This study investigated the impact of cadherin binding differences on both cell sorting and GTPase activation. The use of N-terminal domain point mutants of Xenopus C-cadherin enabled us to quantify binding differences and determine their effects on cadherin-dependent functions without any potential complications arising as a result of differences in cytodomain interactions. Dynamic cell-cell binding measurements carried out with the micropipette manipulation technique quantified the impact of these mutations on the two-dimensional binding affinities and dissociation rates of cadherins in the native context of the cell membrane. Pairwise binding affinities were compared with in vitro cell-sorting specificity and ligation-dependent GTPase signaling. Two-dimensional affinity differences greater than five-fold correlated with cadherin-dependent in vitro cell segregation, but smaller differences failed to induce cell sorting. Comparison of the binding affinities with GTPase signaling amplitudes further demonstrated that differential binding also proportionally modulates intracellular signaling. These results show that differential cadherin affinities have broader functional consequences than merely controlling cell-cell cohesion.
The intercellular adhesion cadherin proteins are essential for maintaining the structural integrity of tissues. During morphogenesis, they are required for cell patterning, and, in mature tissues, they regulate crucial barrier functions (Gumbiner, 2005; Takeichi, 1995). The classical cadherins are the most extensively studied proteins in the cadherin superfamily. There are >20 known subtypes, which exhibit the same overall fold, but differ in their primary structure and tissue expression patterns. A central question is whether subtype-dependent sequence differences alter cadherin-mediated intercellular binding, and the implications of those differences for cadherin-dependent cell functions.
Investigations of differences between cadherin subtypes mainly focused on cadherin-dependent cell segregation. This is in part due to important in vitro studies suggesting that cell sorting depends on both the identities and surface densities of the expressed cadherin subtypes (Nose et al., 1988; Steinberg, 1963; Steinberg, 2007). Those findings suggested that that subtype-dependent differences in intercellular adhesion energies direct cell sorting in vitro and possibly in vivo (Steinberg, 1963). This focused attention on cadherin affinities and their relationship to adhesion energies or surface tension thought to influence cell segregation (Foty and Steinberg, 2005; Steinberg, 1963; Steinberg, 2007).
The N-terminal cadherin domain is the main locus of cadherin binding differences that influence in vitro sorting. In structures of complexes of the extracellular segment of Xenopus C-cadherin (Boggon et al., 2002) and truncated fragments of N-cadherin (Shan et al., 2000; Shapiro et al., 1995) or E-cadherin (Häussinger et al., 2004; Pertz et al., 1999; Tomschy et al., 1996), the W2 on the first extracellular domain (EC1) inserts into a hydrophobic pocket on the EC1 domain of the adjacent cadherin. The high degree of sequence similarity among EC1 domains of type I classical cadherins begs the question of how this conserved binding motif supports cell binding selectivity. Yet, mutations in the W2 binding pocket alter cell-cell cohesion and sorting. Exchanging the N-terminal domain of E-cadherin with that of P-cadherin, or substituting residues 78 and 83 on mouse E-cadherin with the corresponding P-cadherin sequence altered the aggregation specificity of cells expressing the E-cadherin mutants (Nose et al., 1990). The A78M mutation abolished N-cadherin function (Tamura et al., 1998). Despite these qualitative observations, links between sequence differences, quantified affinities, and cadherin-dependent functions have not been established.
Solution binding affinities of recombinant, soluble fragments indicated that affinities differing by at least 5 fold correlated with in vitro cell sorting, assuming similar cadherin expression levels (Katsamba et al., 2009). However, semi-quantitative estimates of relative cell adhesion (Niessen and Gumbiner, 2002), quantified, protein-level adhesion energies (Prakasam et al., 2006b), strengths of single cadherin bonds (Shi et al., 2008), or cohesive energies of cell aggregates (Duguay et al., 2003) do not always correlate with in vitro cell sorting outcomes.
In vivo, the role of cadherin binding differences in cell sorting is less clear. Differential cadherin expression correlates with retinal cell patterning in Drosophila, for example (Hilgenfeldt et al., 2008). Yet, cortical tension, rather than cell cohesion, appears to direct germ cell positioning in zebrafish embryos (Krieg et al., 2008). A possibility is that differential adhesion is unimportant in vivo, but correlations between mutations that impair cadherin adhesion and gastric cancer (Becker et al., 1999; Handschuh et al., 1999; Handschuh et al., 2001) suggest that altered cadherin adhesion also modulates signaling. Differential cadherin adhesion could more broadly influence cell behavior through signaling. For example, affinity-dependent Rho GTPase signal amplitudes (Braga, 2002; Drees et al., 2005; Noren et al., 2003) could also modulate cortical tension, cell cycle progression, and differentiation (Fournier et al., 2008; Levenberg et al., 1999). The broader impact of such binding differences on both sorting and signaling has not been considered.
This study investigated the impact of cadherin binding site mutations on two-dimensional affinities, in vitro cell sorting, and GTPase signaling. Comparisons of ectodomain mutants rather than cadherin subtypes enabled us to focus on affinity differences, independent of differences in cytoplasmic domain interactions. The use of cadherins that cause Chinese Hamster Ovary (CHO) cells to segregate from each other (Shi et al., 2008), also facilitated determinations of the impact of binding differences on cell sorting. Selected Xenopus C-cadherin mutants were based on sequence differences between amino acids near docked W2 in the hydrophobic pocket of N-cadherin. Micropipette measurements then quantified the affinities of full-length C-cadherin mutants in the native context of the cell membrane. These cadherin properties were compared with both in vitro cell sorting outcomes and ligation-dependent GTPase signaling (Becker et al., 1999; Handschuh et al., 1999; Handschuh et al., 2001).
Design and expression of C-cadherin mutants
CHO cells that express the same densities of Xenopus C-cadherin (C-CHO) and chicken N-cadherin (N-CHO) sort out in both hanging drops and in agitated cell suspensions (Shi et al., 2008). Here we used these proteins as models to investigate the impact of binding site mutations on affinities, in vitro cell sorting, and GTPase signaling. On the basis of sequence and structural comparisons of docked W2 at EC1-EC1 interfaces of Xenopus C-cadherin and mouse N-cadherin (Fig. 1A,B), three sites in the EC1 domain of C-cadherin were mutated to the corresponding amino acid in chicken N-cadherin (Fig. 1C). The EC1 domain of mouse N-cadherin (Fig. 1B) is 98% identical to that of chicken N-cadherin. The K8NS10P double mutant potentially alters the docked W2 orientation (Pokutta and Weis, 2007). The other two mutations S78A and M92I involve more polar residues lining the W2 binding pocket that were postulated to play a greater role in modulating the affinity (Patel et al., 2003). Two other mutants Q23G and E83V did not express sufficiently well for these biophysical studies.
Clones that express the C-cadherin mutants were selected according to expression level, by quantitative FACS and by Western blots of cell surface proteins. Comparisons of in vitro cell sorting and quantitative GTPase activation measurements require cell populations that express similar cadherin surface densities. The following clones (cadherin surface densities in parentheses) were selected for these studies: WT C-cadherin (18/µm2), K8NS10P (20/µm2), S78A (22/µm2), M92I (19/µm2), and WT N-Cadherin (16/µm2). The expression levels of WT and mutant C-cadherins were compared to each other and to an actin loading control, by Western blots of the biotinylated surface proteins (Fig. 1D). The Western blots agreed with FACS measurements, which used antibodies against the ectodomains.
C-cadherin mutations alter in vitro cell sorting patterns
In vitro hanging-drop assays (Foty and Steinberg, 2004) showed that C-CHO sorted out from N-CHO in this assay (Fig. 2A,D), when both surface expression levels of cadherin and cell densities (106 cells/ml) were similar (Shi et al., 2008). The K8NS10P mutant did not affect cell sorting relative to the WT C-cadherin: cells expressing K8NS10P sorted from N-CHO but intermixed with C-CHO (Fig. 2B–D). By contrast, the S78A mutation completely switched the sorting specificity, such that cells expressing S78A sorted from WT C-CHO but intermixed with WT N-CHO (Fig. 2B–D). The M92I cells intermixed with both C-CHO and N-CHO (Fig. 2B–D). The Rac1 inhibitor NSC 23766 did not affect the sorting outcomes (Fig. 2E). Although the mechanisms and parameters determining these in vitro assays are not completely defined, the outcomes identify cadherin differences.
Biophysical properties of homophilic and heterophilic cadherin bonds
Micropipette measurements (Fig. 3) quantified the biophysical differences of the cadherin bonds that correlate with the sorting behavior. These measurements, which are not single molecule force measurements, quantify the probability of cell-cell binding as a function of contact time, cadherin surface densities, binding affinities, and dissociation rates. The binding probability P is the number of detected binding events nb divided by the total number of cell-cell touches nb/N. This established approach has been used to quantify the kinetics and two-dimensional affinities of selectin/lectin, T-Cell-receptor/MHC, integrin/ligand, and C-cadherin bonds (Chesla et al., 2000; Chien et al., 2008; Huang et al., 2004; Huang et al., 2007; Huang et al., 2010). A significant advantage of these measurements is the quantification of two-dimensional protein affinities in the native environment of the membrane.
The two-dimensional affinities and dissociation rates for pairwise cadherin interactions were determined from the time-dependent binding probabilities P(t) measured between CHO cells expressing cadherins and red blood cells (RBCs) modified with either WT Fc-tagged C-cadherin ectodomains (CEC1-5-Fc) or WT Fc-tagged N-cadherin ectodomains (NEC1-5-Fc). We previously showed that the initial (<40 seconds) cadherin binding kinetics are independent of the cytoplasmic domain (Chien et al., 2008). Because the two-dimensional affinity and adhesion energy of the Xenopus CEC12-Fc fragment are lower than both CEC1-5-Fc and the EC1245-Fc fragments (Chappuis-Flament et al., 2001; Chien et al., 2008; Zhu et al., 2003), these studies used the full extracellular segment immobilized on RBCs to probe cadherin kinetics.
Fig. 4A shows the binding probability P(t) for homophilic N-cadherin adhesion, as a function of cell-cell contact time. As reported for C-cadherin (Chien et al., 2008), the kinetic profiles exhibits two distinct stages. The first step is a fast rise to a limiting plateau at P1∼0.4 – that is, 40% of cell-cell touches resulted in a binding event (Fig. 4A). This is followed by a 2–5 second lag, and a subsequent rise to a second plateau at a binding probability of P2∼0.7 (Fig. 4A). The time-course for heterophilic C-cadherin/N-cadherin binding is qualitatively similar (Fig. 4B). The differences in the plateau amplitudes, e.g. P1 and P2 reflect differences in the protein surface densities, the two-dimensional affinities, and the dissociation rates.
The previous study mapped the fast, first step to EC1 (Chien et al., 2008), which is the putative specificity-determining region (Harrison et al., 2005; Klingelhöfer et al., 2000; Nose et al., 1990). In this study, we therefore analyzed the affinities and dissociation rates for this first binding step. Although the subsequent steps in the kinetic pathway require additional cadherin domains (Chien et al., 2008), here we address the impact of quantitative changes in EC1-dependent bonds on in vitro cell sorting and signaling.
Plots of −Ln(1−P1) versus mL × mR (Fig. 5) are linear for both homophilic and heterophilic binding by WT N-cadherin and WT C-cadherin for a range of different cadherin densities, e.g. variable mL × mR. This confirms that the fast, first step is described by the mechanism in Eqn 1. The slopes reflect the different affinities of the indicated binding interactions (Eqn 3). The linear plots also confirm that Eqn 2 mathematically describes the first binding step. Consequently, EC1-EC1 binding affinities for different pairwise cadherin interactions can be obtained either from slopes of graphs as in Fig. 5 or from nonlinear least squares fits of Eqn 2 to kinetic profiles (see Fig. 4), without the need for measurements at several different cadherin surface densities.
The weighted, nonlinear least-squares fit of Eqn 2 to the binding probability curves (Fig. 4A–C) gave values of 11±2×10−4 µm2 and 0.6±0.2 seconds−1 for Ka and the dissociation rate, respectively for homophilic C-cadherin bonds (Table 1). The similarly determined values for the homophilic N-cadherin bond are, respectively 1.9±0.3×10−4 µm2 and 1.1±0.4 seconds−1. The lesser N-cadherin affinity relative to the homophilic C-cadherin bond agrees with prior molecular-level adhesion measurements (Prakasam et al., 2006b). The two-dimensional affinity of the heterophilic bond between WT N-cadherin and CEC1-5-Fc is 3.2±0.7×10−4 µm2, and the dissociation rate is 0.8±0.4 seconds−1. These results are summarized in Table 1.
|Expressed cadherin||mR (µm−2)||Cadherin-Fc on red blood cells||mL (µm−2)||kr (second−1)||Ka (×10−4 µm2)||R2|
|Expressed cadherin||mR (µm−2)||Cadherin-Fc on red blood cells||mL (µm−2)||kr (second−1)||Ka (×10−4 µm2)||R2|
C-cadherin binding site mutations alter EC1-dependent affinities and kinetic rates
All binding kinetics measured between different C-cadherin mutants and either CEC1-5-Fc or NEC1-5-Fc on RBCs exhibited two-stage kinetics (cf. Fig. 4C). The effects of the point mutations on the C-cadherin affinities and dissociation rates were then similarly quantified from fits of the first binding step (t<10 seconds) to Eqn 2. Fig. 4C shows the kinetics and model fit for binding between the S78A mutant and CEC1-5-Fc. Relative to the WT C-cadherin, this mutation reduced the two-dimensional affinity for CEC1-5-Fc more than six-fold to 1.65±0.25×10−4 µm2. At 2.35±0.37×10−4 µm2, the two-dimensional affinity between S78A and N-Cadherin was slightly less than the WT C-cadherin/NEC1-5-Fc bond (Table 1).
The affinity of M92I for CEC1-5-Fc was 2.5-fold lower than WT C-cadherin, but the affinity for NEC1-5-Fc was similar to WT C-cadherin, within experimental error. The two-dimensional affinity measured between M92I and CEC1-5-Fc was 4.3±0.5×10−4 µm2, and the dissociation rate was 2.3±0.7 seconds−1 (Table 1). The fitted parameters for M92I binding to NEC1-5-Fc are in Table 1. By contrast, the double mutant K8NS10P, which is postulated to alter the W2 orientation in the binding site, bound CEC1-5-Fc with a statistically similar affinity as WT C-cadherin, at 10.3±0.8×10−4 µm2, and the dissociation rate of 1.2±0.3 seconds−1 was slightly higher. The affinity and dissociation rate for the K8NS10P/NEC1-5-Fc bond are in Table 1.
Ligation-dependent Rac1 activation
Quantitative measurements of Rac1-GTP levels in cells seeded on either CEC1-5-Fc or NEC1-5-Fc substrata demonstrated the correlation between binding affinities and signaling. In these comparisons between signaling amplitudes, the immobilized protein (ligand) densities, the cell type, the overall protein scaffold, including the cytoplasmic domain, the cadherin expression levels, and the measurement time are the same. Therefore, the only known variable is the affinity between the cadherins mediating cell attachment. Rac1-GTP in C-CHO cells increased up to 45 minutes after attachment to CEC1-5-Fc substrata (5×103 cadherin/µm2) (Fig. 6A), similar to prior reports (Noren et al., 2003). By contrast, N-CHO on either CEC1-5-Fc or NEC1-5-Fc (5×103 cadherin/µm2) did not exhibit any change in Rac1-GTP levels. The difference between N-CHO and C-CHO was not due to differences in cadherin expression (Table 1). Due to the robust Rac1 activation in C-CHO at 45 minutes, we compared Rac1-GTP levels 45 minutes after seeding different cells onto either CEC1-5-Fc or NEC1-5-Fc substrata (Fig. 6B).
A key finding is that Rac1-GTP levels increase with the two-dimensional cadherin affinity (Fig. 6C). In C-CHO attached to CEC1-5-Fc, the Rac1-GTP increased by a factor of 6.3±0.3 relative to t = 0 minutes (Fig. 6B). The quantitative data are summarized in supplementary material Table S1. K8NS10P ligation to CEC1-5-Fc increased Rac1-GTP 7±2-fold over initial Rac1-GTP levels (Fig. 6B; supplementary material Table S1). The latter is statistically similar to WT C-cadherin (P = 0.69). The M92I mutant, with an intermediate affinity for WT CEC1-5-Fc between K8NS10P and S78A, triggered a 3.7-fold increase in Rac1-GTP (Fig. 6B; supplementary material Table S1). By contrast, in cells expressing S78A, the Rac1-GTP was 0.6±0.2. Rac1-GTP levels were unchanged in N-CHO seeded on CEC1-5-Fc. Conversely, on NEC1-5-Fc substrata, the Rac1-GTP levels in WT C-CHO and N-CHO were 0.9±0.1 and 0.8±0.3, respectively (Fig. 6B; supplementary material Table S1). The K8NS10P cell adhesion to NEC1-5-Fc similarly did not activate Rac1-GTP. Fig. 6C shows the correlation between C-cadherin affinities and Rac1-GTP activation.
To address possible differences in GTPase signaling at cell-cell junctions, global Rac1 activation was quantified in confluent cell monolayers, following a calcium switch. Although it was only possible to analyze Rac1 activation at homophilic cell-cell junctions, the results were qualitatively similar to signaling triggered by adhesion to immobilized, recombinant ectodomains (supplementary material Fig. S1).
Dynamic fluorescence imaging monitored Rac1-GTP accumulation at nascent cell-cell junctions, by quantifying the localization of the YFP-PBD-PAK reporter for active Rac1 (or Cdc42) at junctions between WT C-CHO, following a calcium switch (supplementary material Movie 1). Although it was not feasible to test heterophilic interactions, the qualitative trends for homophilic cadherin ligation are similar to the Rac1 assays (Fig. 6B; supplementary material Fig. S1). The signal amplitude was greater and persisted longer at C-CHO junctions compared to N-CHO (Fig. 7A,B; supplementary material Fig. S2). Differences in YFP-PBD-PAK accumulation at these junctions were significant (P<0.05) at 40 minutes after calcium stimulation (Fig. 7). In addition to diminished signal amplitudes, N-CHO exhibited more extensive ruffling and increased motility (supplementary material Movie 2). Treatment with NSC 23766, which inhibits Rac1 without affecting Cdc42 or RhoA (Gao et al., 2004), abolished YFP-PBD-PAK localization (supplementary material Movie 3), confirming that YFP-PBD-PAK localization in these experiments reflects Rac1 rather than Cdc42 activity (Gao et al., 2004).
The K8NS10P mutant triggered greater YFP-PBD-PAK accumulation at junctions than the S78A mutant or N-CHO. At 40 minutes, the intensity was less than C-CHO (Fig. 7; supplementary material Fig. S2, Movie 4). By contrast, the S78A mutant exhibited only transient increases in junctional YFP-PBD-PAK (Fig. 7A; supplementary material Fig. S2, Movie 5). Rac1 activity extended throughout the intercellular junction, but the PBD-PAK-YFP localization did not persist over the time-lapse measurement. Signal amplitudes were significantly diminished (P<0.05) in the S78A mutant compared with C-CHO, 40 minutes after calcium stimulation (Fig. 7B). Both N-CHO and the S78A mutant displayed less stable junctions than C-CHO. Junction persistence was reduced or even abolished during observations, as some cell pairs separated from each other. Similarly, at 40 minutes, the Rac1 activity in the S78A mutant was not statistically different than N-CHO.
These aggregate dynamic imaging studies (Fig. 7; supplementary material Fig. S2) are qualitatively similar to results obtained with the Rac1 immuno-capture assay (Fig. 6B). The exception is the K8NS10P mutant, which is discussed below. The dynamic imaging also confirmed that cadherin junctions are loci of Rac1 activation.
Ligation-dependent RhoA activation
Quantitative RhoA-GTP levels in confluent cell monolayers, following a calcium switch, were compared with Rac1-GTP activated under similar conditions (supplementary material Fig. S3). Neither WT C-cadherin nor K8NS10P, which have similar affinities for CEC1-5-Fc, activated significant levels RhoA-GTP (supplementary material Fig. S3). By contrast, nascent N-CHO junctions increased RhoA-GTP 2.5-fold, 90 minutes after calcium addition. Interestingly, the S78A mutant activated similar RhoA-GTP levels to N-CHO, but with slower kinetics. This inverse correlation between Rac1 and RhoA agrees with reports suggesting that RhoA and Rac1 have antagonistic effects (Comunale et al., 2007; Wildenberg et al., 2006).
Magnetic twisting cytometry (MTC) measurements tested whether cadherin ligation might affect the global stiffness of these cells. Probing N-CHO and C-CHO with fibronectin-modified beads interrogated global changes in cytoskeletal tension through integrin bonds (Potard et al., 1997). The results were mixed. The expressed cadherin subtypes altered cell rigidity to different extents (supplementary material Fig. S4). The substratum ligand also influenced cell stiffness, but not in a way that compares simply to ligation-dependent GTPase activity. This suggests a more complex relationship between cadherin ligation and cell mechanics.
The main findings of this study are (1) the identification of a C-cadherin binding site mutation that substantially attenuates the affinity and correspondingly switches cell aggregation specificity, (2) the demonstration that, above an apparent threshold difference in cadherin affinities, in vitro cell sorting correlates with quantitative differences between two-dimensional affinities of cell surface cadherins, and (3) cadherin affinity differences modulate GTPase signaling.
Despite the highly conserved W2 binding site, these kinetic measurements also demonstrate that small sequence differences in the binding pocket can generate substantial affinity differences between cadherin subtypes. The switch in cell sorting specificity by the C-cadherin point mutation S78A correlates with a six-fold decrease in the two-dimensional affinity for WT C-cadherin. By contrast, the modest 2.5-fold decrease in the M92I affinity caused cells to intermix with both WT C-CHO and WT N-CHO. These results suggest that relatively large differences in two-dimensional binding affinities for the same overall protein scaffold, e.g. C-cadherin, correlate with CHO cell sorting in vitro.
The affinity appears to be particularly sensitive to sequence variations at positions 78 and 83; namely, S78A substantially reduced the C-cadherin affinity, A78M ablated N-cadherin adhesive function (Tamura et al., 1998), and mutations at positions 78 and 83 in human E-cadherin altered the in vitro sorting patterns of L1 cells expressing the proteins (Nose et al., 1990). The A78M mutation also allosterically altered epitope accessibility on EC1 (Harrison et al., 2005). Distinct from earlier reports, this study quantified the biophysical differences of the membrane-bound proteins that altered cell adhesion and sorting.
The K8NS10P mutation did not significantly alter the two-dimensional affinity for CEC1-5-Fc, Rac1 activation at 45 minutes, or the RhoA-GTP levels at 45 and 90 minutes. However, there are differences between Rac1 activation following CEC1-5-Fc ligation (Fig. 6B) versus junction activation following a calcium switch (Fig. 7A; supplementary material Figs S1,S2). This difference is attributed to the ligands used in the two assays. In the adhesion assay (Fig. 6B), one protein ligand, or half of the strand dimer, is wild-type C-cadherin, but in the calcium switch assays, both proteins are identical. Prior biophysical measurements of the W2A mutant showed that adhesion between WT C-cadherin and the W2A mutant is intermediate between homophilic WT C-cadherin and homophilic W2A adhesion (Prakasam et al., 2006a). We would thus expect the two-dimensional affinities between CEC1-5-Fc and the mutants to be somewhat higher than between identical mutants. These Rac1 assays are consistent with this trend, and suggest that, although K8NS10P does not detectably alter the affinity for CEC1-5-Fc, it may lower the affinity between identical mutants with a corresponding change in Rac1 signaling.
Prior biophysical studies identified multiple bonds between cadherin ectodomains that require different domains (Bayas et al., 2006; Bibert et al., 2002; Chappuis-Flament et al., 2001; Chien et al., 2008; Perret et al., 2004; Shi et al., 2008; Sivasankar et al., 1999; Sivasankar et al., 2001; Tsukasaki et al., 2007; Zhu et al., 2003). All five EC domains are needed to recapitulate the kinetic signature of the intact protein (Chien et al., 2008). The fast first step requires EC1, but the lag and second step require EC3 (Chien et al., 2008) and are modulated by N-glycosylation on EC2 and EC3 (Langer et al., 2012). These studies therefore focused on the first binding step between EC1 domains. Our results show that two-dimensional affinity differences associated with EC1-dependent binding correlate with in vitro sorting and signaling. Thus, only the initial, fast step was considered here.
The logarithm of these two-dimensional affinities are proportional to the protein adhesion energies (Prakasam et al., 2006b). However, in micropipette measurements (1) the affinities and kinetic rates are not determined by mechanically breaking cadherin bonds, and (2) micropipette measurements probe the full-length cadherins in the native environment of the cell membrane. The micropipette data demonstrate that cadherins with similar two-dimensional affinities (< ∼3-fold differences), and correspondingly similar adhesion energies do not induce cell segregation, at the expression levels considered. This agrees with prior studies (Prakasam et al., 2006b; Shi et al., 2008), and could explain the absence of correlations between adhesion and sorting in some cases (Duguay et al., 2003; Niessen and Gumbiner, 2002; Prakasam et al., 2006b).
Importantly, a 3-fold difference in the three-dimensional (solution) affinity corresponds to a difference in bond energies of ∼1 kcal/mole at 37°C that could be readily offset by cadherin expression levels or cortical tension (Duguay et al., 2003; Kalantarian et al., 2009; Krieg et al., 2008; Manning et al., 2010; Steinberg and Takeichi, 1994; Winklbauer, 2009). The capacity of cadherin affinities to regulate actin organization through GTPase signaling introduces an additional mechanism that could augment or offset differential cadherin adhesion.
The two-dimensional C-cadherin affinities correlate with ligation-dependent Rac1 activation. Thus, binding differences due to mutations or binding to heterophilic ligands would similarly modulate signaling. Because we compared proteins with the same overall backbone and cytoplasmic domain, this correlation might not apply for general comparisons across cadherin subtypes due to possible differences in their interactions with GTPases (Anastasiadis et al., 2000; Boulter et al., 2010). C-cadherin and E-cadherin activate Rac1 (Noren et al., 2003; Yap and Kovacs, 2003), but N-cadherin ligation triggers RhoA activation (Charrasse et al., 2002; Comunale et al., 2007; Marrs et al., 2009; Taulet et al., 2009). Whether the latter is due to low N-cadherin affinity and Rac1/RhoA antagonism (Boulter et al., 2010; Burridge and Doughman, 2006; Comunale et al., 2007; Wildenberg et al., 2006), or to differences in GTPase interactions with N-cadherin complexes (Anastasiadis et al., 2000) remains to be determined.
The two-dimensional affinity measurements were determined from data acquired within 45 seconds of cell-cell contact, while the signaling and sorting assays range from 45 minutes to 48 hours, respectively. Importantly, the measured affinities reflect intrinsic, equilibrium (time-independent) properties of cadherin bonds. This is implicit in Eqn 2 where Ka is defined as the two-dimensional equilibrium binding-constant. Solution binding constants obtained from kinetic, e.g. surface plasmon resonance measurements are similarly intrinsic properties, and are similarly compared with cell sorting on timescales of 4 and 48 hours (Katsamba et al., 2009; Takeichi and Nakagawa, 2001). The same physical chemical justification for such comparisons applies here.
Given the role of cadherin-dependent GTPase activation in cytoskeletal regulation (Takaishi et al., 1997) and cell cycle control (Liu et al., 2006), this connection between cadherin affinities and GTPase signaling is expected to have broader consequences for cadherin-dependent cell functions. Compromised cadherin adhesion is associated with cancer. For example, the E-cadherin exon 8 deletion is associated with human gastric and breast cancers, and it reduces Rac1 signaling, with a corresponding increase in RhoA activity (Deplazes et al., 2009). Our findings indicate that differential cadherin binding affinities play more diverse physiological and mechanical roles than merely modulating cell cohesion.
Materials and Methods
Plasmids and cell lines
The cDNAs for the full length Xenopus C-cadherin and the C-cadherin W2A mutant in pEE14 plasmids were gifts from B. Gumbiner (University of Virginia, Charlottesville, VA). The cDNA encoding the full-length chicken N-cadherin in the pEGFP-N1 plasmid was from Andre Sobel (Institut du Fer a Moulin, Gif-sur-Yvette, France). These plasmids were transfected into Chinese Hamster Ovary (CHO-K1) cells using Lipofectamine2000 (Invitrogen, Carlsbad, CA). CHO-K1 cells stably expressing the full length C-cadherin were selected as described (Brieher et al., 1996). CHO-K1 cells expressing the full-length chicken N-cadherin were cultured in Dulbecco’s Modified Eagle Medium (DMEM) containing 10 v/v% FBS and 400 µg/ml G418 (Sigma-Aldrich, St Louis, MO). For live-cell imaging experiments, CHO cell lines were transfected with a plasmid encoding YFP-PBD-PAK (from Y. Wang, University of Illinois, Urbana, IL) using Fugene 6 (Roche, IN). The YFP-PBD-PAK construct contains the YFP-tagged p21 binding domain (PBD) of p21-activated kinase (PAK). At 12 hours after transfection, cells were plated at low confluence on 35 mm glass bottom dishes (Cell E&G) coated with 20 µg/ml fibronectin (Sigma).
Biotinylation and western blot
Cadherin surface expression was assessed with a standard biotinylation assay. Cells were detached from substrates with 0.01 v/v% trypsin in HBSS containing 2 mM CaCl2. The cells were washed, resuspended in PBS. Then 0.5 ml were treated with 0.5 ml of 1 mg/ml Sulfo-NHS-SS-biotin (Pierce) for 25 minutes at 4°C, and then washed with PBS. Prior to cell lysis, 50 µl of streptavidin-coated beads were washed twice with ice-cold lysis buffer (50 mM Tris-HCl, 10 mM MgCl2, 200 mM NaCl, 1 w/v% Triton X-100, 5 v/v% glycerol, and Roche complete protease inhibitors at pH 7.5) and pelleted by centrifugation at 12,000 g. The beads were stored on the ice prior to adding 1 ml of lysate. The biotinylated cells were lysed, and the lysate was clarified by centrifugation at 4°C. Then 1 ml of supernatant was incubated with streptavidin beads (Sigma) and overnight at 4°C. The beads were then washed with ice-cold lysis buffer, and centrifuged for 1 minute at 12,000 g and 4°C. The collected beads were boiled in SDS-PAGE buffer, and the proteins were separated by SDS-PAGE. In western blots, the primary antibody used was a monoclonal, mouse anti-E-cadherin antibody against the conserved cytoplasmic domain (BD Transduction, Clone 36/E-cadherin). The secondary antibody was horseradish peroxidase conjugated, polyclonal anti mouse IgG (Sigma).
FACS quantification of cadherin surface expression levels
Cadherin surface expression levels were quantified by flow cytometry (Chesla et al., 1998; Chien et al., 2008). Cells were labeled with protein-specific antibodies against the ectodomains. C-cadherin expressing cells were labeled with anti-C-cadherin antibody (C/EP/B-Cadherin (clone xC-12), Santa Cruz Biotechnology, Santa Cruz, CA) followed by the secondary fluorescein isothiocyanate (FITC)-conjugated anti-goat IgG (whole molecule; Sigma). Chicken N-cadherin was detected with monoclonal mouse anti-N-cadherin (Clone GC-4, Sigma) and then fluorescein-isothiocyanate (FITC)-conjugated anti-mouse IgG (whole molecule; Sigma). The antibody labeling was in phosphate buffered saline (PBS) containing 1 w/v% bovine serum albumin (BSA) at pH 7.4. The fluorescence intensities of labeled cells were measured with an LSR II flow cytometer (BD Biosciences) (Zhang et al., 2005). The fluorescence intensity calibration curve was obtained with calibrated FITC-labeled standard beads (Bangs Laboratories, Fishers, IN) (Zhang et al., 2005).
Hanging-drop sorting assay
Cell sorting measurements used the hanging-drop method (Foty and Steinberg, 2005). Briefly, cells with cadherin expression levels within 15% of each other were labeled with DiI or DiO (Molecular Probes, Eugene, OR) by incubating 80% confluent monolayers with growth medium supplemented with 5 µl/ml of the appropriate dye for 60 minutes. After washing to remove excess dye, cells were detached with 0.01% in trypsin in Hank’s Balanced Salt Solution (HBSS) (Invitrogen, Carlsbad, CA) supplemented with 2 mM CaCl2 (Nose et al., 1988). Cells were resuspended in HBSS with 2 mM CaCl2 and 5 v/v% FBS at 1×106 cell/ml. A 10 µl aliquot of each of the two cell suspensions was mixed on the lid of a 10 cm Petri dish, inverted over a dish containing 10 ml of PBS (Invitrogen, Carlsbad, CA). After incubating the hanging drops in an incubator at 37°C under 5% CO2, cell aggregates were imaged after 24 and 48 hours under a 10× objective with a Zeiss Axiovert 200 inverted fluorescence microscope equipped with a Zeiss Axiocam MR camera. To quantify sorting, at least 50 aggregates of three or more cells were scored as containing red cells, green cells, or both red and green cells (Niessen and Gumbiner, 2002).
Surface modification of erythrocytes with oriented cadherin extracellular domains
Erythrocytes were isolated from human whole blood collected from healthy donors. The whole blood was stored in Vacutainers, and proper protocols were followed for handling human-derived materials. The erythrocytes were isolated with Histopaque 1119 (Sigma). To 12 ml of Histopaque 1119 in a 50 ml centrifuge tube, a mixture of 7 ml whole blood and 7 ml of 0.9 w/v% NaCl was slowly transferred to the tube containing Histopaque. The mixture was centrifuged at 800 g for 20 minutes at room temperature. The supernatant was discarded as biological waste, and the remaining cells were resuspended in 7 ml of 0.9 w/v% NaCl prior to the addition of 1.5 ml of 6 w/v% Dextran. The cells were incubated at 23°C for 45 minutes, during which they settled to the bottom of the tube. After discarding the supernatant, the red blood cells (RBC) were washed twice at room temperature with 0.9 w/v% NaCl, and resuspended in 12 ml EAS45 (2.0 mM adenine, 110 mM dextrose, 55 mM mannitol, 50 mM NaCl, 10 mM glutamine and 20 mM Na2HPO4, at pH 8.0) (Dumaswala et al., 1996). The RBC suspension in EAS45 can be stored at 4°C up to 3 weeks, after which the RBCs are treated with bleach and discarded.
Antibodies were covalently coupled to the RBCs using the CrCl3 coupling method (Gold and Fudenberg, 1967; Kofler and Wick, 1977). Approximately 106 RBCs were washed with 0.85 w/v% NaCl, and resuspended in 250 µl of 0.85% NaCl with 1 µg of either goat polyclonal anti-human immunoglobulin G (IgG) Fc or goat polyclonal anti-mouse IgG Fc antibodies (Sigma). The CrCl3 solution was diluted to below 0.01 w/v% with 0.02 mM sodium acetate containing 0.85 w/v% NaCl. A 250 µl aliquot of CrCl3 solution was mixed with 250 µl of the RBC/antibody mixture and incubated at 23°C for 5 minutes. The reaction was stopped with 500 µl of PBS with 0.5 mM EDTA and 1% BSA. The cells were then washed twice. The concentration of CrCl3 determined the density of antibodies immobilized to the surface of the RBCs.
Micropipette measurements of cell binding kinetics
The binding probability was determined as a function of contact time with the micropipette manipulation technique (Chesla et al., 1998; Chien et al., 2008; Evans et al., 2004; Huang et al., 2010; Zhang et al., 2005). The binding probability P(t) is the ratio of the number of binding events nb to the total NT cell-cell touches, nb/NT. A cadherin-expressing CHO cell and a RBC modified with Fc-tagged cadherin were partially aspirated into opposing micropipettes (Fig. 3). The cells were maintained in the chamber with L15 medium (Invitrogen, Carlsbad, CA) supplemented with 1 w/v% BSA. Cells were visualized with a 100× oil-immersion objective on a Zeiss Axiovert 200 microscope, and images were recorded with a DAGE-MTI CCD100 CCD camera (DAGE-MTI, Michigan City, IN). Automated piezo-electric controllers were programmed to cyclically bring the two cells into contact for a defined period. The contact area was controlled at ∼3 µm2 (∼1 µm diameter). Adhesion events are identified from the surface deformation of the RBC during separation and recoil at adhesive failure. Each cell pair was tested for 50 cell-cell touches (NT = 50), and each contact time represents measurements with at least three different cell pairs (N>150). The reported probabilities P are the mean ± s.d. from the mean.
Data fitting and parameter estimation
Rac1 activation assay
Rac1-GTP was determined with a published Rac1-GTP immuno pull-down assay (Benard and Bokoch, 2002; Noren et al., 2003). Cells expressing cadherins at ∼20/µm2 were seeded onto substrata coated with cadherin ectodomains. CEC1-5-Fc and NEC1-5-Fc surface densities, determined by isotope labeling (Yeung et al., 1999), were 5×103 molecules/µm2 and 4×103 molecules/µm2, respectively. To prepare these substrata, 10 cm, non tissue culture polystyrene plates (Fisher Scientific, Pittsburgh, PA) were incubated with 5 ml of 30 µg/ml cadherin EC1-5-Fc in HEPES buffer (20 mM HEPES, 150 mM NaCl, 5 mM CaCl2, 1 mM MgCl2, pH 7.5) for 1 hour at 23°C, and stored overnight at 4°C before use. Controls used plates coated with poly-L-lysine (PLL). At least 30 minutes prior to use, the coated plates were washed with HEPES buffer, and equilibrated with serum-free, phenol red-free DMEM at 37°C.
Cells were maintained at confluence for 2 days before the experiment, detached from tissue culture plates with 0.01% trypsin in 1× HBSS, supplemented with 1 mM CaCl2 (Nose, 1988), and then collected by centrifugation. Cells were then re-suspended in serum-free DMEM prior to seeding at 3–4×106 cells on the coated dishes. At defined intervals, the plates were washed twice with ice-cold HEPES buffer and lysed with 750 µl ice-cold lysis buffer per plate. Cells were removed with a cell scraper, and the lysate was clarified by centrifugation at 14,000 g for 2 minutes at 4°C. Then 20 µl of the clarified supernatant was analyzed for total Rac1 by western blot, with anti-Rac1 antibody (Cytoskeleton, Denver, CO).
The remaining lysate was added to 30 µl of GST-PBD beads. Just prior to use, the beads were washed with ice-cold lysis buffer (50 mM Tris-HCl, 10 mM MgCl2, 200 mM NaCl, 1% v/v Nonidet P-40, 5% v/v glycerol, and Roche complete protease inhibitors at pH 7.5) for 10 minutes with gentle shaking at 4°C. The beads were collected by centrifugation and stored on ice.
Each time point required 30 µl beads. After mixing the lysis buffer and beads, the GST-PBD bead slurry was centrifuged and washed three times with ice-cold lysis buffer. After the final wash, the beads were collected by centrifugation, and boiled in SDS-PAGE buffer. Western blots with anti-Rac1 antibody (Cytoskeleton Inc., Denver, CO) determined the amount of Rac1-GTP in the lysate. The Rac1-GTP was normalized by the Rac1-GTP at t = 0 minutes, defined by cells in suspension, and compared to the total Rac1 in the cells and to the actin loading control. Rac1-GTP levels showed a robust increase at 45 minutes (Fig. 6), so Rac1-GTP was determined 45 minutes after seeding cells on different substrata.
Rac1 activation at junctions following a calcium switch
Rac1-GTP activation at cell-cell junctions was determined with the Rac1-GTP immuno pull-down assay described above. The cells were maintained at confluence for 2 days. Then, 12 hours before the experiment, cells were incubated in medium with 0.05 v/v% FBS and no calcium. After 12 hours, the medium in one flask was switched to DMEM containing 0.05 v/v% FBS and 1.8 mM CaCl2, and incubated for 45 minutes at 37°C. Cells in the second flask (t = 0 minutes) were washed twice with ice-cold HEPES buffer, lysed with 1500 µl ice-cold lysis buffer, and removed with a cell scraper. This step was repeated 45 minutes after the addition of calcium. The relative amounts of total Rac1 and Rac1-GTP in the cell lysates were determined as described above. Actin was used as a loading control.
Cortical tension measurement
Differences in cortical tension were determined by quantifying the global cell stiffness with Magnetic Twisting Cytometry (MTC) and magnetic beads covalently modified with fibronectin (Wang et al., 1993). Glass-bottom Petri dishes were incubated overnight with an anti-immunoglobulin Fc antibody, followed by rinsing and incubation with 0.5 mg of either CEC1-5-Fc or NEC1-5-Fc for 4 hours at 4°C. After rinsing with HEPES buffer, the surfaces were incubated with 1 w/v% BSA at room temperature for 30 minutes. Ferromagnetic beads (4.9 µm; Spherotech), chemically activated with ethyl-3-(dimethylaminopropyl)-carbodiimide and N-hydroxysuccinimide (Prakasam et al., 2006a; Prakasam et al., 2006b), were covalently modified with fibronectin.
Cells stably expressing different cadherins were grown to confluence, detached with PBS containing 3.5 mM EDTA and 1w/v% BSA, collected by centrifugation, and seeded at low density on the cadherin-Fc coated substrata, in medium supplemented with 0.05 v/v% FBS. These measurements focused on isolated cells to eliminate interference from cell-cell adhesion. After 4 hours at 37°C, fibronectin-coated beads were incubated with the cells for 20 minutes. All MTC measurements were performed on an inverted microscope (Leica) using a 20× objective and a cooled charge-coupled device camera (Orca2; Hamamatsu Photonics). After initial bead magnetization, an oscillating magnetic field perpendicular to the bead magnetic moment was applied for a defined period. The bead magnetic moment constant of 0.12 Pa/Gauss was calibrated as described (Wang et al., 1993). The bead displacements were measured and converted to the complex modulus (Wang et al., 1993).
RhoA activation assay
Ligation-activated RhoA-GTP was quantified following cadherin activation with a calcium switch with a commercial G-LISA kit (BK 124; Cytoskeleton, Denver, CO). Confluent cells in 10 mm Petri dishes were serum starved for 24 hours, before addition of 4 mM EGTA, which disrupts cell-cell junctions (Noren et al., 2003). After a 1 hour incubation with EGTA, the medium was exchanged with medium containing 1.8 mM calcium, but lacking serum. Duplicate samples were analyzed at t = 0, 45 and 90 minutes. After specific time periods, plates were immediately treated with ice-cold lysis buffer, as described for Rac1 assays. Cell lysates were scraped into tubes and clarified by centrifugation, before snap-freezing in liquid nitrogen. Lysates at t = 0 minutes were prepared by re-suspending cells in lysis buffer.
The protein concentration in the cell lysates was normalized to the t = 0 minute sample, and equal amounts of total protein were incubated in duplicate in a 96-well GLISA assay plate. RhoA-GTP levels were determined with a microplate spectrophotometer (Molecular Devices SpectraMax M2). The change in RhoA-GTP was normalized relative to the corresponding zero time-point samples.
Dynamic fluorescence imaging
Prior to imaging, cell-cell junctions were disrupted by incubation for 4–6 hours at 37°C with calcium-free culture medium without phenol red (Braga, 2002; Noren et al., 2003). Live cell imaging was performed at 37°C with a Nikon Eclipse Ti inverted microscope equipped with a cooled charge-coupled device camera (QuantEM 512SC; Photometrics) using MetaFluor 6.2 software (Universal Imaging). Time-lapse images were acquired every 1–2 minutes using a 40× or 100× oil objective lens (Nikon). When switching to the calcium-containing medium, imaging was paused to allow replacement of calcium-free medium with standard medium containing 1.8 mM calcium without phenol red. To inhibit Rac1, cells were incubated with 50 µM of sterile NSC 23766 (Tocris Bioscience, Ellisville, MO) for 12 hours at 37°C (Gao et al., 2004). The Rac1 inhibitor was added to the calcium-containing medium used for the calcium switch imaging experiments.
Image processing and the creation of time-lapse movies were done with MetaMorph software (Universal Imaging). In order to quantify YFP-PBD-PAK at cell-cell junctions (Deplazes et al., 2009), the background-subtracted mean fluorescence intensity (MFI) of YFP within the maximal area of a single junction was recorded using ImageJ64 software (National Institutes of Health, USA). Only two cells in contact were examined so that a single cell-cell junction could be clearly defined for analysis. In order to correct for differences in transfection efficiency, each trace was normalized to the average basal fluorescence intensity before calcium stimulation. The normalized YFP intensity at cell-cell junctions was plotted against time using the spline interpolation feature, and curves were smoothed by loess, provided in MatLab software (Mathworks). The standard error of the interpolated mean value was calculated using the Excel function, and a two-tailed Student’s t-test determined whether detected differences were statistically significant.
We thank Saiko Rosenberger and Dr Sandy McMasters for technical assistance; Shaoying Lu for assistance with image analysis; and Yingxiao Wang for discussions, reagents, and microscope and software usage.
This work was supported by the National Science Foundation [grant number CBET 0853705]; the National Institutes of Health [grant numbers R21 HD059002 and HL098472]; and the American Heart Association [grant number 10PRE3840004 to A.K.B.]. Deposited in PMC for release after 12 months.