Analysis of the alpha-actinin/zyxin interaction.

The yeast two-hybrid system was used to search for interaction partners of human zyxin. Screening of two different cDNA libraries, one prepared from human placenta, the other from human heart, yielded several positive clones that occurred in both searches, including clones coding for cyclophilin, nebulette, and alpha-actinin. The zyxin/alpha-actinin interaction was analyzed in detail. By site-directed mutagenesis, a linear motif of 6 amino acids (Phe-Gly-Pro-Val-Val-Ala) present at the N terminus of zyxin was found to play a critical role. Replacement of a single amino acid within this motif abolished binding to alpha-actinin in blot overlays as well as in living cells. On the other hand, the interaction site in alpha-actinin was mapped to a conformational determinant present in the center of the protein as demonstrated by a fragment deletion analysis. This binding site involved a tandem array of two complete spectrin-like domains. Only fragments that were able to dimerize in yeast also bound to zyxin, suggesting that dimerization of alpha-actinin is essential for zyxin binding.

The actin cytoskeleton of a eukaryotic cell plays a central role in a variety of cellular processes, including cell motility, migration, phagocytosis, intracellular transport, and maintenance of cell polarity. To accomplish such critical functions, the assembly and disassembly of the actin filaments must be tightly controlled, both with respect to time and space. Regulatory adapter proteins harboring specific interaction domains (SH2, SH3, PDZ domains) 1 seem to play an important role in the control of actin filament assembly.
One of the adapter proteins that has attracted considerable attention is zyxin (1). Zyxin was originally identified in chicken fibroblasts as a protein associated with focal adhesions, stress fibers, and cell-cell adherence junctions (2). It represents a monomeric protein with an apparent molecular mass of 82 kDa that is phosphorylated at multiple sites. The mammalian homologue of chicken zyxin has recently been cloned from a subtracted cDNA library by virtue of its reduced expression in SV40-transformed human fibroblasts (3). At the same time, human and mouse zyxin cDNA clones were isolated from normal cDNA libraries by cross-hybridization (4). Sequencing studies demonstrated that the avian and the mammalian zyxin display a similar, modular structure with a proline-rich N terminus, a nuclear export signal, and three C-terminal LIM domains, although they share less than 60% sequence identity.
Zyxin has been localized to the barbed ends of actin filaments, especially in lamellipodia, filopodia, and focal adhesions, suggesting that it is involved in the organization of the actin cytoskeleton (5). This notion is supported by its structural relationship with the protein ActA from the bacterial pathogen Listeria monocytogenes (6). This microorganism invades eukaryotic cells and exploits the actin cytoskeleton of the host for its own motility (7). Reorganization of the cytoskeleton seems to be accomplished by the bacterial surface protein ActA, which enhances the nucleation of new actin filaments. The structural and functional similarities of zyxin with the central and Cterminal region of ActA suggest that zyxin might also be involved in the assembly and control of the actin cytoskeleton (6).
Zyxin interacts with a variety of cytoskeletal and regulatory proteins, and some of these interactions have been mapped to individual domains of the zyxin polypeptide. It binds to ␣-actinin, an actin cross-linking protein enriched in focal adhesion plaques (8). The interaction site has recently been identified as the extreme N terminus of zyxin by deletion analysis and peptide inhibition studies (9,10). When the ␣-actinin binding site was deleted, the association of zyxin with focal adhesion plaques was largely impaired, emphasizing the physiological role of the zyxin/␣-actinin interaction (10). Zyxin also binds to members of the Ena/VASP family of proteins, which are known to control microfilament organization (11,12). A cluster of four proline-rich motifs present in the N-terminal domain of zyxin seems to be responsible for this VASP-zyxin interaction (13). The proline-rich region of zyxin has also been demonstrated to act as a docking site for the proto-oncogene product Vav (14). Vav is a regulatory protein controlling the activity of small GTP-binding proteins in blood cells. Finally, zyxin binds to members of the cysteine-rich protein family CRP (5). This interaction is accomplished by the first of the three LIM domains found at the C terminus of zyxin (15).
The two-hybrid system offers an elegant approach to analyze protein-protein interactions in yeast (16). It is based on the modular structure of the yeast transcription factor GAL4, which, if reconstituted by interaction of two fusion proteins, will activate transcription of suitable reporter genes (HIS3, lacZ). The advantage of the system is that large numbers of cDNA clones for potential ligands can be screened and that the interactions are tested under physiological conditions within eukaryotic cells.
We have made extensive use of the yeast two-hybrid system to search for additional binding partners of zyxin. In this way we were able to identify a number of potential ligands, including proteins from the cytoskeleton and the transcription machinery, which might be directly or indirectly involved in the formation of actin filaments and focal adhesion plaques.

EXPERIMENTAL PROCEDURES
Yeast Two-Hybrid System-Yeast two-hybrid screenings were performed utilizing the matchmaker 2 system of CLONTECH Laboratories (Palo Alto, CA). Selected regions of the human zyxin cDNA (GenBank accession number X95735 (3)) were prepared by restriction enzyme digestion or by PCR utilizing the primers outlined above in Table I and cloned into the bait vector pAS2-1. The plasmids were transfected by the lithium acetate method into the yeast reporter strain Y190 together with a cDNA library prepared from human placenta (CLONTECH, HL4025AH) or human heart (HL4042AH) in the prey vector pACT2. Selection for HIS3 reporter gene activation was performed on agar plates lacking histidine, tryptophan, and leucine. Colonies appearing after 5-10 days at 30°C were assayed for ␤-galactosidase activity utilizing the colony-lift filter assay. For quantitative data, the colonies were grown in liquid medium and assayed for ␤-galactosidase activity using O-nitrophenyl ␤-D-galactopyranoside as substrate. The plasmid DNA of positive colonies was isolated with phenol and glass beads as suggested by the supplier of the system. Positive two-hybrid protein interactions were verified by transfection of the plasmids back into the reporter strain Y190 together with the original bait or with selected controls. The plasmids of positive colonies were amplified in Escherichia coli XL-1 blue and sequenced by the dideoxy chain termination method with Sequenase 2.0 (Amersham Pharmacia Biotech). All sequences were analyzed with the GCG computer program package of the University of Wisconsin.
Selected fragments of the ␣-actinin cDNA (GenBank accession number X15804 (17)) were prepared in a similar way by PCR (Table I) and cloned into the prey vector pACT2. Two-hybrid interactions with zyxin were assayed as described above. To test for self-interaction (dimerization), the cDNAs for several ␣-actinin fragments were cloned into the bait vector pAS2-1 and cotransfected into yeast with the prey vector pACT2 carrying the same fragment.
Site-directed Mutagenesis-The codons for individual amino acids were mutated by the ExSite PCR-based mutagenesis method (18) em-ploying the cDNA of zyxin ligated into pUC19 as a template. The forward primer harbored the desired mutation, whereas the reverse primer was phosphorylated at its 5Ј-end and selected in a way that it annealed directly adjacent to the 5Ј-end of the forward primer (Table I). After amplification by PCR, the maternal DNA was removed by digestion with the restriction enzyme DpnI (Roche Molecular Biochemicals, Switzerland). The ends of the linear products were joined by ligation with T4 DNA ligase (Roche Molecular Biochemicals) and the nicked plasmids were transfected into competent bacteria (E. coli XL-1 blue). Authenticity and reading frame of all mutated clones were verified by DNA sequencing.
GST Fusion Protein Expression and Blot Overlays-The cDNA sequences for wild type and mutated zyxin (residues 1-42) were subcloned into the BamHI/XhoI restriction site of the expression vector pGEX-5X-2 (Amersham Pharmacia Biotech) downstream of the gst gene and transfected into competent bacteria (E. coli BL21). Fusion proteins were expressed after induction with isopropylthio-␤-galactoside as suggested by the supplier of the gst gene fusion system (Amersham Pharmacia Biotech). The bacteria were collected by centrifugation and lysed by sonication. Fusion proteins were purified from the lysates by affinity chromatography on reduced glutathione-Sepharose and analyzed on 15% SDS-polyacrylamide gels. After transfer to nitrocellulose by electroblotting, the polypeptides were detected with the GST detection module (Amersham Pharmacia Biotech) using goat anti-GST antibodies, followed by alkaline-phosphatase-conjugated secondary antibodies (Sigma Chemical Co.). The color reaction was performed with bromochloroindolyl phosphate and nitroblue tetrazolium as substrate. A similar blot prepared in parallel was blocked with bovine serum albumin and incubated with radioiodinated ␣-actinin in 10 mM NaCl, 1 mM EDTA, 1% Nonidet P-40, 0.1% 2-mercaptoethanol, 20 mM HEPES, pH ␣-Actinin/Zyxin Interaction 7.5, as described previously (8,10). After 4 h at room temperature, the blot was washed twice with the same buffer and exposed to BioMax MS film (Eastman Kodak Co.).
Cell Culture and GFP Fusion Protein Expression-Primary chicken fibroblasts were prepared from tendons of 17-day-old chicken embryos with the help of collagenase (Roche Molecular Biochemicals) as described previously (19). COS-1 cells (CRL-1650) were purchased from the American Type Culture Collection (Manassas, VA). The cells were cultivated under an atmosphere of 5% CO 2 in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 100 units/ml penicillin, and 100 g/ml streptomycin.
Wild type and mutated zyxin sequences were cloned into the EcoRI/ BamHI site of the expression vector pEGFP-C3 (CLONTECH) downstream of the GFP reporter gene. The plasmids (1 g/well) were mixed with 100 l of Opti-MEM 1 (Life Technologies) containing 3 l of FuGENE-6 reagent (Roche Molecular Biochemicals) and added to the cells grown to 60% confluence in 6-well plates. One day after transfection, the cells were washed with phosphate-buffered saline and inspected under a Zeiss Axiovert microscope equipped with epifluorescence optics.
Transfected cells were examined for the distribution of endogenous zyxin, VASP, and vinculin by indirect immunofluorescence as described previously (10). Rabbit antisera against human VASP (M4) and porcine zyxin (11) were kindly provided by Dr. M. Reinhard (University of Wü rzburg, Germany) and used at a dilution of 1:500 and 1:100, respectively. A monoclonal antibody against human vinculin was purchased from Sigma (St. Louis, MO) and used at a dilution of 1:400.

Two-Hybrid
Screening-Initial studies utilizing the complete zyxin cDNA for two-hybrid analysis suggested that zyxin itself must have transcription-activating properties. Yeast cells transfected only with the bait plasmid coding for the DNA binding domain of GAL4 fused to the full-length zyxin cDNA activated transcription of the HIS3 and the lacZ reporter genes and grew on selective agar plates. The zyxin sequence was therefore cut into several fragments, and each fragment was tested separately for its autonomous activating potential (Fig.  1). Two regions were found to induce autonomous reporter gene activation. One region was situated between amino acid residues 345 and 362 and harbored the nuclear export signal KEVEELEQL, which was previously demonstrated to play a role in the nuclear-cytoplasmic shuttling of zyxin (20). The other region was situated between residues 110 and 167 and contained two proline clusters.
The four fragments of zyxin that did not cause autonomous activation were used to screen a cDNA library prepared from human placenta in the prey vector pACT2. With the N-terminal fragment spanning residues 1-42, a large number of positive colonies were obtained that grew on histidine-deficient agar plates. In contrast, no meaningful clones were obtained with the other three fragments spanning residues 43-127, 289 -346, and 370 -572.
We therefore focused on the N terminus (residues 1-42) of human zyxin. Screening of 3 ϫ 10 6 transformants with this bait led to the isolation of 68 putative positive colonies that grew on histidine deficient plates and that transcribed the lacZ reporter gene as demonstrated by the colony-lift filter assay. The corresponding pACT plasmids were isolated and transfected back into competent cells along with the original bait plasmid, or alternatively, with an unrelated control plasmid that coded for lamin 5 fused to the DNA binding domain of GAL4. Of the 68 candidates tested, 47 induced reporter gene activation exclusively when cotransfected with the zyxin bait plasmid, but not the lamin 5 plasmid, suggesting that they represent truly positive clones.
Sequencing studies demonstrated that the 47 plasmids coded for a total of 12 independent proteins. These proteins could be grouped according to their putative function and their subcellular distribution in eukaryotic cells (Table II). Three proteins, namely ␣-actinin 1, nebulette, and the zyxin-related protein ZRP, are involved in the assembly of the cytoskeleton and the formation of focal adhesion plaques. One protein, GRIP 1, is associated with the plasma membrane. Two proteins, fibulin 2 and fibronectin, are typical members of the extracellular matrix. Cyclophilin B is a peptide isomerase that assists in the cis-trans isomerization of proline-rich segments. Sorting nexin 5 is involved in intracellular protein trafficking. Finally, PPAR FIG. 1. Autonomous transactivation of various zyxin fragments in the yeast two-hybrid system. DNA sequences coding for the zyxin fragments as indicated were ligated into the bait vector pAS2-1. The resulting plasmids were transfected into the yeast reporter strain Y190. Reporter gene activation was analyzed by growth on histidine-, tryptophan-, and leucine-deficient plates and by the colony-lift filter assay.
␥2 and BSPRY (35) are transcription factors localized in the cell nucleus.
One clone coded for a novel protein whose sequence has not yet been deposited in public data banks. Characterization of this protein will have to await the cloning of its complete mRNA. Our partial sequences indicate that it is related to the cytoskeletal proteins nebulin and nebulette.
A similar two-hybrid search was also performed with the N-terminal zyxin bait (amino acids 1-42) and 2 ϫ 10 6 transformants from a human heart library. In this case, seven positive colonies were obtained, which coded for a total of four different proteins. Sequencing studies revealed that these proteins were either identical or related to proteins identified above from the placenta cDNA library. These clones coded for ␣-actinin 2, nebulette, fibulin 2, and cyclophilin A (Table II).
Site-directed Mutagenesis-The isolation of several fulllength clones for ␣-actinin offered the possibility of analyzing the zyxin/␣-actinin interaction in greater detail. In a first set of experiments we focused on the binding site in zyxin, which according to our previous studies (10) should be situated within the N-terminal region comprising residues 21-42. Utilizing in vitro mutagenesis, individual amino acids of this region were replaced by serine, which carries an uncharged, hydrophilic side chain, and each construct was tested for its activity to interact with ␣-actinin in the two-hybrid system (Fig. 2). A quantitative assay was employed in which the accumulation of the reporter gene product ␤-galactosidase was determined colorimetrically. Mutation of Phe-26 3 Ser as well as Val-29 3 Ser yielded fusion proteins that were no longer capable of activating GAL4-based transcription, suggesting a complete loss of the zyxin/␣-actinin interaction. Likewise, mutation of Ala-31 3 Ser almost completely abolished the interaction, whereas replacement of amino acids at other positions between residues 17 and 39 barely compromised the interaction (Fig. 2). In fact, mutation of Phe-39 to Ser increased, rather than decreased, the Yeast reporter strain Y190 was cotransfected with the bait vector pAS2-1 containing the sequence for zyxin residues 1-42 and the prey vector pACT2 containing the sequence for ␣-actinin 1. The codons for some residues of wild type zyxin were replaced by codons for serine by site-directed mutagenesis as indicated. Colonies that grew on tryptophan-and leucine-deficient plates were analyzed for transcription of the ␤-galactosidase reporter gene using a quantitative, colorimetric assay. The results are expressed relative to the wild type zyxin sequence (100%) and represent the means with standard deviation from three independent determinations.   ␣-Actinin/Zyxin Interaction relative transcriptional activity. These results indicate that amino acids 26 -31 (Phe-Gly-Pro-Val-Val-Ala) represent the critical residues involved in the zyxin/␣-actinin interaction. Independent Verification-The results obtained by the twohybrid system were confirmed by the blot overlay technique (Fig. 3). The wild type N terminus of zyxin and its mutated forms were expressed as GST fusion proteins in bacteria, resolved on a polyacrylamide gel, and transferred to nitrocellulose. When incubated with radiolabeled ␣-actinin, the probe bound specifically to the wild type zyxin fusion protein, but barely to the mutated form in which Phe-26 had been replaced by Ser (Fig. 3). Consistent with the results from the quantitative two-hybrid analysis, mutation of Phe-39 to Ser appeared to enhance the relative binding activity to ␣-actinin. Thus, replacement of a single amino acid (Phe-26 3 Ser) inhibits the zyxin/␣-actinin interaction not only in yeast cells but also in a direct, biochemical assay.
To examine whether a single amino acid substitution would abolish the interaction also under physiological conditions, we prepared full-length fusion constructs with green fluorescent protein (GFP), which allowed the visualization of the subcellular distribution of zyxin in living cells (Fig. 4). Transfection of the wild type zyxin sequence into chicken fibroblasts yielded a positive signal along stress fibers and at focal adhesion sites consistent with the published distribution of zyxin (5, 10). When Phe-26 was replaced by Ser, the specific staining was abolished and the fusion protein distributed evenly throughout the cytoplasm of the cells (Fig. 4). Analogous results were obtained with COS cells. When Phe-39 was mutated to Ser, results similar to those obtained with the wild type sequence were obtained. In some experiments the staining at focal contacts appeared to be more strongly pronounced in agreement with the enhanced interaction of this construct with ␣-actinin in blot overlays. Thus, replacement of a single amino acid at the N terminus of zyxin (Phe-26 3 Ser) not only abolishes the binding of zyxin to ␣-actinin but also impairs the recruitment of zyxin to its normal subcellular sites.
In a separate set of experiments we examined whether transfection of zyxin in its wild type or mutated form (Phe-26 3 Ser) would lead to any alteration in the subcellular distribution of endogenous focal adhesion proteins (data not shown). To this end, cells were transfected with various zyxin constructs and stained, after fixation and permeabilization, with antibodies specific for zyxin, the zyxin-associated protein VASP, or the focal adhesion protein vinculin. No obvious changes in the distribution of these proteins could be detected, neither after transfection of the wild type nor the mutated construct.
Binding Site in ␣-Actinin-Next we focused on the other binding partner and analyzed the interaction site in the ␣-actinin molecule. A full-length cDNA that had been obtained during the initial two-hybrid screening was successively truncated from the 5Ј-end or the 3Ј-end, and the resulting constructs were analyzed for their potential interaction with zyxin by the two-hybrid system (Fig. 5). ␣-Actinin has a modular structure consisting of two calponin homology (CH) domains, four spectrin like (SPEC) repeats, and two C-terminal EF hands. A cDNA construct coding for the N-terminal half of the molecule but lacking SPEC domains 3 and 4 as well as the C-terminal EF hands was inactive. In contrast, a construct lacking the two CH domains but containing all the SPEC ␣-Actinin/Zyxin Interaction domains and the EF hands was active in the two-hybrid analysis. When the first SPEC domain was deleted, the interaction with zyxin persisted. However, binding was abolished when also the second SPEC domain was deleted. The zyxin binding site of ␣-actinin must therefore comprise part of the central SPEC domains. Constructs were then prepared that coded for two consecutive SPEC domains (Fig. 5). A positive reaction was obtained with SPEC domains 2 and 3 but not with the tandem arrays 1 and 2 or 3 and 4. Likewise, no interaction with zyxin was observed with any of the single-SPEC domains. Thus, zyxin binds to the center of the ␣-actinin molecule containing SPEC domains 2 and 3.
Binding to the ␣-Actinin Dimer-␣-Actinin is known to form aligned, anti-parallel dimers in which the four SPEC domains interact with each other via three extended ␣-helices (21,22). The resulting dimeric rod is extremely stable. Even fragments consisting only of SPEC domains 2 and 3 have been demonstrated to form correct dimers in vitro (21). Hence there are two possibilities to explain why zyxin interacts with a tandem array of SPEC domains 2 and 3 but not with the individual SPEC domains 2 or 3. Either zyxin binds to a sequence in the central region comprising part of SPEC domain 2 and part of SPEC domain 3, or alternatively zyxin interacts exclusively with the dimeric conformation of the SPEC domains. To distinguish between these two possibilities, we prepared several subfragments of the two SPEC domains and analyzed their interaction with zyxin by the two-hybrid system (Fig. 6). In a parallel series of experiments, the dimerization potential of the subfragments was tested by measuring their self-interaction employing the quantitative ␤-galactosidase assay. We found that zyxin interacted exclusively with the complete tandem array of SPEC domains 2 and 3, but not with any of the truncated forms derived thereof. As soon as a short piece of SPEC domain 2 or SPEC domain 3 was deleted, binding to zyxin was abolished. Consequently, it is unlikely that zyxin recognizes a sequence situated in the central region of the SPEC domains, because this sequence was preserved in each of the subfragments tested. On the other hand, the full tandem array of SPEC domains 2 and 3 was the only construct that formed stable dimers in yeast as indicated by the quantitative ␤-galactosidase assay (Fig. 6). When the first ␣-helix of SPEC domain 2 or the third ␣-helix of SPEC domain 3 was deleted, dimerization was grossly impaired. These results demonstrate that zyxin binds specifically to the dimeric assembly of the central ␣-actinin rod but not to any of the monomeric fragments derived thereof. DISCUSSION Zyxin is a cytoskeletal protein that plays an important role in the organization of actin filaments at focal adhesion sites. To fulfill this function zyxin makes multiple contacts with structural and regulatory proteins, but thus far only the interactions with the actin cross-linking protein ␣-actinin, the regulatory protein VASP, and the LIM protein CRP have been described in some detail.
In this study, we have made extensive use of the yeast two-hybrid system to search for additional interaction partners of zyxin. During the initial experiments it became evident that zyxin is a difficult bait to work with, because it has autonomous transactivation properties. By a fragment analysis approach, we could identify two regions of the protein that are responsible for this autonomous activation. One region contains two proline clusters but otherwise does not exhibit any structural motif that would explain its interaction with nuclear proteins of the RNA polymerase complex. The other region contains a nuclear export signal that has previously been identified by Nix and Beckerle (20).
Our search had therefore to be restricted to fragments that did not induce autonomous transactivation in the two-hybrid system. Searches with fragments comprising the three LIM domains or portions of the proline-rich N terminus have not been fruitful so far. Our search, however, was successful with the N-terminal fragment of human zyxin spanning residues 1-42. To facilitate the identification of truly positive clones, two different cDNA libraries, one prepared from human placenta, the other from human heart, were screened in parallel. Among all the positive clones obtained, there were four that coded for the same or the homologous protein in the two tissues, namely fibulin, cyclophilin, nebulette, and ␣-actinin. Fibulin is a typical protein of the extracellular matrix. Because zyxin is normally not found in the extracellular matrix, this two-hybrid interaction must be fortuitous, although the majority of independent clones coded for fibulin. Cyclophilin is a cis/trans isomerase that promotes the isomerization of peptide bonds at prolyl residues. This protein might fulfill an important function during the folding of the proline-rich N terminus of zyxin. It is therefore plausible that zyxin possesses a docking site for cyclophilin at its N terminus where the enzyme is actually needed. Five independent clones for cyclophilin B were obtained from the placenta library, whereas one clone for cyclophilin A was retrieved from the heart library. Studies are now in progress to verify this interaction by biochemical experiments. Nebulette is a cytoskeletal protein expressed in heart muscle. Similar to nebulin it is involved in the assembly of the sarcomere, where actin filaments are anchored to the plasma membrane and the Z-disc. Zyxin and nebulette could therefore play an important role in the anchorage of microfilaments to FIG. 6. Interaction of zyxin with the ␣-actinin dimer. Two-hybrid interactions were analyzed in yeast as described in the legend to Fig. 5. A quantitative colorimetric assay was used to determine the expression of the reporter gene ␤-galactosidase. The results are expressed in enzyme units and represent the means with standard deviation from three independent determinations. The pACT2 vector contained various cDNA fragments for the second and third SPEC domain of ␣-actinin as indicated. The pAS2-1 vector contained the same fragments (when self-interaction was analyzed) or the cDNA for zyxin residues 1-42 (when the ␣-actinin/ zyxin interaction was analyzed). Helixes 1-3 refer to the ␣-helices of the SPEC domains as described (21). n.d., not done. the plasma membrane, an attractive hypothesis that can be tackled now by biochemical experiments.
The last binding partner of zyxin that was identified in both screens was ␣-actinin. From the placenta library, ␣-actinin 1 was obtained, whereas from the heart library ␣-actinin 2 was recovered. This is consistent with the relative expression of the isoforms in the two tissues. The zyxin/␣-actinin interaction is thus far the only interaction that has been verified by biochemical experiments (8,10) and that has also been documented in living cells under physiological conditions (9,10). Because several full-length cDNAs for ␣-actinin were obtained during our screenings, we were able to map the interaction site in minute detail.
Our results in combination with data from the literature provide compelling evidence that a linear epitope of zyxin binds to a conformational epitope of ␣-actinin. By site-directed mutagenesis the critical amino acids of zyxin that are involved in the interaction were identified as residues 26 -31 (Phe-Gly-Pro-Val-Val-Ala). Replacement of a single amino acid (Phe-26 3 Ser) within this motif abolished ␣-actinin binding in the twohybrid system and in blot overlays. The same mutation also prevented the normal subcellular distribution of zyxin in fibroblasts, emphasizing the physiological importance of the binding motif. The critical six residues are part of a larger domain that has previously been identified as the major ␣-actinin binding site. Using a crude deletion analysis (10) we have mapped this site to a fragment spanning zyxin residues 21-42 and demonstrated that this fragment is both necessary and sufficient for ␣-actinin binding. On the other hand, Drees et al. (9) have developed a specific peptide inhibitor comprising zyxin residues 16 -33 that blocked the interaction of zyxin with ␣-actinin. Injection of this peptide displaced zyxin from its normal subcellular location and perturbed cell migration. Taken together all these data suggest that zyxin contains a linear epitope exposed at its surface that is specifically recognized by ␣-actinin. The most critical amino acids of this epitope are represented by the sequence motif Phe-Gly-Pro-Val-Val-Ala. This motif also occurs in LPP (23) but not in the related protein ZRP (24). It will therefore be of interest to investigate whether ␣-actinin does also interact with LPP but not with ZRP.
The situation seems to be quite different in the case of the other interaction partner, ␣-actinin. Deletion analyses indicated that zyxin binds to the central region of ␣-actinin made up of SPEC domains 2 and 3. There is good evidence that a conformational, rather than a linear determinant is recognized in this case. When the tandem array of the two SPEC domains is truncated by deletion of short pieces from the N terminus or from the C terminus, binding to zyxin is abolished. The size of the zyxin fragment used in our studies does not allow it to bind simultaneously to the N terminus and to the C terminus of the tandem array. It is therefore likely that the deletions cause a major change in the conformation of the tandem repeat, which is no longer compatible with binding. In this context, it is important to remember that ␣-actinin forms rod-like dimers in which two anti-parallel molecules align in register. The tandem array of SPEC domains 2 and 3 appears to be the minimal fragment that forms stable dimers in vitro (21). As soon as a short piece is removed from the N terminus or from the C terminus of this tandem array, dimerization is prevented as demonstrated by our two-hybrid analysis. The loss of dimerization might therefore explain the loss of zyxin binding. Thus, we propose that zyxin binds into a groove formed by the ␣-actinin dimer where it interacts simultaneously with both anti-parallel chains. This interpretation is consistent with our preliminary findings that ␣-actinin does not interact with zyxin in blot overlays after heat denaturation. 2 Our results are at variance with the findings of Crawford et al. (8), who demonstrated binding of zyxin to the N-terminal CH domain of ␣-actinin. When thermolysin-derived fragments of ␣-actinin were blotted onto nitrocellulose and probed with native, radiolabeled zyxin, the 27-kDa fragment corresponding to the N-terminal domain reacted with the probe but not the 53-kDa fragment corresponding to the SPEC repeats. One possible way to reconcile these conflicting results is that ␣-actinin contains more than one binding site for zyxin. One site would be the conformational determinant in the center of ␣-actinin as identified in our report, the other would be an unrelated determinant at the N terminus of ␣-actinin. Another possibility is that native zyxin is extremely sticky and produces unspecific staining when used as a radiolabeled probe in blot overlays. At any rate, binding of radiolabeled zyxin to the dimeric form of ␣-actinin as proposed in our report cannot be demonstrated by the blot overlay technique, because ␣-actinin will not form dimeric assemblies after denaturation and separation on SDSpolyacrylamide gels.
A great variety of molecules are now known that interact with ␣-actinin (reviewed in Ref. 22). Some of these molecules, including actin (22), CRP (25), and ERK (26), interact with the N-terminal CH domain of ␣-actinin. Other proteins such as titin (27) and ZASP (28) have been reported to interact specifically with the C-terminal domain containing the EF hands. Again some other proteins bind, in a way similar to zyxin, to the central SPEC repeats of ␣-actinin, namely ␣-catenin, the Z-disc protein ALP, and its related protein CLP-36, the protein kinase PKN and the methyl-aspartate receptor NMDA. Binding of ␣-catenin involves the central region of SPEC domains 2 and 3 (residues 479 -529) and does not seem to depend on the dimerization of ␣-actinin (29). ALP binds via its PDZ domain to SPEC repeat 3, which by itself does not form dimers (30). The interaction of CLP-36 with ␣-actinin has not yet been analyzed in detail (31). Based on its structural homology with APL, it is likely that CLP-36 binds to SPEC repeat 3, too. The same domain has also been identified as the major target for protein kinase PKN (32). In the case of the methyl-aspartate receptor NMDA, SPEC domain 4 appears to be essential, although a longer fragment is required for strong binding (33). Thus, all the interactions listed above do not seem to depend on the integrity of the tandem array of SPEC domains 2 and 3. In contrast, zyxin is an example that requires the dimeric, antiparallel conformation of ␣-actinin for interaction. It would be interesting to investigate whether this interaction is affected by binding of other ligands in close vicinity. This is conceivable for the regulatory proteins ALP (31) and CLP-36 (30), because these proteins possess, in addition to their PDZ domain, a C-terminal LIM domain that could interact with the C-terminal LIM domains of zyxin. In this way, ALP and CLP-36 could modulate the interaction of zyxin with ␣-actinin and control the recruitment of zyxin to focal adhesion sites.