Suppressor Mutations Suggest a Surface on PAT-4 (Integrin-linked Kinase) That Interacts with UNC-112 (Kindlin)*

Background: Interaction between PAT-4 (ILK) and UNC-112 (kindlin) is required for localization of each protein to integrin adhesions. Results: PAT-4 missense mutations were identified that restore binding of a mutant UNC-112 protein and its proper localization. Conclusion: Residues mutated in the suppressor mutants cluster in two regions on the surface of a homology model of PAT-4. Significance: An interaction surface between PAT-4 and UNC-112 is suggested. Caenorhabditis elegans striated muscle cells attach to basement membrane and transmit the force of muscle contraction through integrin adhesion complexes. The cytoplasmic tail of β-integrin (PAT-3) is associated with a conserved four-protein complex that includes UNC-112 (kindlin), PAT-4 (integrin-linked kinase), PAT-6 (α-parvin/actopaxin), and UNC-97 (PINCH). The proper localization of UNC-112 to muscle integrin adhesion sites requires PAT-4. A recent report (Qadota, H., Moerman, D. G., and Benian, G. M. (2012) A molecular mechanism for the requirement of PAT-4 (integrin-linked kinase (ILK)) for the localization of UNC-112 (kindlin) to integrin adhesion sites. J. Biol. Chem. 287, 28537–28551) suggests a possible molecular mechanism for this requirement: that UNC-112 exists in closed inactive and open active conformations, and conversion to the open active form is promoted by binding to PAT-4 (ILK). Previously, we also reported identification of a single missense mutation in UNC-112, D382V, which abolishes both binding to PAT-4 and normal localization to integrin adhesion sites in vivo. In this report, we describe isolation and characterization of PAT-4 missense mutations that permit binding with UNC-112 D382V and place nine affected residues on a homology model of PAT-4. These nine residues cluster in two regions on the surface of PAT-4, do not overlap the likely binding surface for PAT-6 (α-parvin), and therefore may reside along the interaction surface of PAT-4 for UNC-112 (kindlin). We also show that one of these PAT-4 mutations restores the ability of UNC-112 D382V to localize to integrin adhesions and participate in complex formation.

Integrin adhesion sites consist of the transmembrane heterodimeric protein integrin and Ͼ100 different associated pro-teins (2)(3)(4). These protein complexes have an important role in cell crawling; as a cell moves forward, new adhesion sites are assembled at the leading edge, and old adhesion sites are disassembled at the back end of the cell. Integrin adhesion sites are also important in transmitting the force of muscle contraction to the outside of a mammalian striated muscle cell. These cells contain many myofibrils mechanically liked by intermediate filaments, and at the periphery of these cells, myofibrils are attached to the muscle cell membrane via integrin adhesion complexes called costameres (5,6).
In the model genetic organism Caenorhabditis elegans, the striated muscle contains a narrow zone of myofibrils adjacent to the cell membrane along the outer side of the muscle cell (7,8). As in mammalian striated muscle, the thin filaments are attached to Z-disks (called dense bodies in nematodes), and the thick filaments are organized around M-lines. In addition, all of the dense bodies and M-lines are attached to the muscle cell membrane, and thus most of the force of muscle contraction is transmitted through these structures. The membrane-proximal regions of the dense bodies and M-lines consist of integrin and associated proteins (9 -11). The cytoplasmic tail of ␤-integrin is associated with four conserved proteins: UNC-112 (kindlin), PAT-4 (integrin-linked kinase, ILK 2 ), PAT-6 (actopaxin/ ␣-parvin), and UNC-97 (PINCH). PAT-4 has been shown to interact with UNC-112, PAT-6, and UNC-97 (12)(13)(14)(15). Each of these proteins is required for myofibril assembly in embryonic muscle (15)(16)(17).
In contrast to C. elegans, which has one kindlin (UNC-112), humans have three kindlins, each encoded by a separate gene (18). Human kindlin 1 is expressed primarily in epithelial cells such as keratinocytes and intestinal epithelial cells. Kindlin 2 is expressed everywhere except for hematopoietic cells and is the only kindlin expressed in heart muscle. Kindlin 3 is expressed in hematopoietic cells, with highest expression in megakaryocytes (which give rise to platelets). Inherited recessive mutations in kindlin 1 result in a serious skin disease (Kindler syndrome), and recessive mutations in kindlin 3 result in severe dysfunction of both platelets and leukocytes (leukocyte adhesion deficiency type III). Nematode UNC-112 is equally similar to human kindlin 1 and kindlin 2, each sharing 42% identical residues. UNC-112 is less homologous to kindlin 3; there is an internal span of about 130 residues in which there is little sequence similarity (supplemental Fig. 1).
In C. elegans, the relationship between UNC-112 and PAT-4 has been analyzed in several ways. The proper localization of UNC-112 to the base of M-lines and dense bodies requires PAT-4; the converse is also true. In addition, the protein kinase domain of PAT-4 interacts with the N-terminal half of UNC-112 (12,13).
Our recent report (1) provides a possible molecular explanation for the PAT-4 dependence of UNC-112 localization. We demonstrated that the cytoplasmic tail of ␤-integrin (PAT-3) binds to UNC-112, but not to PAT-4, and that the N-and C-terminal halves of UNC-112 bind to each other. PAT-4 was shown to compete with the C-terminal half of UNC-112 for binding to the N-terminal half of UNC-112. From these protein interaction data we proposed the following model: UNC-112 exists in closed inactive and open active conformations, and conversion to the open active form is promoted by binding to PAT-4 (ILK). Support of this model came from analysis of two types of UNC-112 mutants; one type of mutant, UNC-112 D382V cannot bind to PAT-4; another type of mutant, represented by UNC-112 T346A or E349K, does not permit interaction of the N-and C-terminal halves of UNC-112. Whereas UNC-112 D382V does not localize to integrin adhesion sites (similar to deficiency of pat-4), UNC-112 double mutants (D382V,T346A or D382V,E349K) do localize properly, suggesting that detachment of the two halves of UNC-112, perhaps in an "open" conformation, favors interaction with ␤-integrin. In this report, we describe isolation and characterization of PAT-4 missense mutations that permit binding with UNC-112 D382V and place the affected residues on a homology model of PAT-4. We also demonstrate that one of these PAT-4 mutations restores the ability of UNC-112 D382V to localize to integrin adhesions and participate in complex formation.

EXPERIMENTAL PROCEDURES
Nematode Strains and Culture-Wild-type, strain N2, was used in these studies. Nematodes were grown at 20°C on NGM agar plates with Escherichia coli strain OP50 (19).
Homology Modeling-Domains within PAT-4 were defined by use of PFAM. Molecular modeling of C. elegans PAT-4 (ILK, protein kinase domain, residues 260 -460) was completed using Modeler (version 9v7, University of California San Francisco (22)) for model construction and the UCSF Chimera package (version 1.6.2, University of California San Francisco (23)) for visualization, structure comparison, and analysis. The PAT-4 model was built and minimized based on the crystal structure of human ILK (Protein Data Bank ID code, 3KMW) chain A (24). Initial sequence alignment between human ILK (residues 185-452) and C. elegans PAT-4 (residues 199 -464) was generated from NCBI Blast. The final alignment is shown in Fig. 1B. Overlay of the structures (C. elegans PAT-4 with ␣-parvin) was completed in Chimera using the Matchmaker- Structure Comparison tool (Tools for integrated sequencestructure analysis with UCSF Chimera (25)). Structural similarity of the ankyrin repeats of ILK (PAT-4) to p19 INK4d was deter-mined by the Dali Protein Structure 3D Alignment Server (26). Fig. 6 shows the structural and sequence alignment of ankyrin repeats and p19 INK4d . The sequence represented for ILK is the sequence of the protein expressed and crystallized, as reported by Fukuda et al. (24). In the row between the two sequences, identical residues are indicated, and ϩ indicates conservation. Over the span of 266 residues, 59% are identical (157 of 266). Yellow highlighting denotes the residues affected in the nine PAT-4 suppressor mutations.
Detection of Complex Formation from Worm Lysates-Nematode-integrated transgenic lines sfIs3 and sfIs5 (1) were induced to express HA-tagged UNC-112 wild type or UNC-112 D382V, respectively, by heat shocking mostly adult animals by exposure to 30°C for 3 h, and worm lysates were prepared as described by Qadota et al. (1). To 500 l of each worm lysate was added 50 l of a 1:1 slurry of monoclonal anti-HA-agarose beads (Sigma-Aldrich A2095) and mixing at 4°C for 2 h. The beads were pelleted and washed three times with lysis buffer, transferred to new tubes, pelleted, and as much liquid was removed as possible. The proteins from these beads were eluted with 33 l of 2ϫ Laemmli sample buffer, and portions were separated on 10% SDS-polyacrylamide gels, and blotted. Sepa- rate lanes were reacted with the following antibodies: rabbit anti-HA (Sigma-Aldrich H6908) at 1/200 dilution; rabbit anti-PAT-4 (1), affinity-purified, at 1/100; rat anti-PAT-6 (32), affinity-purified, at 1/200 dilution; and rabbit anti-UNC-97 (20), affinity-purified, at 1/100. Reactions were detected by ECL.
Co-immunoprecipitation of HA-UNC-112 and myc-PAT-4-Transgenic lines carrying integrated arrays described above were induced to express HA-tagged UNC-112 D382V and either myc-tagged PAT-4 WT (sfIs8) or P257L (sfIs10) by heat shock, and worm lysates were prepared as described above. Immunoprecipitation using anti-HA beads and gel analysis was carried out as described in the previous section. Separate lanes were reacted with rabbit anti-HA (Sigma-Aldrich H6908) at 1/200 dilution, and rabbit anti-myc (Sigma-Aldrich H4439) at 1/200 dilution. Reactions were detected by ECL.
Calculation of Molar Ratios of PAT-4 to UNC-112 in Co-IP Experiments-Varying amounts (in nanograms) of recombinant GST-3ϫHA-myc (GST fused to three copies of HA and one copy of myc) on Western blots were reacted with either anti-HA or anti-myc, using the same concentration of antibodies and ECL exposure times as used in the immunoprecipitation experiments. Similarly, varying amounts of recombinant MBP-PAT-4 (1) on Western blots were reacted with anti-PAT-4. The intensity of ECL bands obtained from immunoprecipitations was matched with intensity of ECL bands obtained with known quantities of recombinant proteins. Molar ratios of PAT-4 to UNC-112 in the immunoprecipitation experiments were then calculated using these estimates of quantities of proteins in nanograms and their known molecular masses.

RESULTS
To obtain further information about the interaction between UNC-112 and PAT-4, we isolated PAT-4 mutants that could bind to UNC-112 D382V. Previously, we demonstrated that although wild-type UNC-112 binds to PAT-4, UNC-112 D382V fails to bind to PAT-4 (1). Using random PCR mutagenesis and the yeast twohybrid system, we identified 37 PAT-4 mutants that can interact with UNC-112 D382V (Table 1). DNA sequencing revealed that 6 of the 37 PAT-4 mutants contained single amino acid changes, with one of them appearing twice independently (Table 1 and Fig.  1). An additional set of 4 single amino acid changes was found to be responsible for suppression by testing individual mutated residues from clones that had 2 mutations. Interestingly, although we mutagenized the entire pat-4 cDNA, all 9 single amino acid changes reside in the protein kinase domain of PAT-4, the region of PAT-4 demonstrated previously to be sufficient for interaction with UNC-112 (13). All 9 of these PAT-4 mutants also bind to wild-type UNC-112, suggesting that the structural changes in these mutants are minimal. PAT-4 also interacts with UNC-97 and PAT-6 (13,14). Significantly, the mutations do not affect binding of PAT-4 to UNC-97 or PAT-6 (data not shown), further indicating the unique effect the mutations have on UNC-112 D382V binding, suggesting localization on the UNC-112 binding surface.
Fukuda et al. (24) reported the crystal structure of human ILK kinase domain complexed with ␣-parvin. As shown in Fig.  1B, the pseudokinase domains of human ILK and C. elegans PAT-4 have extensive similarity, with 59% identical residues (157 of 266). Of the 9 PAT-4 suppressor mutations (indicated by yellow in Fig. 1B), 5 reside at conserved residues. (These conserved residues are underlined in Fig. 1A.) We used the crystal structure of human ILK to create a homology model of nematode PAT-4 kinase domain, and located the 9 residues mutated in our PAT-4 mutations on this structure. As shown in Fig  This is even more apparent when PAT-4 is viewed in space filling mode (Fig. 2C). The residues mutated in the suppressors are available to form a binding surface for another protein, perhaps UNC-112. That UNC-112 and PAT-6 bind along different surfaces of PAT-4 is consistent with the previously reported ternary complex of UNC-112, PAT-4, and PAT-6 (14).
To examine the in vivo significance of the ability of mutant PAT-4 proteins to interact with UNC-112 D382V, we created transgenic lines expressing HA-tagged UNC-112 D382V and either myc-tagged wild-type PAT-4 or myc-tagged P257L mutant PAT-4. P257L was chosen because, as noted above, it is one of the five mutations that reside on a potential binding surface for UNC-112 and is also conserved in human PAT-4. The transgenic lines were designed for expression of the two tagged proteins by a heat shock promoter in a wild-type background. As shown in Fig. 3A, for each line, both the HA-tagged UNC-112 and myc-tagged PAT-4 are expressed only upon heat shock. Previously, we have shown that expression of HA-UNC-112 D382V in a wild-type background does not allow localization of the tagged protein to muscle integrin adhesion sites (dense bodies and M-lines) (1). Consistent with this result, even in the presence of ectopic myc-tagged wild-type PAT-4, HA-UNC-112 D382V fails to localize to these sites (Fig. 3B, upper  row). However, co-expression of myc-tagged PAT-4 P257L allows localization to these structures (Fig. 3B, lower row).
These results indicate that PAT-4 P257L binds to UNC-112 D382V, and this is consistent with our model that PAT-4 binding to UNC-112 is required for the localization of UNC-112 to integrin adhesion sites ( Ref. 1 and see Fig. 5).
An additional function of PAT-4 is to recruit PAT-6 (␣-parvin) and UNC-97 (PINCH) to integrin adhesion sites. Previously, it was reported that UNC-112 binds to PAT-4 (13), that PAT-4 binds to PAT-6 (14), and that PAT-4 binds to UNC-97 (13,15). In addition, yeast three-hybrid assays suggest ternary complexes in the following combinations: UNC-112, PAT-4, PAT-6 (14); UNC-112, PAT-4, UNC-97 (15); and UNC-97, PAT-4, PAT-6 (15). From these data, we hypothesized the existence of a four-protein complex (Refs. 10, 30 and see Fig. 5). To obtain in vivo evidence for this four-protein complex, we used a co-immunoprecipitation approach from lysates of worms expressing HA-tagged UNC-112. Anti-HA beads were used to immunoprecipitate HA-UNC-112, and Western blotting was used to detect the presence of endogenous PAT-4, PAT-6, and UNC-97. When the experiment was conducted with HA-UNC-112 WT, PAT-4, PAT-6, and UNC-97 could be detected in the precipitate (Fig. 4A, left). However, when HA-UNC-112 D382V was precipitated, less PAT-4 was detected, and neither PAT-6 nor UNC-97 was detected (Fig. 4A, right). The difference in the amounts of endogenous PAT-4 co-immunoprecipitated by WT versus D382V UNC-112 can be compared by calculating the molar ratios using a method described under "Experimental Procedures." As indicated on Fig. 4A, there is a 10-fold difference in these molar ratios. These results are consistent with the model that in nematode muscle, UNC-112 binds to PAT-4, and that PAT-4 binds to PAT-6 and UNC-97, and forms a stable four-protein complex (Fig. 5A).
We next examined the ability of PAT-4 P257L to restore the ability of UNC-112 D382V to bind to PAT-4 in a similar co-  bound to the ankyrin repeats as in 1BLX. The N terminus and C terminus of each protein is marked with blue and red, respectively. E, sequence alignment between p19 INK4d and ankyrin repeats. Helices from the structures are outlined in black, identical residues are marked with an asterisk (*). Region of high conservation is highlighted in orange (ϳ28% identity). ability of UNC-112 D382V to localize to integrin adhesion sites also restores the ability to bind to UNC-112 D382V in an in vivo protein complex.

DISCUSSION
Beginning with a mutant version (D382V) of UNC-112 that cannot bind to wild-type PAT-4, we isolated missense mutants of PAT-4 that can bind to this mutant UNC-112. Of 37 mutant clones isolated, 5 of them had single amino acid changes. The others had multiple changes (2-5 mutations each). Site-directed mutagenesis permitted us to identify the single missense mutations responsible for the phenotype in 4 additional mutant clones. All together, we identified 9 single missense mutations that permit PAT-4 to bind to UNC-112 D382V (Fig. 1). All 9 of these single amino acid changes reside in the protein kinase domain of PAT-4, the region of PAT-4 shown previously to be required for binding to UNC-112. All 9 residues affected by the suppressor mutations cluster in two regions on the surface of a homology model structure of PAT-4, based on the crystal structure of its human ortholog, ILK. The reported ILK structure is actually a complex of ILK bound to ␣-parvin, the ortholog of nematode PAT-6. Significantly, when PAT-4 was substituted for ILK in the ILK⅐␣-parvin complex, ␣-parvin does not overlap or cover either cluster of suppressor residues on PAT-4. Thus, we suggest that the surfaces of PAT-4 on which the suppressor mutations reside constitute an independent binding surface for UNC-112 (Fig. 2).
Although it has been reported that a UNC-112⅐PAT-4⅐PAT-6 ternary complex can form, our structural insight suggests that a single molecule of PAT-4 can interact simultaneously with one molecule each of PAT-6 and UNC-112. Verification of this model awaits co-crystallization of PAT-4 with UNC-112.
Our previous data support a model in which one of the functions of the interaction of PAT-4 with UNC-112 is to promote opening up of UNC-112 so that it can bind to the cytoplasmic tail of PAT-3 (␤-integrin) (1). In agreement with this model, we have shown here that in vivo, expression of one of the PAT-4 missense mutants permits UNC-112 D382V to localize to integrin adhesion sites (Fig. 3B). Previous reports also suggest that another function of the interaction of PAT-4 with UNC-112 is to recruit PAT-6 and UNC-97. Here, we show that the immunoprecipitation of HA-tagged UNC-112 from nematode lysates co-precipitates PAT-4, PAT-6, and UNC-97 (Fig. 4A). This is the first evidence of formation of this four-protein complex in vivo. In contrast, immunoprecipitation of HA-tagged UNC-112 D382V co-precipitates 10-fold less PAT-4 and no detectable PAT-6 or UNC-97. However, co-expression of UNC-112 D382V together with one of the PAT-4 missense mutants (P257L) restores UNC-112⅐PAT-4 complex formation. The recruitment of UNC-97 and PAT-6 by PAT-4 is essential for formation of huge protein complexes at the base of the integrin adhesion sites. We have already described that downstream of UNC-97 there are proteins that bind directly or indirectly to myosin at the M-line (20,30). For example, UNC-97 interacts with UNC-98, and UNC-98 interacts with myosin (20); UNC-97 interacts with LIM-9 (FHL), LIM-9 interacts with UNC-96, and UNC-96 interacts with myosin (30). Recently, we have described that downstream of PAT-6 is CPNA-1, which links to UNC-89 and several other M-line proteins (32). At this time, we have not yet identified proteins that reside downstream of either UNC-97 or PAT-6 exclusively at the dense bodies.
Chiswell et al. (33) have reported the crystal structure of the ankyrin repeats of human ILK (PAT-4) bound to the first LIM domain (of five total LIM domains) of PINCH (UNC-97). We have noticed that the ILK ankyrin-PINCH LIM1 complex is similar in size and shape to the cell cycle inhibitor p19 INK4d (34) (Fig. 6). Actually, the reported structure is of p19 INK4d bound to cyclin D-dependent kinase Cdk6. Although there is no evidence that the ankyrin repeats of ILK (PAT-4) are bound to the kinase domain of ILK (PAT-4), we thought it would be reasonable to substitute our homology model of PAT-4 kinase for CDK6 and the ankyrin-PINCH LIM1 complex for p19 INK4d (Fig. 7A). Note that the ankyrin-PINCH LIM1 complex (and thus orthologous structures from PAT-4 and UNC-97) do not overlap or cover the sites of the PAT-4 suppressor mutations. However, the LIM1 domain of PINCH (UNC-97) is "near" the suppressor mutation residue Phe-262. When ␣-parvin (PAT-6) is docked back onto this model of PAT-4, we note that 7 of 9 of the residues mutated in the suppressor mutants (Pro-257, Ile-261, Phe-262, Ala-274, Asn-275, Lys-356, and Asn-453) are on the surface between ␣-parvin and PINCH, on the face opposite from the binding site of the ankyrin repeats (Fig. 7, B and C). Therefore, even with these additional proteins bound to PAT-4, the binding surface defined by the PAT-4 suppressor mutations will be unobstructed to bind to UNC-112. Again, all of this speculation awaits confirmation from PAT-4⅐UNC-112 and larger complex crystal structures.