Turn-off, drop-out: Functional state switching of cadherins



The classic cadherins are a group of calcium dependent, homophilic cell–cell adhesion molecules that drive morphogenetic rearrangements and maintain the integrity of cell groups through the formation of adherens junctions. The formation and maintenance of cadherin-mediated adhesions is a multistep process and mechanisms have evolved to regulate each step. This suggests that functional state switching plays an important role in development. Among the many challenges ahead is to determine the developmental role that functional state switching plays in tissue morphogenesis and to define the roles of each of the several regulatory interactions that participate in switching. One correlate of the loss of cadherin-mediated adhesion, the “turn-off” of cadherin function, is the exit, or “drop-out” of cells from neural and epithelial layers and their conversion to a motile phenotype. We suggest that epithelial mesenchymal conversions may be initiated by signaling pathways that result in the loss of cadherin function. Tyrosine phosphorylation of β-catenin is one such mechanism. Enhanced phosphorylation of tyrosine residues on β-catenin is almost invariably associated with loss of the cadherin-actin connection concomitant with loss of adhesive function. There are several tyrosine kinases and phosphatases that have been shown to have the potential to alter the phosphorylation state of β-catenin and thus the function of cadherins. Our laboratory has focused on the role of the nonreceptor tyrosine phosphatase PTP1B in regulating the phosphorylation of β-catenin on tyrosine residues. Our data suggest that PTP1B is crucial for maintenance of N-cadherin-mediated adhesions in embryonic neural retina cells. By using an L-cell model system constitutively expressing N-cadherin, we have worked out many of the molecular interactions essential for this regulatory interaction. Extracellular cues that bias this critical regulatory interaction toward increased phosphorylation of β-catenin may be a critical component of many developmental events. © 2002 Wiley-Liss, Inc.


The principle players, the Cadherins, are an ancient family of adhesion molecules, possibly even represented in the social amoeba of Dictyostelium (Wong et al., 1996; see also Grimson et al., 2000), and in the primitive metazoan Hydra (Hobmayer et al., 1996). The extent and diversity of this family has only recently been fully appreciated with the completion of the C. elegans, Drosophila, and human genomes, (Hynes and Zhao, 2000; Nollet et al., 2000; Hill et al., 2001). Members of the cadherin family are divided in subfamilies, based on structural differences. Several excellent articles have recently appeared reviewing the structure and function of this superfamily (Tepass et al., 2000; Yagi and Takeichi, 2000; Angst et al., 2001). The so-called type I cadherins are the most thoroughly studied, and are the group to which E- and N-cadherin (cadherins 1 and 2) belong, the two cadherins that we will focus on. These two cadherins establish stable adhesions between the cells that comprise epithelial and neuroepithelial layers. The emigration of cells from epithelial layers and conversion to a motile phenotype—epithelial mesenchymal transformations—is correlated with down-regulation of cadherin itself (Burdsal et al., 1993; Nakagawa and Takeichi, 1998). This has been more thoroughly analyzed during the transition of tumors to an invasive, motile phenotype. In this case, conversion to an invasive state has also been correlated with loss of cadherin function, as distinct from loss of cadherin itself (reviewed in: Birchmeier et al., 1996; Bracke et al., 1996; Hirohashi, 1998). However, pertinent data on cadherin function during normal epithelial mesenchymal transformations is lacking. Despite the scarcity of hard data, it is a reasonable hypothesis that loss of adhesive function plays a role in some, or all, epithelial mesenchymal transformations. In fact, several mechanisms appear to have evolved to regulate cadherin function independent of transcriptional regulation.

The extracellular domain of cadherins forms the intercellular bonds—cadherins on one cell binding to similar or identical cadherins on an adjacent cell—and determines the specificity of the interaction (Nose et al., 1988, 1990). However, the adhesive function of extracellular domain on its own is compromised and the cytoplasmic tail is essential for strong, stable cadherin-mediated adhesions (Nagafuchi and Takeichi, 1988; Ozawa et al., 1990; Kintner, 1992). The importance of the cytoplasmic domain is emphasized by the fact that the sequence similarity in this region is greater than the sequence similarity of the extracellular domain: 43% vs. 70% for N- and E-cadherin. Even these numbers do not do justice to the degree of conservation in the cytoplasmic domain, as stretches of 90–100% sequence identity are interspersed with nonsimilar regions. The presence of these highly conserved regions suggests that these are sequences where common adaptors and effectors bind and that the signal generation and regulatory functions of this group of cadherins has been highly conserved.


Regulatory Interactions at the Cytoplasmic Domain of Type I Cadherins

The discovery of most of the participants and interactions among regulatory components at the cytoplasmic domain of type I cadherins has been based on biochemical and molecular characterization of interacting molecules, and on recapitulation of the formation of adhesive bonds among single cells in vitro by using mutant proteins or in the presence of perturbational agents. This is clearly not what is going on in an embryo. Thus we know a great deal about the potential molecular and biochemical interactions that regulate cadherin function, without understanding in any detail at all the role such regulatory interactions play in tissue morphogenesis.

The formation of cadherin-mediated adhesions is thought to involve several steps: the formation of cis dimers, association with the actin-containing cytoskeleton, and actin-driven clustering (Adams and Nelson, 1998). The detailed mechanisms underlying each of these steps, as well as the mechanistic relationship between each step, remain to be worked out; however, many of the players have been identified. The protein-protein interactions involved in each step are points of potential regulation of cadherin function. Thus cis cadherin interaction, the protein interactions essential for connection with the actin cytoskeleton—cadherin-β-catenin, β-catenin-α-catenin, α-catenin-actin—as well as the regulated polymerization of actin are points of potential regulation (Fig. 1), switching cadherin between a functional and a nonfunctional state.

Figure 1.

Points at which cadherin function may be regulated. Shown are the basal set of cadherin associated proteins and those proteins mediating attachment to the actin-containing cytoskeleton. Asterisks indicate sites of potential regulation. Note that cadherin dimerization and actin assembly, as well as the interaction of effector proteins with cadherin are important potential sites for regulation of cadherin function.

The focus of this mini-review is on the role of tyrosine phosphorylation of β-catenin as a means of epigenetic regulation of the association of β-catenin with cadherin, and therefore, the association of cadherin with the cytoskeleton and thus function. However, other important, nontranscriptional regulatory interactions that control cadherin function also occur and some idea of the multiplicity of interactions adds perspective to the more detailed presentation of a subset of these interactions.

The catenin p120 has emerged as a key intermediate regulating cadherin function, and may be involved in several distinct steps (reviewed in Anastasiadis and Reynolds, 2000). The juxtamembrane cadherin sequence to which p120ctn binds (Thoreson et al., 2000) is within a region required for cis clustering (Yap et al., 1998) and is adjacent to a sequence that has the potential to directly dimerize (Arregui et al., 2000; see Fig. 2). The proximity of these domains suggests that p120ctn has the potential to regulate dimerization and/or lateral clustering. Indeed, in cotransfectants expressing both N- and VE-cadherin, p120ctn preferentially associates with VE-cadherin and N-cadherin is excluded from areas of cell–cell contact (Navarro et al., 1998). This suggests that in this case p120ctn may be a positive regulator of cis associations. Additionally, phosphorylation on either tyrosine or serine/threonine residues may determine whether p120ctn is a positive or negative regulator of cadherin-mediated adhesion (reviewed in Anastasiadis and Reynolds, 2000; see also Aono et al., 1999; Ozawa and Ohkubo, 2000; Zondag et al., 2000).

Figure 2.

Known effector-binding sites on the cytoplasmic domain of cadherin. The target regions in cytoplasmic domain of cadherin for interacting proteins are shown as colored boxes. The position of two peptides (JMP and CBP; Arregui et al., 2000) that perturb cadherin function is shown as a bold line above the sequence. Additional features such as the dileucine basolateral targeting signal (Miranda et al., 2001) and the ubiquitin-proteasome targeting signal (Huber et al., 2001) are underlined. For the purposes of this figure the sequence used is that of chicken N-cadherin.

p120ctn also alters the relative activities of the small G-proteins Rho A, Rac, and Cdc42 (Anastasiadis et al., 2000; Grosheva et al., 2000; Noren et al., 2000). This may well be related to changes in the activities of these proteins initiated by adhesion (Nakagawa et al., 2001; Noren et al., 2001) and involved in actin-driven cadherin clustering (Braga et al., 1997, 1999; Jou and Nelson, 1998; Takaishi et al., 1997).

The function of p120ctn may further be modulated by presenilin through direct competition with p120ctn for binding to cadherin (Baki et al., 2001; see Fig. 2). Presenilin also stimulates the interaction of E-cadherin with β-catenin and the cytoskeleton; the ultimate effect is to enhance adhesion (Baki et al., 2001; Georgakopoulos et al., 1999). The relationship between competition for binding of p120ctn, enhanced binding of β-catenin to cadherin, and the interaction of presenilin with the cytosolic β-catenin complex (Murayama et al., 1998; Yu et al., 1998) remains to be clarified. This complex web of p120ctn interactions and their ramifications, as well as its potential role in transcription (Daniel and Reynolds, 1999; Prokhortchouk et al., 2001), suggests that, like β-catenin, p120ctn is a key regulatory molecule integrating cell adhesion and transcription during development.

The interaction of β-catenin with α-catenin appears to be regulated by the small GTPase activating protein IQGAP (reviewed in Kuroda et al., 1999; see also Kuroda et al., 1998; Fukata et al., 1999, 2001; Li et al., 1999). It is interesting to note that the stability of this interaction has also been correlated with tyrosine phosphorylation of β-catenin (see below; Ozawa and Kemler, 1998).

The association of β-catenin with the cytoplasmic domain of cadherin has recently been shown to be regulated by the GTP-bound form of Gα subunit of heterotrimeric G proteins. Gα12/13 associates with the cytoplasmic domain of E-, N-, and cadherin 14 (Meigs et al., 2001; Kaplan et al., 2001; see Fig. 2) and overexpression of the constitutively active versions of these proteins in HEK cells results in dissociation of β-catenin from cadherin. Coupled with the observation that a constitutively active form of Gαq disrupts ectodermal cell adhesion during late gastrulation in Xenopus embryos (Rizzoti et al., 1998), these observations suggest an additional means of functionally regulating the association of cadherin with the cytoskeleton and thus function. It is interesting to note that the target site for Gα on the cytoplasmic domain overlaps considerably with the target site for PTP1B, thus binding of Gα may displace PTP1B and vice-versa (see below and Fig. 2).

The multiple interactions that occur at the cytoplasmic domain of possibly all type I cadherins are mapped to their respective sequence targets in Figure 2. Several points immediately emerge from this figure: First, the extent of a binding domain varies depending on the methods used for mapping; this is most apparent for β-catenin. Second, the overlapping nature of many of the targets emphasizes the notion that each of these regulatory interactions is most likely brought to bear in a specific tissue, and/or under a special set of circumstances. Third, like the well-characterized role of β1-integrins, the type I cadherins are a nexus, transducing signals in both directions, from the oustside in, and the inside out. Fourth, the proximity and overlapping nature of effector target sites makes functional analysis of deletion mutants complex. This is particularly apparent in the relationship between the tyrosine phosphatase PTP1B, β-catenin and Gα12/13, and between presenilin and p120 catenin (see below and Fig. 3). Finally, putting this machinery in its proper developmental and physiological context presents a real challenge.

Figure 3.

β-catenin and PTP1B binding domains on N-cadherin and conservation of the PTP1B site. The core β-catenin binding sequence defined by deletions (Kaplan et al., 2001 and Simcha et al. 2001) and that defined by peptide competition (Arregui et al., 2000) are compared in Figure 2. In this figure we use the region defined by the peptide competitor. The overlapping target sites on N-cadherin for beta catenin and PTP1B is shown in green, and the corresponding unique binding sites are shown in blue (β-catenin) and yellow (PTP1B). The PTP1B binding domain is highly conserved among type I and type II cadherins. Identical residues are shown in blue and conserved substitutions in red.


Phosphorylation May be a Key Regulatory Step Coordinating and Integrating the Function of β-Catenin in Adhesion and Transcription

β-catenin is a key cellular regulatory element and occupies a central position in cell physiology and development as both a transcriptional co-activator regulated by Wnt signaling (reviewed in Peifer and Polakis, 2000; Seidensticker and Behrens, 2000), and as the bridge between the cytoplasmic domain of cadherins and the actin-containing cytoskeleton (reviewed in Wheelock et al., 1996). It is notable that in C. elegans each of these functions is mediated by a distinct β-catenin (Korswagan et al., 2000).

Phosphorylation of tyrosine residues on β-catenin has repeatedly been correlated with loss of cadherin adhesive function (see below). This, in turn, is correlated with instability of the β-catenin-cadherin bond (Roura et al., 1999), uncoupling cadherin from the actin-containing cytoskeleton, and an increase in the free cytosolic pool of tyrosine phosphorylated β-catenin (Kinch et al., 1995; Balsamo et al., 1996, 1998). Dephosphoryation of β-catenin also promotes reassembly of complexes containing E-cadherin, β-catenin and α-catenin (Hu et al., 2001). Direct in vitro measurements of the interaction of β-catenin with the cytoplasmic domain of cadherin reveal that phosphorylation on tyrosine residues does indeed reduce binding (Roura et al., 1999; Piedra et al., 2001). Interestingly, increased phosphorylation of β-catenin has also been correlated with loss of the α-catenin-β-catenin bond concomitant with loss of adhesion (Ozawa and Kemler, 1998), suggesting that interactions with α-catenin are also compromised by phosphorylation of β-catenin.

There is indeed a large body of literature consistent with, and supporting a role for β-catenin phosphorylation in regulating its interaction with cadherin and therefore, cadherin-mediated adhesion. There are, however, some exceptions. One key set of observations often referred to is based on the introduction of v-src into cells expressing a fusion of E-cadherin with α-catenin (Takeda et al., 1995). In these cells cadherin-mediated adhesions are converted from a “strong” to a “weak” state under conditions where β-catenin does not play a role in bridging between cadherin and α-catenin. This report has been used to suggest that loss or weakening of adhesions “cannot” be explained by tyrosine phosphorylation of β-catenin. This is clearly an overstatement; as briefly reviewed above, there are several pathways that affect the state of cadherin-mediated adhesions. This particular report predates the emergence of p120ctn as a key regulator of cadherin function, and the reduction in cadherin-mediated adhesion may well be due to v-src-mediated tyrosine phosphorylation of p120ctn, a notion consistent with the work of Ozawa and Ohkubo (2000). Given the reduced affinity of tyrosine-phosphorylated β-catenin for cadherin, it is difficult to explain other exceptions to the general rule (e.g., Calautti et al., 1998) other than through technical differences in measurement of adhesion and/or differences in the complexity of protein-protein interactions at de novo formed cadherin-mediated adhesions versus fully formed adherens junctions. Fully formed adherens junctions are indeed a complex web of many interacting components and may contain more than one system of cell–cell adhesion (Tachibana et al., 2000).

Tyrosine phosphorylation of β-catenin may have additional effects on cell function as it also increases the interaction of β-catenin with the TATA-box Binding Protein (TBP) and increases transcriptional activity of the β-catenin/Tcf complex (Piedra et al., 2001). Furthermore, the same tyrosine residue, 654, is critical for both instability of the β-catenin-cadherin bond and for enhanced binding to TBP (Roura et al., 1999, Piedra et al., 2001), suggesting that the two functions of β-catenin may be coordinated through the creation of a pool of free β-catenin after tyrosine phosphorylation and dissociation of the β-catenin-cadherin link, and a concomitant increase in the potential of this pool to participate in transcription. This suggests that under certain conditions regulated changes in cadherin-mediated adhesion through phosphorylation of β-catenin may precede transcriptional changes mediated by the β-catenin-Tcf/LEF complex.

It is notable that loss of β-catenin from the cytoplasmic domain of cadherin may have multiple ramifications on cadherin; besides loss of the cytoskeletal connection and thus immediate function, it leaves the cytoplasmic domain of cadherin in an unstructured state and potentially enhances degradation (Huber et al., 2001). Thus loss of function by means of disruption of the cadherin-cytoskeletal connection is immediate and long-term. Furthermore, this regulatory switch operates at an epigenetic level, that is, transcription plays no role, and as such, this mechanism has the potential to execute very rapid changes in cadherin function, and thus initiate rapid morphogenetic responses to environmental cues.

Maintaining Stable Adhesions: The Nonreceptor Tyrosine Kinase PTP1B Maintains β-Catenin in a Dephosphorylated State

We have used embryonic chick neural retina cells, PC12 cells, and L-cells constitutively expressing N-cadherin (LN-cells) as model systems to characterize the role of PTP1B in regulating tyrosine phosphorylation of β-catenin and thus, N-cadherin-mediated adhesion and neurite outgrowth. Introduction into LN-cells of a catalytically inactive, dominant-negative construct of PTP1B (C215S), which displaces endogenous PTP1B bound to N-cadherin (see below), results in hyperphosphorylation of β-catenin, an increase in the free cytosolic pool of tyrosine phosphorylated β-catenin, and dissociation of the cadherin-actin connection, concomitant with loss of cadherin function (Balsamo et al., 1996, 1998). This dominant-negative PTP1B construct also inhibits neurite extension by PC12 cells on N-cadherin substrates, as does down-regulation of PTP1B by means of antisense oligonucleotides (Pathre et al., 2001). Consistent with these observations, PTP1B is present at adherens junctions and localizes to growth cones (Balsamo et al., 1998; Pathre et al., 2001).

PTP1B associates directly with the cytoplasmic domain of N-cadherin, as does the dominant-negative construct (Balsamo et al., 1996, 1998; Rhee et al., 2001, Xu et al., 2002). We have recently determined the binding site on the cytoplasmic domain to which PTP1B is targeted (Xu et al., 2002). The site is adjacent to and partially overlaps the core-binding domain for β-catenin determined both by using in vitro binding assays (Kaplan et al., 2001; see Fig. 2) and in situ binding assays (Simcha et al., 2001; see Fig. 2). Despite the partial overlap of binding domains, β-catenin and PTP1B do not compete with each other for binding to the cytoplasmic domain of N-cadherin in vitro. This appears to be due to the ability of PTP1B to interact with the nonoverlapping portion of the target site; when the nonoverlapping portion of the target site is deleted, β-catenin does indeed compete for PTP1B binding (Xu et al., 2002). It is interesting to note that the target site for PTP1B also overlaps with the target site for Gα12 (Kaplan et al., 2001; see Fig. 2), potentially adding another layer of complexity to the regulation of cadherin function. The target binding domain for PTP1B in cadherin is highly conserved among both type I and type II cadherins (Fig. 3).

N-cadherin lacking the portion of the PTP1B binding domain that does not overlap with the core β-catenin binding target sequence (see Figs. 2, 3) fails to mediate adhesion among L-cells. Similarly, introduction into embryonic chick neural retina cells of a cell permeable, “Trojan” peptide bearing the sequence of the cadherin region deleted above abolishes N-cadherin-mediated adhesion and neurite outgrowth (Xu et al., 2002). In these cells, PTP1B and β-catenin fail to associate with N-cadherin, and N-cadherin fails to associate with the actin-containing cytoskeleton. Thus, PTP1B must be continually associated with the cytoplasmic domain to maintain the integrity of the cadherin complex.

The combined binding region for β-catenin and PTP1B contains 10 serines in N-cadherin (8 in E-cadherin, Fig. 3). In vitro phosphorylation of these serine residues enhances the binding of β-catenin, but has no effect on PTP1B binding (Xu et al., 2002). This is consistent with prior data indicating that serine phosphorylation enhances binding of β-catenin to cadherin (Stappert and Kemler, 1994; see also Huber and Weis, 2001). In this context it is interesting to note that casein kinase 2-mediated phosphorylation of E-cadherin results in an increase in the association of β-catenin with E-cadherin and “strengthens” cell–cell adhesions (Lickert et al., 2000) whereas a decrease in CK2-mediated phosphorylation is associated with disruption of E-cadherin based adherens junctions (Serres et al., 2000).

PTP1B is a ubiquitous enzyme present in many cellular locations and interacting with many different intracellular partners. Targeting is thus crucial for its proper functioning within cells. Interaction with N-cadherin depends on phosphorylation of tyrosine 152 in PTP1B (Rhee et al, 2001); alanine substitution of this residue abolishes the dominant-negative effect of the C215S mutant PTP1B and prevents localization at cell boundaries. Thus, an additional feature of this set of regulatory interactions is the potential to regulate the phosphorylation of PTP1B and, therefore, its association with cadherin and the content of phosphorylated tyrosine residues on β-catenin.

Targeting of PTP1B to other sites similarly depends on specific protein-protein interactions. Tyrosine 66 is the major target for phosphorylation of PTP1B by the insulin receptor, creating a site essential for downstream signaling (Bandyopadhyay et al., 1997). The carboxy terminus of PTP1B directs its localization to the cytosolic face of the endoplasmic reticulum (Frangioni et al., 1992). In platelets and activated T cells, proteolytic cleavage in the ER targeting domain results in translocation of PTP1B to the cytoskeletal/membrane fraction (Frangioni et al., 1993; Ezumi et al., 1995; Rock et al., 1997). This cleavage is dependent on integrin engagement, resulting in increased Ca2+ levels and, consequently, activation of calpain. We also find that PTP1B associated with N-cadherin in vivo migrates faster on SDS-PAGE than the intact ∼50 kD enzyme, suggesting cleavage (Balsamo et al., 1996, 1998). Elimination of the ER localization signal does not alter the interaction of PTP1B with N-cadherin, indicating that targeting of PTP1B to the N-cadherin complex does not depend on prior targeting to the ER. Furthermore, targeting to specific plasma membrane locations does not appear to depend on cleavage of the ER targeting sequence, as the PTP1B associated with focal adhesion complexes (Arregui et al., 1998) and the insulin receptor (Bandyopadhyay et al., 1997) has an apparent molecular mass of ∼50 kD, that of the intact protein. Interaction with focal adhesion complexes is most likely through interaction with p130cas and is mediated by a proline rich, SH3-binding domain in PTP1B (Liu et al., 1996). The role of PTP1B in focal adhesion complexes appears to be activation of Src (Arregui et al., 1998; Bjorge et al., 2000) and failure to activate Src leads to impaired integrin-mediated adhesion and signaling (Arregui et al., 1998). Even though PTP1B is a ubiquitous enzyme, it plays a pivotal role in regulating many cellular functions through specific targeting to distinct cellular locations.

It is interesting to note that tyrosine phosphorylation of PTP1B targets it to the cadherin complex, and therefore, favors stability of the adhesion complex whereas tyrosine phosphorylation of β-catenin favors instability of the complex. This suggests that the two phosphorylations are mediated by distinct kinases and/or that the two events are spatially and temporally distinct. Because PTP1B can dephosphorylate itself, under conditions where stable adhesions are maintained phosphorylation of PTP1B is either continuous, maintaining a population of PTP1B bound to cadherin, or its self-dephosphorylation must be inhibited. We have yet to distinguish these alternatives. In contrast, enhanced phosphorylation of tyrosine residues on β-catenin, either through activation of a kinase or inactivation of a phosphatase, most likely occurs only in response to specific extracellular cues initiating developmentally significant events. Figure 4 diagrammatically represents the role of PTP1B in regulating the stability of cadherin-mediated adhesion by altering the stability of the β-catenin-cadherin bond.

Figure 4.

Model showing the different players and functional relationships among tyrosine kinases and phosphatases that have the potential to regulate cadherin function. Protein tyrosine kinases (PTKs) activated by extracellular cues, for example neurocan acting through its N-acetylgalactosaminylphosphotransferase receptor (PTase), may lead to β-catenin tyrosine phosphorylation, β-catenin-cadherin uncoupling, and loss of cadherin function. PTP1B, which is targeted to the cytoplasmic domain of cadherin by tyrosine phosphorylation, constitutively dephosphorylates β-catenin thus, promoting cadherin function by stabilizing the cadherin-cytoskeletal connection. Other protein tyrosine phosphatases and kinases that may play a role in regulating the phosphotyrosine content of β-catenin are indicated.

Destabilizing Adhesions: Altering the Balance Between Phosphorylation and Dephosphorylation of β-Catenin

Regulation of the phosphotyrosine content of β-catenin is a function of the balance between the accessibility of β-catenin to tyrosine kinases and phosphatases and their activity. Several kinases and phosphatases have been shown to have the potential to alter the phosphotyrosine content of β-catenin and thus cadherin-mediated adhesion. This suggests that multiple kinases and phosphatases regulate cadherin function, possibly each playing a role in a specific tissue and at a specific time during development. Overexpression of the nonreceptor tyrosine kinases Src (Matsuyoshi et al., 1992; Hamaguchi et al., 1993; Behrens et al., 1993), and Fer (Rosato et al., 1998) has been demonstrated to increase tyrosine phosphorylation of β-catenin (and p120ctn). Furthermore, a dominant-negative Src that interferes with function, or a Src-specific tyrosine kinase inhibitor induce cell–cell adhesion (Owens et al., 2000), further suggesting that Src plays an in vivo role in regulating the phosphotyrosine content of β-catenin and that suppression of tyrosine phosphorylation of β-catenin promotes formation of cadherin-mediated adhesions. Two transmembrane tyrosine kinases, the EGF receptor (Hoschuetzky et al., 1994; Ochiai et al., 1994; Kanai et al., 1995) and hepatocyte growth factor/scatter factor receptor c-met (Shibamoto et al., 1994) have also been shown to target β-catenin. Treatment of a squamous carcinoma cell line (Fujii et al., 1996) or a mammary carcinoma cell line (Hazan and Norton, 1998) with EGF results in reduced cadherin-mediated adhesion and increased phosphorylation of β-catenin. Furthermore, suppression of the association of the EGF receptor with β-catenin increases the association of β-catenin with cadherin (Bonvini et al., 2001) and suppresses in vitro and in vivo invasion of a gastric cancer cell line (Shibata et al., 1996). Finally, activated Ras, often coupled to transmembrane tyrosine kinase signaling, also results in an increase in the phosphorylation of β-catenin and reduces the stability of the cadherin-β-catenin bond (Kinch et al., 1995; Kinch and Burridge, 1995).

There are also several phosphatases that have the potential to alter the state of phosphorylation of β-catenin. Members of three distinct families of receptor PTPs have been reported to alter either β-catenin phosphorylation and/or be correlated with the state of phosphorylation of cadherin itself: LAR-PTP, the chondroitin sulfate proteoglycan PTPβ/ζ, and three members of the MAM (Meprin/A5/Mu) domain containing family: κ, γ, and μ.

LAR-PTP (Kypta et al., 1996; Müller et al., 1999) has been shown to interact with and dephosphorylate β-catenin. Additionally, overexpression of LAR-PTP correlates with prevention of β-catenin phosphorylation and inhibition of epithelial cell migration (Müller et al., 1999). Similarly, PTPβ/ζ interacts with and dephosphorylates β-catenin (Meng et al., 2000). Furthermore, interaction of PTPβ/ζ with its ligand, pleiotrophin, results in inactivation of its intrinsic catalytic activity and enhanced tyrosine phosphorylation of β-catenin (Meng et al., 2000). The RPTPs κ, γ, and μ are closely related (Neel and Tonks, 1997) and all are able to mediate homophilic adhesions (Brady-Kalnay et al., 1993, Gebbink et al., 1993; Sap et al., 1994; Brady-Kalnay and Tonks, 1994; Cheng et al., 1997). Thus it is interesting that there are two apparently different roles with respect to cadherin function within this structurally related group. PTPκ (Fuchs et al., 1996) and PTPγ (Cheng et al., 1997) interact directly with β-catenin and PTPκ has actually been shown to dephosphorylate β-catenin (Fuchs et al., 1996). In contrast to its close relatives, PTPμ interacts with and potentially dephosphorylates cadherin (Brady-Kalnay et al., 1995, 1998).

PTPμ binds directly to the carboxy-terminal 38 amino acids of E-cadherin, a region that includes most of the minimal PTP1B binding site (Fig. 2). This might suggest that the interaction of PTPμ and PTP1B are mutually exclusive and may, under certain circumstances, play the same role, maintaining the stability of cadherin adhesions. This is further suggested by the fact that downregulation of either PTP1B (Pathre et al., 2001) or PTPμ (Burden-Gulley and Brady-Kalnay, 1999), mediated by antisense oligonucleotides or a dominant-negative construct, blocks N-cadherin-mediated neurite outgrowth. However, it is not clear whether the effect of blocking PTPμ function is on adhesion per se or on other events essential to N-cadherin-mediated neurite outgrowth. The absence of PTPμ is correlated with increased phosphorylation of cadherin itself, not β-catenin (Brady-Kalnay et al., 1998). Since increased tyrosine phosphorylation of N-cadherin has been associated with increased turnover of N-cadherin through cleavage and release of a 90 kD extracellular fragment (Lee et al., 1997), PTPμ may regulate cadherin turnover and therefore, the relative amount at the cell surface. A further possibility is that PTPμ alters the affinity of p120ctn for cadherin through dephosphorylation (Zondag et al., 2000) and that this affects cadherin-mediated adhesion.

Ligation of receptor tyrosine kinases and phosphatases is an obvious possibility for altering the state of β-catenin; however, other environmental cues may also set in motion signaling paths that affect the phosphorylation state of β-catenin. Our laboratory has demonstrated that the binding of the chondroitin sulfate proteoglycan neurocan to its GalNAcPTase cell surface receptor (PTase) initiates a signal that results in hyperphosphorylation of β-catenin on tyrosine residues, functional inactivation of N-cadherin and disassembly of the cadherin complex of proteins releasing both β-catenin and PTP1B (Li et al., 2000; see Fig. 4). It will be particularly interesting to determine the intermediate steps through which neurocan influences the phosphorylation of β-catenin.


Epithelial Mesenchymal Transformations: Breakin'Up Is Hard to Do

The extensive and potentially overlapping set of regulatory interactions that impinge on cadherin function, particularly the tyrosine phosphatases and kinases that regulate the phosphorylation of β-catenin and, therefore, the stability of cadherin-mediated adhesions suggests that this is a physiologically and developmentally important process. Yet we have little insight into when, and under what circumstances, this set of regulatory interactions are called into play. One possibility is in the conversion of epithelial or neuroepithelial cells into migratory cells. Cadherin-mediated adhesion at adherens junctions is an essential requisite for the integrity of epithelial layers (however, see Costa et al., 1998). It is only at prescribed developmental times and places, or during pathologic changes, that cells are set free of their neighbors to migrate and participate in new tissue formation. This is clearly a complicated process involving many different cellular changes; repression of cadherin transcription is clearly one aspect of epithelial mesenchymal transitions (Cano et al., 2000, Pérez-Moreno et al., 2001) and conversion to an invasive phenotype (Batlle et al., 2000; Comijn et al., 2001; Poser et al., 2001). However, transcriptional regulation will be slow as adherens junctions are rather stable with a half life of 5 to 10 hr (McCrea and Gumbiner, 1991; Shore and Nelson, 1991), creating a rather long lag time in developmental terms between transcriptional down-regulation and the actual event. We suggest that functional inactivation of cadherin plays a role in the rapid and timely onset of cell movement. In this regard, it is noteworthy that, on emigration from the neural tube, neural crest cells retain expression of N-cadherin (Monier-Gavelle and Duband, 1995); this suggests that loss of function precedes loss of protein. This is further substantiated by the observation that neural crest cells do not exit the neural epithelium when N-cadherin is overexpressed (Nakagawa and Takeichi, 1998).

Perturbation of the balance between phosphorylation of β-catenin and its dephosphorylation has the potential to effect rapid changes in cadherin function. Furthermore, in the absence of β-catenin the cytoplasmic domain is unstructured and subject to degradation (Huber et al., 2001). We have also demonstrated that, among L-cells constitutively expressing N-cadherin lacking the PTP1B binding site, β-catenin fails to bind and cadherin is subject to enhanced degradation (Xu et al., 2002). Thus, the immediate loss of β-catenin from the cadherin complex will potentiate the more permanent loss of cadherin itself. Tyrosine phosphorylation of β-catenin initiates a rapid loss of adhesion and rapid turnover of existing cadherin molecules, whereas transcriptional regulation yields a more permanent suppression of new synthesis, all promoting the rapid and continuous acquisition of a nonepithelial, potentially migratory phenotype.

In both cases, transcriptional repression and altering the balance of tyrosine phosphorylation, setting the process in motion, is clearly initiated by environmental cues. Given the ubiquity of epithelial mesenchymal transitions during development, it is not surprising that many different growth factors and environmental cues have been shown to be involved (reviewed in Hay, 1995; Savagner, 2001). This suggests that different cues may set in motion different pathways, some or all of which alter the balance of phosphorylated tyrosine residues in β-catenin.

Activation through ligand binding to the EGFR or c-met has been shown to be critical to epithelial mesenchymal transformations (reviewed in Boyer et al., 2000) and to have the potential to alter the balance of phosphorylated tyrosine residues on β-catenin (see above). The presence of constitutively active growth factor receptors has repeatedly been correlated with aggressively invasive carcinomas, a type of epithelial mesenchymal transformation (Kolibaba and Druker, 1997). Similarly, multiple signals involved in epithelial mesenchymal transformations have been shown to activate Src (reviewed in Boyer et al., 2000; Hay, 1995) and, therefore, may also change the balance of phosphorylated tyrosine residues on β-catenin. In some cases these two kinases may act in the same pathway; in at least one case EGF stimulated cell scattering has been shown to be mediated by activation of src and to be independent of transcriptional activity (Boyer et al., 1997). However, going beyond correlations and in vitro model systems to actually defining a role for functional inactivation of cadherin through tyrosine phosphorylation of β-catenin during development remains a challenge.

Although the role of growth factors and their kinase receptors in tissue remodeling is well documented, a potential role of phosphatases has not been examined. As reviewed above, some of these enzymes are associated with cadherin adhesion complexes through direct interaction with cadherin (PTPμ, PTP1B) or with β-catenin (κ, γ, LAR, β/ζ, PTP1B). The transmembrane tyrosine phosphatases have the potential to promote stability, through direct dephosphorylation of β-catenin (or cadherin itself for PTPμ) as has been demonstrated for the nonreceptor tyrosine phosphatase PTP1B (Balsamo et al., 1996, 1998). These phosphatases may become associated with the cadherin complex at specific times and places during development and promote stability of adhesions through dephosphorylation of β-cateinin. Thus loss of this association or down regulation of activity, as has been noted after binding of pleotrophin to PTPβ/ζ (Meng et al., 2000), would promote instability of adhesions through retention of phosphorylated tyrosine residues on β-catenin.

The remaining acts are yet to be written, however, it is clear that epigenetic regulation of the function of cadherin adhesion molecules in response to environmental cues is a rapid means of initiating morphogenetic changes, whether in normal developing embryos or among tumor cells. There are undoubtedly several avenues through which this can occur, many of which impinge on the association of the cytoplasmic domain of cadherins with the actin-containing cytoskeleton. In this review we have focused on the phosphorylation of β-catenin as one key regulatory element in altering the integrity of cadherin-mediated adhesions. Although it is clear that the mechanisms exist to effect these rapid changes in cadherin-mediated adhesion, we have yet to define their role in specific developmental situations. This is clearly the next step. Because some of the proteins regulating cadherin functional switching are ubiquitous signaling molecules or adaptors, developing specific methods and probes for in situ perturbation of these regulatory networks without altering other cell functions will present significant challenges. Additionally, epithelial mesenchymal transformations involve only a limited number of cells in an epithelial or neuroepithelial layer and determining how this population is selected will also present significant challenges.