Allosteric Regulation of E-Cadherin Adhesion*

Background: Measured binding kinetics between Colo 205 cells tested the postulate that E-cadherin adhesion is allosterically regulated. Results: p120 catenin dephosphorylation or cell treatment with activating anti-E-cadherin antibody increased E-cadherin binding affinities by 2–3-fold. Conclusion: Perturbations that do not directly affect the binding site enhanced the adhesive function of E-cadherin. Significance: These biophysical measurements demonstrated that E-cadherin is allosterically regulated. Cadherins are transmembrane adhesion proteins that maintain intercellular cohesion in all tissues, and their rapid regulation is essential for organized tissue remodeling. Despite some evidence that cadherin adhesion might be allosterically regulated, testing of this has been hindered by the difficulty of quantifying altered E-cadherin binding affinity caused by perturbations outside the ectodomain binding site. Here, measured kinetics of cadherin-mediated intercellular adhesion demonstrated quantitatively that treatment with activating, anti-E-cadherin antibodies or the dephosphorylation of a cytoplasmic binding partner, p120ctn, increased the homophilic binding affinity of E-cadherin. Results obtained with Colo 205 cells, which express inactive E-cadherin and do not aggregate, demonstrated that four treatments, which induced Colo 205 aggregation and p120ctn dephosphorylation, triggered quantitatively similar increases in E-cadherin affinity. Several processes can alter cell aggregation, but these results directly demonstrated the allosteric regulation of cell surface E-cadherin by p120ctn dephosphorylation.

Cadherins are transmembrane adhesion proteins that maintain intercellular cohesion in all tissues, and their rapid regulation is essential for organized tissue remodeling. Despite some evidence that cadherin adhesion might be allosterically regulated, testing of this has been hindered by the difficulty of quantifying altered E-cadherin binding affinity caused by perturbations outside the ectodomain binding site. Here, measured kinetics of cadherin-mediated intercellular adhesion demonstrated quantitatively that treatment with activating, anti-Ecadherin antibodies or the dephosphorylation of a cytoplasmic binding partner, p120 ctn , increased the homophilic binding affinity of E-cadherin. Results obtained with Colo 205 cells, which express inactive E-cadherin and do not aggregate, demonstrated that four treatments, which induced Colo 205 aggregation and p120 ctn dephosphorylation, triggered quantitatively similar increases in E-cadherin affinity. Several processes can alter cell aggregation, but these results directly demonstrated the allosteric regulation of cell surface E-cadherin by p120 ctn dephosphorylation.
Cadherins mediate cell-cell cohesion in all tissues and are indispensable for morphogenesis, the maintenance of tissue barriers, and regulated tissue remodeling. Intercellular interactions are not static, and many critical biological processes, such as collective migration (1,2) or endothelial barrier disruption during leukocyte extravasation (3), require dynamic cadherin regulation for facile cell detachment and reorganization. Adhesion strength is a function of E-cadherin affinity and surface expression, which is under transcriptional control and modu-lated by trafficking and endocytosis (4). Mechanical factors, such as the stiffness of the cell cortex or increased cytoskeletal interactions can influence adhesion strength (5)(6)(7). Additional evidence suggests that inside-out signaling may also allosterically regulate E-cadherin adhesive activity (4, 8 -11).
Inside-out/outside-in signaling typically involves allosteric coupling between binding sites and distal effector sites on opposite sides of the membrane (12). Thus, altered cadherin binding caused by perturbations at sites away from the homophilic binding site would evince the allosteric regulation of cadherin adhesion. Integrins are prototypical, allosteric transmembrane adhesion proteins (13), and cytoplasmic perturbations alter both integrin adhesion and clustering (14 -16). Although cadherins are also transmembrane proteins, studies of cadherin binding mechanisms have largely focused on constitutively active recombinant extracellular domains (17). There are few tools able to quantify binding affinities of membrane-bound adhesion proteins, and cadherins are further complicated by the possible formation of both adhesive (trans) and lateral (cis) bonds with cadherins on the same membrane (17)(18)(19)(20), such that both interactions could be regulated.
Circumstantial evidence for the allosteric regulation of cadherins includes the modulation of cadherin adhesion upon association with other membrane proteins, such as protocadherins (10,21,22). Additionally, point mutations and antibody binding at epitopes away from the N-terminal binding site abrogated adhesion (23)(24)(25). Conversely, the binding site mutation W2A altered epitope exposure in an ectodomain region near the membrane (26).
E-cadherin-specific monoclonal antibodies (mAbs) were recently shown to activate the E-cadherin-mediated aggregation of Colo 205 cells (8). Colo 205 cells express E-cadherin but do not aggregate unless treated with trypsin or the kinase inhibitor staurosporine (11). The activating antibodies triggered Colo 205 aggregation and tight cell compaction and also decreased the phosphorylation of p120 ctn , a cytoplasmic protein that binds the cadherin cytodomain (17). Conversely, the expression of a phosphorylation-deficient p120 ctn mutant constitutively stimulated Colo 205 cell aggregation (8).
An important conceptual advance of this study is the direct demonstration that four distinct perturbations, which did not target the N-terminal binding site, quantitatively enhanced the affinity of membrane-bound E-cadherin. Intercellular adhesion frequency measurements (39) were used to quantify the binding kinetics and two-dimensional affinity of full-length E-cadherin expressed on Colo 205 cells. These adhesion frequency (kinetic) measurements have been used extensively to quantify the affinities of several different cell surface adhesion receptors, including cadherins (39 -49). We used this approach to establish the biophysical basis of altered Colo 205 aggregation and corresponding changes in the phosphorylation status of p120 catenin, which binds the cytoplasmic domain of E-cadherin. The results demonstrated that four different treatments that altered p120 catenin phosphorylation had quantitatively similar effects on the E-cadherin-mediated binding kinetics of Colo 205 cells, increasing the E-cadherin binding affinity ϳ3-fold. Superresolution imaging confirmed that these treatments did not alter the size distributions of E-cadherin clusters at the resolution of the measurements. These results thus provide direct biophysical evidence for the allosteric regulation of E-cadherin adhesive function.

Experimental Procedures
Plasmids, Cell Lines, and Antibodies-All cell lines used were from the American Type Culture Collection (Manassas, VA). Cells were cultured in Dulbecco's minimum Eagle's medium (DMEM) containing 10% fetal bovine serum (FBS) (Life Technologies, Inc.) in a 5% CO 2 atmosphere at 37°C. The activating antibody 19A11 (whole and Fab fragments) and the neutral antibody 76D5 (whole and Fab fragments) as well as the generation of Colo 205 cells infected with mouse p120 ctn retroviral constructs were described previously (8). Inhibitory antibody rat uvomorulin anti-E-cadherin IgG (DECMA-1 clone) was purchased from Sigma-Aldrich.
Erythrocyte Isolation and Modification-The surfaces of the erythrocytes used to probe the cadherin-mediated adhesion were covalently modified with oriented, immunoglobulin Fctagged ectodomains of canine E-cadherin (E-Cad-Fc) 3 or with hexahistidine-tagged ectodomains of mouse E-cadherin (47). The erythrocytes were isolated from human whole blood collected from healthy subjects by informed consent. The erythrocyte surfaces were modified with either anti-Fc or anti-hexahistidine antibodies, as described (53). The thus immobilized antibodies were used to capture Fc-tagged or hexahistidinetagged cadherin ectodomains.
Treatment of Red Blood Cells with E-cadherin Ectodomains and with Anti-E-cadherin Antibodies-C-terminal Fc-tagged or hexahistidine-tagged E-cadherin ectodomains were bound and oriented on red blood cells (RBCs) modified with anti-Fc or anti-hexahistidine antibody, respectively. When the E-cadherin-modified RBCs were treated with anti-E-cadherin antibodies, excess cadherin was first removed, by centrifuging the modified RBCs, followed by resuspension in Ca 2ϩ -containing PBS. Then 19A11 mAb or its Fab fragments, 76D5 Fab, or DECMA-1 mAb, each at 2 g/ml, was incubated with the RBC cell suspension at 4°C for 45 min.
Quantification of Cadherin Surface Expression Levels-Flow cytometry measurements quantified the cadherin densities on cell surfaces (cadherins/m 2 ) (41). E-cadherin-expressing cells were labeled with the primary, anti-E-cadherin antibody DECMA-1 (Sigma-Aldrich). DECMA-1 recognizes both the canine and human E-cadherin used in these studies (25). The secondary antibody was fluorescein isothiocyanate (FITC)conjugated anti-rat IgG (whole molecule; Sigma-Aldrich). The antibody labeling was done in PBS containing 1% (w/v) BSA at pH 7.4. Calcium was omitted at this step in order to prevent cell aggregation. The fluorescence intensities of labeled cells were measured with an LSR II flow cytometer (BD Biosciences). The calibration curve for the fluorescence intensity was generated with calibrated FITC-labeled standard beads (Bangs Laboratories, Inc., Fishers, IN).
Micropipette Measurements of Cell Binding Kinetics-The intercellular binding probability was measured as a function of contact time, using micropipettes to manipulate interacting cell pairs (39). The binding probability, P(t), is the ratio of the number of binding events, n b , to the total (N T ) cell-cell touches, n b /N T , and is a function of the number of cell-to-cell bonds (39). In these measurements, a cadherin-expressing cell and an RBC with surface-bound, Fc-tagged cadherin ectodomains were partially aspirated into opposing glass micropipettes (Fig. 1, A  and B). The experimental chamber contained L15 medium (Invitrogen) supplemented with 1% (w/v) BSA and 2 mM CaCl 2 and diluted 1:1 with deionized water. This hypoosmotic solution keeps the RBCs rounded. Cells were observed with a ϫ100 oil immersion objective on a Zeiss Axiovert 200 microscope, and images were recorded with a Manta G201B camera (AVT Technologies) interfaced with a high resolution (1080 ϫ 720 pixels), flat screen monitor. The contact time was manipulated with automated piezo-electric controllers programmed to repeatedly bring the two cells into contact for defined intervals. The contact area was controlled at 6 Ϯ 1 m 2 during a single set of measurements. Binding events were identified from surface deformations of the RBCs during separation and the recoil at bond failure. Each cell pair was tested for 50 repetitive cell-cell touches (N T ϭ 50), and each contact time represents measurements with at least three different cell pairs. The mean and S.E. of each set of 50 tests was determined from the Bernoulli distribution. The probabilities (p) shown in the graphs are the average of measurements with at least 3 cell pairs, and the error bars indicate the S.E. of the three sets of measurements with different cell pairs, at each time point.
Treatment of Colo 205 Cells with Activating Antibodies, Staurosporine, or LiCl-For studies with staurosporine-treated cells, Colo 205 cells were incubated with DMEM containing 7 nM staurosporine (in DMSO) for 4 h, at 37°C in 5% CO 2 . These cells were then collected by vigorous pipetting and washed with PBS containing BSA and 2 mM Ca 2ϩ at least three times before use in kinetic measurements.
Colo 205 cells were treated with either activating or neutral antibody. Additionally, we compared results obtained with the whole antibody versus the Fab fragment. Antibodies were incubated overnight with Colo 205 cells in DMEM, at a final concentration of 2 g/ml. One hour before the measurements, the cells were gently triturated, collected, and washed twice with PBS containing 1% (w/v) BSA and 2 mM CaCl 2 . The same antibody used for overnight incubation was then re-added to the washed cells at a final concentration of 2 g/ml in order to ensure that antibody remained bound to the cadherin ectodomains. These cells were gently agitated on a rocker at 4°C for 45 min before use.
Colo 205 cells were incubated with DMEM containing 60 mM LiCl for 2.5 h (37°C, 5% CO 2 ). After treatment, the cells were gently trypsinized with TrypLE for less than 2 min, collected, and resuspended in DMEM containing 60 mM LiCl, and allowed to recover under gentle agitation at room temperature for 1 h.
Colo 205 cells were also stably transfected with wild type mouse p120 ctn or its variants (6S,T3 A; mouse p120 WT or neomycin vector) (8); only the adherent Colo 205 cells were used. p120 ctn expression was verified by immunostaining, which required adherent cells (Fig. 4A). Thus, we did not use floating Colo 205 cells, because their p120 ctn expression could not be verified. Nevertheless, our kinetic data did not detect any differences in the two-dimensional affinities or the kinetic signatures measured with the floating or adherent cells (data not shown). The adherent Colo 205 cells were harvested with TrypLE, which does not affect cadherin surface expression levels, as verified by quantitative flow cytometry.
Western Blotting-Cells were collected in 1% Triton X-100 with 50 mM Tris, pH 7.4, 150 mM NaCl, 2 mM EDTA, Complete miniprotease inhibitors EDTA-free (Roche Applied Science), and PhosSTOP phosphatase inhibitor (Roche Applied Science). Cells were incubated on ice for 10 min and then vortexed for 5 s to break up cell clumps. Insoluble material was pelleted at 10,000 ϫ g at 4°C in a tabletop centrifuge. The supernatant was collected and boiled for 5 min in SDS-DTT sample buffer. Samples were run on a 6% SDS-polyacrylamide gel and transferred to a PVDF membrane for 3 h at 60 V. The membrane was blocked for 30 min with 5% milk plus PBST (phosphate-buffered saline with Tween 20) and incubated with primary antibodies overnight at 4°C. After washing, the membrane was incubated with secondary antibody for 45 min at room temperature. The blot was washed three times in PBST and then imaged on a LI-COR Odyssey system.
Superresolution Structured Illumination Microscopy (SIM)-Colo 205 cells were immunostained with an Alexa 568-labeled, neutral 76D5 Fab fragment, at a ratio of 2:1 (dye/antibody). The antibody was incubated with the cells at 37°C at a final concentration of 2 g/ml. After the 45-min incubation and just before imaging, the cells were washed three times with phenol red-free medium and imaged using a Zeiss ELYRA 700 microscope with a ϫ64 oil immersion lens. The images were processed using the Structured Illumination module of the Zeiss software to obtain the superresolved images.
Three-dimensional Superresolution Imaging (Stochastic Optical Reconstruction Microscopy (STORM))-The details of the STORM setup were described previously (54,55). Briefly, we used a microscope (Olympus IX-71 with an Olympus ϫ100, numerical aperture 1.4 SaPo oil immersion objective) with red (647 nm, 100 milliwatts, Crystalaser, DL640-100-AL-O) and violet lasers (405 nm, 10 milliwatts, SpectraPhysics, Excelsior). The lasers were combined through dichroic mirrors and expanded 7.5ϫ with a beam expander. The expanded and collimated beams were directed to a total internal reflection lens, which focuses the beams at the back focal plane in the microscope, with an angle slightly smaller than the total internal reflectance angle. This reduced the background while illuminating several hundred nm along the z axis. A dichroic mirror (Semrock FF408/504/581/667/762-Di01-25X36) reflected the laser lines to the objective. The emission signals were collected by the same objective, passing through an emission filter (Semrock FF01-594/730-25) and two additional notch filters (Semrock NF01-568/647-25X5.0 and NF01-568U-25) and finally imaged on a 512 ϫ 512 Andor EMCCD camera (DV887ECS-BV, Andor Technology). For three-dimensional imaging, a cylindrical lens with a 2-m focal length (SCX-50.8-1000.0-UV-SLMF-520-820, CVI Melles Griot) was inserted in the emission path to generate astigmatism (54). Because image acquisition requires tens of minutes, it was essential to correct for horizontal and vertical drift. The z-drift was fixed by pairing the PI piezo-objective (P-721.10) and ASI CRISP (Applied Scientific Instrumentation CRISP system). The horizontal drift was corrected later by data analysis. A home-written program (Cϩϩ) controlled data acquisition. The data analysis algorithm was provided by Xiaowei Zhuang (55) and modified for three-dimensional imaging (54).
Cluster Analysis-E-cadherin clusters in STORM images were analyzed using the DBSCAN (density-based spatial clustering of applications with noise) approach (56). Based on previous reports of superresolution imaging of cadherin and other proteins (57, 58), we fixed two parameters required for cluster allocation and analysis. First, the minimum distance between two points in a cluster (Eps) was set at twice the theoretical resolution (40 nm in this case). Second, the minimum number of points to define a cluster (Nps) was set at 20. From the resulting cluster analysis and cluster allocation, we obtained the 1) total number of clusters and total number of spots in clusters on individual cells; 2) diameter and number of points in individual clusters; 3) area of individual cells; and 4) center coordinates of individual cells. Here the cluster "diameter" is defined as twice the average distance between the center of the cluster and every point in the cluster and can be considered to be the average diameter of a cluster.

Activating but Not Neutral Antibody Increases Cadherin trans-Dimerization Affinity-
The present studies used the mAb 19A11, which binds a discontinuous epitope between extracellular domains 1 and 2 (EC1 and EC2) of the E-cadherin ectodomain (EC1-EC5). The 19A11 antibody significantly enhanced Colo 205 aggregation and tight cell compaction and adhesion to E-cadherin surfaces in fluid shear assays (8). Controls used the neutral antibody 76D5, which did not significantly alter Colo 205 cell aggregation or Madin-Darby canine kidney cell adhesion to substrates coated with E-Cad-Fc, in fluid shear assays (8).
Adhesion frequency measurements (40) revealed antibodyinduced changes in the two-dimensional affinities of E-cadherin on Colo 205 cells. Fig. 1C compares the binding probability (the number of binding events normalized by the number of cell-cell contacts) measured between Colo 205 cells and RBCs modified with oriented, canine E-Cad-Fc, with or without treatment with the activating 19A11 mAb.
The binding probability measured with untreated Colo 205 cells rapidly increased to an initial plateau at P1 ϳ0.3 and then gradually increased to ϳ0.4 over ϳ20 s (Fig. 1C). In sharp contrast, treatment with 19A11 mAb altered the kinetics qualitatively and quantitatively. The binding probability rapidly increased to a higher initial plateau at P1 ϳ0.5. This was followed by a short 2-5-s lag and then a further increase to a final, steady state plateau at P2 ϳ0.8 within 20 s (Fig. 1C). There was no further change, at least up to ϳ30 s (Fig. 1C). Prior Western blots and determinations of cadherin surface expression levels by quantitative flow cytometry showed that treatment with either activating or neutral antibody did not statistically alter E-cadherin expression on Colo 205 cells and was 44 Ϯ 4 cadherins/m 2 in both cases.
EC1-dependent trans-dimerization is described by a simple receptor-ligand binding reaction.
The analytical expression for the time-dependent binding probability, P(t), for the above reaction is as follows (39).
Here, m L and m R are the receptor and ligand surface densities (number/m 2 ) on the two cells, A c is the contact area (m 2 ), K 2D is the two-dimensional binding affinity (m 2 ), and k off is the off rate (s Ϫ1 ). The ligand densities (number/m 2 ) and contact areas are known. The two-dimensional affinity K 2D and k off for trans-dimerization were then estimated from fits of Equation 1 to the data corresponding to the first, trans-binding step (i.e. the rise to P1) (41,48).
To determine the trans-binding parameters, we first parsed the kinetic data between the initial binding step (rise to P1) and the subsequent rise to P2, using a non-linear lack-of-fit F-test (48). The F-test compares the least squares residuals of data fits to the model (Equation 1) with the intrinsic variability in the data. If the test statistic exceeded the critical value for a given time point, then the time point did not fit the trans-dimerization model (Reaction 1) and was attributed to the second phase. In this study, trans-binding, as modeled by Equation 1, described the rise to P1 but not the subsequent increase to P2 at times (t) Ͼϳ10 s. The dissociation rate and two-dimensional trans-binding affinity were then determined by fitting Equation 1 to the maximum number of points in each data set that did not fail the lack-of-fit test. The lines in Fig. 1 (and subsequent graphs) are the resulting weighted, nonlinear least squares fits (OriginLab version 9.0, Northampton, MA), and the thus determined best-fit parameters are given in Table 1.
Data fits showed that the 19A11 mAb treatment increased the trans-binding affinity of E-cadherin on Colo 205 cells ϳ3-fold, relative to the untreated control. The best-fit K 2D for binding between E-Cad-Fc-modified RBCs and Colo 205 cells was (1.6 Ϯ 0.3) ϫ 10 Ϫ4 m 2 , and k off was 1.0 Ϯ 0.2 s Ϫ1 . With untreated cells, the best-fit value of K 2D was (0.7 Ϯ 0.1) ϫ 10 Ϫ4 m 2 and k off was 1.5 Ϯ 0.4 s Ϫ1 . The apparent difference in k off for antibody-treated Colo 205 cells is insignificant at the 95% confidence level (n ϭ 18, p ϭ 0.2). Statistical significance is defined by p values of Ͻ0.05. However, the difference between K 2D values is statistically significant (p ϭ 0.002, n ϭ 18).
The enhanced affinity of E-cadherin on Colo 205 cells was not due to cadherin cross-linking by whole antibody. Measurements with the Fab fragment of 19A11 tested whether the enhanced affinity could be due to cadherin cross-linking by the whole 19A11 mAb (Fig. 1D). Treatment with the 19A11 Fab gave quantitatively similar results; namely, the best-fit values for K 2D and k off were (2.1 Ϯ 0.3) ϫ 10 Ϫ4 m 2 and 0.9 Ϯ 0.2 s Ϫ1 , respectively. The latter values are statistically similar to those measured after treatment with the whole antibody (for K 2D , p ϭ 0.19, n ϭ 18; for k off , p ϭ 0.7, n ϭ 18). The best-fit parameters are summarized in Table 1.
In controls, the Fab fragment of the neutral antibody 76D5, which did not induce Colo 205 aggregation, had a small effect on the kinetics, relative to untreated Colo 205 cells (Fig. 1E). The amplitude of the first plateau P1 increased slightly, and there was a slow, small further increase that slightly exceeded the binding probability of untreated Colo 205 cells at 20 s. The fitted values for K 2D and k off were (1.2 Ϯ 0.2) ϫ 10 Ϫ4 m 2 and 1.3 Ϯ 0.3 s Ϫ1 , respectively (Table 1). These values are statistically similar to the untreated Colo 205 cells (p ϭ 0.06 for K 2D and p ϭ 0.7 for k off ). K 2D is statistically different from the value measured after treatment with the 19A11 Fab (p ϭ 0.02). Results obtained with the whole 76D5 mAb were similar to results obtained with untreated Colo 205 cells (Table 1).
Antibodies Do Not Significantly Affect the Affinity of Recombinant E-cadherin Ectodomains or Adhesion-competent E-cadherin on MCF7 Cells-We tested whether the activating or neutral antibodies altered the kinetics of adhesion-competent E-cadherin extracellular domains. Bulky N-glycans on EC2 and EC3 appeared to alter the second binding step (P2), without affecting K 2D , and the change was attributed to altered cis-interactions (48). The neutral antibody 76D5 binds near the EC3-EC4 junction and could similarly perturb the kinetics in ways that might not be obvious in cell aggregation studies.
In the absence and presence of 19A11 or 76D5 Fabs, the kinetics were statistically similar to control measurements ( Fig.  2A). Best-fit values for K 2D and k off for the trans-dimerization of untreated E-Cad-Fc ectodomains were (2.7 Ϯ 0.4) ϫ 10 Ϫ4 m 2 and 1.1 Ϯ 0.3 s Ϫ1 , respectively (Table 1). After treatment with activating 19A11 mAb, the fitted parameters were (3.0 Ϯ 0.4) ϫ 10 Ϫ4 m 2 and 0.9 Ϯ 0.2 s Ϫ1 for K 2D and k off , respectively. Similarly, after treatment with the neutral 76D5 Fab, the fitted values for K 2D and k off were (2.9 Ϯ 0.4) ϫ 10 Ϫ4 m 2 and 1.0 Ϯ 0.2 s Ϫ1 . For all three pairwise comparisons, the affinities were statistically similar (p Ͼ 0.5), as were the off rates (p Ͼ 0.7). These results confirm that the antibodies do not alter the trans-binding parameters or the overall two-stage binding kinetics of adhesion-competent E-cadherin ectodomains.
Unlike Colo 205 cells, MCF7 cells express constitutively active E-cadherin and exhibit robust cell cohesion. Treatment with 19A11 antibody did not enhance the affinity of adhesioncompetent E-cadherin on MCF7 cells (Fig. 2B). The two-dimensional E-cadherin binding affinity was (3.3 Ϯ 0.4) ϫ 10 Ϫ4 m 2 . Following 19A11 Fab treatment, the affinity was (4.1 Ϯ 0.5) ϫ 10 Ϫ4 m 2 . The fitted values are not statistically different from those of untreated cells at the 95% confidence level (p ϭ 0.2). The k off values of 1.2 Ϯ 0.4 and 0.9 Ϯ 0.2 s Ϫ1 were similar for the treated and untreated MCF cells, respectively (p ϭ 0.4). By contrast, treatment with the inhibitory antibody DECMA-1 reduced K 2D to (0.5 Ϯ 0.1) ϫ 10 Ϫ4 m 2 , which is slightly lower than the value measured with untreated Colo 205 cells (Table 1).
Colo 205 Treatment with Staurosporine Increases the E-cadherin Affinity-The broad-spectrum kinase inhibitor staurosporine was shown to activate Colo 205 cell aggregation and p120 ctn dephosphorylation (8,11). Here, the binding kinetics of staurosporine-treated Colo 205 cells (Fig. 3C) was quantitatively similar to 19A11 Fab-treated cells; namely, K 2D increased 2-fold to (1.5 Ϯ 0.2) ϫ 10 Ϫ4 m 2 (p ϭ 0.02), but the dissociation rate of 0.9 Ϯ 0.2 s Ϫ1 was statistically similar to that of untreated cells (p ϭ 0.19) ( Table 1). In controls, staurosporine had no effect on binding between E-Cad-Fc-modified RBCs (not shown). This result confirmed that the kinase inhibitor does not affect the ectodomains directly or alter the RBCs.

Lithium Chloride (LiCl) Decreases p120 Phosphorylation and Increases E-cadherin Affinity on Colo 205 Cells-Different
kinases, including glycogen synthase kinase 3␤, regulate serine/ threonine phosphorylation of p120 ctn at various sites on the protein (59). Treatment with the glycogen synthase kinase 3␤ inhibitor LiCl (60, 61) significantly increased Colo 205 cell aggregation and compaction, relative to control NaCl-treated cells, which remained rounded and dispersed (Fig. 3A). West-ern blots confirmed that LiCl treatment resulted in p120 ctn dephosphorylation (Fig. 3B). Further Western blots with phosphospecific mAb to residue Thr-310 of p120 ctn (8,59), one of eight major residues known to be phosphorylated (62), revealed decreased overall phosphorylation at that site.
LiCl treatment of Colo 205 cells increased the E-cadherin affinity, relative to controls. As with staurosporine-treated cells, the initial binding probability increased to an initial plateau (P1), followed by a short lag, and an increase to a higher, steady state probability P2 (Fig. 3C) ( Table 1), relative to untreated cells. Western blots and flow cytometry measurements confirmed that LiCl treatment did not significantly alter E-cadherin expression levels (Fig. 3B). Thus, glycogen synthase kinase 3␤ specifically phosphorylates p120 ctn , and the p120 ctn phosphorylation status allosterically alters the E-cadherin affinity, on Colo 205 cells. p120 ctn Dephosphorylation Enhances the E-cadherin Affinity on Colo 205 Cells-Kinetics measurements with Colo 205 cells expressing p120 ctn phosphorylation mutants confirmed the causal relationship between p120 ctn phosphorylation and E-cadherin affinity. Fig. 4B shows the binding kinetics of Colo 205 cells stably expressing either WT mouse p120 ctn or a multisite N-terminal phosphorylation mutant, 6S,T3 A, in which serine and threonine phosphorylation sites were mutated to alanine. Immunofluorescence imaging confirmed that heterologous wild type mouse p120 ctn localized to intercellular junctions (Fig. 4A). Negative controls used cells infected with the neomycin vector. Colo 205 expression of the 6S,T3 A mouse p120 ctn mutant altered both the magnitude of the binding prob-ability P1 and the qualitative shape of the time course relative to cells expressing either WT mouse p120 ctn or the neomycin vector (Fig. 4B). Importantly, model fits of data for the initial, trans-binding step (Fig. 4B and Table 1) showed a more than 3-fold increase in K 2D to (2.3 Ϯ 0.3) ϫ 10 Ϫ4 m 2 by cells expressing the 6S,T3 A mutant, relative to K 2D of (0.7 Ϯ 0.1) ϫ 10 Ϫ4 m 2 for cells transfected with the neomycin vector. There was a greater than 2-fold increase relative to the K 2D (1.1 Ϯ 0.2) ϫ 10 Ϫ4 m 2 determined with Colo 205 cells expressing WT mouse p120 ctn (p Ͻ 0.001). The values of k off were statistically the same, within error, for all three conditions (p Ͼ 0.3 in all cases). These results demonstrated that p120 ctn dephosphorylation increased the E-cadherin affinity on Colo 205 cells and confirm that the affinity changes are due to allosteric regulation of E-cadherin binding by both p120 ctn -dependent, inside-out signaling and by activating antibodies. The nearly identical kinetic effects of all four treatments are also compelling evidence that both the activating antibody and p120 ctn dephosphorylation alter the E-cadherin affinity by the same mechanism.
The E-cadherin Mutant L175D Does Not Alter the trans-Binding Affinity-Cadherin clustering appears to involve both the extracellular and cytoplasmic domains (18,19,37,58). Here we tested how disrupting a putative cis-interaction (63) between extracellular domains affected the fitted kinetic parameters. We reasoned that, if cis-dimerization between the ectodomains enhanced the affinity, then its disruption would lower K 2D .
Based on STORM images, clusters of the mouse E-cadherin L175D cis mutant expressed on A431 cells were significantly smaller than WT E-cadherin (58). Fig. 5 compares the kinetics measured with RBCs modified with ectodomains of either the WT or L175D mutant. In contrast to WT mouse E-cadherin, the kinetics of the mutant exhibited a single exponential rise to a limiting plateau (Fig. 5). The amplitudes of P1 for the mutant and WT proteins were the same, at the same cadherin surface density. The L175D mutant exhibited a subtle, slow increase at longer contact times, but disrupting the putative cis-interac-   Table 1. Error bars, S.E.
tion between ectodomains essentially eliminated the second kinetic step. This result is consistent with the previous attribution of the second kinetic rise (to P2) to lateral cadherin interactions (37). The fitted K 2D and k off values for trans-dimerization were statistically the same for both proteins, in agreement with measured solution binding affinities (63). Thus, altering the observed lateral ectodomain interaction mediated by L175 did not affect the trans-binding affinity of E-cadherin. Superresolution Imaging of E-cadherin Distributions on Colo 205 Cells-To further test the possibility that the altered E-cadherin affinities were due to differences in clustering, we imaged E-cadherin clusters on unmodified Colo 205 cells, Colo 205 cells expressing the p120 ctn 6S,T3 A mutant, and cells expressing WT p120 ctn , using both SIM and STORM.
SIM images of E-cadherin on live cells at 37°C revealed punctate E-cadherin clusters at a resolution of ϳ130 nm on all three indicated Colo 205 cell types (Fig. 6A, i-iii). The latter images were obtained, after directly staining E-cadherin with the neutral 76D5 Fab labeled with Alexa Fluor 568. The use of labeled Fab fragments eliminated the possibility of antibody cross-linking, and SIM imaging of live cells eliminated fixation artifacts. Visual comparisons of the three images showed that there was no apparent difference in the E-cadherin clusters on individual live cells, at the ϳ130 nm resolution of SIM, despite significant differences in cell aggregation and E-cadherin affinity.
The three-dimensional STORM images of fixed cells stained with whole primary and secondary antibodies were qualitatively similar to the SIM images (Fig. 6B, i-v). At a resolution of ϳ20 nm, the expression of the p120 ctn 6S,T3 A mutant did not alter the size distribution of E-cadherin clusters, relative to unmodified cells or cells expressing WT p120 ctn (Fig. 7). E-cadherin clusters were distributed uniformly over the cell surfaces (Fig. 6B), and their size distributions were similar on all three cell types, with the most probable cluster size (mode of the distribution) being ϳ100 nm in all cases (Fig. 7, A and E). In control images of cells incubated with secondary antibody only, the number of clusters was much lower (Fig. 7C), and the mode of the distribution was ϳ70 nm (Fig. 7, A and E). However, nonspecific aggregates associated with the secondary antibody appeared to have similar numbers of points per cluster, as per our analysis, but were more variable than the other samples ( Fig. 7B). However, to rule out sample fixation artifacts, we also performed the aforementioned SIM imaging with live, unfixed cells. We also tested the impact of varying the cluster parameters (Eps ϭ 20 -60, Nps ϭ 10 -40) used to analyze the results. Although varying Eps and Nps did affect the total number of clusters identified, the cluster distributions were similar for each of the Colo 205 cell lines (data not shown).

Discussion
Kinetic studies together with biochemical treatments provide direct biophysical evidence for the allosteric regulation of E-cadherin binding, both by activating antibodies and by inside-out signaling modulated by the phosphorylation status of p120 ctn . The ability of micropipette measurements to quantify changes in ectodomain binding due to perturbations away from the N-terminal binding site uniquely enabled the quantitative demonstration of the allosteric regulation of E-cadherin.
Receptor accumulation at cell-cell contacts (changes in m L and m R ) would not account for the increased affinities. According to the model (Equation 1), this would require a ϳ3-fold increase in the overall cadherin density in the contact area. The sparse cadherin distributions on both cells exclude significant accumulation, within the 20-s cell-cell contact time. This conclusion is supported by experimental measurements of N-cadherin accumulation at smaller intercellular contacts, where a ϳ3-fold increase in the local N-cadherin density required 15 min (64). The latter time scale is 60 times greater than that of these adhesion frequency measurements. Thus, cadherin accumulation through diffusion and kinetic trapping could not explain the affinity differences.
These results support the postulate that treatments triggered Colo 205 aggregation by allosterically regulating the cadherin binding affinity. The precise allosteric mechanism is presently not known, but it could involve induced conformational changes, altered configurational entropy (12,65), the activation of cis-dimerization, or a combination of these mechanisms.
Increased cadherin clustering could alter the measured affinity by constraining cadherins near ligands on the opposing cell. In adhesion frequency measurements of selectin dimers, for example, ligand binding by the first selectin constrained the second and thereby enhanced binding to the second ligand (66). The thermodynamics of multivalent cooperativity was addressed previously (67,68), and recent simulations demonstrated such cis-trans cooperativity within adhesion zones between model cell membranes (69).
In these studies, perturbing a postulated cis-dimerization interface between extracellular domains did not, however, alter the trans-binding affinity. The K 2D values of WT and L175D mouse E-cadherin were the same, although L175D reduces the sizes of cadherin clusters larger than 20 nm (58).
Although the L175D mutant targeted extracellular domain interactions, antibody binding and/or p120 ctn dephosphorylation could also impact transmembrane or cytoplasmic domain associations (37, 57, 70 -73). There were no observed differences in clusters on treated versus untreated Colo 205 cells, so that any dimerization changes would be below the resolution of STORM. Nevertheless, altered dimerization and/or changes in intrinsic cadherin affinities in response to the treatments described in this study would necessarily require allosteric regulation to account for these findings.
Consistent with an allosteric mechanism, none of the four treatments that activated Colo 205 aggregation involved the E-cadherin trans-binding interface directly, and all generated quantitatively similar changes in cadherin binding kinetics. The activating antibody 19A11 binds a discontinuous epitope at the EC1-EC2 junction (8) and increased K 2D . Antibody binding altered neither the cadherin surface densities nor the two-stage kinetics of adhesion-competent ectodomains. These results are consistent with the allosteric activation of the adhesive function of E-cadherin on Colo 205 cells. The 19A11 antibody also altered p120 ctn phosphorylation, by outside-in signaling (8). Staurosporine, lithium chloride treatment, or the expression of the mouse p120 ctn 6S,T3 A mutant similarly reduced p120 ctn phosphorylation (8), all three treatments generated kinetic changes that were similar to those induced by the 19A11 antibody. The LiCl treatment shows that this allosteric regulation of E-cadherin activity requires a signaling pathway, which involves glycogen synthase kinase 3␤ as a regulator of p120 catenin phosphorylation.
All four treatments increased the E-cadherin affinities on Colo 205 cells 2-3-fold, but they did not alter k off . Analysis of variance confirmed that the apparent differences in k off values obtained with 19A11-treated cells (Table 1) are not statistically significant, at the 95% confidence level. Thus, all four treatments investigated appear to activate E-cadherin by enhancing its association rate.
Activating antibodies did not enhance the affinity of adhesion-competent, soluble extracellular domains. This result indicates that the isolated fragment assumes the fully active conformation, which can be regulated at the cell surface, by cytoplasmic binding partners. Depending on the expressed kinases and their activation states, distributions of p120 ctn phosphorylation states within a cell could generate distributions of active and inactive cell surface cadherins, such that activating antibodies could further increase adhesion and p120 ctn dephosphorylation, as observed with A431 cells (8). A high level of constitutive E-cadherin activity would explain the small effect of 19A11 mAb on MCF7 cells. The possibility that other factors could regulate p120 phosphorylation levels (and the relative E-cadherin adhesive activity in different cell lines) remains to be explored.
The activation of cell aggregation coincident with this modest 2-3-fold increase in affinity is not surprising because aggregate size is not thermodynamically controlled and does not scale with either affinity or adhesion energy. The onset of aggregation indicates that the intercellular adhesion energy exceeds the threshold ambient (background) thermal energy (74). Importantly, the kinetic measurements quantified binding affinities, not adhesion energies. Our results leave open the possibility that factors in addition to cadherin affinity modulation could augment Colo 205.
The neutral Fab 76D5, which did not induce Colo 205 aggregation (8), modestly altered the dissociation rate, k off . However, the antibodies were classified based on qualitative cell aggregation or shear flow assays (8), and the small kinetic change induced by 76D5 mAb may not cause detectable differences in cell adhesion or aggregation.
Evidence for the allosteric regulation of cadherin adhesion is currently based on functional data. Although structures might reveal the basis of E-cadherin regulation, increasing experimental evidence exposed the limitations of classical, structurebased models of allostery (12,65,75,76). It is often unclear how different conformations are functionally interconnected. In the context of "dynamic allostery" (12,77), altered conformational dynamics and associated entropic changes could also perturb binding, without changing the average protein structure. Indeed, structural fluctuations and their suppression appear to influence cadherin interactions at cell surfaces (38,69,78). Thus, structures can support, but not prove, allostery. Within this current view, functional data are necessary and sufficient evidence for allosteric regulation.
In summary, adhesion frequency measurements enabled the unique, quantitative demonstration of allosteric regulation of homophilic E-cadherin binding by p120 ctn phosphorylation. Fig. 8 illustrates a postulated mechanism, based on the available data. Here, the initially low E-cadherin K 2D results in low binding probabilities (Fig. 8A). Activating antibody allosterically alters the cytoplasmic domain, resulting in p120 ctn dephosphorylation by outside-in signaling (8) (Fig. 8B). Inside-out signaling then increases the cadherin affinity, which increases the measured cell-cell binding probability (Fig. 8C). The antibodies could also allosterically influence the binding site directly. The activating antibody had no detectable effect on isolated cadherin ectodomains, as expected, if the fragments adopted a constitutively activate conformation.
Allosteric regulation is an efficient mechanism for propagating signals through large cadherin structures. Intriguingly, the large number of cadherin superfamily members whose cytoplasmic domains and binding partners are kinase targets suggests that allosteric regulation may be a more common mechanism than previously thought, with important consequences for morphogenesis and the controlled regulation of tissue functions.  Low Affinity E-Cadherin p120 P p120 P p120 P p120 P High Affinity p120 p120 p120 p120 Y Y mAb FIGURE 8. Proposed model for the allosteric activation Of E-cadherin. A, the low E-cadherin affinity on untreated Colo 205 results in little bond formation and low cell-cell binding probabilities. B, activating antibody treatment or the expression of the phosphorylation mutant mouse p120 ctn 6S,T3 A decreases p120 ctn phosphorylation and activates E-cadherin-mediated cell aggregation. C, the substantially higher E-cadherin following activating antibody treatment or expression of the p120 ctn phosphomutant 6S,T3 A increases the frequency of homophilic E-cadherin bonds and the measured cell-cell binding probabilities.