophilic interactions. Unlike NCad W161A , which binds Fbxo45 and rescues multipolar migration (28), the mutant NCad that does not bind Fbxo45 did not rescue multipolar migration.
These results suggest that secreted Fbxo45 may stimulate NCad-dependent neuron migration during brain development.

BioID identifies Fbxo45 and MycBP2 as extracellular NCad-interacting neuronal proteins.
We used two approaches to identify NCad binding proteins. First, we adapted proximitydependent biotin identification (BioIDf) for extracellular use (33). BioID uses the BirA mutant (R118G, BirA*) to generate reactive biotinyl-AMP from biotin and ATP. Biotinyl-AMP rapidly reacts with nucleophiles in the immediate surroundings, including ε-amino groups of lysine residues on nearby proteins (33). BioID has been used to detect binding partners of relatively insoluble proteins including those at cell junction complexes (34-37). The working distance of BirA* is ~10 nm (34), which is about half the length spanned by the five EC repeats that comprise the NCad ectodomain (38) (diagrammed in Figure 1A). To detect proteins that might interact with either end of the NCad ectodomain, we inserted Myc-BirA* into HA-tagged NCad (NCad-HA) either between EC5 and the transmembrane domain (TM) (N5-BirA*) or in the middle of EC2 (N2-BirA*, Figure 1A). Insertion of small protein domains at these sites does not to interfere with cadherin function (38). As a control, the EC1-5 region of N5-BirA* was replaced with EC1-5 of ECad, to create E5-BirA*.
We first tested whether cadherin BirA* proteins were expressed and trafficked to the surface. Western blotting showed that all BirA* proteins were expressed, contained Myc and HA tags, and transferred biotin to cell proteins when provided with ATP and biotin in the media (Supplementary Figure 1A). Immunofluorescence of cadherin-BirA* fusion proteins confirmed that they were transported to the surface and localized to cell-cell junctions in HEK293 cells, resembling wildtype NCad (38) (Figure 1A and Supplementary Figure 1B). Moreover, NCad was biotinylated when cells co-transfected with NCad and N5-BirA* were labeled with ATP and biotin ( Figure 1B). These results suggest that N5-BirA* is in close proximity to NCad in cell-cell adhesions.
We then used N5-BirA* and N2-BirA* in BioID experiments to identify neuronal proteins that interact with the NCad extracellular domain. Figure 1C shows the experimental design. N5-BirA* and N2-BirA* were expressed in HEK293 cells that were then cultured with rat embryonic cortical neurons and labeled with biotin and ATP. Cell lysates were collected and samples were analyzed for protein expression (top, Figure 1D) and biotinylation (bottom, Figure 1D). The remaining sample was purified using high-affinity Streptavidin beads and digested with trypsin before LC-MS/MS analysis. The experiment was performed twice, once with GFP as a negative control (Experiment 1), and once with E5-BirA* as a negative control (Experiment 2). The peptides that were identified were searched against the human and rat protein databases. Since the neurons originated from rat embryos, we were able to identify neuronal proteins by looking for peptide sequences that differ between rat and human. The top ten rat proteins that were detected with N5-BirA* or N2-BirA* but not with GFP or E5-BirA* in both experiments are listed in Figure 1E, and a full list provided in Table 1. MycBP2 and Fbxo45 were the highest-ranked rat neuron proteins in both experiments (Figure 1E and Supplementary Table 1). MycBP2 also was the highest-ranked human protein in both experiments (Supplementary Table 2). The results suggest that Fbxo45 and MycBP2 are major proteins expressed by rat neurons that interact with the ectodomain of NCad.

Affinity purification of neuronal proteins that bind to NCad.
As an independent approach to detect proteins that might regulate neuron migration, we affinity purified proteins from embryonic mouse brain homogenate using recombinant NCad W161A ectodomain, which does not form strand-swap dimers, as bait. The ectodomain was fused to the human immunoglobulin constant region (Fc), expressed in HEK293T cells, and purified from the culture supernatant using a Protein A/G (PAG) column. An ECad-Fc fusion protein was prepared similarly as a negative control. Neuronal proteins were then purified as illustrated in Figure 2A. Mouse embryonic forebrains were homogenized, insoluble material was removed by centrifugation, and the supernatant passed over PAG. Flow through from the PAG column was then passed over ECad-Fc, to remove proteins that bind to ECad. The unbound material was then passed over NCad W161A -Fc, to select proteins that bind to NCad W161A . All three columns were washed and eluted at low pH. Samples of various fractions were visualized by silver staining after SDS-PAGE ( Figure 2B). As expected, PAG retained large amounts of ~55 and 25 KDa proteins, corresponding in size to immunoglobulin heavy and light chains (IgH and IgL, lane 7). ECad-Fc and NCad W161A -Fc eluates contained a spectrum of proteins of various sizes (lanes 9 and 11), including major amounts of ~110 kDa proteins corresponding to the cadherin-Fc fusions. The ECad-Fc and NCad W161A -Fc eluates were then subjected to preparative SDS polyacrylamide gel electrophoresis and the gel stained with colloidal Coommassie (Supplementary Figure 1C). Gel regions above and below the ~110 KDa cadherin-Fc bands were excised, trypsinized and analyzed by LC-MS/MS. The most abundant proteins that bound to NCad W161A but not to ECad are listed in Figure 2C, and a complete list is provided in Supplementary Table 3. MycBP2 and Fbxo45 were the major proteins. Skp1 was also detected.
Both our screens complement the findings of (32), who detected NCad as a major interacting protein for Fbxo45, and are consistent with an extracellular Fbxo45-Skp1-MycBP2 complex.

The EC1-2 region of NCad interacts with the SPRY domain of Fbxo45.
To confirm that Fbxo45 binds to NCad but not ECad, Fbxo45 was tagged with T7 at the N terminus and co-transfected with C-terminally HA-tagged cadherins. As expected, Fbxo45 coimmunoprecipitated with NCad but not ECad ( Figure 3A). When the ectodomain of NCad was deleted (∆EC) or the EC1-2 domains of NCad were switched to ECad (EN), Fbxo45 binding was inhibited ( Figure 3B). Moreover, NCad EC1-2 was sufficient to bind Fbxo45 ( Figure 3C). Taken together, our data suggest that the NCad but not ECad EC1-2 domains are necessary and sufficient to bind Fbxo45. This finding is consistent with our observation that Fbxo45 was more efficiently biotinylated if BirA* was fused into EC2 than into the juxtamembrane region (compare N2-BirA* and N5-BirA*, Table 1), and with a previous report (32).
We next investigated the NCad-binding region of Fbxo45. Fbxo45 contains an F-box domain (33-83) and a SPRY domain (92-283), which binds ubiquitin ligase substrates (30). We deleted different regions of the SPRY domain and assayed binding to NCad ( Figure 3D).
Remarkably, none of the mutants tested was able to bind NCad, even when we deleted as little as 10 residues from the C terminus (mutant 276) or five residues between 225 and 229 (∆225-229). These regions correspond to beta strands 11 and 15 in the predicted structure of the Fbxo45 SPRY domain (39). Our results implicate the folded structure of the Fbxo45 SPRY domain in binding to NCad.

SPRY-binding motifs in NCad
Since Fbxo45 binds the EC1-2 region of NCad but not ECad (Figure 3), we reasoned that the Fbxo45 SPRY domain makes contact with EC1-2 residues that differ between NCad and ECad. Three sequences resembling the canonical SPRY domain-binding consensus (40)) are present in this region, and all three differ between NCad and ECad ( Figure 4A). Chung et al (32) previously reported that deletion of motifs 2 and 3 abolished Fbxo45 binding. However, motifs 2 and 3 also fit the calcium-binding consensus (DXNDN (25,41)) and chelate calcium ions in the NCad ectodomain crystal structure (PDB: 3Q2W (42)). Therefore, deleting motifs 2 and 3 might interfere with calcium-dependent folding and indirectly disrupt Fbxo45 binding. To avoid disrupting calcium binding, we substituted each SPRY motif with the corresponding sequence from ECad, generating NCad mutants NC1, NC2, NC3, NC12 and NC123. These mutants, all tagged with HA, were then tested by co-expression with T7-Fbxo45 and assaying for co-immunoprecipitation ( Figure 4B). Substitution of motif 1 (D242S, I243S, N246E and Q247A) almost completely inhibited interaction, while substitution of motif 2 (M261Q) or motif 3 (V376I and P380A) was less inhibitory. Accordingly, the NC12 and NC123 compound mutants did not bind Fbxo45.
To confirm that SPRY motif substitution did not alter NCad folding or homophilic interactions, we co-expressed HA-tagged NCad wildtype or NC123 with T7-tagged NCad or Fbxo45 and tested for co-immunoprecipitation. The result showed that NC123-HA binds NCad-T7 but not T7-Fbxo45 ( Figure 4C). This suggests that NC123 folds normally and can form homophilic interactions. Moreover, NC123 could support calcium-dependent cell-cell adhesion.
wildtype or mutant NCad were expressed with a GFP marker in CHO-K1 cells, which lack cadherins (43). Transfected cells were incubated under gentle agitation in the presence or absence of calcium. Calcium-dependent aggregation was stimulated by wildtype NCad but not by NCad W161A (Figure 4D), as expected (26). Importantly, NC123 also stimulated calciumdependent cell aggregation ( Figure 4D and Supplementary Figure 2A). These results confirmed that NC123 can form trans-homophilic interactions and NCad W161A cannot. To investigate whether NC123 can also bind in trans to wildtype NCad, we repeated the cell aggregation assay using two cell populations marked by different fluorescent markers. CHO-K1 cells were separately transfected with NCad WT and GFP, or with NCad WT or mutants and mCherry. Equal numbers of transfected cells were allowed to aggregate with or without calcium. GFP+ NCad WT cells aggregated equally with mCherry+ NCad WT or NC123 but not with NCad W161A cells when calcium was added (Supplementary Figure 2B). These results indicate that NC123 and NCad W161A are complementary mutations, with NC123 binding to NCad but not Fbxo45 and NCad W161A binding to Fbxo45 but not NCad. Thus, the NC123 and NCad W161A mutations completely separate two functions of the NCad extracellular domain.

Fbxo45 reaches the cell surface through a non-classical secretion pathway
The binding of Fbxo45 to EC1-2 of NCad raises the question of how an ostensibly intracellular Fbxo45 reaches the extracellular domain of NCad. When cells are lysed for immunoprecipitation, proteins are released from intracellular compartments, from the cell surface, and from the lumen of secretory compartment, which is topologically outside the cell.
We tested whether Fbxo45 is on the cell surface using immunofluorescence of nonpermeabilized cells. HeLa cells were co-transfected with Flag-Fbxo45 and GFP-NCad (GFP was inserted between EC5 and TM). Anti-Flag and anti-GFP antibodies were added to non- To test whether Fbxo45 also localizes to the surface of neurons, primary mouse hippocampal neurons were transfected with Flag-Fbxo45. GFP was co-transfected to show the cell outlines. Flag-Fbxo45 was detected using anti-Flag antibodies with or without permeabilization ( Figure 5B). As expected, Fbxo45 was detected in permeabilized neurons throughout the cytoplasm and neurites. Importantly, Fbxo45 was detected on the surface of non-permeabilized cells, where it formed patches ( Figure 5B). The non-uniform distribution may be partly due to inaccessibility of some regions due to overlaying non-transfected neurons, or to uneven binding of Fbxo45 to the surface.
Fbxo45 could be actively secreted from cells or released from damaged cells. Secreted proteins typically contain N-terminal signal sequences for co-translational translocation into the lumen of the endoplasmic reticulum (ER) (44). Fbxo45 lacks a consensus signal peptide, and an N-terminal Flag tag did not inhibit its appearance on the surface ( Figure 5). However, some proteins that lack signal peptides are secreted by unconventional mechanisms (45)(46)(47)(48). While conventional secretion is inhibited by Brefeldin A (BFA), unconventional secretion is insensitive to BFA (49). We therefore tested whether BFA inhibits secretion of Fbxo45. As expected, BFA inhibited secretion of a similar-sized N-terminal signal sequence protein, NCad EC1-T7. In contrast, BFA slightly increased secretion of T7-Fbxo45 ( Figure 6A). BFA did not induce release of a cytoplasmic protein, tubulin. This suggests that Fbxo45 is secreted by a BFA-insensitive unconventional route.
To test whether untagged, endogenous Fbxo45 is secreted by neurons, primary mouse cortical neurons were incubated with serum-free Neurobasal media with or without BFA for 20-24 hrs. Media were collected and Fbxo45 detected with a specific antibody (32). Fbxo45 but not tubulin was detected in the neuron culture media, and secretion was increased by BFA, suggesting unconventional secretion ( Figure 6B).
NCad could potentially bind Fbxo45 in an intracellular vesicle and help ferry Fbxo45 to the surface. Therefore, we tested whether co-expression of NCad with Fbxo45 would stimulate Fbxo45 secretion. For this purpose we used full-length NCad, which travels through the conventional secretion pathway. As a negative control, we used NC123, which does not bind Fbxo45 ( Figure 4B). NCad did not affect secretion of T7-Fbxo45 from HeLa cells ( Figure 6C), or from CHO-K1 cells, which lack endogenous cadherins ( Figure 6D). This suggests that NCad does not regulate Fbxo45 secretion.
Taken together these results suggest that Fbxo45 is secreted by a non-classical, signalindependent pathway, and likely associates with NCad only after it reaches the cell surface.

Extracellular Fbxo45 regulates dendrite arborization.
Since NCad is required for dendrite morphogenesis (50-52), we tested if Fbxo45 regulates dendritic arborization and neurite outgrowth. GFP and vector or GFP and Flag-Fbxo45 We also tested if extracellular Fbxo45 regulates dendritic arborization and neurite outgrowth. Fbxo45 or control medium was prepared by from HEK293 cells that had been transfected with T7-Fbxo45 or control vector, and added to DIV5 hippocampal neurons that had been transfected that day with GFP. On DIV7, after 2 additional days of incubation, neurons were fixed and visualized. Secreted Fbxo45 increased arborization between 20 and 100 µm from the soma ( Figure 7F-H). The increase was statistically significant at 25 µm (p= 0.0252), 35 µm (p=0.0045) and 65 µm (p=0.0185) from the soma (Two-way ANOVA, Boneferroni's multiple comparison test). In addition, there were significant increases in the number of branches (p=0.0146, two-tailed t-test, n = 52 for control, n = 51 for secreted Fbxo45, Figure 7I) and the number of neurites (p=0.0211, two-tailed t-test, n = 52 for control, n = 51 for secreted Fbxo45,

Neuron migration through the intermediate zone requires NCad SPRY motifs
Since NCad and NCad W161A rescue migration of multipolar neurons (28) and bind to Fbxo45 (Table 3), we reasoned that NCad interaction with Fbxo45 may stimulate radial migration. We first tested whether Fbxo45 knockdown or over-expression regulates migration.
Fbxo45 knockdown inhibited neurogenesis or viability, so we were unable to test whether Fbxo45 is needed cell-autonomously for migration (data not shown). In contrast, over-expressed Fbxo45 inhibited migration (Supplementary Figure 5). This was unexpected if Fbxo45 interacts with NCad to stimulate migration. Since Fbxo45 is secreted, it is possible that over-expression alters the environment in a way that overwhelms directional information.
As an alternative approach, we asked whether NC123, which binds NCad but not Fbxo45, could support neuron migration in vivo. Rap1 blockade with Rap1GAP inhibits NCad upregulation on the neuron surface and delays migration in the multipolar zone (53). This migration delay can be rescued by over-expressed NCad or NCad W161A . We tested whether NC123 could also rescue the migration delay in this assay. Embryos were electroporated in utero at E14.5 to express GFP, Rap1GAP and NCad-HA or NC123-HA. Three days later, embryos were euthanized and the positions of the GFP-expressing neurons were visualized. As shown in Figure 8A and B, Rap1GAP inhibited neuron migration and this was rescued by NCad, as expected. However, NC123 did not rescue migration, suggesting that NCad Fbxo45-binding site is needed for migration from the IZ to the cortical plate. To confirm that NC123 was expressed in this experiment, sections were stained for the HA tag on NCad-HA and NC123-HA. The wildtype and mutant proteins were expressed at similar levels ( Figure 8C). Taken together with the finding that NCad W161A -HA rescues migration (28), these results suggest that NCad interaction with Fbxo45 or other proteins through its SPRY motifs is required to orient multipolar neuron migration.

DISCUSSION
NCad plays important roles during brain development, including stabilizing apical junctional complexes in the ventricular zone, regulating the multipolar to bipolar transition in the intermediate zone, stimulating terminal translocation at the top of the cortical plate, and coordinating axonal, dendritic and synaptic differentiation (6, 7, 11-13, 22, 53). While many of these functions presumably involve trans-homophilic interactions between NCad molecules on the surfaces of neighboring cells, some may not. Indeed, we recently found that transinteractions are not needed for NCad to stimulate radial migration in the intermediate zone (28).
In that study, we found that NCad binds, stabilizes, and activates fibroblast growth factor receptors (FGFRs), and that FGFR stabilization is needed for neuron migration. In the present study we identified Fbxo45 and MycBP2/PAM as major extracellular NCad-binding partners by two unbiased proteomic approaches. Fbxo45 is expressed in the developing brain and mutation affects cortical layering and axonogenesis (30, 31). MycBP2/PAM has no known role in cortical neuron migration, although it does regulate axonal pathfinding (54). We found that Fbxo45 is secreted by an unconventional mechanism, circumventing the usual ER to Golgi pathway, binds to the cell surface, and stimulates dendritic arborization in vitro. Moreover, a SPRY motif mutant NCad that cannot bind Fbxo45, NC123, but can form trans-homophilic interactions is unable to induce the radial migration of multipolar neurons. NC123 can still bind FGFRs (data not shown) suggesting that NCad may interact with both Fbxo45 and FGFRs during neuron migration.
A previous proteomics screen for Fbxo45-binding proteins detected interaction with the NCad extracellular domain and proposed that the interaction occurred inside the cell, before NCad reached the cell surface (32). However, the question of how Fbxo45, which lacks a classic N-terminal signal sequence, would translocate across the ER membrane and come into contact with the NCad extracellular domain was not addressed. We found that the NCad EC1-2 region is necessary and sufficient to bind Fbxo45 and that Fbxo45 is actively secreted from the cell by a non-classical, signal peptide-independent mechanism that is insensitive to BFA.
Several other proteins undergo unconventional secretion, including cytokine interleukin-1 (55), thioredoxin (56), fibroblast growth factors FGF-1 (57) and FGF-2 (58), α-Synuclein (59, 60), Galectins (61), phosphoglucose isomerase/autocrine motility factor (62) and heat shock proteins (48). However, the mechanisms and functions of unconventional secretion are still unclear. The lack of conserved signals, relatively small amounts of proteins secreted, and apparently different cell-type mechanisms have presented technical challenges (46,48,63). One mechanism, known as autophagy-mediated secretion, involves chaperone-mediated autophagy (CMA) to translocate proteins into lysosomes, followed by lysosome fusion with the plasma membrane (64). A loose consensus for CMA targeting contains a glutamine preceded or followed by four residues including at least one basic (K or R), one acidic (D or E), and one hydrophobic (I, L, V or F) residue. Fbxo45 contains such a sequence (residues 225-229: QIGER). However, deleting this sequence did not reproducibly inhibit Fbxo45 secretion (data not shown). It is possible that Fbxo45 secretion involves CMA and that variability in secretion may stem from variable levels of autophagy between experiments.
Because unconventional secretion by-passes the usual ER-Golgi route, Fbxo45 may not come into contact with NCad until after both proteins reach the surface. Indeed, cooverexpressing NCad had no effect on the quantity of Fbxo45 secreted ( Figure 6C and 6D).
However, Fbxo45 could be detected on the cell surface after secretion ( Figure 5). Curiously, the binding was not uniform. Extracellular Fbxo45 formed patches, co-localizing with NCad on parts of the cell surface but not others. It is possible that Fbxo45 is only secreted across sub-regions of the surface and binds immediately to NCad in those regions before it can diffuse.
Alternatively, some other molecule may be unevenly distributed and either masks or increases binding of Fbxo45 in those regions. The polarized binding of Fbxo45 may be important for its proposed function in guiding neuronal migration.
Deletion of the entire Fbxo45 SPRY domain was previously found to inhibit Fbxo45-NCad interaction (32). We found that deletion of only five (∆225-229) or ten (∆277-286) residues from the SPRY domain of Fbxo45 totally disrupted the interaction with NCad ( Figure 3D). These residues comprise parts of the β-sheet structure of the SPRY domain, which is critical for displaying loops that mediate protein-protein interactions (39). One caveat is that these deletions may have inhibited Fbxo45 secretion, thereby preventing access to the NCad extracellular domain. However, we found that NCad and Fbxo45 bind efficiently in vitro after cell lysis (unpublished results). Therefore, even if a mutant Fbxo45 is not secreted, it should be able to bind NCad in the cell lysate. Taken together, it is likely that Fbxo45 forms a conventional SPRY fold, and contacts SPRY-binding motifs in NCad.
We tested candidate Fbxo45-binding SPRY motifs in NCad by replacing NCad sequences with corresponding ECad sequences. Replacement of residues in motifs 3, 2 and 1 increasingly inhibited Fbxo45 binding without inhibiting cell-cell adhesion. Motif 1 is potentially most interesting, since its replacement strongly inhibits Fbxo45 binding ( Figure 4). It lies on the opposite surface of EC1 from the EC1-EC1 interface in the strand-swap NCad dimer (42), so Fbxo45 binding is unlikely to alter NCad dimerization. However, this surface has been implicated in cis interactions between adjacent cadherins on the same membrane (42).
Mutations here cause measurable changes in mechanical coupling to the actin cytoskeleton but do not affect cell-cell adhesion (65). The same region of ECad is important for a proposed conformational change involved in inside-out activation (66,67). Mutation of a nearby residue in ECad is found in some cancers and also inhibits inside-out activation (67). Therefore, trans interactions with Fbxo45 may affect cis interactions between NCad molecules on the same cell or modulate NCad conformational activation.
Previous studies showed that Rap1 activation by Reelin in multipolar neurons increases surface expression of NCad and initiates radial migration (53). In addition, Reelin-dependent increases in NCad stimulate terminal translocation at the top of the cortical plate (11)(12)(13). It is challenging to study NCad function in migrating neurons because NCad gene silencing (knockdown) disrupts the neurogenic ventricular zone (6). We therefore evaluated NCad function in migration by inhibiting Rap1 in post-mitotic cells and rescuing with wildtype or mutant NCad.
The rescue of migration by NCad W161A (28) suggests that stable NCad-NCad interaction is not required, and raises the possibility that another extracellular protein, like Fbxo45, may be involved. However, attempts to rescue neuron migration by over-expressing Fbxo45 were unsuccessful because ectopic expression of Fbxo45 alone inhibited migration. This may reflect local saturation of NCad and loss of directional information. Furthermore, gene silencing (knock down) of Fbxo45 decreased the number of migrating neurons, perhaps secondary to altered neurogenesis (32). We therefore generated a NCad mutant that retains homophilic interactions but does not interact with Fbxo45. This mutant, NC123, does not rescue migration. Together, our results suggest that NCad-NCad interaction is neither necessary nor sufficient for migration, and that migration correlates with the ability to bind Fbxo45.
It remains a puzzle that Fbxo45 has intracellular and extracellular functions. Inside the cell, Fbxo45 binds MycBP2 and Skp1 to stimulate ubiquitylation and turnover of Par-4, mTOR, Munc-13 and p73 (31, [68][69][70]. This intracellular function is thought to regulate the epithelialmesenchymal transition and synaptic function (31, 71). However, Fbxo45 is also secreted and binds to NCad from outside the cell. This allows Fbxo45 to have cell non-autonomous functions.
Other unconventionally secreted proteins have different functions inside and outside the cell.
Heat shock protein 70 (Hsp70) is an intracellular chaperone, but stress stimulates Hsp70 secretion and extracellular Hsp70 activates macrophages (48,72). Vasohibins catalyze tubulin detyrosination but when secreted regulate angiogenesis (73,74). As another example, phosphoglucose isomerase is a key metabolic enzyme, but when secreted it is known as autocrine motility factor and stimulates cell migration by binding to specific cell surface receptors (62). By analogy, we can hypothesize that Fbxo45 is secreted under specific biological conditions and that secreted Fbxo45 regulates NCad non-adhesive functions.
In conclusion, Fbxo45 is secreted by a non-classical mechanism. Outside the cell, it binds to SPRY motifs in EC1-2 of NCad. Mutating these motifs inhibits NCad function in neuron migration without affecting cell-cell adhesion. Therefore, secreted Fbxo45, or a protein with overlapping binding requirements, likely regulates NCad during neuron migration. This suggests that Fbxo45 has different functions depending on whether it is intra-and extra-cellular. More studies are warranted to characterize the mechanism and function of the secreted form.

Immunoprecipitation and pull down
Cells were washed with cold PBS two times and harvested in 1% Triton X-100 in PBS with protease/phosphatase inhibitors. The lysates were centrifuged at 14,000 rpm for 15 minutes and supernatant mixed with T7 Tag Antibody Agarose (EMD Millipore) for immunoprecipitation or NeutrAvidin (Thermo Fisher Scientific) for the streptavidin pull down for 2 hours. Agarose beads were washed three times with lysis buffer, eluded with SDS sampling buffer, and analyzed by SDS-PAGE and western blotting.

HEK293-neuron trans-biotinylation (BioID)
HEK293 cells were transfected with pCAG plasmids encoding GFP (experiment 1), E5-BirA* Beads were centrifuged at 4,000 rpm for 2 minutes and the supernatant was transferred to a new tube. Beads were rinsed twice more with water and all supernatants were combined and lyophilized. The peptides were cleaned using ZipTip micro-C18 (Millipore Corporation) and analyzed by LC-MS/MS.

Cadherin-Fc fusion proteins
DNA sequences encoding the ectodomains of ECad and NCad W161A were PCR amplified and cut with MfeI and AgeI (ECad) or EcoRI and AgeI (NCad) and cloned into a plasmid (pcDNA3.1-ApoER2-Fc (15)) that had been cut EcoRI and AgeI. The murine ECad template was a kind gift of Masatoshi Takeichi (75) and murine NCad W161A template was from (28). The plasmids were confirmed by sequencing. Fusion proteins were produced in HEK293T cells. For each construct, five 6-cm plates were transfected with a total of 120 µg of DNA using the calcium phosphate method. Media were removed 6 hours later and gently replaced with serum-free DMEM. Two days later, media were harvested and concentrated using Amicon Centricon YM-100 centrifugal filters. Protein concentration and purity were estimated by SDS PAGE with a BSA standard.
Approximately 250 µg of each fusion protein was mixed with 250 µL packed volume Protein A/G PLUS-Agarose (Santa Cruz Biotechnology) and unbound proteins washed away using PBS.

LC-MS/MS analysis was performed with an Eksigent NanoLC-2D system (Eksigent/AB Sciex)
coupled to an LTQ Orbitrap mass spectrometer (Thermo Scientific). The LC system configured in a vented format (77)   Trypsin was set as the enzyme with maximum missed cleavages set to 2. The precursor ion tolerance was set to 10 ppm, and the fragment ion tolerance was set to 0.6 Da. SEQUEST (78) was used for search, and search results were run through Percolator (79) for scoring.

Preparation of secreted proteins and total cell lysates
To collect secreted proteins, cells were incubated in serum-free DMEM (for HeLa) or Neurobasal Media (for primary cortical neurons) for 20-24 hours at 37 °C with 5 % CO2. The conditioned medium was centrifuged at 3,000 rpm for 5 minutes to remove cell debris.
Supernatants were collected and concentrated using Amicon Ultra, 10 kDa NMWL (EMD Millipore). After removing the conditioned medium, cells were washed with cold PBS and harvested in 1% Triton X-100 in PBS buffer with protease/phosphatase inhibitors. Tritoninsoluble material was removed at 14,000 rpm for 15 minutes. Samples of the conditioned medium and cell lysate were analyzed by SDS polyacrylamide gel electrophoresis and Western blotting.

Short-term aggregation assay
CHO-K1 cells were co-transfected with GFP or mCherry and indicated NCad constructs. After 24 hours, transfected cells were washed twice with pre-warmed HCMF (137 mM NaCl, 5.4 mM KCl, 0.63 mM Na2HPO4, 5.5 mM Glucose, 10 mM HEPES, pH 7.4) and placed in suspension using 0.01 % trypsin, 1 mM CaCl2 in HCMF. Suspended cells were centrifuged at 1,000 rpm for 5 minutes, resuspended in 0.5 % soybean trypsin inhibitor in HCMF, then washed three times with cold HCMF and counted. 200,000 cells in HCMF were added to 1 % BSA-coated 24-well plates with and without 2 mM CaCl2. Cells were shaken at 80 rpm for 20 minutes at 37 °C and images were collected using 2X or 4X objectives. Cell aggregation assays were performed three biological replicates. Data were quantified using Analyze Particles in Image J (80).

In utero electroporation
In utero microinjection and electroporation was performed at E14.5 essentially as described (81) using timed pregnant CD-1 mice (Charles River Laboratories). In brief, mice were anesthetized and the midline incision the uterine horns were exposed. Plasmid solution was injected into the lateral ventricle using needles for injection that were pulled from Wiretrol II glass capillaries (Drummond Scientific) and calibrated for 1-µl injections. DNA solutions were mixed in 10 mM Tris, pH 8.0 with 0.01% Fast Green. The embryo in the uterus was placed between the forcepstype electrodes (Nepagene) with 5-mm pads and electroporated with five 50-ms pulses of 45 V using the ECM830 electroporation system (Harvard Apparatus). The uterine horns were then placed back into the abdominal cavity to continue normal brain development. Three days later, mice were euthanized and embryo brains were sectioned and GFP visualized.
Immunohistochemistry to detect HA-tagged NCad was performed essentially as described (53).

Dendrite Morphology
Images of GFP-expressing mouse hippocampal neurons were acquired as a z-stack (3 sections, 0.5 um per section) on the Deltavision deconvolution microscope (GE Life Sciences,).
A maximum intensity projection was created from the z-stack. Images were converted to binary and analyzed using the Sholl Analysis Plugin (82) with the concentric ring size of 10 µm.                   Rat proteins detected by LC-MS/MS from two independent experiments. The numbers represent the score and the data was sorted by the score of N2-BirA* in experiment 2 as a descending manner. The score is the sum of all peptide Xcorr values above the specified score threshold. The score threshold is calculated as follows: 0.8 + peptide charge X peptide relevance factor. 0.4 was used for the peptide relevance factor. For each spectrum, only the highest-scoring match is used. Blank cells indicate that no peptide was detected in the sample.

Supplementary Table 2. Human proteins detected by BioID with NCad-BirA*.
Human proteins detected by mass spectrometry from the same two experiments as Table 1. Proteins from gel regions above and below of the cadherin-Fc bands in Figure 1F were digested with trypsin and analyzed by LC-MS/MS. The numbers represent the score calculated as described in Table 1