Identification of an N-cadherin motif that can interact with the fibroblast growth factor receptor and is required for axonal growth.

In this study, we show that the neurite outgrowth response stimulated by N-cadherin is inhibited by a recently developed and highly specific fibroblast growth factor receptor (FGFR) antagonist. To test whether the N-cadherin response also requires FGF function, we developed peptide mimetics of the receptor binding sites on FGFs. Most mimetics inhibit the neurite outgrowth response stimulated by FGF in the absence of any effect on the N-cadherin response. The exceptions to this result were two mimetics of a short FGF1 sequence, which has been shown to interact with the region of the FGFR containing the histidine-alanine-valine motif. These peptides inhibited FGF and N-cadherin responses with similar efficacy. The histidine-alanine-valine region of the FGFR has previously been implicated in the N-cadherin response, and a candidate interaction site has been identified in extracellular domain 4 of N-cadherin. We now show that antibodies directed to this site on N-cadherin inhibit the neurite outgrowth response stimulated by N-cadherin, and peptide mimetics of the site inhibit N-cadherin and FGF responses. Thus, we can conclude that N-cadherin contains a novel motility motif in extracellular domain 4, and that peptide mimetics of this motif can interact with the FGFR.

In this study, we show that the neurite outgrowth response stimulated by N-cadherin is inhibited by a recently developed and highly specific fibroblast growth factor receptor (FGFR) antagonist. To test whether the N-cadherin response also requires FGF function, we developed peptide mimetics of the receptor binding sites on FGFs. Most mimetics inhibit the neurite outgrowth response stimulated by FGF in the absence of any effect on the N-cadherin response. The exceptions to this result were two mimetics of a short FGF1 sequence, which has been shown to interact with the region of the FGFR containing the histidine-alanine-valine motif. These peptides inhibited FGF and N-cadherin responses with similar efficacy. The histidine-alanine-valine region of the FGFR has previously been implicated in the N-cadherin response, and a candidate interaction site has been identified in extracellular domain 4 of N-cadherin. We now show that antibodies directed to this site on N-cadherin inhibit the neurite outgrowth response stimulated by N-cadherin, and peptide mimetics of the site inhibit N-cadherin and FGF responses. Thus, we can conclude that N-cadherin contains a novel motility motif in extracellular domain 4, and that peptide mimetics of this motif can interact with the FGFR.
N-cadherin is a member of the classical cadherin family of transmembrane glycoproteins that mediate cell-to-cell adhesion via a homophilic binding mechanism (1). Similar to other members of this family, the extracellular portion of the molecule is composed of five cadherin domains with a large body of evidence, suggesting that the homophilic binding site resides in the first extracellular domain (ECD1) 1 (2,3). In the nervous system, N-cadherin function has been implicated in a number of key events that range from the control of axonal growth and guidance (4,5) to synapse formation and synaptic plasticity (6 -9).
In addition to homophilic binding, cadherins have been shown to interact with a number of adaptor or signaling mol-ecules. For example, the interaction of the cytoplasmic domain of the classical cadherins with the actin-based cytoskeleton, which is important for adhesion, is mediated by the catenins (10,11). An interaction with a cell surface N-acetylgalactosaminylphosphotransferase has also been reported and implicated in N-cadherin function (12). More recently, the non-receptor protein tyrosine phosphatase PTP1␤ and the receptor protein tyrosine phosphatase PTP have been shown to bind directly to the cytoplasmic domain of N-cadherin and other cadherins and modulate function by regulating tyrosine phosphorylation of the cadherin and/or catenins (13,14). The notion, that the activation of signaling cascades in cells rather than adhesion per se might drive some N-cadherin responses, is supported by the a recent observation that a soluble form of N-cadherin can promote axonal growth (15).
A number of groups have implicated the fibroblast growth factor receptor (FGFR) in N-cadherin function. For example, neurite outgrowth stimulated by N-cadherin has been shown to be inhibited by a wide variety of agents that inhibit FGFR function in neurons (16), including the expression of a dominant negative FGFR (15,17,18). In addition, N-cadherin can promote "contact-dependent" survival of ovarian granulosa cells in an FGFR-dependent manner (19). More recently, Ncadherin has been reported to promote the motility of cancer cells with some data suggesting that the FGFR might be involved in this response (20,21). However, it remains unclear as to whether N-cadherin homophilic binding modulates FGF interactions with the FGFR (20), or whether the FGFR requirement reflects a more direct interaction among these molecules (22). In support of the latter possibility, it has recently been shown that the antibody clustering of N-cadherin in cells is associated with the co-clustering of the FGFR (15), and that the FGFR and N-cadherin will co-precipitate from cells (23,24).
We have used the structures of a number of FGF/FGFR crystals to develop peptide mimetics of the receptor binding sites on FGFs. The majority of these peptides inhibit the neurite outgrowth response stimulated by FGF in the absence of any effect on the same response stimulated by N-cadherin. The exceptions to this finding are two mimetics of a short sequence on FGF1 that interacts with the histidine-alanine-valine (HAV) region of the FGFRs; these peptides inhibited the FGF and N-cadherin responses with similar efficacy. Previous studies have implicated the HAV region of the FGFR in the Ncadherin response (16) and identified a candidate interaction motif in ECD4 of N-cadherin (22). In this study, we show that antibodies that bind to this motif in neurons specifically inhibit N-cadherin function. Furthermore, peptide mimetics of the Ncadherin motif inhibit N-cadherin and FGF function with sim-ilar efficacy. Thus, we have defined a novel functional motif in the ECD4 of N-cadherin and provided a clear framework for understanding, at least in part, the molecular basis for an N-cadherin/FGFR interaction. Cell Culture-BT-549, MCF-7, MDA-MB-436, and MDA-MB-231  breast carcinoma cell lines, COS-7 cells, the L6 muscle cell line, and various 3T3 cell lines were maintained in Dulbecco's modified Eagle's medium (DME) supplemented with 10% fetal calf serum (FCS). L6 cells expressing the full-length human FGFR1(25) (denoted as L6* cells in this study) were maintained in DME/10% FCS containing 0.72 mg/ml G418.

EXPERIMENTAL PROCEDURES
Neurite Outgrowth Assays-Co-cultures of cerebellar neurons on monolayers of parental 3T3 cells or established transfected 3T3 cell lines that express physiological levels of chick N-cadherin (the LK8 cell line), human L1, or human NCAM were established as described previously (16). ϳ80,000 3T3 cells in DME/10% FCS were plated into individual chambers of an 8-chamber tissue culture slide coated with poly-L-lysine and fibronectin. After 24 h to allow for monolayer formation, the medium was removed, and 3000 dissociated cerebellar neurons (taken from postnatal day 2 rats) were plated into each well in SATO medium supplemented with 2% FCS. Test reagents including FGF2 at 1 ng/ml were added as indicated in the text, and the co-cultures were maintained for 18 h. The co-cultures were then fixed and stained for GAP-43 immunoreactivity. The mean length of the longest neurite/cell was measured for ϳ120 -150 neurons in each population, again as described previously (for review see Ref. 16).
Generation of an Antibody to a Motif in ECD4 of N-cadherin-A rabbit antiserum was raised against a synthetic peptide that corresponds to a region of N-cadherin ECD4 (RYTKLSDPANWLKIDPVNG-QIT). The antiserum was affinity purified against the peptide immunogen and stored as a stock at 40 g/ml in PBS/glycerol at Ϫ20°C (50:50). A monovalent F(abЈ) fraction of the whole antiserum was also prepared by standard methods and purified by high pressure liquid chromatography and stored as a stock at 33 mg/ml at 4°C. The titer of the affinity purified antibody against native N-cadherin was characterized by enzyme-linked immunosorbent assay. 96-well microtiter plates were coated with 2 g/ml N-cadherin-Fc chimera (15) or a recombinant version of the ECD1 of N-cadherin (26) in DME for 1 h at 37°C. The plates were then washed twice with DME followed by incubation with 1% bovine serum albumin/DME at 37°C for 30 min to block nonspecific protein binding sites. The plates were incubated with a range of concentrations of the affinity purified anti-N-cadherin antibody in DME/ 10% FCS overnight at 4°C. The plates were washed 3 times with DME and bound antibody detected after incubation with an anti rabbithorseradish peroxidase conjugate (1:1000 in DME/10% FCS for 1 h at room temperature). The amount of bound horseradish peroxidase was determined using SIGMA FAST TM OPD peroxidase substrate according to manufacturer instructions.
Western Blotting-For Western blot analysis, confluent cell monolayers were washed with PBS and lysed in immunoprecipitation buffer (50 mM Tris, pH 8.0, 150 mM NaCl, 10 mM MgCl 2 , 2 mM CaCl 2 , 5% glycerol, 1% Triton X-100, 1 mM Na 3 VO 4 , 10 mM NaF, 1 mM phenylmethylsulfonyl fluoride, and "complete" protease inhibitors from Roche Molecular Biochemicals) for 30 min with agitation. Insoluble material was removed by centrifugation at 14,000 ϫ g for 5 min and the supernatants were used for Western blot analysis or immunoprecipitation. Protein concentration was determined with the BCA protein assay kit (Pierce). Cell lysates were resolved by SDS-polyacrylamide gel electrophoresis, and separated proteins were transferred to nitrocellulose. Membranes were blocked for 1 h with 5% milk/PBS and then incubated for 1 h with the affinity purified fraction of the rabbit polyclonal antibody raised against the N-cadherin ECD4 peptide (see above) at a dilution of 1:1000 (dilutions of stock in 2% milk/0.05% Tween 20/PBS). Membranes were then washed with 0.05% Tween 20/PBS and then incubated for 1 h with a 1:3000 dilution of an anti-rabbit horseradish peroxidase conjugate in 2% milk/0.05% Tween 20/PBS. The membranes were washed 3 times with 0.05% Tween 20/PBS before probing with ECL or ECL PLUS (Amersham Pharmacia Biotech).
Protein Data Bank Files Accessed in This Study-In this study, we have considered five FGF/FGFR crystal complexes, and the respective Protein Data Bank files are 1CVS (27), 1FQ9 (28), 1EV2 (29), 1EVT (29), and 1DJS (30). We have used computer algorithms to interrogate these files to generate amino acid contact matrices. The algorithms determine whether each amino acid is present within a fixed distance (7 Å) of any amino acid in the second member of the pair.
Peptide Synthesis and Purity and PD 173074 -Synthetic peptides were obtained from commercial suppliers (AbCam Ltd, Mimotopes, or Multiple Peptide Systems). All peptides were purified by reverse-phase high pressure liquid chromatography and obtained at the highest level of purity (generally Ͼ95%). Where peptide sequences are underlined, this denotes a peptide that has been cyclized via a disulphide bond among the given cysteine residues. PD 173074 (31,32) was prepared according to the general procedures described by Hamby et al. (33).

RESULTS
The N-cadherin Response Is Dependent on the Catalytic Activity of the FGFR-A large body of evidence supports the view that neurite outgrowth stimulated by N-cadherin requires the presence of a functional FGFR in neurons (34). However, the validity of conclusions based on the use of kinase-deleted "dominant negative" forms of the FGFR have recently been challenged (35). To test whether the catalytic activity of the FGFR is required for the N-cadherin response, we took advantage of the recent development of a highly specific antagonist (PD 173074) that binds to the ATP binding pocket of the FGFR (31). We cultured rat cerebellar granule cells for ϳ18 h over monolayers of control 3T3 cells and monolayers of transfected 3T3 cells that express physiological levels of N-cadherin (the LK8 cell line). As expected, the neurons extended longer neurites on the N-cadherin-expressing monolayers ( Fig. 1). This response was fully inhibited in a dose-dependent manner by PD 173074 (Fig. 1A). When cerebellar granule cells are grown over monolayers of 3T3 fibroblasts that express physiological levels of two additional cell adhesion molecules (NCAM and L1) or in the presence of FGF2, a neurite outgrowth response that is similar to the N-cadherin response is found (16,36). PD 173074 also inhibited the neurite outgrowth response stimulated by NCAM, L1, and FGF2 with a dose-response curve that is similar to the inhibition of the N-cadherin response (Fig. 1B). Arachidonic acid is a key second messenger in the signal transduction cascade that couples the activated FGFR to the neurite outgrowth response (37). When tested at up to 1 M, PD 173074 does not inhibit the neurite outgrowth response stimulated by arachidonic acid (32), and this demonstrates that it has no nonspecific effects on neurite outgrowth at any downstream step in the pathway that couples the activated FGFR to the response.
The N-cadherin Response Does Not Require FGF Function-The fact that the FGFR is required for the N-cadherin response begs the question as to whether FGFs are also required for this response. Recent observations from a number of FGF/FGFR crystal complexes has led to a consensus on how FGFs interact with the receptor (27)(28)(29)(30). The contact profile for the FGF1/ FGFR2 interaction is shown in Fig. 2 with an essentially identical profile found for the other FGF/FGFR interactions (29). Four major linear clusters of amino acids on FGF1 account for most of the binding to the FGFR. It follows that peptide mimetics of these clusters might be capable of inhibiting FGF function by binding back to the receptors. In this study, we made linear peptide mimetics to three partially overlapping sites from cluster 1 and single peptide mimetics for clusters 2, 3, and 4. In addition, we made cyclic versions of two of the cluster 1 peptides. All of the peptides were initially tested at 100 g/ml for their ability to inhibit the neurite outgrowth response stimulated by FGF2 with the results summarized in Table I. Although all of the peptides showed significant inhibitory activity, peptide mimetics of cluster 1 and cluster 2 were the most active in the assay, and therefore, full dose response curves to these peptides were obtained (Fig. 3). The results show that these peptides can inhibit the FGF2 response with IC 50 values ranging from ϳ20 to 65 M. The data also clearly show that the peptides do not inhibit basal neurite outgrowth over 3T3 cell monolayers, and we can, therefore, conclude that they do not have any nonspecific effects on neurons. Arachidonic acid is a key second messenger in the pathway that couples the activation of the FGFR to the neurite outgrowth response (36), and at the highest concentration tested (100 g/ml), none of the above peptides inhibited the neurite outgrowth response stimulated by 10 M arachidonic acid (data not shown). Thus, we can conclude that the above peptides inhibit FGF2 responses at the level of ligand interaction with the FGFR.
In contrast to their effects on the neurite outgrowth response stimulated by FGF2, 6 out of the 8 peptides had no effect on the N-cadherin response (Table I). Of the eight peptide mimetics of the FGFR binding sites, two inhibited the N-cadherin response just as well as the FGF2 response (Table I and Fig. 4A). We analyzed the FGF/FGFR crystals to identify potential binding sites for the peptides. In this context, both peptides (N-Ac-YCSNGGHF-NH 2 and N-Ac-YCSNGGHFC-NH 2 ) are mimetics of a short FGF1 sequence that makes extensive and relatively exclusive contacts with the HAV region of the FGFRs (see Fig.   4B for details). In contrast, the other peptides are mimetics of sequences that make extensive contacts with the linker region between the Ig domain 2 and 3 or with Ig domain 3 (data not shown). Thus, we can conclude that although FGF-derived peptides that target the HAV region of the FGFR can inhibit both FGF and N-cadherin responses, a number of other peptides that target other regions of the FGFR exclusively inhibit the FGF response.
Identification of a Novel Functional Motif in ECD4 of Ncadherin-Based on homology with peptide sequences that interact with the HAV region of N-cadherin, we have identified the IDPVNGQ motif from ECD4 of N-cadherin as a candidate binding motif for the HAV region of the FGFR (22). If the IDPVNGQ motif within ECD4 of N-cadherin is a functional motif capable of interacting with the HAV region of the FGFR, then peptide mimetics of the sequence would also be expected to inhibit N-cadherin and FGF responses by binding back to the HAV region of the receptor. To test this, we made a short synthetic peptide (N-Ac-IDPVNGQ-NH 2 ) and tested it in the neurite outgrowth assays. The results in Fig. 5A demonstrate that this peptide can inhibit the N-cadherin and FGF responses with similar efficacy. Again, the peptide had no effect on basal neurite outgrowth over 3T3 monolayers. We also tested a longer version of the peptide primarily in the N-cadherin assay (N-Ac-WLKIDPVNGQI-NH 2 ). This peptide was moderately better than the shorter peptide at inhibiting the N-cadherin response (IC 50 ϭ ϳ60 M as compared with ϳ100 M, see Fig.  5B). When we substituted a single isoleucine to alanine (N-Ac-WLKADPVNGQI-NH 2 ) or proline to alanine (N-Ac-WLKIDA-VNGQI-NH 2 ), we saw an approximate 3-fold reduction in the efficacy of the longer peptide (IC 50 ϭ ϳ180 M). Finally, when we tested the double mutant (N-Ac-WLKADAVNGQI-NH 2 ), we found essentially a complete loss of inhibitory activity (Fig. 5B). The longer peptide also inhibited the FGF2 response, and this activity was lost with the double mutation (data not shown).
An Antibody Targeted to the ECD4 Motif Inhibits the Ncadherin Response-To further test whether the ECD4 of Ncadherin contains a functional motif required for the neurite outgrowth response, a rabbit antiserum was raised against a synthetic peptide that contains the inhibitory peptide motif (RYTKLSDPANWLKIDPVNGQIT), and an affinity purified fraction was prepared. We used the affinity purified antibody to immunoblot lysates from a variety of cell lines (Fig. 6A). The antibody showed weak reactivity against 3T3 cell lysates, however, a clear band at the appropriate molecular weight was observed in transfected 3T3 cells that express chick N-cadherin (the LK8 cell line). An appropriate band was also detected in the BT-549 and MDA-MB436 tumor cell lines that express human N-cadherin and in COS-7 cells. In contrast, no immunoreactivity was found in cell lines that are known not to express N-cadherin (MCF-7 and MDA-MB231 cells) or in parental or FGFR-expressing L6 cells. Some of the N-cadherinnegative cell lines express E-cadherin and Cadherin-11, and the BT-549 cells express P-cadherin, which runs at a slightly lower molecular weight than does N-cadherin (data not shown, but for details, see Ref. 21). Thus, we can conclude that the antibody reacts specifically with N-cadherin and does not crossreact with E-cadherin, Cadherin-11, P-cadherin, or the FGFR. The affinity purified fraction of the antiserum also showed good reactivity against native N-cadherin as determined by enzymelinked immunosorbent assay (Fig. 6B) in the absence of any detectable binding to recombinant ECD1 of N-cadherin (data not shown).
The affinity purified antibody was tested for its effects on the neurite outgrowth response stimulated by N-cadherin. When used at 67 ng/ml, the antibody completely inhibited the N- cadherin response in the absence of any significant effect on the response stimulated by FGF2 (Fig. 7). As an additional control, we also cultured neurons over monolayers of 3T3 fibroblasts that express physiological levels of the NCAM or L1 adhesion molecules, because both of them can also stimulate neurite outgrowth via an FGFR-dependent mechanism (16). The antibody to N-cadherin had no effect on these responses (Fig. 7). To rule out the possibility that the antibody was inhibiting function by clustering N-cadherin in the neurons and perhaps inducing internalization, we prepared a monovalent F(abЈ) fraction from the whole antiserum. This reagent substantially inhibited the N-cadherin response in the absence of any effect on basal neurite outgrowth over 3T3 monolayers or neurite outgrowth stimulated by NCAM or L1 (Fig. 7). To determine whether the antibody was inhibiting function by binding to N-cadherin in the neuron and/or the substrate, we pretreated the neurons and/or monolayers with antibody prior to co-culture. The pretreatment of neurons with the monovalent F(abЈ) inhibited the N-cadherin response, whereas the pretreatment of the substratum had no effect (Fig. 8). DISCUSSION

A large body of evidence suggests that some N-cadherin responses are dependent on FGFR function in cells (Introduction).
A key result that supports this conclusion is the observation that when expressed in neurons, a kinase-deleted dominant negative form of the FGFR inhibits the neurite outgrowth response stimulated by N-cadherin. The validity of the conclusions from the dominant negative FGFR experiments has recently been challenged, based on the fact that the remaining cytoplasmic portion of the kinase-deleted receptor still contained a binding site for the adapter molecule FRS2 (35). The recent availability of a highly selective inhibitor of the tyrosine kinase activity of the FGFR (PD 173074) has allowed us to revisit this important question. This compound was found to inhibit the neurite outgrowth response stimulated by N-cadherin and FGF2 with similar efficacy. This compound binds to the ATP-binding pocket of the FGFR (31), and although this pocket is relatively well conserved in closely related tyrosine kinase receptors at the concentration used, this compound has no effect on the insulin-like growth factor, platelet-derived growth factor, nerve growth factor, BDNF, CNTF, GDNF, and epidermal growth factor receptors (31,32). Given the concurrence of the results obtained with the dominant negative FGFR (17) and the PD 173074 compound described in this study, we can conclude that the neurite outgrowth response stimulated by N-cadherin does indeed depend on FGFR activity in neurons. Likewise, neurite outgrowth stimulated by NCAM and L1 are also further established to be dependent on FGFR activity in neurons.
N-cadherin might act as a surrogate ligand for the FGFR and/or sensitize the FGFR to low levels of an FGF present in the cultures. Indeed, using the induction of the matrix metalloproteinase MMP-9 as a read-out, it has been shown that the expression of N-cadherin in the MCF-7 breast cancer cell line leads to a dramatic increase in the sensitivity to FGF2, however, the FGF2 was exogenous rather than endogenous (20). In the context of the neurite outgrowth response stimulated by N-cadherin, all of the evidence suggests that the cultures do not FIG. 2. Contact profile for FGF binding to the FGFR. The plot shows the average number of FGF receptor residues (y axis) within 7 Å of each FGF residue (x axis). The profile is averaged over a 3-amino acid window. The profile shows that there are four major linear clusters of amino acids that make extensive contacts with the FGFR as indicated. The data are extracted from the PDB file for the FGF1/FGFR2 crystal (see under "Experimental Procedures"), however, an essentially identical profile is obtained with FGF2/FGFR1 crystals. The lines beneath the profile indicate receptor binding sequences that have been made as synthetic peptides. All of the peptides, with the exception of 1b, were mimetics of the FGF1 sequence. The peptide mimetic of 1b was from the FGF2 sequence (see Table I for details).

TABLE I The effects of peptide mimetics of FGFR binding sites on the neurite outgrowth response stimulated by FGF and N-cadherin
Neurons were cultured on parental 3T3 monolayers in control medium or medium containing FGF2 (1 ng/ml) or on monolayers of 3T3 cells that express physiological levels of transfected N-cadherin (the LK8 cell line). The medium was then further supplemented with 100 g/ml peptide mimetic of the given binding motif. In all cases, linear peptide mimetics of the binding motif were synthesized, acetylated, and amide-blocked (e.g. N-Ac-HFKDPKRLY-NH 2 ). In two instances, we also tested cyclic mimetics of the binding motifs (N-Ac-CGNYKKPKC-NH 2 and N-Ac-YCSNGGHFC-NH 2 ). After ϳ18 hours, the cultures were fixed, and the mean length of the longest neurite was determined by sampling ϳ120 neurons in replicate wells. The results show the percentage inhibition of the response stimulated by FGF2 (1 ng/ml) and N-cadherin mean Ϯ S.E. from at least three experiments. The peptides had no effect on basal neurite outgrowth over 3T3 monolayers, and typical control responses to N-cadherin and FGF are shown in Figs. 1 and 3. In the case of the 21 GNYKKPK 27 and 30 YCSNGGHF 37 motifs, the results obtained with the linear and cyclic peptides were not distinguishable, and these have been pooled for inhibition of the FGF2 response. contain functional levels of an endogenous FGF (e.g. the FGFR antagonists have no effect on basal neurite outgrowth). Nonetheless, the possibility of sensitization to a "subthreshold" level of an FGF is difficult to discount. To try to resolve this issue, we used a structural bioinformatics approach to design a range of novel FGFR antagonists. We reasoned that peptide mimetics of receptor binding motifs might be able to bind back to the receptor and antagonize FGF function. Given that all of the FGFs are thought to interact with common regions that are conserved among all four FGFRs (29), such agents would be expected to act as general FGF antagonists. A total of eight peptide mimetics based on four linear amino acid clusters were shown to be able to inhibit the neurite outgrowth response stimulated by FGF with varying degrees of efficacy, and control experiments pointed to the fact that these peptides were indeed acting at the level of the ligand interacting with the receptor.
Most of the peptides that inhibited the FGF response had no effect on the N-cadherin response, and this fact clearly suggests that the latter is not dependent on FGF function. The peptides that failed to inhibit the N-cadherin response are all mimetics of FGF sequences that are known to interact with the linker region between Ig domains 2 and 3 or with Ig domain 3 of the receptor. This finding suggests that if N-cadherin interacts with the FGFR, then it does so in a manner that does not fully FIG. 3. The effect of FGFR binding motif peptide mimetics on the neurite outgrowth response stimulated by FGF2. Neurons were cultured on parental 3T3 monolayers in control medium or medium containing FGF2 (1 ng/ ml), and the medium was then further supplemented with up to 100 g/ml a peptide mimetic of the given interacting motif. A, we tested a linear mimetic of the functional motif (N-Ac-HFKDP-KRLY-NH 2 ). B, we tested both a linear (N-Ac-GNYKKPK-NH 2 ) and a cyclic (N-Ac-CGNYKKPKC-NH 2 ) mimetic of the functional motif. C, we tested both a linear (N-Ac-YCSNGGHF-NH 2 ) and cyclic (N-Ac-YCSNGGHFC-NH 2 ) mimetic of the functional motif. D, we tested a linear mimetic of the functional motif (N-Ac-LSAESVGEVY-NH 2 ). After ϳ18 h, the cultures were fixed, and the mean length of the longest neurite was determined by sampling ϳ120 neurons in replicate wells. The results show absolute neurite length, and each value is the group mean pooled form 2-5 independent experiments. Bars show Ϯ S.E. in those instances where more than two experiments were performed. In B and C, we have pooled the results with the cyclic and linear peptides, because they did not differ in activity.

FIG. 4. A peptide that "targets" the HAV region of the FGFR inhibit the neurite outgrowth response stimulated by N-cadherin.
A, neurons were cultured on monolayers of control 3T3 cells or monolayers of N-cadherin expressing 3T3 cells in medium supplemented with the N-Ac-YCSNGGHF-NH 2 peptide as indicated. This peptide had no effect on basal neurite outgrowth but inhibited the N-cadherin response in a dose-dependent manner. Results show the mean neurite length (Ϯ S.E.) from 4 -6 independent experiments. B, we analyzed FGF1/FGFR2 and FGF1/FGFR1 crystals (see under "Experimental Procedures" for details) to identify the receptor binding sites for the YCSNGGHF motif in FGF1. These sites are given as the inserts. overlap with the binding sites for FGFs. Interestingly, two peptide mimetics of a small motif in FGF1 ( 30 YCSNGGHF 37 ) were as effective at inhibiting the N-cadherin response as the FGF2 response. An analysis of crystal structures shows that this sequence within FGF1 and the homologous sequence in FGF2 make extensive and relatively exclusive contacts with the HAV region of the FGFRs. Interestingly, other data support a function for this region of the FGFR in N-cadherin responses. For example, antibodies directed to this site on the receptor and peptide mimetics of this region of the receptor both inhibit the neurite outgrowth response stimulated by N-cadherin (16). Thus, if the inhibitory peptides are acting as true mimetics of the 30 YCSNGGHF 37 motif, they most probably inhibit both FGF and N-cadherin function by binding to the HAV region of the FGFR.
The IDPVNGQ motif from ECD4 of N-cadherin has previ-ously been identified as a candidate binding motif for the HAV region of the FGFR based on sequence homology with the motifs within N-cadherin (INPISGQ) and R-cadherin (ID-PVSGR) that interact with the HAV region of N-cadherin (22,38). To determine whether the ECD4 motif has the potential to interact with the FGFR we initially adopted a peptide competition approach, as the clear prediction is that peptide mimetics based on this motif should be able to inhibit both FGF and N-cadherin responses. We found that a long peptide mimetic (N-Ac-WLKIDPVNGQI-NH 2 ) and short peptide mimetic (N-Ac-IDPVNGQ-NH 2 ) could both inhibit the neurite outgrowth responses stimulated by N-cadherin and FGF. The fact that the cadherin-derived peptide can antagonize the FGF response provides evidence, albeit indirect, that this motif can interact with the FGFR.
To further test the function of the candidate FGFR binding site in ECD4 of N-cadherin, an antibody was raised against a peptide (RYTKLSDPANWLKIDPVNGQIT) that contained the putative binding motif. The antibody was fully characterized and shown to be specific to N-cadherin. An affinity purified fraction of the antiserum was shown to fully inhibit the Ncadherin response in the absence of any effect on an FGF2 response or an NCAM and L1 response. Furthermore, a monovalent F(abЈ) fraction of the antiserum also specifically inhibited the N-cadherin response. The possibility of the antibody inhibiting cadherin function by inhibiting homophilic binding is improbable for a number of reasons. Firstly, recent experimental evidence on N-cadherin-stimulated neurite outgrowth (38,39) concurs with the structural evidence that points to a direct interaction between the ECD1 domains as mediating the FIG. 5. Peptide mimetics of an N-cadherin ECD4 motif inhibit the neurite outgrowth response stimulated by FGF and N-cadherin. Cerebellar neurons were cultured on 3T3 monolayers in control medium or medium containing 1 ng/ml FGF2 or on monolayers of 3T3 cells expressing N-cadherin. We evaluated the ability of five peptides to inhibit the FGF and/or N-cadherin responses. A, the effect of the N-Ac-IDPVNGQ-NH 2 peptide on the N-cadherin and FGF response is shown. B, the effect of the N-Ac-WLKIDPVNGQI-NH 2 peptide and three "mutated" versions of this peptide on the N-cadherin response is shown. The sequence of the I/A peptide was N-Ac-WLKADPVNGQI-NH 2 . The sequence of the P/A peptide was N-Ac-WLKIDAVNGQI-NH 2 , and the sequence of the I/A and P/A peptide was N-Ac-WLKADAVNGQI-NH 2 . The results show the percent inhibition of the neurite outgrowth response stimulated by N-cadherin and FGF2 (typical control responses are shown in Figs. 1 and 3), and each value is the mean Ϯ S.E. determined from three independent experiments. None of the peptides had any significant effect on the basal neurite outgrowth over parental 3T3 monolayers (data not shown).

FIG. 6. Specificity of the N-cadherin ECD4 peptide antiserum.
A, lysates from a variety of cell types as indicated were resolved by SDS-polyacrylamide gel electrophoresis and immunoblotted with the affinity purified fraction of the antiserum (1/1000 dilution of stock) raised against the N-cadherin ECD4 peptide. B, the affinity purified antiserum tested for reactivity to native N-cadherin by enzyme-linked immunosorbent assay titration against an N-cadherin-Fc chimera.
homophilic binding interaction (26,40). Secondly, our antibody only inhibits function when bound to N-cadherin in the neurons (as compared with the substrate), and this fact is difficult to reconcile with the antibody inhibiting homophilic binding. Finally, it has recently been reported that when a 69-amino acid sequence from ECD4 of N-cadherin, which contains the above motif, is swapped with the corresponding segment of E-cadherin, the new chimeric version of E-cadherin acquires the ability of N-cadherin to promote the migration of a cancer cell line (41). An antibody that interacts specifically with the Ncadherin sequence and has no effect on N-cadherin-mediated adhesion was able to fully inhibit the motility response. Thus, it appears probable that the ECD4 contains a motility motif that can function independently from the homophilic binding site. In terms of the axonal growth response, our data point to this motif mediating a functional interaction with the FGFR.
What mechanism might account for an N-cadherin reliance on FGFR function but not FGF function? The answer to this question might be related to the fact that FGFRs can be activated in a ligand-independent manner simply by overexpressing them in cells (42). In this context, when N-cadherin is clustered with antibodies, the FGFR is found concentrated in the clusters (15). Furthermore, N-cadherin co-immunoprecipitates with the FGFR in ovarian cells (23). Interestingly, in pancreatic tumor cells the co-precipitation of N-cadherin with the FGFR has been reported to be dependent on NCAM expression in the cells (24), however, we have found co-precipitation among these molecules in cells that do not express NCAM. 2 Nonetheless, we have as yet been unable to inhibit the coimmunoprecipitation of N-cadherin and the FGFR with the ECD4 antibody, one possible explanation being that additional sites and/or molecules contribute to the interaction. 3 Irrespective of whether the N-cadherin interacts directly with the FGFR or indirectly as part of a larger complex, cadherin homophilic binding and the consequent lateral dimerization (40) provide a basis for the relocalization of the FGFR in the cell membrane and its consequent activation in a manner that need not rely on the function of the more conventional receptor ligands. FIG. 7. The effects of an N-cadherin ECD4 antiserum on the neurite outgrowth responses stimulated by N-cadherin, FGF2, NCAM, or L1. Cerebellar neurons were cultured for 18 h on monolayers of 3T3 cells in control medium or medium containing 1 ng/ml FGF2 or on monolayers of 3T3 cells expressing N-cadherin, NCAM, or L1 (as indicated by the filled bars). Sister cultures were supplemented with 67 ng/ml affinity purified polyclonal antibody to ECD4 of N-cadherin (hatched bars) or with 80 g/ml monovalent F(abЈ) fraction prepared from the whole serum (open bars). In the case of the 3T3, N-cadherin, and FGF2 responses, the results show the mean length of the longest neurite/cell pooled from three independent experiments. In the case of the NCAM and L1 responses, the results are from a single representative experiment. The bars show the mean Ϯ S.E. for the three independent experiments or for the population of neurons from the single experiment.
FIG. 8. The ECD4 antiserum acts on the neuron. Cerebellar neurons were either left in control medium (filled bars) or treated with 80 g/ml F(abЈ) fraction of the N-cadherin-ECD4 antibody for 1 h at 37°C before being washed and cultured on monolayers of parental 3T3 cells or monolayers expressing N-cadherin (hatched bars). Alternatively, the control and N-cadherin expressing monolayers were treated with 80 g/ml F(abЈ) fraction antibody for 1 h at 37°C before three washes with control medium and the subsequent addition of untreated cerebellar neurons (open bars). After 18 h, the cultures were fixed, and the mean neurite length was determined. The results show that a pretreatment of neurons, but not substrate, with the F(abЈ) fraction of the N-cadherin-ECD4 antibody is sufficient to inhibit the neurite outgrowth response stimulated by N-cadherin. The results show the absolute neurite length; each value is the mean Ϯ S.E. measured from 100 -150 neurons sampled from replicate cultures.