Cell–cell adhesion in metazoans relies on evolutionarily conserved features of the α-catenin·β-catenin–binding interface

Stable tissue integrity during embryonic development relies on the function of the cadherin·catenin complex (CCC). The Caenorhabditis elegans CCC is a useful paradigm for analyzing in vivo requirements for specific interactions among the core components of the CCC, and it provides a unique opportunity to examine evolutionarily conserved mechanisms that govern the interaction between α- and β-catenin. HMP-1, unlike its mammalian homolog α-catenin, is constitutively monomeric, and its binding affinity for HMP-2/β-catenin is higher than that of α-catenin for β-catenin. A crystal structure shows that the HMP-1·HMP-2 complex forms a five-helical bundle structure distinct from the structure of the mammalian α-catenin·β-catenin complex. Deletion analysis based on the crystal structure shows that the first helix of HMP-1 is necessary for binding HMP-2 avidly in vitro and for efficient recruitment of HMP-1 to adherens junctions in embryos. HMP-2 Ser-47 and Tyr-69 flank its binding interface with HMP-1, and we show that phosphomimetic mutations at these two sites decrease binding affinity of HMP-1 to HMP-2 by 40–100-fold in vitro. In vivo experiments using HMP-2 S47E and Y69E mutants showed that they are unable to rescue hmp-2(zu364) mutants, suggesting that phosphorylation of HMP-2 on Ser-47 and Tyr-69 could be important for regulating CCC formation in C. elegans. Our data provide novel insights into how cadherin-dependent cell–cell adhesion is modulated in metazoans by conserved elements as well as features unique to specific organisms.

Stable intercellular adhesions that maintain tissue integrity are critical for morphogenetic movements during metazoan development (1). Maintaining such adhesions is also important for adult organisms, in which defects can lead to tumorigenesis and metastasis (2,3). One crucial mediator of intercellular adhesion is the adherens junction, which contains a highly conserved cadherin⅐catenin complex (CCC). 4 Intercellular adhesions are mediated by the CCC through calcium-dependent homophilic interactions of transmembrane cadherins (4); the intracellular tail of cadherins binds to p120-catenin and ␤-catenin (5). ␣-Catenin, which binds to ␤-catenin, acts as a physical linker connecting the CCC at the membrane to the F-actin cytoskeleton (6 -8).
␣-Catenin is an actin-binding and -bundling protein consisting of a series of linked ␣-helical bundles (9 -12). The N-terminal domain of mammalian ␣Eand ␣N-catenins contains overlapping ␤-catenin binding and homodimerization sites (7,13). The C-terminal domain of ␣-catenin binds to F-actin (the actin-binding domain) (6,14). The middle (M) domain of ␣-catenin is composed of three four-helix bundles designated M1, M2, and M3 (11). The central two helices of the M1 bundle contain the vinculin-binding site (15,16). These helices dissociate or "unfurl" from the bundle to bind vinculin. This unfurling is inhibited by the M3 bundle, but mechanical tension alters the relative positions of the M subdomains to enable vinculin binding, which further strengthens the CCC-F-actin linkage (11,12,15,17,18).
␣-Catenin is recruited to the CCC by ␤-catenin. Via its central armadillo (Arm) repeats, ␤-catenin binds the highly conserved intracellular domain of E-cadherin, and the interaction is strengthened by phosphorylation of a serine-rich region in cadherin (19 -22). The region of ␤-catenin N-terminal to the Arm repeats binds to the N-terminal domain of ␣-catenin (23,24). In vitro reconstitution of the CCC demonstrates that these three proteins form a stable complex (25). Although binding to ␤-catenin weakens the affinity of ␣E-catenin for F-actin in solution (26,27), ␣E-catenin maintains association with both of these binding partners when tension is applied to the complex (8). Previous structural studies of ␣Eand ␣N-catenin indicate that ␣-catenin interacts with ␤-catenin mainly via its N-terminal four-helix bundle (N1 bundle), which is bridged to the second four-helix bundle (N2 bundle) by one continuous ␣ helix (␣4). The two four-helix bundles of ␣-catenin move with respect to each other to accommodate insertion of an ␣ helix from ␤-catenin (13). Whether these N1 interactions and structural changes in the N domain are evolutionarily conserved has not been established nor has the functional significance of these interactions been examined in an in vivo setting.
Another feature of adherens junctions in vivo is their ability to assemble, disassemble, and reassemble dynamically during morphogenesis (28). Because post-translational modifications to the cadherin⅐catenin complex are known to modulate the ability of CCC components to bind one another (19, 20, 29 -34), such modifications represent a potential mechanism for regulating junctional stability. In particular, perturbing phosphorylation of key residues in E-cadherin and ␤-catenin has been shown to alter adhesion in cultured cells (35)(36)(37)(38) and to modify junctional morphology during embryonic development (19,39,40).
Phosphoregulation of the ␤-catenin⅐␣-catenin association has been less well-studied. Phosphorylation on Tyr-142 of vertebrate ␤-catenin inhibits its ability to bind ␣-catenin (30,32,41). The ␤-catenin residue Tyr-142 is a target of multiple kinases, including Fer and Fyn, that are recruited to adherens junctions by p120-catenin (32). Accordingly, a phosphomimetic Y142E transgene added to cells that lack endogenous ␤-catenin does not confer adhesion in cultured mouse cells (42), and Tyr-142 phosphorylation by focal adhesion kinase increases vascular permeability and the accompanying junctional breakdown in vascular endothelial cells (43). Although the effects on CCC-mediated adhesion play a role in both phenotypes, Tyr-142 phosphorylation of ␤-catenin also increases its nuclear localization and ability to participate in transcriptional coactivation (30,41), which limits the interpretation of these experiments. Moreover, the role of these residues in regulating the ␣-catenin/␤-catenin interaction during embryonic morphogenesis has not been assessed. Protein kinase D1 (PKD1) has been shown to phosphorylate ␤-catenin at Thr-120 in cultured cells, resulting in decreased nuclear ␤-catenin localization and transcriptional function (44) and increased plasma membrane and trans-Golgi network localization (45,46). The significance of phosphorylation of Thr-120 for association with ␣-catenin and its role in an intact organism have not been examined.
Caenorhabditis elegans provides a unique model to explore evolutionary conservation of mechanisms that mediate the ␣-catenin⅐␤-cateninbinding interface and to probe requirements for elements of that interface in vivo. C. elegans has conserved homologs of each component of the CCC as follows: HMR-1/cadherin, HMP-2/␤-catenin, and HMP-1/␣-catenin (47,48). Moreover, HMP-1 is the sole ␣-catenin homolog in C. elegans, removing any issues of redundancy. In addition, HMP-2 does not normally play a role in the nucleus, and its functions are restricted to the CCC, simplifying interpretation of experiments involving ␤-catenin.
Our results show that the HMP-1⅐HMP-2 complex adopts a five-helix bundle structure. Deleting the first ␣ helix within the HMP-1 N domain reduces its ability to bind HMP-2. We also show that phosphomimetic substitutions at HMP-2 Ser-47 or HMP-2 Tyr-69, which are homologous to ␤-catenin Thr-120 and Tyr-142, respectively, decrease its binding to HMP-1. These studies test requirements for key elements within ␣and ␤-catenin in maintaining their strong association during morphogenetic events in vivo. Our results highlight structurally diverse yet biologically convergent evolutionary solutions that metazoans have adopted to stabilize the ␣-catenin⅐␤-catenin association, which is crucial in all multicellular organisms.

HMP-1 N1 domain is required to suppress the latent ability of HMP-1 to form homodimers
C. elegans HMP-1/␣-catenin and HMP-2/␤-catenin contain key conserved features found in their vertebrate counterparts ( Fig. 1A and supplemental Figs. S1 and S2). Morphogenesis of the C. elegans embryo requires adherens junctions to be capable of withstanding substantial tensile forces (48 -50), so we compared the binding affinity of HMP-1 and HMP-2 to their vertebrate counterparts to see whether there are features of the worm complex that reflect adaptation to such physical demands. HMP-1/␣-catenin is homologous to mammalian ␣-catenin, consisting of the N-terminal HMP-2/␤-cateninbinding domain, M domain, and the C-terminal actin-binding domain (Fig. 1A). The N-terminal regions of ␣E-catenin and ␣N-catenin are responsible for homodimerization as well as ␤-catenin binding; they form either homodimers or heterodimers with ␤-catenin (7,13,51). Previously, native gel-shift datashowedthatunlike␣E-catenin,HMP-1ispresentasamonomer in solution even after a 1-h incubation at 25°C (49). Here, we confirmed that purified HMP-1 is homogeneously monomeric in solution by multiangle light scattering (MALS) (Fig.  1B), and we showed that HMP-1 remains monomeric even after overnight incubation at 28°C at a concentration of 80 M (Fig.  1C). To test whether HMP-1 has latent potential to form a dimer, we made a construct that deletes 70 amino acids at the N terminus (HMP-1N⌬70), which corresponds to a dimeric ␣E-catenin construct (␣E-catN⌬81 (7)), and we identified the oligomeric state of this mutant by gel-filtration chromatography. Similar to ␣E-catN⌬81, HMP-1N⌬70 was predominantly dimeric in solution even at 10 M and at 4°C (Fig. 1C), suggesting that the third and fourth helices of HMP-1 N1 can form a homodimer if the HMP-1 N1 bundle is disassembled. Because wild-type HMP-1 is a monomer, these results imply that the HMP-1 N1 forms a more stable four-helix bundle than that in ␣Nand ␣E-catenins.
Monomeric forms of ␣Nand ␣E-catenins have similar affinity toward ␤-catenin, with dissociation constants in the range of 15-25 nM (13). HMP-1 was recently shown to interact with the ␤-catenin homolog, HMP-2, with K d of ϳ1 nM (52), but its minimal binding site with full affinity has never been studied.

Helix ␣1 of HMP-1 is crucial for formation of the HMP-1⅐HMP-2 complex
To understand the high-affinity interaction between HMP-1 and HMP-2, we determined a crystal structure of the HMP-1⅐HMP-2 complex. We purified the complex of HMP-1(2-274) and HMP-2(36 -79), the minimal regions required for full binding affinity, crystallized it, and determined its structure at 1.6 Å resolution. The structure reveals that the four helices of the HMP-1 N1 subdomain rearrange to accommodate a single ␣ helix formed by HMP-2 residues 46 -69, forming a five-helix bundle ( Fig. 2A). Structural alignment of HMP-1 with the unbound ␣N-catenin monomer shows that the position of ␣1 is almost identical in both structures but that helices ␣2 and ␣3 are located in different positions (Fig. 2B) and are repositioned upon HMP-2 binding; the ␣1 helix, which makes direct con- ␣-Catenin⅐␤-catenin binding in C. elegans tacts with the N2 bundle and the ␣4 helix, remains stationary. The 11-amino acid linker between the HMP-1 ␣1 and ␣2 helices is long enough to accommodate the structural rearrangement from a four-to five-helix bundle upon HMP-2 binding. The C-terminal region of the HMP-2 helix interacts with the linker. Although this linker conformation is stabilized by crystal contacts in our structure (supplemental Fig. S5), computational analysis suggests that these interactions exist in the absence of crystal contacts (supplemental Fig. S6).

␣-Catenin⅐␤-catenin binding in C. elegans
HMP-2 with an affinity of 24 nM (Table 1 and supplemental Fig. S3), comparable with the affinity of mammalian ␣-catenin for ␤-catenin. Thus, these additional polar interactions appear to contribute to the higher affinity between HMP-1 and HMP-2 relative to ␣and ␤-catenin.
Remarkably, the change from a four-to five-helix bundle is very similar to that binding of talin to vertebrate vinculin, which is a paralog of ␣-catenin (53,54). Superposition of the HMP-1⅐HMP-2 complex with those of vinculin⅐talin (e.g. PDB codes 1T01 and 1RKC) reveals very similar structures (Fig. 2D).
To test the importance of helix ␣1 to the affinity of HMP-1 for HMP-2, we made an ␣1 deletion mutant (HMP-1(N⌬44)). We found that the solubility of HMP-1(N⌬44) was much lower than wild-type HMP-1N, likely because hydrophobic residues on helix ␣2 are exposed to solvent by dimerization. Indeed, the gel-filtration profile of HMP-1N⌬44 suggested it dimerizes (supplemental Fig. S7). Although HMP-1N⌬44 does bind to HMP-2, the affinity was decreased dramatically by more than 2000-fold (K d of 2.2 M; Table 1), indicating that helix ␣1 is essential for tight binding of HMP-1 to HMP-2.
We also evaluated the effects of HMP-2 binding on HMP-1 stability. First, we measured the melting temperature (T m ) of the HMP-1 N domain in the absence and presence of HMP-2 using circular dichroism (CD) spectroscopy. Unexpectedly, HMP-1 N showed two transitions, one at 42°C and the other at 67°C, implying separate melting of the two bundles (Fig. 4A). In contrast, the HMP-1⅐HMP-2 complex has a single transition at 55°C, with no obvious second transition, although the CD signal changes slightly around 98°C. The shift of the initial transition to a higher temperature when bound to HMP-2 could imply a more stable structure. This was tested by limited proteolysis using endoproteinase GluC, which selectively cleaves peptide bonds C-terminal to glutamic acid residues. When the HMP-1 and the HMP-1⅐HMP-2 complex were incubated with GluC under the same conditions, more digestion was observed in HMP-1 alone (Fig. 4B), although only 2 of 14 possible cleavage sites are located close to HMP-2 in the complex. Interestingly, one cleavage site is present at the linker between the ␣1 and ␣2 helices, and in the absence of HMP-2, this site could be cleaved readily, which may reflect lower stability.
Deleting any of the helices ␣2, ␣3, and ␣4 should affect the overall structural stability of HMP-1, and it would be expected to abolish its binding to HMP-2. Consistent with this prediction, constructs carrying deletions of each of the remaining ␣-helices (␣2, ␣3, and ␣4) all failed to rescue hmp-1(zu278) mutants (supplemental Fig. S10).

␣-Catenin⅐␤-catenin binding in C. elegans
In summary, the detailed functional analysis we performed, based on the HMP-1⅐HMP-2 complex structure, confirmed a key role for helix ␣1 of an ␣-catenin for the first time in vivo.
As shown in Table 2, the two non-phosphorylatable mutants (S47 and Y69F) did not show any difference in binding affinity for HMP-1, with a K d of ϳ1 nM. The phosphomimetic mutants (S47E and Y69E), however, showed a dramatically decreased binding affinity for HMP-1. The S47E mutant shows ϳ40-fold reduced affinity to HMP-1, which could be caused by charge repulsion with Asp-141 of HMP-1 as well as Glu-50 of HMP-2 (Fig. 6B). The affinity of the HMP-2 Y69E mutant for HMP-1 was reduced by 100-fold, which is likely caused by introduction of a charged residue at the hydrophobic core (Fig. 6C). Given the solvent exposure of the Tyr-69 OH group, it is possible that a Glu substitution has other consequences, such as effects on packing interactions; nevertheless, these results are consistent with a role for pSer-47 and pTyr-69 in negative regulation of the binding of HMP-2 to HMP-1.

␣-Catenin⅐␤-cateninbinding interface exhibits evolutionary diversity but functional convergence
Structural, biochemical, and in vivo studies of HMP-1/␣catenin in C. elegans provide a unique opportunity to assess structural diversity yet functional conservation of the ␣-catenin⅐␤-cateninbinding interface in metazoans. There is a basic similarity of the ␣-catenin/␤-catenin interaction across the animal kingdom; HMP-2/␤-catenin forms an amphipathic helix that packs into the HMP-1/␣-catenin N1 bundle. However, the overall architecture of the worm and mammalian complex differs; the C. elegans complex involves converting the NI four-helix bundle into a five-helix bundle, whereas the ␣1 helix in mammalian ␣-catenin is displaced from the bundle and instead contributes to the interaction by binding along the outside of the bundle and forms additional contacts with another ␤-catenin helix that is absent in HMP-2.
The similarity of the five-helix bundles in the HMP-1⅐HMP-2 and vertebrate vinculin⅐talin complexes is consistent with the evolution of these two proteins from a common ancestor. The differences in the architecture of these complexes with that of the mammalian ␣-catenin⅐␤-catenin complex likely reflects further divergence of ␣-catenin and vinculin functions during the evolution of more complex tissue architectures.
The structural features of the HMP-2⅐HMP-1 complex may also explain differences we recently noted in the binding affinities of these two proteins compared with their vertebrate counterparts. Whereas E-cadherin binding to ␤-catenin increases the affinity of ␤-catenin for ␣-catenin, with K d of ϳ1 nM (13), which is similar to the affinity between HMP-1 and HMP-2, we recently found that binding of HMR-1, an E-cadherin homolog, did not change the affinity of HMP-2 for HMP-1 (52). This ␣-Catenin⅐␤-catenin binding in C. elegans difference may reflect the longer tail of E-cadherin versus HMR-1 (Ref. 21 and discussed in Ref. 52). The structural features of the HMP-2⅐HMP-1-binding interface, including the polar interactions between HMP-2 and the HMP-1 ␣1 helix, may enable constitutively strong binding in the presence or absence of HMR-1. The result is that in both the nematode and mammalian systems ␣-catenin binds to the cadherin⅐␤-catenin complex with single nanomolar affinity, but the binding energetics are encoded differently.

Latent ability of HMP-1 to homodimerize provides insights into ␣-catenin evolution
The evolutionary origins of ␣-catenin functions beyond binding to ␤-catenin are unclear, especially regarding the functions of ␣-catenin homodimers (58). Our results shed light on the diversification of ␣-catenins with regard to homodimerization, which competes with ␤-catenin binding. Mammalian ␣E-catenin has a strong propensity to homodimerize, and there is evidence that homodimeric ␣E-catenin has roles away from junctions,includingsuppressionofArp2/3-mediatedactinpolymerization (26,59). Drosophila ␣-catenin likewise seems to exist predominantly in a homodimeric form (60). In contrast, ␣E-catenin from zebrafish is monomeric (61), as is Dictyostelium ␣-catenin (62). ␣N-catenin seems to have its own characteristic feature, temperature-dependent dimerization. At 37°C, ␣N-catenin readily forms a homodimer, although it can be purified as a monomer at 4°C (13). Our results shed light on the structural requirements for constitutive homodimerization. C. elegans HMP-1 is predominantly monomeric even after overnight incubation at physiological temperature; however, HMP-1N⌬44 or HMP-1N⌬70, in which the N-terminal fourhelix bundle is disrupted, shows a strong tendency to homodimerize in solution. Unlike mammalian ␣⌭-catenin, whose homodimeric form does not measurably bind to ␤-catenin, HMP-1N⌬44 binds weakly to HMP-2, with K d of ϳ2 M, and HMP-1⌬2-44::GFP is still able to rescue hmp-1(zu278), albeit weakly. Taken together, results across the animal kingdom suggest that ␣-catenins may have evolved independently multiple times to acquire their homodimerization abilities to fit specific demands in different organisms and that the ␣1 helix is crucial for masking this latent capability.

Conserved phosphorylatable residues in ␤-catenin are required for association with ␣-catenin
In addition to the important role of the ␣1 helix in mediating the ␣-catenin⅐␤-cateninbinding interface, our work also demonstrates the crucial role of two specific amino acids in ␤-catenin in vivo. HMP-2 Ser-47 and Tyr-69 are homologous to the vertebrate ␤-catenin Thr-120 and Tyr-142, respectively. Phosphorylation at Tyr-142 of ␤-catenin has been shown to inhibit its ability to bind ␣-catenin (30,32), likely because the introduction of a negatively charged phosphate group disrupts the predominantly hydrophobic interface between the two molecules (11,12,23). In contrast, the role of phosphorylation of Thr-120 has not been examined. Our work shows that phosphorylation of the amino acid homologous to Thr-120 of ␤-catenin could disrupt the interaction with ␣-catenin. Ser-47 and Tyr-69 of HMP-2 are located at the N-and the C-terminal ends, respectively, of the HMP-2 helix in the HMP-1⅐HMP-2 complex, similar to the positions of Thr-120 and Tyr-142 in the ␤-catenin⅐␣-catenin complex (7,13). It is compelling to speculate that this positioning allows for access by kinases to disrupt the HMP-2⅐HMP-1 association and therefore modulate CCC function. It is interesting that phosphomimetic mutants (S47E or Y69E) of HMP-2 show more severe defects in vivo than an HMP-1 ⌬␣1 deletion mutant, whose affinity is ϳ20 -50-fold weaker than that between a phosphomimetic mutant of HMP-2 and HMP-1 in vitro. This suggests that phosphorylation of HMP-2 plays other important regulatory roles besides the direct effects on HMP-1⅐HMP-2 binding during embryogenesis.
Although endogenous phosphorylation has not yet been observed at HMP-2 Ser-47 or Tyr-69 (63), it may be extremely difficult to detect whether it is only required transiently during dynamic junctional remodeling. The observation that HMP-2(S47A)::GFP and HMP-2(Y69F)::GFP constructs exhibit different junctional dynamics from wild-type HMP-2::GFP is consistent with the possibility that HMP-2 is endogenously phosphorylated at these sites. Although it is possible that the S47A and Y69F mutations may subtly alter HMP-2 protein folding in such a way as to reduce its junctional mobility, the fact that both non-phosphorylatable constructs show full affinity for HMP-1 in vitro and rescue hmp-2(zu364) to viability in vivo suggests that conformation of the mutant protein is not severely altered.
In conclusion, our in vitro and in vivo functional studies have definitively demonstrated the importance of ␣1 helix interactions in stabilizing the ␣-catenin⅐␤-cateninbinding interface, and how phosphorylation of ␤-catenin can regulate its binding to ␣-catenin. Our results will enable future studies of the

ITC
Isothermal titration calorimetry was done using Nano ITC (TA Instruments, Inc.) at 25°C in a buffer consisting of 20 mM HEPES, pH 8.0, 150 mM NaCl, and 1 mM DTT. For each mea-

Circular dichroism spectroscopy
CD spectra were measured using a J-815 CD spectrometer (Jasco Analytical Instruments, Easton, MD). For thermal melting experiments, heat-induced changes were monitored at 222 nm for each of HMP-1 and HMP-1⅐HMP-2 complex samples in PBS buffer at a concentration of 0.2 mg/ml. Data were mea- ␣-Catenin⅐␤-catenin binding in C. elegans sured between 20 and 99°C at a scan rate of 1°C/min. All the spectra were corrected for solvent contribution.
Crystals of the HMP-1N⅐HMP-2(36 -79) complex were cryoprotected by perfluoropolyether oil. A 1.6 Å resolution diffraction data set was collected at the Stanford Synchrotron Radiation Lightsource (SSRL) beamline 12-2 and processed using XDS and SCALA as described earlier (64). The structure was solved by molecular replacement using Phaser. The search model was the mouse ␤␣-catenin chimeric protein structure (PDB code 1DOW), and a solution was obtained and refined to give initial R work and R free values of 49 and 54%, respectively. Several cycles of refinement and manual rebuilding were performed using PHENIX and Coot, respectively. The refinement statistics are shown in Table 3. The final model consists of HMP-1 residues 13-260 and HMP-2 residues 46 -72. Coordinates and structure factors have been deposited in the Protein Data Bank under accession code 5XA5.

Computational modeling
The HMP-1⅐HMP-2 complex structure without neighboring molecules that make crystal contacts was relaxed by iterative short molecular dynamics simulations with subsequent sidechain repacking steps. The energy function used for relaxation comprised molecular mechanics energy terms and knowledgebased terms (65,66). Additional restraints were applied to avoid drift away from the initial structure, as in typical relaxation (65)(66)(67). After relaxation, the lowest-energy structure among the generated 48 models was selected.

Imaging
Embryos were isolated from gravid hermaphrodites, mounted on a 5% agarose slide, and aged at 20 -25°C until the onset of morphogenesis. For four-dimensional differential interference contrast microscopy, embryos were imaged using 1-m slice spacing at 3-min intervals using a Nikon Eclipse E600 microscope with a ϫ60/1.45 NA oil objective at 20°C with a Macintosh computer running ImageJ using custom macros/plugins. For fluorescent imaging, a PerkinElmer Life Sciences UltraView spinning disk confocal microscope and Micromanager software, using a Nikon Eclipse E600 microscope and Hamamatsu ORCA-ER camera, were used to collect images of GFP-expressing embryos, using 0.5-m slices at 3-min intervals with a ϫ60/1.45 NA oil objective at 20°C. Antibody staining (0.6-m slices) images were collected with the same confocal microscope using a ϫ100/1.45 NA oil objective.

FRAP
Transgenic embryos were isolated from gravid hermaphrodites, mounted on 5% agarose slide, and aged at 20°C for 4 h or until the onset of elongation. The FRAP experiments were then performed as described in our previous research (50).