Specific Conserved C-terminal Amino Acids of Caenorhabditis elegans HMP-1/α-Catenin Modulate F-actin Binding Independently of Vinculin*

Background: α-Catenin is a crucial link between adherens junctions and F-actin. Results: C-terminal amino acids in HMP-1/α-catenin quantitatively modulate its ability to bind F-actin, but a putative vinculin-binding domain is not required in vivo. Conclusion: Key C-terminal residues in α-catenin modulate its ability to bind F-actin. Significance: This is the first genetic dissection of the ability of α-catenin to bind F-actin. Stable intercellular adhesions formed through the cadherin-catenin complex are important determinants of proper tissue architecture and help maintain tissue integrity during morphogenetic movements in developing embryos. A key regulator of this stability is α-catenin, which connects the cadherin-catenin complex to the actin cytoskeleton. Although the C-terminal F-actin-binding domain of α-catenin has been shown to be crucial for its function, a more detailed in vivo analysis of discrete regions and residues required for actin binding has not been performed. Using Caenorhabditis elegans as a model system, we have characterized mutations in hmp-1/α-catenin that identify HMP-1 residues 687–742 and 826–927, as well as amino acid 802, as critical to the localization of junctional proximal actin during epidermal morphogenesis. We also find that the S823F transition in a hypomorphic allele, hmp-1(fe4), decreases actin binding in vitro. Using hmp-1(fe4) animals in a mutagenesis screen, we were then able to identify 11 intragenic suppressors of hmp-1(fe4) that revert actin binding to wild-type levels. Using homology modeling, we show that these amino acids are positioned at key conserved sites within predicted α-helices in the C terminus. Through the use of transgenic animals, we also demonstrate that HMP-1 residues 315–494, which correspond to a putative mechanotransduction domain that binds vinculin in vertebrate αE-catenin, are not required during epidermal morphogenesis but may aid efficient recruitment of HMP-1 to the junction. Our studies are the first to identify key conserved amino acids in the C terminus of α-catenin that modulate F-actin binding in living embryos of a simple metazoan.

Stable intercellular adhesions formed through the cadherincatenin complex are important determinants of proper tissue architecture and help maintain tissue integrity during morphogenetic movements in developing embryos. A key regulator of this stability is ␣-catenin, which connects the cadherin-catenin complex to the actin cytoskeleton. Although the C-terminal F-actin-binding domain of ␣-catenin has been shown to be crucial for its function, a more detailed in vivo analysis of discrete regions and residues required for actin binding has not been performed. Using Caenorhabditis elegans as a model system, we have characterized mutations in hmp-1/␣-catenin that identify HMP-1 residues 687-742 and 826 -927, as well as amino acid 802, as critical to the localization of junctional proximal actin during epidermal morphogenesis. We also find that the S823F transition in a hypomorphic allele, hmp-1(fe4), decreases actin binding in vitro. Using hmp-1(fe4) animals in a mutagenesis screen, we were then able to identify 11 intragenic suppressors of hmp-1(fe4) that revert actin binding to wild-type levels. Using homology modeling, we show that these amino acids are positioned at key conserved sites within predicted ␣-helices in the C terminus. Through the use of transgenic animals, we also demonstrate that HMP-1 residues 315-494, which correspond to a putative mechanotransduction domain that binds vinculin in vertebrate ␣E-catenin, are not required during epidermal morphogenesis but may aid efficient recruitment of HMP-1 to the junction. Our studies are the first to identify key conserved amino acids in the C terminus of ␣-catenin that modulate F-actin binding in living embryos of a simple metazoan.
The cadherin-catenin complex (CCC) 4 comprises a highly conserved set of proteins that mediates adhesion between contacting cells at adherens junctions. Proper development requires that cadherin-based adhesions be highly adaptable; they must be stable enough to transmit forces to the cell surface that drive cell shape changes and tissue architecture, but at the same time dynamic enough to allow the cell rearrangements and morphogenetic movements required to shape the embryo (1,2). Maintenance of intercellular adhesions is also important in adult organisms, where loss of CCC proteins can result in tumorigenesis and metastasis (3,4).
The CCC mediates intercellular adhesion through the calcium-dependent homophilic interactions of transmembrane cadherin receptors (5). Adhesive connections are stabilized through the association of the cadherin intracellular tail with p120-catenin, ␤-catenin, and ␣-catenin (6 -10). ␣-Catenin localizes to the CCC through an N-terminal ␤-catenin binding site (11)(12)(13)(14)(15), whereas a C-terminal actin-binding domain (ABD) associates the CCC with the F-actin cytoskeleton (16 -18). ␣-Catenin has three domains with significant homology to the protein vinculin, and are therefore termed vinculin homology (VH) domains (19). The ABD of ␣-catenin includes the VH3 domain. Although several studies have shown the absolute requirement of the ␣-catenin ABD in conferring full adhesive activity and cytoskeletal interaction to the CCC (16, 17, 20 -23), it is unclear if F-actin attachment occurs mainly through direct binding to ␣-catenin or through other actin-binding proteins in an ␣-catenin-dependent fashion. Several other actin-binding proteins bind to vertebrate ␣-catenin, including vinculin (16, 24 -28), ␣-actinin (15,16,29), ZO-1 (16,30), afadin (18), formin-1 (31), and EPLIN (32). Binding sites on ␣-catenin have been partially mapped for several of these proteins; some lie outside of the C terminus, whereas others overlap with the ABD (15, 16, 18, 24 -32). It is thus possible that interactions between ␣-catenin and the actin cytoskeleton occur through both direct and indirect mechanisms, and the contribution of each depends on the cellular context (33). Recently ␣E-catenin has been shown to be a coactivator of vinculin (26), and the binding of vinculin to ␣E-catenin has been found to be force-dependent, suggesting a role for ␣E-catenin in mechanotransduction (22). However, it is unclear whether or not this function is conserved among other ␣-catenins.
Detailed in vivo analysis of ␣-catenin function in mammals has been difficult due to its requirement in early development (23). Cre-mediated recombination has allowed several studies to temporally control conditional ablation of ␣-catenin in specific tissues (34,35); however, this method has only been used to create null mutants rather than to examine requirements for specific domains or single amino acids of ␣-catenin.
Caenorhabditis elegans provides a unique opportunity to study the requirements for ␣-catenin function in a simple metazoan. This powerful invertebrate model system is amenable to mutagenesis and genetic manipulation, and its transparency allows the visualization of complex morphogenetic movements during development. There are conserved C. elegans homologs for each of the main CCC proteins: HMR-1/cadherin, JAC-1/ p120-catenin, HMP-2/␤-catenin, and HMP-1/␣-catenin (36,37). HMP-1 is 35-38% identical to vertebrate and invertebrate ␣-catenins (36). Moreover, unlike vertebrates, HMP-1 is the sole ␣-catenin homolog in C. elegans, removing any issues of functional redundancy. Mutations in hmp-1 result in a characteristic defect during epidermal morphogenesis, when Rho-dependent actomyosin-mediated contractile forces are transmitted along circumferential actin filament bundles (CFBs) to cell-cell junctions to drive the cell shape changes necessary for elongation of the embryo. Strong zygotic loss-of-function (LOF) hmp-1 mutants exhibit loss of junctional proximal actin and detachment of CFBs, which leads to the formation of dorsal folds in the epidermis (the Humpback phenotype). Removal of both maternal and zygotic HMP-1 via RNAi or through germline mosaics results in a more severe phenotype, in which ventral epidermal cells fail to properly adhere to one another. This results in anterior rupture of the embryo at the onset of elongation, mimicking mutations in hmr-1/E-cadherin (36).
Here we identify discrete regions and single amino acids in the HMP-1/␣-catenin ABD that modulate attachment to F-actin. Through characterization of hmp-1 alleles isolated in a previous EMS mutagenesis screen (36), we have identified residues 687-742, 826 -927, and 802, all within the ABD, as critical to proper HMP-1 function and actin binding during epidermal morphogenesis. Utilizing a hypomorphic allele, hmp-1(fe4), we have also used EMS mutagenesis to identify second site mutations in the ABD that alleviate the fe4 weak LOF phenotype. Through in vitro actin cosedimentation assays, we find that these mutations are able to rescue the ability of HMP-1 to bind F-actin directly. Through the use of transgenic deletion constructs, we also show that residues 315-494 of HMP-1, the putative vinculin-binding domain, are not essential for embryonic development. Fluorescence recovery after photobleaching (FRAP) analysis of embryos expressing a construct lacking this region nevertheless suggests that this domain may be required for efficient targeting of HMP-1 to the CCC.

EXPERIMENTAL PROCEDURES
Strains and Alleles-C. elegans strains were cultured using standard protocols (38). The Bristol strain N2 was used as wildtype. The strong zygotic loss-of-function hmp-1 mutants were derived from a previous EMS mutagenesis screen (36). For a complete list, see supplemental "Experimental Procedures." To determine the ability of different domains of HMP-1 to rescue hmp-1(zu278), hmp-1::gfp deletion constructs were microinjected (39) at 1 ng/l along with rol-6(su1006) (79 ng/l) and noncoding DNA (F35D3, 20 ng/l) into the gonads of either N2 or SU370 (ϩ/nT1[qIs5I] IV; hmp-1(zu278)/ nT1[qIs5I] V) hermaphrodites. Extrachromosomal arrays maintained in N2 were crossed into SU370 hermaphrodites later. If hmp-1(zu278) mutant embryos were rescued to viability, a separate line was maintained. For a complete list of strains, see supplemental "Experimental Procedures." hmp-1(fe4) suppressors were isolated from nonclonal screens by mutagenizing hmp-1(fe4) homozygotes with 50 mM EMS and starting cultures with individual mutagenized adults. Once the F2 generation had reached adulthood, plates were screened by eye for morphologically wild-type animals. In some cases, the animals picked still segregated a proportion of Vab progeny, suggesting either that they were heterozygous and that the mutations might be dominant/semi-dominant, or that the suppressor effect was incompletely penetrated. In each case we were able to establish strains that segregated almost completely wild-type animals, indicating that the former explanation was more likely. Subsequent outcrosses with unmutagenized hmp-1(fe4) homozygotes confirmed this assessment.
Allele Sequencing-For strong LOF alleles, individual homozygous hmp-1 mutant embryos were isolated based on phenotype and treated for 2 min with a 20 mg/ml of chitinase solution. If hatchoids were chosen, no chitinase was used. Embryos were transferred to 5 l of single worm lysis buffer and frozen for 15 min at 80°C, then lysed at 65°C for 1 h followed by 95°C for 15 min. Embryonic lysate was used to PCR amplify fragments of hmp-1 and the products were cloned using the pCR8/GW/TOPO-TA cloning kit (Invitrogen, K250020). Clones were miniprepped and DNA was sequenced via Sanger sequencing through the University of Wisconsin-Madison Biotechnology Center.
Staining-Embryos were isolated from plates and gravid hermaphrodites with 0.5% NaOCl in 250 mM NaOH for 5 min followed by three washes in distilled deionized water. For phalloidin staining, embryos were mounted on poly-L-lysine-coated ring slides and fixed with 4% paraformaldehyde, 0.1 mg/ml of lysolecithin, 48 mM PIPES, pH 6.8, 25 mM HEPES, pH 6.8, 2 mM MgCl 2 , and 10 mM EGTA for 20 min and then washed three times with PBS. Embryos were incubated in the dark with 1:20 Alexa 555:phalloidin in PBST overnight at 4°C, then washed three times with PBS before being covered with antifade reagent and sealed with nail polish.
For antibody staining, embryos were mounted on poly-L-lysine-coated ring slides and covered with a coverslip. Slides were quick frozen for 10 min on dry ice, then the coverslips were removed with a razor blade. Slides were immediately transferred to methanol at Ϫ20°C for 5 min, acetone at Ϫ20°C for 5 min, and PBST at room temperature for 5 min, followed by two additional PBST washes. Embryos were incubated with 1:3000 rabbit anti-HMP-1 and 1:20 mouse anti-AJM-1 (MH27) in 0.5% milk/PBST in the dark overnight at 4°C. Slides were washed three times with PBST then incubated in the dark for 4 h at room temperature with 1:50 anti-rabbit Texas Red and 1:50 anti-mouse FITC in 0.5% milk/PBST. Slides were washed three times with PBST before being covered with antifade reagent and sealed with nail polish.
Mutant Constructs-Using an existing hmp-1::gfp containing the endogenous promoter, 5 targeted in-frame deletions were created using site-directed mutagenesis. Forward and reverse primers flanking the region to be deleted were used to create a linear PCR product. Subsequent phosphorylation and ligation created a circular plasmid that was transformed into DH5␣ competent cells. Deletion-positive clones were identified via gel electrophoresis and sequencing.
For the actin cosedimentation assays, a previously published gst::hmp-1(677-927) (21) construct was used to clone the hmp-1(fe4) mutation with Phusion polymerase via a circle PCR method. The second site intragenic suppressor mutations were cloned into the hmp-1(fe4) construct. The mutations were confirmed through sequencing after the mutant constructs were transformed into BL21(DE3) protein-expressing cells (Stratagene).
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 (available at worms.zoology.wisc.edu/research/4d/4d.html). For fluorescent imaging, a Perkin-Elmer UltraView spinning disk confocal microscope and software, using a Nikon Eclipse E600 microscope and Hamamatsu ORCA-ER camera, was used to collect images of GFP expressing embryos, using 0.8-m slices at 3-min intervals with a ϫ60/ 1.45 NA oil objective at 20°C. Phalloidin (0.2 m slices) and antibody staining (0.6 m slices) images were collected with the same confocal microscope using a ϫ100/1.45 NA oil objective.
Fluorescence Recovery After Photobleaching-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. An Olympus Fluoview 1000 confocal microscope with SIM scanner was used to bleach and image GFP expressing embryos at Ͻ25°C with a ϫ60 oil objective at zoom 3, with images taken every 1.1 s. At 1.25-1.5-fold elongation, the seamseam junction of a seam cell was bleached and allowed to recover. Collected images were analyzed in ImageJ using the FRAP profiler plug-in (available at worms.zoology.wisc.edu/research/4d/4d.html). Normalized data were further analyzed in F calc 6 and Microsoft Excel. A Student's t test was used to determine statistical significance.
Actin Assays-G-actin from rabbit skeletal muscle (Cytoskeleton, Inc., AKL99) was polymerized into F-actin for 1 h at RT in actin polymerization buffer (5 mM Tris, pH 8.0, 0.2 mM CaCl 2 , 0.5 mM DTT, 50 mM KCl, 2 mM MgCl 2 , 1.2 mM ATP). 5 M proteins of interest were incubated with 2 or 10 M F-actin in reaction buffer (20 mM Tris, pH 8.5, 150 mM NaCl, 2 mM EDTA, 1 mM DTT, 10% glycerol, 0.2 mM ATP, 0.2 mM CaCl 2 ) for 30 min at RT. Reactions were pelleted at 100,000 ϫ g for 20 min. Supernatant and pellet samples were prepared with the same dilution in Laemmli sample buffer and run on an 8% SDS-PAGE gel, then stained with Coomassie Blue. Band intensities were measured and quantified using ImageJ, then analyzed in Microsoft Excel. A Student's t test was used to determine statistical significance.
Homology Modeling-HMP-1 amino acids 677-927 were used to perform a BLAST search against the Protein Data Bank (PDB) to identify a suitable template structure for homology modeling. The tail of metavinculin had the highest homology (PDB code 3MYI, Chain A). This structure was used as a template to model HMP-1 amino acids 699 -864 with SWISS-MODEL. The resulting PDB file was then viewed and analyzed using MacPyMOL (66).

Sequencing of Strong LOF hmp-1 Alleles Identifies Lesions
That Perturb the C-terminal ABD-To begin dissecting HMP-1/␣-catenin function in vivo, we obtained 8 largely uncharacterized hmp-1 alleles from a previous EMS mutagenesis screen (36). These alleles include zu242, zu244, zu349, zu402, zu365, zu399, zu404, and zu406. Using four-dimensional differential interference contrast confocal microscopy, every allele was found to exhibit a strong zygotic LOF Humpback phenotype as previously described (36) (supplemental Fig. S1). On rare occasions, ruptures from the ventral surface were observed for hmp-1(zu242), hmp-1(zu365), and hmp-1(zu399) (data not shown). In these cases, there appeared to be a failure of ventral cells to properly adhere at the ventral midline. Late posterior ruptures were also occasionally observed for hmp-1(zu365) and hmp-1(zu399), in which cells begin to leak out of the posterior end of the embryo after the embryo has formed dorsal humps (data not shown).
Each of the alleles originally identified by Costa et al. (36) is recessive and zygotic lethal; however, we noticed that for certain alleles, a portion of homozygous mutant progeny were no longer contained in an eggshell (supplemental Fig. S2A) and in some cases exhibited muscle activity (data not shown). As embryos mature in the eggshell, they begin to secrete enzymes that aid in hatching (40). Although such embryos therefore appear to have lived long enough to degrade the eggshell, careful observation over time confirmed these hatched embryos ("hatchoids") eventually die. The percentage of hmp-1 homozygous mutants that become hatchoids varies among alleles (supplemental Fig. S2B) and indicates that there are subtle differences in the overall time of death.
Next we sequenced the hmp-1 locus in homozygous hmp-1 mutants to determine the genetic lesion associated with each allele ( Fig. 1 and supplemental Table 1). Despite analyzing the genomic region 5 kbp upstream and 0.3 kbp downstream of hmp-1, we were unable to identify the molecular lesions that correspond to hmp-1(zu242), the only allele for which no protein expression was detectable (see Fig. 2), or hmp-1(zu349). We presume that hmp-1(zu242) and hmp-1(zu349) involve mutations in enhancer elements that lie far from the hmp-1 locus itself.
Several new truncation mutants were identified through sequencing. hmp-1(zu404) and hmp-1(zu406) both have nonsense mutations in the N-terminal half of hmp-1 and are predicted to express truncated proteins at amino acids 433 and 187, respectively (Fig. 1). A nonsense mutation was also identified in hmp-1(zu399), which is predicted to create a truncated protein in the ABD at amino acid 794 (Fig. 1). This mutation is one residue before the truncation previously reported for hmp-1(zu278) (21). The most C-terminal mutation we identified was hmp-1(zu365) (Fig. 1), and is predicted to encode another premature stop codon at residue 826. It is clear from these truncation mutants that the ABD of HMP-1 is critical for proper epidermal morphogenesis, especially the C-terminal tail, from amino acids 827 to 927.
Only one missense mutation was identified among the alleles, hmp-1(zu402) (Fig. 1). This mutation is also in the ABD (Fig. 1) and is upstream of another known missense mutation, hmp-1(fe4) (37). Because the S823F mutation in hmp-1(fe4) was found to be in a conserved SLIQ motif (37), we used a multiple sequence alignment to determine whether the V802M mutation in hmp-1(zu402) also lies in a conserved motif. Indeed, this mutation occurs at a valine residue in an SKVKA motif that is absolutely conserved in ␣-catenins from five diverse species, including humans, and is also conserved in human vinculin (supplemental Fig. S3). Another intriguing lesion was found in hmp-1(zu244); a 56-amino acid in-frame deletion at the extreme N terminus of the ABD, ⌬687-742 (Fig. 1). This 56amino acid deletion leaves the C-terminal tail intact, yet homozygotes display 100% embryonic lethality. This is the first evidence that the N-terminal portion of the ABD is critical to ␣-catenin function. In summary, HMP-1 residues 687-742 and 802, as well as the C-terminal tail, residues 827-927, are important for proper epidermal morphogenesis.
Strong LOF hmp-1 Alleles Express Mutant Protein-Our sequence analysis suggests that the HMP-1 ABD is essential for function, as various mutations or deletions in the C terminus affect epidermal morphogenesis. It is possible that nonsensemediated decay mechanisms prevent some of these deleterious mutant proteins from being expressed, or that the protein products are mislocalized, so we next examined protein expression and localization in the mutant alleles. Using anti-AJM-1 antibody as a marker for cell-cell contacts (Fig. 2, A-E and K-N), double antibody staining of hmp-1 homozygous mutants with anti-HMP-1 polyclonal antibodies revealed variations in HMP-1 protein expression and localization among alleles (Fig.  2, F-J and O-R). In a WT embryo (Fig. 2, A and F), both AJM-1 ( Fig. 2A) and HMP-1 (Fig. 2F) localize to seam-dorsal, seamseam, and seam-ventral contacts, although AJM-1 is basal to HMP-1 (41,42). In hmp-1(zu242) (Fig. 2, B and G), punctate Although two lesions lie upstream of the ABD (zu406 and zu404), both of these are premature stop codons that would presumably delete the C terminus without affecting the localization signal (␤-catenin binding site). The lesions in all other alleles lie within the ABD and include not only nonsense mutations (zu399 and zu365), but also a small deletion (zu244) and a missense mutation (zu402).
Intragenic Suppressors of hmp-1(fe4) All Map to the ABD-Homozygotes of a hypomorphic allele, hmp-1(fe4), exhibit only ϳ70% embryonic lethality compared with 100% observed with strong LOF mutants. This feature has made hmp-1(fe4) a valuable tool in identifying genetic interactors through suppressor and enhancer screens (37,(43)(44)(45)(46). Using EMS mutagenesis, we identified several dominant intragenic suppressors of hmp-1(fe4) (Fig. 4 and supplemental Table S2). Suppressor mutations were outcrossed using a strain homozygous for both hmp-1(fe4) and an integrated GFP transgene (mls10), both on chromosome V. Although some suppressors are clearly extragenic, extragenic suppressors partially characterized thus far appear to have no phenotypes and have not been analyzed further. 7 Many suppressors, however, mapped to the same genetic region as hmp-1. Outcrossing and subsequent sequencing confirmed these lesions are not revertants of hmp-1(fe4) and that they are located in the C-terminal coding region of hmp-1 (Fig.  4). Although most of the intragenic suppressors are missense mutations, a few alleles are nonsense mutations C-terminal to the VH3 domain (Fig. 4). Multiple sequence alignment reveals a cluster of suppressors in a conserved FTRG motif, but the rest appear to be scattered among other conserved residues in the ABD (supplemental Fig. S3).
Lesions in hmp-1(fe4) and the Intragenic Suppressors Identify Specific Residues That Modulate HMP-1 F-actin Bindinghmp-1(fe4) encodes a missense mutation, S823F, which does not affect accumulation of junctional proximal actin in mutant embryos but does cause thicker and sparser CFBs (37). The equivalent mutation in vertebrates, S817F, was reported to increase the avidity of the ␣-catenin ABD for F-actin by 37% in a high speed pelleting assay (47). This is surprising given the strong evidence that hmp-1(fe4) is a weak LOF mutant (37,(43)(44)(45)(46). If HMP-1 protein from hmp-1(fe4) homozygous mutants indeed has a higher binding affinity for F-actin than WT, and this is the cause of the thicker CFBs, then the intragenic suppressor mutations would be predicted to decrease the binding affinity to near WT levels. Conversely, if the hmp-1(fe4) mutation decreases F-actin binding affinity, the suppressors would be predicted to increase binding affinity. To determine how these secondary mutations restore proper HMP-1 function, we purified bacterially expressed protein consisting of HMP-1 amino acids 677-927 fused to GST, or the same fragment carrying the hmp-1(fe4) S823F mutation alone, as well as in tandem with the second site mutations identified in two of the intragenic suppressor mutants (N853K, found in hmp-1(fe29); F735C, found in hmp-1(fe27)). We showed previously that a recombinant protein containing HMP-1 residues 677-927 binds avidly to F-actin in a high speed actin cosedimentation assay (21), so we employed the same technique here to determine the avidity of these mutant constructs for F-actin (Fig. 5, A and B). We found a statistically significant decrease in the percentage of HMP-1 ABD bound to F-actin (pellet fraction ϭ P; Fig. 5A) when the S823F mutation found in hmp-1(fe4) is present compared with WT (Fig. 5B). This is most prominent at 2 M F-actin, where binding is decreased an average of 34.1%, but even under saturating conditions (10 M F-actin), binding is still decreased an average of 7.3%. When the hmp-1(fe4) S823F mutation is in tandem with either the hmp-7 C. Lockwood and J. Hardin, unpublished data.  -1(zu242) and hmp-1(zu406) mutant embryos. A-I, staining with fluorophore-conjugated phalloidin to visualize F-actin organization in homozygous hmp-1 embryos. Anterior is left, dorsal is up. Scale bar ϭ 10 m. Yellow carets indicate the underlying muscle; red arrows point to junctional proximal actin near adherens junctions or a lack thereof; blue arrowheads mark gaps between CFBs. In the wild-type embryo (A), parallel CFBs are visible as well as junctional proximal actin, which has concentrated along the seam-dorsal boundary (red arrow). B-I, in all of the hmp-1 mutants there is a lack of junctional proximal actin, leaving gaps at the seam-dorsal and seam-ventral boundaries (red arrows). These gaps are especially pronounced in B (right red arrow) and I. The mutants also display large gaps between CFBs (blue arrowheads) as well as thicker bundles. Thicker bundles or large accumulations of F-actin appear more pronounced in B. FEBRUARY 22, 2013 • VOLUME 288 • NUMBER 8 JOURNAL OF BIOLOGICAL CHEMISTRY 5699 1(fe29) N853K mutation or the hmp-1(fe27) F735C mutation, there is also a statistically significant increase in the amount of HMP-1 ABD bound to F-actin (P, Fig. 5A) when compared with hmp-1(fe4) S823F alone (Fig. 5B). At 2 M F-actin, the hmp-1(fe29) N853K and hmp-1(fe27) F735C mutations increase actin binding an average of 15.3 and 16.1%, respectively. Although at this concentration the intragenic suppressor mutations do not restore actin binding to fully wild-type levels, at 10 M F-actin the double mutants are indistinguishable from wildtype. In summary, hmp-1(fe4) S823F decreases the ability of HMP-1 to bind F-actin, and the intragenic suppressor mutations revert this binding to nearly wild-type levels. Based on these results, we conclude that the fe4 intragenic suppressors identify amino acids that are capable of modulating the ability of HMP-1 to bind F-actin.

Structure-Function Analysis of HMP-1/␣-Catenin
Homology Modeling Identifies Conserved C-terminal Amino Acids Required for HMP-1 to Bind Actin-Locating the residues in the C terminus of HMP-1 that we identified genetically within the context of a three-dimensionally folded protein would provide significant further insight into how these residues function in vivo. No crystal structure exists for the ABD of ␣-catenin or HMP-1; however, homology of the C terminus relative to the solved structure of the vinculin tail domain has proven useful in the past (47), so we pursued this approach with HMP-1 using SWISS-MODEL (48). We used the protein sequence of the HMP-1 ABD domain to search the Protein Data Bank for a suitable template for homology modeling of the C terminus. Not surprisingly, our search recognized the tail of metavinculin, a muscle-specific isoform of vinculin, as the highest homology sequence. The tails of both vinculin and metavinculin form a five-helix antiparallel bundle (49,50). Our homology model of the HMP-1 VH3 domain also identifies five ␣-helices (Fig. 6). In Fig. 6, these helices are numbered 1-5 from the N to C terminus and colored as blue (1), cyan (2), green (3), orange (4), and red (5).
We then mapped the C-terminal amino acids mutated in the strong LOF hmp-1 alleles onto our homology model (Fig. 6, A  and B). Residue 742 (gray, Fig. 6A) marks the last C-terminal amino acid deleted in hmp-1(zu244). Both helices 1 and 2 (Fig.  6A) would be deleted in these allele, whereas leaving the rest of the ABD intact. Interestingly, the missense mutation identified at residue 802 in hmp-1(zu402) lies adjacent to residue 823, the site of the missense mutation in hmp-1(fe4) (37) (cyan and purple, respectively; Fig. 6B). This implicates these specific residues in proper actin binding and HMP-1 function. The premature stop codon found at amino acid 794 in hmp-1(zu399) would completely remove this region, as well as the rest of the C terminus (magenta, Fig. 6B). Another nonsense mutation, identified at residue 826 in hmp-1(zu365), lies directly C-terminal to the hmp-1(fe4) site and may also perturb this region directly (yellow, Fig. 6B). Moreover, the model suggests a possible explanation for why the hmp-1(zu244) mutation affects actin bind-  ing: if this face of the C terminus (helices 4 and 5; Fig. 6B) acts as a discrete binding site for F-actin, deletion of the N-terminal helices 1 and 2, as occurs in hmp-1(zu244) (Fig. 6A), may abrogate actin binding by changing the spatial orientation of the C-terminal helix 5 (Fig. 6B). It is also possible that N-terminal helices 1 and 2 are part of a continuous binding site distributed among the C-terminal helices 4 and 5 (Fig. 6B), and its removal abrogates strong F-actin binding.
We also used our homology model to investigate why second site, intragenic missense and nonsense mutations alleviate the phenotypic defects of hmp-1(fe4) mutants (Fig. 6C). Although the hmp-1(fe4) mutation at residue 823 (purple), and one suppressor mutation at 821 (magenta), lie on helix 5, a cluster of intragenic suppressor sites are located directly adjacent, at the end of helix 2 (Fig. 6C). As suggested for the LOF hmp-1 alleles (Fig. 6, A and B), mutations in the N-terminal helix 2 may alter the position of the C-terminal helix 5 sufficiently to restore stable actin binding. Alternatively, N-terminal helix 2 may also be directly involved in binding to actin, and these mutations may increase the avidity of HMP-1 for F-actin enough to compensate for the original defect.
The two amino acids identified as nonsense mutations in hmp-1(fe9), hmp-1(fe24), and hmp-1(fe26), Trp-861 and Arg-862, are both located in the C-terminal arm past helix 5 that lies above N-terminal helices 1 and 2 (Fig. 6C). In vinculin, this C-terminal arm binds to acidic phospholipids and causes a conformational change in the tail that potentially relieves inhibition by the head domain, which may activate its ability to bind other proteins, such as talin (49). Removal of this C-terminal arm when only the vinculin tail is present does not affect actin binding (49), indicating that the conformational change in the tail apart from its context in the whole protein is not sufficient to affect actin binding. If the hmp-1(fe4) S823F mutation perturbs F-actin binding due to increased steric hindrance or blockage of other needed residues, then truncation of the C-terminal arm may result in a constitutively active HMP-1 with a more flexible ABD that allows a reorientation of residue 823. Binding of the C-terminal arm of HMP-1, which lies outside the VH3 domain, by a currently unknown ligand might also result in release of the intramolecular inhibition we previously reported for full-length, wild-type HMP-1 in vitro (21). It should be noted that an ABD construct lacking this same C-terminal arm region of ␣E-catenin(678 -864) was not able to bind F-actin in vitro (18), which may indicate HMP-1 has retained functional similarity to vinculin in this capacity compared with ␣E-catenin.
The Putative Vinculin-binding Domain of HMP-1 Is Not Required for Epidermal Morphogenesis But Does Affect HMP-1 Mobility-The results we have reported thus far and those reported previously (21) clearly show that the HMP-1 ABD is critical for proper C. elegans development, but the function of more N-terminal domains is not well understood. We have previously shown that expression of a hmp-1::gfp construct driven by the endogenous hmp-1 promoter can rescue a strong LOF mutant, hmp-1(zu278), which contains a Q795Stop mutation. However, deletion of the putative ␤-catenin binding site in this construct (HMP-1(⌬13-185)::GFP) results in its mislocalization to the cytoplasm, and as a result, it is not able to rescue homozygous hmp-1(zu278) mutants (21). To determine the importance of other regions, the same full-length hmp-1::gfp was used to create targeted in-frame deletions of other internal domains of HMP-1. These deletion constructs were expressed in the same strong LOF background, hmp-1(zu278), and homozygous mutants were identified by the absence of a pharyngeal GFP balancer. Transgenic mutant embryos were then analyzed using four-dimensional spinning disc confocal microscopy (Fig. 7). FIGURE 6. Homology modeling of the HMP-1 VH3 domain provides insights into the possible function of certain regions and residues in the C terminus. A-C, using the metavinculin VH3 domain crystal structure as a template (PDB code 3MYI, Chain A), the HMP-1 VH3 domain was modeled using SWISS-MODEL (48). Images were created using MacPyMOL (66). The tertiary structure of the VH3 domain contains a series of ␣-helices. The helices are numbered 1-5 from the N to C terminus and colored as blue (1), cyan (2), green (3), orange (4), and red (5), respectively. The residues mutated in hmp-1 alleles are shown as space-filled molecules labeled with the amino acid number. A, residues mutated in identified strong LOF hmp-1 alleles. Amino acid 742 marks the C-terminal end of the deletion found in hmp-1(zu244). The N-terminal limit of this deletion is beyond the VH3 domain, thus both helices 1 and 2 would be missing in this allele. B, a second view rotated around the vertical axis of the structure to more clearly show the locations of residues mutated in other strong LOF alleles. Note here that the missense mutation at residue 802 identified in hmp-1(zu402) is adjacent to another well characterized missense mutation at residue 823, hmp-1(fe4). C, residues mutated in hmp-1(fe4) and the intragenic suppressor alleles. Residue 879 (fe30) is not shown because it lies C-terminal to the sequence we were able to model. Note the cluster of suppressor mutation sites between helices 2 and 3, which lie behind the hmp-1(fe4) mutation site at residue 823. In addition, several suppressor sites are located on a C-terminal arm after helix 5 that lies above helix 1/2. Each of these clusters may indicate that the orientation of helix 5 is important for proper F-actin binding.
At the onset of elongation, all of the transgenes localize to seam-seam, seam-ventral, and seam-dorsal junctions (Fig. 7, A-D and I-K), unlike HMP-1(⌬13-185)::GFP, which does not (21). We next assessed the ability of the constructs to rescue beyond the 1.5-fold stage of elongation. For comparison, a mutant embryo expressing the full-length HMP-1::GFP elongates 2-fold and maintains continuous transgene localization along cell-cell contacts (21) (Fig. 7E). Not surprisingly, constructs without the ABD are not able to rescue hmp-1(zu278) mutants, and localization becomes discontinuous as junctions become destabilized (Fig. 7, F-H). Two N-terminal regions not predicted to be involved in binding to ␤-catenin were also deleted, residues 13-72 and 194 -306. Constructs lacking either of these regions properly localize to junctions (Fig. 7, J, K, M, and N), but surprisingly, are not able to rescue the mutant phenotype, even though they contain a functional ABD (Fig. 7,  M and N). Because HMP-1(⌬13-185)::GFP does not localize to junctions (21), but HMP-1(⌬13-72)::GFP does, we conclude that residues 73-185 are required for proper localization of HMP-1.
The vinculin-binding domain of ␣E-catenin has been the subject of intense interest, because evidence suggests it is only exposed in response to tension (22). Interestingly, deletion of the corresponding domain of HMP-1, residues 315-494, did not effect the ability to rescue homozygous hmp-1 mutants; localization of the transgene remains continuous as the mutant embryo elongates to 2-fold (Fig. 7L), similar to full-length HMP-1::GFP (Fig. 7E). The rescue by HMP-1(⌬315-494)::GFP is complete, and the transgenic homozygous mutants survive to adulthood and have viable progeny. Like HMP-1::GFP (21), HMP-1(⌬315-494)::GFP rescued adults exhibit an uncoordinated phenotype, but are otherwise developmentally normal with no body morphology defects (data not shown). We conclude that residues 315-494 are not essential for HMP-1 function during epidermal morphogenesis.

DISCUSSION
We have previously shown that HMP-1/␣-catenin has conserved functions in adhesion-dependent processes during C. elegans embryogenesis: the N terminus is required for localiza- tion to cell-cell contacts and the C-terminal ABD is essential for F-actin recruitment (21). Here we extended this analysis to identify other essential domains and important residues required for HMP-1 function during epidermal morphogenesis. Determining the molecular basis of ␣-catenin function in a simple metazoan may in turn help to elucidate its functions in higher organisms.
Our analysis of strong LOF hmp-1 alleles confirms and extends previous studies. Truncation mutants that delete portions of the ABD at amino acids 794, 795, and 826, as well as the missense mutations at 802 and 823, confirm that the C-terminal portion of the ABD is critical for proper F-actin attachment at adherens junctions. We also identified an in-frame deletion (amino acids 687-742) that leaves the rest of the C terminus intact but fails to function in vivo. This is the first evidence for a crucial role for the N-terminal-most region of the ABD. Although the available truncation mutants indicate that at least residues 827-927 are important for proper HMP-1 function, our previous in vitro work has shown that removal of the nonconserved tail of HMP-1, residues 905-927, from an ABD construct (677-904) has no effect on actin binding in a high speed pelleting assay (21). The novel missense mutation at amino acid 802, in combination with the most C-terminal truncation at 826, suggests that a more discrete region within residues 802-904 may facilitate actin binding in combination with residues 687-742. It is possible that within the tertiary structure of the protein, these two regions form one continuous actin-binding interface. No crystal structures for the C terminus of HMP-1 or ␣E-catenin have been solved; however, our homology model of the HMP-1 VH3 domain supports this hypothesis, showing that these regions are adjacent (Fig. 6, A and B). It is also of interest to note that the hmp-1(fe4) mutation lies within one region, whereas a cluster of intragenic suppressor mutations lies near the other (Fig. 6C).
Our homology model also supports another, more speculative possibility that could also explain how regions 802-904 and 687-742 facilitate actin binding. Helices 1 and 2 (Fig. 6A), containing residues 687-742, may be one of two discrete F-actinbinding sites, the other comprising residues 802-904. In this case, the full function of HMP-1 would require both sites, not only to allow binding of F-actin, but also bundling. Previous studies have shown the ability of both the N and C termini of ␣E-catenin to bind F-actin (51), and ␣E-catenin is known to homodimerize (14,52,53), providing two mechanisms for its bundling activity in a full-length protein. We have previously shown, however, that full-length HMP-1 does not form a homodimer in vitro (21), making HMP-1 distinct from mammalian ␣E-catenin, which binds actin more avidly as a homodimer (52). The possibility of dual actin binding sites within the HMP-1 C terminus may explain why HMP-1 does not need to homodimerize. Alternatively, a currently unidentified event may be needed to relieve the intramolecular inhibition observed in full-length HMP-1 in vivo (21). Although dual actin binding sites in the tail have been suggested for vinculin (54 -56), this mechanism has never been suggested for ␣-catenin; further experiments would be needed to adequately test this theory.
Recently it has been shown that mammalian vinculin binds to ␣E-catenin in a force-dependent manner (22). Under lowstress conditions, an inhibitory region of ␣E-catenin blocks the vinculin-binding site. Upon F-actin binding to the C terminus of ␣E-catenin and subsequent actomyosin contraction, this inhibitory region is displaced and vinculin is able to bind (22). Subsequent studies in cultured endothelial cells do not support a major role for such tension-dependent vinculin recruitment, however, and even show that the vinculin-binding site of ␣-catenin is not required for basic adhesion (57). Our studies are the first analysis to address the importance of this conserved domain of ␣-catenin in an intact organism, and show that HMP-1 residues 315-494 are not necessary for viability. DEB-1/vinculin is not expressed in the epidermis of C. elegans (58), which had previously suggested that this domain is not involved in vinculin recruitment in a more simple metazoan. Our results confirm that this domain plays a relatively unimportant role during morphogenesis. It is not clear at this time if the proposed force-dependent regulation of ␣E-catenin is conserved among the other mammalian isoforms of ␣-catenin, but it may be that ␣E-catenin has acquired specializations not found in other vertebrate ␣-catenins or those of other metazoans.
The domain of ␣E-catenin that corresponds to amino acids 315-494 of HMP-1 can bind several other actin-related proteins, including ␣-actinin (15, 29), formin-1 (31), and afadin Although there is no statistically significant difference in the percent mobile fraction for the two transgenes (p ϭ 0.72), there is a significant difference in the half-lives as determined by a two-tailed t test (p Ͻ 0.01). Although the transgenes eventually exhibit the same level of recovery, the HMP-1(⌬315-494)::GFP recovers at a much slower rate. (18). Like vinculin, the C. elegans homolog of ␣-actinin, ATN-1, is not expressed in the epidermis (59). There are several formin candidates in C. elegans, including FHOD-1, FHOD-2, INFT-1, and FRL-1. Knockdown of each homolog by RNAi does not result in embryonic lethality or body shape defects, and neither does double knockdown of fhod-2 and frl-1 (60), suggesting functional redundancy among formins in C. elegans. However, fhod-1(RNAi) was found to enhance the lethality of weak let-502/ROCK mutants and suppress the lethality of mel-11/myosin phosphatase mutants, two genes with critical functions during embryonic elongation (61-63). 8 fhod-1(RNAi) does not enhance the penetrance or lethality of hmp-1(fe4), 9 and recent evidence suggests that FHOD-1 is muscle-specific (64). Moreover, the in vivo relevance of formin-1 in vertebrates is unclear. Recent evidence in mice suggests that the endogenous protein is not in fact localized to adherens junctions, but instead to the cytoplasm and microtubules (65). Therefore, thus far there is no evidence to suggest that the function of formin-1 depends on HMP-1/␣-catenin.
RNAi knockdown of the only afadin homolog in C. elegans, afd-1, results in very low lethality in a WT background, but enhances the penetrance and lethality of hmp-1(fe4) to near 100% (45). AFD-1 is also expressed in the epidermis and localizes to cell-cell contacts. However, localization of a significant fraction of AFD-1 depends on the inverted MAGUK protein, MAGI-1, which localizes to a distinct domain basal to the cadherin-catenin complex (46): it is unclear if some AFD-1 physically interacts with HMP-1. In any case, our finding that amino acids 315-494 in HMP-1 are not required suggests that if HMP-1 and AFD-1 interact via this domain, this interaction is quantitatively minor. The ability of this domain in vertebrates to bind other actin-binding proteins may have become more important later in evolution to further stabilize attachment of F-actin to adherens junctions.
Although amino acids 315-494 are not essential, we have shown impaired mobility of HMP-1(⌬315-494)::GFP at the junction compared with full-length HMP-1::GFP. This result suggests that amino acids 315-494 may modulate the efficiency of targeting HMP-1 to sites of cell-cell contact. Although it is possible that deletion of this domain simply causes misfolding of the protein, thereby impeding its mobility, the fact that this construct can completely rescue all lethal developmental defects in hmp-1(zu278) homozygotes suggests that this is not the case. If amino acids 315-494 do facilitate targeting of HMP-1, our FRAP results indicate that they only affect the rate of deposition of mobile HMP-1, not the overall percentage of mobile protein.
In conclusion, our studies have substantially refined our understanding of the function of conserved residues in the C terminus of ␣-catenin family members that are required for F-actin binding and epithelial morphogenesis. Our results in turn will allow further focus on the detailed mechanisms by which this key family of proteins links the cadherin complex to the actin cytoskeleton in all metazoans.