CLP-36 PDZ-LIM protein associates with nonmuscle alpha-actinin-1 and alpha-actinin-4.

The PDZ-LIM family of proteins (Enigma/LMP-1, ENH, ZASP/Cypher, RIL, ALP, and CLP-36) has been suggested to act as adapters that direct LIM-binding proteins to the cytoskeleton. Most interactions of PDZ-LIM proteins with the cytoskeleton have been identified in striated muscle, where several PDZ-LIM proteins are predominantly expressed. By contrast, CLP-36 mRNA is expressed in several nonmuscle tissues, and here we demonstrate high expression of CLP-36 in epithelial cells by in situ hybridization analysis. Our subcellular localization studies indicate that in nonmuscle cells, CLP-36 protein localizes to actin stress fibers. This localization is mediated via the PDZ domain of CLP-36 that associates with the spectrin-like repeats of alpha-actinin. Interestingly, immunoprecipitation and matrix-assisted laser desorption/ionization time-of-flight mass spectrometry analysis indicate that both nonmuscle alpha-actinin-1 and alpha-actinin-4 form complexes with CLP-36. The high expression of alpha-actinin-4 in the colon, together with these results, suggests a specific function for the alpha-actinin-4-CLP-36 complex in the colonic epithelium. More generally, results presented here demonstrate that the association of PDZ-LIM proteins with the cytoskeleton extends to the actin stress fibers of nonmuscle cells.

PDZ-LIM proteins have been suggested to act as adapters between kinases and the cytoskeleton (3,6). This is based on two lines of studies indicating that PDZ-LIM proteins associate on one hand to the cytoskeleton via their PDZ domain (3)(4)(5)(6), and on the other hand to kinases via their LIM domains (6,11,12,15). The LIM-kinase interaction has been demonstrated with three PDZ-LIM proteins, mostly using yeast two-hybrid screens. Enigma was identified by virtue of association of its third LIM domain to the cytoplasmic tail of the insulin receptor; this interaction is apparently important for endocytosis of the receptor (11). The association of the second LIM domain of Enigma with Ret/ptc2 is required for the plasma membrane localization and mitogenic activity of Ret/ptc2 (15,16). ENH was discovered through its binding to protein kinase C (12), and subsequently, Cypher1 was also found to associate with protein kinase C (6). The association of the LIM domain of RIL with the second and fourth PDZ domains of protein tyrosine phosphatase PTP-BL in a yeast two-hybrid screen (17) suggests that the LIM domain interactions of PDZ-LIM proteins may not be limited to kinases.
The association of PDZ domains with the cytoskeleton in PDZ-LIM proteins has been mostly studied in muscle due to their high specific expression in this tissue (3)(4)(5)(6). Enigma, ALP, and ZASP/Cypher1 proteins localize to the Z line of striated muscle. This localization is mediated by the association of Enigma PDZ domain with ␤-tropomyosin (3) or by the association of ALP and ZASP/Cypher1 PDZ domains with ␣-actinin-2 (4 -6).
␣-Actinin-2 and ␣-actinin-3 are the muscle-specific ␣-actinins forming part of the contractile machinery anchoring actin thin filaments at the Z lines and dense bodies in striated and smooth muscle, respectively (18). Cellular ␣-actinin exists as an antiparallel dimer with a globular head domain, spectrinlike repeats and EF-hands (reviewed in Ref. 19). Dimerization of ␣-actinin is mediated by the spectrin-like repeats of ␣-actinin (20 -22). This rod domain mediates the association with the PDZ domain of ALP, whereas ZASP binds to a 155-amino acid C-terminal region of ␣-actinin-2 (5,6).
During studies on proteins associated with a partially characterized novel kinase, 1 we became interested in the CLP-36 PDZ-LIM protein. CLP-36 (also called hCLIM1; Ref. 9) was initially identified as a rat gene down-regulated during hypoxia in hepatocytes (8). CLP-36 mRNA is expressed in several non-muscle tissues (8,9), and here we demonstrate high expression of CLP-36 in epithelial cells by in situ hybridization analysis. Moreover, CLP-36 was found to localize to actin stress fibers via its PDZ domain through its association with cellular ␣-actinin-1 and ␣-actinin-4, suggesting that CLP-36 acts as an adapter between stress fibers and LIM-binding proteins in nonmuscle cells.

EXPERIMENTAL PROCEDURES
In Situ Hybridization-Embryos of CBA ϫ NMRI mice at stages E7, E9, E11, E15, and E17.5 were timed by both vaginal plugs of mothers and by morphological criteria. The experiments were approved by the Animal Welfare Committee of the Haartman Institute, University of Helsinki. Tissue preparation and in situ hybridization using a antisense or sense RNA probe generated from human CLP-36 cDNA (nucleotides 87-1504 in GenBank TM accession number U90878) were performed as described (30).

Generation of Recombinant Proteins and Solution Binding
Assays-For expression of GST-CLP-36, 2 a NotI fragment from 38/pAMC was subcloned into pAcGHLT-A (Pharmingen, San Diego, CA) baculovirus transfer vector, which was introduced into Sf9 insect cells together with BaculoGold baculovirus DNA as recommended by the manufacturer. For protein expression, the baculovirus was propagated in Hi5 insect cells (Invitrogen) for 48 h. The GST-CLP-36 protein was purified using glutathione-Sepharose and eluted with glutathione. The bacterially produced chicken GST-␣-actinin proteins (a kind gift from Dr. Critchley; Ref. 35) were purified as described (32).
For detection of association between cellular proteins and GST-CLP-36, 200 g of COS-7 cell extracts in ELB were incubated with the indicated amounts of GST-CLP-36 for 2 h at ϩ4°C prior to adding 7.5 l (packed volume) of glutathione-Sepharose beads for 1 h. Subsequently, the beads were washed four times with ELB and subjected to SDS-PAGE and Coomassie staining (see Fig. 6). For mapping of the CLP-36-␣-actinin association, 200 g of transfected COS-7 cell lysate in ELB were incubated with 4 g of GST-␣-actinins for 2 h at ϩ4°C prior to anti-Myc immunoprecipitation, SDS-PAGE, and Western blotting analysis (see Fig. 4B).
Mass Spectrometry-Matrix-assisted laser desorption/ionization time-of-flight mass spectrometry was performed on a Biflex TM time-offlight instrument equipped with a nitrogen laser operating at 337 nm. CLP-36 associated ␣-actinin band from either U2OS or COS7 cells was cut out from the Coomassie-stained gel, digested with trypsin, eluted, and analyzed in the linear positive ion delayed extraction mode. The output was analyzed using ProFound.

High Expression of CLP-36 in Embryonal and Adult
Epithelial Cells-Northern blot analyses of CLP-36 indicated expression in various tissues with some variability between rat (8) and human (9) tissues. To determine the pattern of expression of CLP-36 in adult tissues and during embryogenesis, in situ hybridization analysis with an antisense CLP-36 probe was performed. Hybridization of a sagittal section of a E11 mouse embryo (Fig. 1A) indicated expression in several tissues, including dorsal root ganglia, liver, and arteries. Prominent expression was also detected throughout the skin epithelium (Fig. 1A, SE). Analysis of earlier stage embryos indicated that the skin epithelium had a strong hybridization signal even in E9 stage embryos, in which the skin is at the single layered ectodermal stage (not shown). At stage E17.5 expression in the skin and the oral cavity epithelium (Fig. 1, B and C, OC) localized predominantly to the basal cell layer, whereas expres- sion in the outer dental epithelium was low (Fig. 1, B and C,  DE).
CLP-36 expression was also noted in the mucosal epithelium of the developing gut from embryonal stage E15 onwards (data not shown) and in adult intestine, as demonstrated in strong epithelial expression in the small intestine (Fig. 1, D and E,  IE), whereas CLP-36 expression was undetectable in the surrounding smooth muscle layers (SM). In adult mouse tissues, a prominent expression was also noted in the epithelium of esophagus and urinary bladder, whereas expression in skeletal muscle was undetectable.
CLP-36 Is Localized to Actin Stress Fibers-The expression of CLP-36 in epithelium and other nonmuscle tissues prompted us to study its subcellular localization in cell lines of nonmuscle origin. To this end, a plasmid expressing a Myc epitope-tagged CLP-36 (Myc-CLP-36) or vector control was transiently transfected into U2OS osteosarcoma (Fig. 2) or COS-7 green monkey kidney epithelial cells (not shown). Double fluorescence analysis with anti-Myc and rhodamine-labeled phalloidin from vector transfected or Myc-CLP-36 transfected cells indicated partial colocalization of CLP-36 with filaments resembling actin stress fibers (Fig. 2A). To verify this colocalization, similar coverslips were subjected to detergent extraction or treated with cytochalasin B prior to fixation. As demonstrated in Fig.   2B (Triton X-100), detergent extraction revealed colocalization of CLP-36 and actin stress fibers. If microfilament formation was inhibited by cytochalasin B prior to fixation, CLP-36 was partially accumulated with the disrupted actin-containing filaments (Fig. 2B, cytochalasin B). These experiments indicate that a significant fraction of CLP-36 is associated with actin stress fibers.
CLP-36 Associates with Cellular ␣-Actinin via Its PDZ Domain-During the immunoprecipitation studies, it was noted that whenever Myc-CLP-36 was purified from cell lysates, a prominent associated protein of approximately 100 kDa was detected when staining for total protein. The size of the polypeptide together with the immunofluorescence pattern of CLP-36 suggested that the band represented ␣-actinin. Moreover, based on interactions of the other PDZ-LIM proteins (3)(4)(5)(6), the association was further predicted to be mediated by the PDZ domain of CLP-36. To test this hypothesis, immunoprecipitates from cells expressing either Myc-CLP-36 or a mutant lacking part of the PDZ domain (Myc-CLP⌬1-24) were analyzed by Western blotting with an ␣-actinin antibody (33). ␣-Actinin was only detected in the lane immunoprecipitated with the full-length CLP-36 (Fig. 3A), indicating that the first 24 amino acids of CLP-36 are required for ␣-actinin binding and thus suggesting that an intact PDZ domain is required for binding. The inability of Myc-CLP⌬1-24 to associate with ␣-actinin was associated with loss of localization to stress fibers (Fig. 3B). This indicates that CLP-36 localization to stress fibers is mediated through its association with ␣-actinin.
Association of the endogenous ␣-actinin with CLP-36 was also noted in all cases, as expected (Fig. 4B, ␣-actinin). However, the level of associated ␣-actinin decreased in the presence of GST-R1-R4 indicating competitive binding of these with CLP-36. As endogenous ␣-actinin and GST-R1-R4 were detected using the same antibody, the results also demonstrated that the total amount of CLP-36-associated GST-R1-R4 was significantly higher than CLP-36-associated endogenous ␣-actinin. This indicates that association of GST-R1-R4 with CLP-36 is not mediated indirectly through dimerization with endogenous ␣-actinin. As the levels of CLP-36-associated GST-ABD/R1-R2 were lower than those of endogenous ␣-actinin, the same conclusion cannot be made for GST-ABD/R1-R2. Taken together, these results indicate that the efficient association of CLP-36 with ␣-actinin is mediated via the spectrin-like repeats.
To estimate the stoichiometry of binding of ␣-actinin to CLP-36, varying amounts of the GST-CLP-36 protein were incubated with 200 g of COS-7 cell lysate and subsequently purified using glutathione-Sepharose beads and analyzed with a Coomassie stain of SDS-PAGE (Fig. 6). A prominent band of approximately 100 kDa representing ␣-actinin was the only protein detected in addition to GST-CLP-36. An approximate mass ratio of 2:3 for GST-CLP-36 and ␣-actinin was linear until ␣-actinin in the lysate became limiting. Considering the molecular masses of GST-CLP-36 (66 kDa) and ␣-actinin (103 kDa) this mass ratio indicates an equimolar complex, and together with the higher affinity of CLP-36 to actinin dimers, it suggests that two molecules of CLP-36 are associated with an ␣-actinin dimer. The efficient purification of soluble ␣-actinin from the cell extracts using GST-CLP-36 also enabled analysis of this protein by matrix-assisted laser desorption/ionization time-offlight mass spectrometric analysis, which indicated the presence of both ␣-actinin-1 and ␣-actinin-4 in the 100-kDa band. in the colonic epithelium.
The expression pattern observed in the mouse in situ hybridization analyses is in concordance with previous Northern blotting data from rat tissues (8), both indicating that CLP-36 is not expressed in skeletal muscle. However, a positive signal in Northern blotting analysis from human skeletal muscle (9) suggests either species-specific expression differences or the presence of a closely related human gene. Analysis of expressed sequence tag data base content did not reveal closely related cDNAs, and 8 of 244 (3.3%) human CLP-36 expressed sequence tags were from skeletal muscle, supporting low relative expression of human CLP-36 in this tissue.
Previous studies have implicated roles for PDZ-LIM proteins in skeletal muscle. This is based both on the muscle-oriented approaches used in prior studies (3)(4)(5)(6) and also on the predominant expression of ALP and ZASP/Cypher in skeletal muscle by Northern blotting analysis (4 -6). However, results from this study, as well as from those with Enigma (11), ENH (12,36), and RIL (7,37), indicate that the PDZ-LIM family is similarly involved in mediating interactions of LIM-binding proteins with the cytoskeleton in other tissue types.
In this regard, it is also interesting to note that the mRNA expression of RIL is concentrated in the epithelial cells (7). However, there is a notable difference in tissue distribution between RIL and CLP-36. RIL is highly expressed in the epithelium of brain, testis, mid-size bronchii, uterine tubae and stomach, whereas CLP-36 mRNA expression in these tissues is low or undetectable. Although the subcellular localization and the PDZ domain associated protein of RIL have not been identified, it is interesting to speculate that these two close relatives might share a common function specific for a particular type of epithelium. An extension of this hypothesis includes ALP, the third close relative of CLP-36 and RIL, with a nonoverlapping expression pattern (5). ALP could mediate a similar or related function in skeletal muscle.
CLP-36 was predominantly localized to stress fibers in an apparently PDZ domain-dependent fashion. CLP-36 also directly associated with ␣-actinin in an apparently PDZ domaindependent fashion, indicating that CLP-36 localization to stress fibers is mediated via ␣-actinin. Interaction of CLP-36 with ␣-actinin is not limited to stress fibers, as CLP-36 efficiently coimmunoprecipitated with cellular ␣-actinin from nonionic, low detergent cell lysates. Analysis of soluble versus insoluble fraction of such lysates indicated that roughly equal amounts of both ␣-actinin and CLP-36 are present in soluble and insoluble fractions (data not shown). Based on these results, it is interesting to speculate that the soluble ␣-actinin may represent a free pool to be used for rapid reorganization of actin stress fibers in nonmuscle cells and that the associated CLP-36 could be involved in this function.
CLP-36 associated efficiently with an ␣-actinin fragment containing all four spectrin-like repeats. Previously, all four repeats were demonstrated to be required for a maximally stable ␣-actinin dimer (21,22). Taken together, these results suggest that CLP-36 associates with ␣-actinin dimers. This fits well with the observed stoichiometry of the CLP-36-␣-actinin complex, indicating that two molecules of CLP-36 are associated with an ␣-actinin dimer. The identification of PDZ domain interactions has revealed the association of several PDZ-LIM proteins with the microfilaments (3)(4)(5)(6). This has led to the suggestion that the primary function of these proteins is to act as adapters between the cytoskeleton and LIM-binding proteins (3,6). So what are these LIM-binding proteins? The identified interactions of some LIM domains with serine/threonine kinases (6,12) suggests that similar partners may be found for the orphan LIM domains as well. This is supported by our original identification of the LIM domain of CLP-36 in a yeast two-hybrid screen using a novel partially characterized serine/threonine kinase. 2 This would point to a role for CLP-36 in recruiting kinase(s) to actin stress fibers via ␣-actinin binding.