The LIM-only Protein DRAL/FHL2 Binds to the Cytoplasmic Domain of Several α and β Integrin Chains and Is Recruited to Adhesion Complexes*

LIM proteins contain one or more double zinc finger structures (LIM domains) mediating specific contacts between proteins that participate in the formation of multiprotein complexes. We report that the LIM-only protein DRAL/FHL2, with four and a half LIM domains, can associate with α3A, α3B, α7A, and several β integrin subunits as shown in yeast two-hybrid assays as well as after overexpression in human cells. The amino acid sequence immediately following the conserved membrane-proximal region in the integrin α subunits or the C-terminal region with the conserved NXXY motif of the integrin β subunits are critical for binding DRAL/FHL2. Furthermore, the DRAL/FHL2 associates with itself and with other molecules that bind to the cytoplasmic domain of integrin α subunits. Deletion analysis of DRAL/FHL2 revealed that particular LIM domains or LIM domain combinations bind the different proteins. These results, together with the fact that full-length DRAL/FHL2 is found in cell adhesion complexes, suggest that it is an adaptor/docking protein involved in integrin signaling pathways.

Integrins are key regulators of cell adhesion and migration, processes that play a crucial role in cell survival and differentiation (1)(2)(3)(4). Upon ligand binding, integrins cluster and many intracellular components, including adaptors, structural proteins, and kinases, are recruited into large multimolecular complexes where actin microfilaments anchor. As integrins do not possess intrinsic enzymatic activity, proteins associated with the complexes activate signaling pathways (1, 4 -6). However, the precise molecular mechanisms of signal transmission still remain elusive. The cytoplasmic domain of integrin ␤ subunits is assumed to play an active role in transducing the signals, whereas those of the integrin ␣ subunits are thought to be modulators. These different functions are probably medi-ated by diverse proteins that have been shown to interact, for most of them in vitro, with the cytoplasmic domains of the ␣ or ␤ integrin subunits. Talin, ␣-actinin, paxillin, focal adhesion kinase (for review see Ref. 7), filamin (8), ␤ 3 -endonexin (9), integrin-linked kinase (10), cytohesin-1 (11), integrin cytoplasmic domain-associated protein 1 (ICAP-1; see Refs. 12 and 13), receptor for activated protein kinase C (Rack-1; see Ref. 14), and WD protein interacting with integrin tails (WAIT-1; see Ref. 15) were shown to interact with the cytoplasmic tail of either specific or various ␤ integrin subunits. Similarly, several proteins have been shown to bind the cytoplasmic domain of integrin ␣ subunits. Five of them, calreticulin (16), mammalian suppressor of secretion (Mss4), Bridging Integrator protein-1 (also called box-dependent Myc interaction protein-1 or BIN1), and ␣ integrin-binding proteins 63 (AIBP63) and 80 (AIBP80) (17) interact with several integrin ␣ subunits and bind to the conserved sequence proximal to the transmembrane domain. Another, the calcium-and integrin-binding protein, has a more restricted binding pattern (18). With the exception of calreticulin, which has been shown to modulate cell adhesion, the biological significance of these interactions is still unknown.
Such proteins are likely involved in the control of multiprotein complex remodeling. The ␣ 3 ␤ 1 integrin transdominantly regulates the clustering of other integrins, including ␣ 6 ␤ 1 (19 -21), ␣ 5 ␤ 1 , and ␣ 2 ␤ 1 (22) and of intracellular proteins associated with adhesion complexes. To identify proteins involved in the transdominant control exerted by the ␣ 3 ␤ 1 integrin, a yeast two-hybrid screen was performed using the cytoplasmic domain of the ␣ 3A integrin subunit. Of five proteins identified in the screen, four bind the cytoplasmic domain conserved from several integrin ␣ subunits, whereas the fifth protein interacted with the unique, more distal sequence of the integrin ␣ 3A cytoplasmic domain (17). This fifth protein is the four and a half LIM protein 2 (FHL2) of the LIM-only family, which was originally identified as DRAL (for down-regulated in rhabdomyosarcoma LIM protein) by subtractive cloning (23). To follow the recommended nomenclature (HUGO/GDB Nomenclature Committee) and to acknowledge the original discovery, we will refer to the protein as DRAL/FHL2. LIM domains are double zinc finger motifs, defining the expanding family of LIM proteins involved in protein-protein interactions and transcriptional regulation (24 -26). They are composed of either exclusively of LIM domains, the LIM-only proteins, or of LIM domains associated with homeodomains, kinase domains, or other functionally active sites, the LIM-plus proteins. Of the latter several proteins such as paxillin, zyxin, CRP1, abLIM, and the paxillin analog in platelets (Hic-5) are associated with the cytoskeleton (25,(27)(28)(29)(30)(31) and could function as nuclear-cytoplasmic shutters (32). DRAL/FHL2 is one of five known LIM-only proteins with four and a half LIM domains (33)(34)(35). We report here that DRAL/FHL2 and specific subdomains thereof have the capacity to interact with several ␤ integrin subunits, a restricted number of integrin ␣ subunits, and with integrin-binding proteins. Further DRAL/FHL2 can localize to cell adhesion complexes.

EXPERIMENTAL PROCEDURES
DNA Constructs-cDNA fragments encoding the complete cytoplasmic domain of integrin ␣ 2 , ␣ 3A , ␣ 5 , ␣ 6A , and ␤ 1A subunits were amplified by reverse transcriptase-PCR 1 using as template total RNA from the human mammary epithelial cell line HBL100. They were inserted into EcoRI/BamHI sites of the pAS2-1 vector (CLONTECH. Heidelberg, Germany) as fusion proteins with the GAL4 DNA binding domain as described previously (17). cDNA fragments containing the partial or complete cytoplasmic domains of the integrin ␣ 3B , ␣ 6B , ␣ 7A , ␣ 7B , ␤ 1D , ␤ 2 , and ␤ 3A subunits were amplified by PCR using the full-length cDNA as a template and appropriate specific sense and antisense primers containing restriction site tags. The constructs for the cytoplasmic domain of the integrin ␣ 1 and ␤ 6 subunits were provided by Drs. B. Eckes and D. Petersohn (Department of Dermatology, University of Cologne, Germany) and Dr. S. Spong (Lung Biology Center, University of California, San Francisco), respectively. All integrin ␣ 3A and ␤ 1A deletion mutants were derived by PCR amplification using the appropriate cDNA constructs and inserted into the pAS2-1 vector as above. DRAL/ FHL2, AIBP63, AIBP80, Mss4, and BIN1 were available from our previous study (17) as clones in the pACT2 vector (CLONTECH). Deletion mutants of DRAL/FHL2 generated by PCR were cloned into BamHI/XhoI sites of the pACT2 vector. AIBP80 and DRAL/FHL2 were subcloned and inserted into EcoRI/BamHI sites of the pAS2-1 vector.
For expression in mammalian cells, the cytoplasmic domains of integrin ␣ 2 (aa 1126 -1152), ␣ 3A (aa 1015-1051), ␣ 7A (aa 1060 -1115), and ␤ 1A (aa 752-798) subunits were cut from the pAS2-1 construct using NdeI and SpeI and, after fill-in of the NdeI overhang, the inserts were cloned in-frame into the GST-vector pEBG (a gift of Dr. A. Kalmes, University of Wü rzburg, Germany), which had been cut and filled in at the BamHI site and then digested with SpeI. The full-length human DRAL/FHL2 was cut from pACT2 construct by NheI/AvrII and inserted into the SpeI site of the pEBG vector. For the generation of Myc-tagged full-length or truncated DRAL/FHL2 and AIBP80, the cDNAs encoding these proteins were cut from pACT2 constructs (17) by SalI/XhoI and XbaI/XhoI, respectively, and cloned in-frame into the pCS2ϩMT vector (a gift of Dr. A. Kalmes). Full-length or truncated DRAL/FHL2 was inserted into the XhoI site, and AIBP80 was inserted after filling in of the existing overhangs into the StuI site. Full-length cDNA coding for the entire integrin ␣ 3A subunit in the Bluescript vector (a generous gift of Dr. M. E. Hemler, Dana-Farber Cancer Institute, Boston) was cut by XbaI and subcloned into the XbaI site of the pcDNA3 vector (Invitrogen, Groningen, The Netherlands).
Yeast Two-hybrid Library Screening, Mating, and Transformation Assays-Yeast cultures were grown under standard conditions in liquid or on solid media using YPD or minimal SD media. The yeast strain Y190 (CLONTECH) was transformed sequentially with the pAS2-1 plasmid coding for the cytoplasmic domain of the integrin ␣ 3A subunit as bait and then with a pACT2 plasmid containing the placenta cDNA library (CLONTECH). Transformants were grown on SD medium lacking the amino acids leucine, tryptophan, and histidine in the presence of 25 mM 3-amino-1,2,4-triazole. On day 5 the colonies were tested for the activity of the lacZ reporter gene in a ␤-galactosidase filter assay.
To remove the bait cDNA, positive clones were recultured on SD medium without tryptophan in the presence of 10 g/ml cycloheximide. The cycloheximide-resistant Y190 yeast clones were verified in a mating assay with yeast strain Y187 expressing the pAS2-1 plasmid with either the cytoplasmic domain of the integrin ␣ 3A subunit, the unrelated protein lamin C (CLONTECH), or the nonfused GAL4 DNA binding domain as baits. Clones were scored as positive when the His ϩ and LacZ ϩ phenotype of yeast cells was dependent on the co-expression of only the cytoplasmic domain of the integrin ␣ 3A subunit as bait. Such clones were retested in a co-transformation assay with purified plasmid cDNA using the same controls as in mating assays. For direct two-hybrid binding assays, yeast Y190 cells were co-transfected with DRAL/FHL2 or its deletion mutants fused to the GAL4 transactivation domain in the pACT2 vector and with one of the cDNA constructs coding for either integrin cytoplasmic domains, mutants thereof, or DRAL/FHL2 itself, fused to the GAL4-DNA binding domain in the pAS2-1 vector. In other experiments, yeast cells were co-transfected with DRAL/FHL2 in pAS2-1 and with AIBP63, AIBP80, Mss4, BIN1, or DRAL/FHL2 itself in pACT2. In all cases, positive clones were scored as described above.
Expression of Proteins in Mammalian Cells-For transient expression, human embryonic kidney 293 (HEK293) cells (American Type Culture Collection) or mouse 3T3 fibroblasts (kindly provided by Dr. U. Rapp, University of Wurzburg) were grown for 24 h in 6-well plates (2.5 ϫ 10 5 cells/35 mm-diameter well) prior to transfection with plasmid DNA (2 g/well) using Superfect Transfection mixture (Qiagen, Hilden, Germany) according to the manufacturer's instructions. When cells were co-transfected with different DNAs, the DNA content was equalized with appropriate amounts of empty expression vectors. For stable expression, HEK293 cells were transfected with the pTracer-CMV vector (Invitrogen) containing the full-length DRAL/FHL2 cDNA with an N-terminal insertion of the nine amino acid hemagglutinin tag (HA tag). The cells were trypsinized 48 h after transfection and selected further in medium containing 25 g/ml zeocin. After 2 weeks, single colonies were picked, cells were cultured, and the expression of HAtagged DRAL/FHL2 was analyzed by immunoblotting using HA tagspecific monoclonal antibody 12CA5 (a generous gift from Dr. U. Rapp, University of Wü rzburg, Germany).
Cell Cultures and Subcellular Fractionation-Normal human skin fibroblasts and the epithelial cell line HaCat were provided by Dr. H. Smola (Department of Dermatology, University of Cologne, Germany). Established lines of human lung fibroblasts (Wi26), embryonic kidney (293), fibrosarcoma (HT1080), mammary epithelia (HBL100), mammary epithelial carcinoma (MCF-7), epidermoid carcinoma (A431), and ductal mammary carcinoma (T47D) have been previously described (36 -38). CaCo2 cells were newly purchased from the ATCC. Rat PC12 and mouse NIH 3T3 cells were provided by Dr. U. Rapp (University of Wurzburg) and human RD-9 cells by Dr. A. Hoffmann (Department of Biochemistry, University of Cologne). All cells were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 2 mM glutamine, a mixture of antibiotics, and 10% heat-inactivated fetal calf serum (Seromed/Biochrom, Berlin, Germany). Cells were fractionated as described previously (39). Cell monolayers were washed twice with phosphate-buffered saline, pH 7.4, and scraped in 1 ml of hypotonic lysis buffer (1 mM EGTA, 1 mM EDTA, 10 mM ␤-glycerophosphate, 2 mM MgCl 2 , 10 mM KCl, 1 mM sodium vanadate, 1 mM phenylmethylsulfonyl fluoride, 5 g/ml leupeptin, 5 g/ml aprotinin, 2 mM pepstatin, pH 7.2). After incubation on ice for 30 min, the cells were homogenized with a tight-fitting pestle and loaded onto 1 ml of 1 M sucrose in lysis buffer. The nuclear fraction was collected by centrifugation (1,600 ϫ g, 10 min). The pellet was washed once with 1 M sucrose in lysis buffer. The supernatant was further centrifuged (150,000 ϫ g, 30 min), and the resulting pellet and supernatant was taken as membrane and cytosolic fraction, respectively. Cytosolic proteins were precipitated by the methanol/chloroform method (40). Pellets of all three fractions were dissolved in 100 l of electrophoresis sample buffer, and 30 l were used for SDS-PAGE analysis.
Immunoprecipitation and Immunoblotting-Transiently transfected cells were washed twice with phosphate-buffered saline and lysed in 25 with DRAL/FHL2 in a yeast two-hybrid binding assay Yeast Y190 cells were co-transformed with the indicated combinations of bait and prey. The activation of the first reporter gene was determined by growth on His Ϫ medium and expression of the second reporter gene, lacZ, evaluated in a ␤-galactosidase filter assay. The interaction was scored as negative (Ϫ) when no blue colonies were visible after 8 h; the interaction was scored as weak (ϩ), intermediate (ϩϩ), or strong (ϩϩϩ) when blue colonies became visible after 8, 4, or 1 h, respectively. The ubiquitously expressed lamin C was used as a negative control.

Bait Prey
Reporter gene His expression lacZ activity mM Hepes, pH 7.5, 137 mM NaCl, 1 mM MgCl 2 , 1% Brij 98, or 1% Triton X-100, 2% glycerol, 1 mM sodium vanadate, 1 mM phenylmethylsulfonyl fluoride, 5 g/ml leupeptin, 5 g/ml aprotinin, 2 mM pepstatin at room temperature for 20 min. The lysates were cleared by centrifugation (10,000 ϫ g, 10 min) at 4°C, and the supernatants were incubated for 3 h at 4°C with antibodies against the Myc tag (clone 9E10, Oncogene distributed by Calbiochem, Schwalbach, Germany) and protein G-con-jugated agarose (Roche Molecular Biochemicals) or with glutathioneconjugated Sepharose beads (Amersham Pharmacia Biotech) when precipitating GST-tagged proteins. The complexes were washed three time with lysis buffer, suspended in electrophoresis sample buffer, and heated at 95°C for 3 min. The samples were resolved by SDS-PAGE on 10% acrylamide gels and electrophoretically transferred onto nitrocellulose membrane. Proteins were detected with goat polyclonal antibodies against a synthetic peptide corresponding to the cytoplasmic domain of the integrin ␣ 3A subunit (Santa Cruz Biotechnology) or against GST (Life Technologies, Inc.), mouse monoclonal antibodies against Myc (clone 9E10), or rabbit polyclonal antibodies against recombinant GST-DRAL/FHL2 fusion protein and partially purified by affinity chromatography on a GST column (a generous gift of Dr. B. Schä fer, University of Zurich, Switzerland), followed by appropriate horseradish peroxidase-coupled secondary antibodies (DAKO, Glostrup, Denmark) and the ECL detection system (Amersham Pharmacia Biotech). Immunofluorescence Staining of Cell Adhesion Complexes-Cells were cultured either overnight on uncoated or for 60 min on fibronectincoated (10 g/ml) glass coverslips in DMEM containing 10% fetal calf serum. The cells were fixed with 2% paraformaldehyde for 15 min, permeabilized with 0.2% Triton X-100 for 2 min, and blocked with 1% BSA (fraction V, Sigma). The cells were processed for immunofluorescence staining with mouse monoclonal antibody F-VII against human vinculin (a gift from Dr. M. Glukhova, Institut Curie, Paris, France), K20 (Immunotech, Marseille, France), or TS2/16 (41) against the human ␤ 1 integrin subunit, 9E10 against the Myc tag, or rabbit polyclonal After 48 h, cell lysates were divided into two parts. From 1 aliquot, proteins were precipitated with the antibody 9E10 against the Myc tag followed by protein G-conjugated agarose for precipitation of Myc-tagged proteins. With the other aliquot, proteins were precipitated with glutathioneconjugated Sepharose beads for GST-tagged proteins. After SDS-PAGE and immunoblotting, the cytoplasmic domain of the integrin ␣ 3A subunit was only detected in precipitates from cells where the integrin peptide had been co-expressed with Myc-tagged DRAL/FHL2 (left blot, lane 4). In the reverse experiment (glutathione-precipitated complexes), Myc-tagged DRAL/FHL2 was detected only when co-expressed with GST-tagged ␣ 3A peptide (right blot, lane 4) but not with GST alone. After the first immunodetection, the blots were stripped and redeveloped with polyclonal antibody against recombinant DRAL/FHL2 (bottom, left blot) or against GST (bottom, right blot) to ascertain that Myc-tagged DRAL/FHL2 and GST-tagged ␣ 3A peptide were equivalently precipitated. B, association of DRAL/FHL2 with the full-length integrin ␣ 3A subunit. HEK293 cells were transiently transfected with different cDNA constructs as indicated. After cell lysis, Myc-tagged DRAL/FHL2 was immunoprecipitated (IP) with antibody against Myc and the precipitated proteins were visualized by immunoblotting (IB) with antibodies against the cytoplasmic tail of the integrin ␣ 3A subunit (upper blot), or against Myc (lower blot).

FIG. 2.
Identification of a DRAL/FHL2-binding site in the unique amino acid sequence of the integrin ␣ 3A subunit cytoplasmic domain. Yeast Y190 cells were transformed with pAS2-1 plasmid expressing the indicated deletion mutants of the integrin ␣ 3A subunit and with DRAL/FHL2 in the pACT2 vector. The interaction was determined and scored as described in the legend to Table I. Numbers in parentheses refer to amino acid positions. The conserved membrane-proximal region is marked. antiserum against recombinant GST-DRAL/FHL2, followed by Cy3conjugated second antibodies against mouse or rabbit immunoglobulins (Jackson Immunoresearch Laboratories distributed by Dianova, Hamburg, Germany) together with fluorescein isothiocyanate-conjugated phalloidin (Sigma). For analysis of cell footprints the cells were lysed by osmotic shock with or without cross-linking with 1 mM dithiobis(succinimidyl propionate) (DSP) (Pierce) as described previously (38) and were then processed for immunofluorescence labeling as detailed above. After mounting, the samples were observed with an Axiophot microscope (Zeiss, Oberkochen, Germany) equipped with epifluorescence optics.

Identification of DRAL/FHL2 as a Protein
Binding the Cytoplasmic Domain of the Integrin ␣ 3A Subunit-To identify proteins that bind to the cytoplasmic domain of the ␣ 3A integrin subunit, a human placenta library (Ͼ5 ϫ 10 6 independent clones) was used in a yeast two-hybrid screen with the Cterminal cytoplasmic part of the ␣ 3A integrin subunit (aa 1015-1051) as bait (17). Out of 84 His ϩ -and LacZ ϩ -positive clones (17), 28 contained cDNAs with an identical open reading frame coding for the 279 amino acid residues of DRAL/FHL2. The specificity of this interaction was confirmed in a direct twohybrid binding assay (Table I). Transformation of yeast cells with DRAL/FHL2 in pACT2 vector alone or together with the GAL4-DNA binding domain in pAS2-1 vector or together with an unrelated protein, lamin C, fused to the GAL4-DNA binding domain instead of ␣ 3A , did not activate the His and lacZ reporter genes ( Table I). The predicted amino acid sequence contains four double zinc finger LIM domains at the C-terminal end and a half-LIM domain at the N terminus. It corresponds to FHL2 (23,42,43).
DRAL/FHL2 Interacts with the Cytoplasmic Domain of the Integrin ␣ 3A Subunit when Co-transfected into Human Cells-To test if DRAL/FHL2 binds the integrin ␣ 3A subunit also in mammalian cells, HEK293 cells were co-transfected with the GST-tagged ␣ 3A cytoplasmic domain (aa 1015-1051) and with Myc-tagged DRAL/FHL2. These were tested for complex formation in co-precipitation experiments. The GSTtagged integrin ␣ 3A cytoplasmic domain precipitated in immune complexes formed in the presence of antibodies against the Myc tag as shown by immunoblotting with specific antibodies (Fig. 1A). Reciprocally, the presence of Myc-tagged DRAL/ FHL2 in complexes specifically precipitated with antibodies against the cytoplasmic domain of the ␣ 3A integrin subunit was dependent on the expression of GST-tagged ␣ 3A cytoplasmic domain but not on the expression of GST alone (Fig. 1A). Thus, consistent with the two-hybrid data, the LIM-only protein DRAL/FHL2 specifically interacts with the cytoplasmic domain of integrin ␣ 3A subunit in mammalian cells.
To analyze whether DRAL/FHL2 also binds the full-length recombinant ␣ 3A subunit, the Myc-tagged DRAL/FHL2 and the full-length integrin ␣ 3A subunit were transfected in HEK293 cells that normally express only small amounts of these polypeptides (data not shown, but see Fig. 7). Only upon coexpression with Myc-tagged DRAL/FHL2, was the integrin ␣ 3 chain detected in Myc-containing immunocomplexes (Fig. 1B).
Mapping of the DRAL/FHL2 Binding Site in the Cytoplas-  integrin ␣ and ␤ subunits in a yeast two-hybrid assay Interactions between the cytoplasmic domain of integrin subunits in pAS2-1 vector and DRAL/FHL2 in pACT2 vector were determined as described in the legend to Table I. The deletion mutants of ␤ 2 cyto-1 and ␤ 3A cyto-1 regions are based on the corresponding integrin subunit cytoplasmic domains and contain, respectively, the last 23 and 24 C-terminal amino acid residues, including the cyto-2 and cyto-3 regions, but lacking cyto-1. In the ␤ 6 -11 construct, the last 11 C-terminal amino acid residues of the integrin ␤ 6 subunit cytoplasmic domain were deleted, but the cyto-3 region is preserved. All other constructs represent the full-length cytoplasmic domains. Scoring was as described in the legend to Table I.  (17) that DRAL/FHL2 does not bind to the stretch of highly conserved amino acids KXGFFKR proximal to the transmembrane domain and common to all integrin ␣ subunits. Here we show also no interaction with a similar construct containing three additional residues at the C terminus, which are crucial for integrin function (44,45). Furthermore, no binding occurred to a construct containing the last 18 C-terminal residues. However, there was a positive reaction with all constructs that contain a stretch of 12 amino acid residues beyond the conserved KCG-FFKR motif irrespective of other deletions (Fig. 2). Binding Specificity of DRAL/FHL2 for ␣ or ␤ Integrin Subunits and Mapping of Its Binding Site within the Integrin ␤ 3A Chain-To investigate the specificity of DRAL/FHL2 binding, direct two-hybrid interaction tests were performed between DRAL/FHL2 and the cytoplasmic domain of nine different ␣ integrin subunits: the ␣ 1 and ␣ 2 integrin chains of collagen receptors, the ␣ 5 chain of the fibronectin receptor, and the A and B variants of the ␣ 3 , ␣ 6 , and ␣ 7 chains of laminin receptors (Table II). In addition to the integrin ␣ 3A subunit cytoplasmic domain, ␣ 3B and ␣ 7A subunits interacted with DRAL/FHL2 (Table II). Similar experiments with the cytoplasmic domains of different integrin ␤ subunits indicated that DRAL/FHL2 interacted with all ␤ integrin chains, including ␤ 1A , ␤1D, ␤ 2 , ␤ 3A , and ␤ 6 ( Table II).
The cytoplasmic domains of integrin ␤ subunits have a stretch of conserved amino acid residues proximal to the transmembrane domain and additionally share the conserved cyto-1, cyto-2, and cyto-3 regions, which are important for integrin function (46 -48). Cyto-2 and cyto-3 contain typical proteinbinding sequences, NPXY and NXXY, respectively. Six different deletion mutants containing one or more of the characteristic motifs of the integrin ␤ 1A cytoplasmic domain were constructed and tested for DRAL/FHL2 binding in the yeast two-hybrid assay. The results showed that the last nine Cterminal amino acids, including the NXXY (NPKY) sequence of the cyto-3 region, are necessary for interaction between DRAL/ FHL2 and the cytoplasmic domain of integrin ␤ 1A subunit (Fig.  3). The importance of this C-terminal sequence was supported by data obtained with deletion mutants of the integrin ␤ 2 , ␤ 3A , and ␤ 6 chains in which the cyto-3 region was preserved (Table II).
DRAL/FHL2 Interacts with the Cytoplasmic Domain of the ␣ 7A or ␤ 1A Integrin Subunits in Co-transfected Human Cells-To analyze whether the interactions reported above occur also in mammalian cells, HEK293 cells were transiently co-transfected with the GST-tagged cytoplasmic domain of three additional integrin subunits, ␣ 2 , ␣ 7A , or ␤ 1A , and with Myc-tagged DRAL/FHL2. As with the integrin ␣ 3A subunit, the Myc-tagged DRAL/FHL2 was precipitated together with GST-tagged ␣ 7A or ␤ 1A integrin cytoplasmic domains when using glutathione-conjugated Sepharose beads (Fig. 4). As expected from the two-hybrid data, upon co-expression of the integrin ␣ 2 cytoplasmic domain and DRAL/FHL2, the GST-tagged ␣ 2 integrin subunit was not associated with DRAL/FHL2 in the complexes (Fig. 4). The reciprocal immunoprecipitation test with antibodies against Myc confirmed these results (not shown).
DRAL/FHL2 Interacts with Itself and with Other Integrinbinding Proteins-Several LIM domain-containing proteins have been suggested to dimerize, thus enhancing their capacity for complex formation (49). DRAL/FHL2, which contains four and a half LIM domains, could therefore represent an adaptor protein that could interact in multimolecular complexes with other integrin-binding proteins such as those that we have recently identified (17). This possibility was tested in direct two-hybrid interaction experiments. The results showed that DRAL/FHL2 is able to interact with itself and with two additional integrin-binding proteins, AIBP80 and BIN1 (Table III). To test the interactions in vivo, we constructed expression vectors for GST-tagged DRAL/FHL2 and for Myc-tagged AIBP80. HEK293 cells co-expressing GST-DRAL/FHL2 and Myc-DRAL/FHL2 or Myc-AIBP80 were lysed, and proteins were precipitated with either anti-Myc antibodies or with glutathione-coupled Sepharose beads. Immunoblot analysis of the precipitates showed that also in human cells DRAL/FHL2 selfinteracts and that AIBP80 associates with DRAL/FHL2 (Fig. 5).
Molecular Dissection of Binding Sites in DRAL/FHL2-To identify the LIM domains of DRAL/FHL2 responsible for the diverse interactions described in this study, cDNA sequences for single LIM modules were cloned in the pACT2 vector and were tested in direct two-hybrid binding assays with the other proteins, including DRAL/FHL2, cloned in pAS-2 vector. Surprisingly, only two out of six proteins analyzed interacted with particular single LIM domains; the cytoplasmic domain of the integrin ␣ 7A subunit had affinity for the LIM 2 domain, whereas DRAL/FHL2 itself bound the C-terminal LIM 3 and LIM 4 domains (Fig. 6A). Tests with additional DRAL/FHL2 mutants, in which one, two, or three LIM domains were deleted from either the C or N terminus confirmed that the LIM 2 domain is responsible for binding to the integrin ␣ 7A subunit and that LIM 3 and LIM 4 modules are involved in DRAL/ FHL2 self-association (Fig. 6A). Additionally, they revealed that both LIM 1 and LIM 2 N-terminal domains are required for interaction with AIBP80, whereas binding of the cytoplasmic domains of the integrin ␣ 3A or ␣ 3B subunits needs the last 2, 3, and 4 LIM domains (Fig. 6A). Furthermore, deletion of any LIM domain prevented the binding of DRAL/FHL2 to the cytoplasmic domain of the integrin ␤ 1A subunit (Fig. 6A), suggesting that the three-dimensional structure of DRAL/FHL2 is important for this interaction. The major binding sites of the analyzed proteins are summarized in Fig. 6B.

DRAL/FHL2 Is Expressed by Normal Human Skin Fibroblasts and Is Mainly Localized in the Cell Nucleus and
Cytosol-DRAL/FHL2 was initially described as a protein preferentially localized in nuclei after overexpression in NIH 3T3 cells, whereas it was distributed uniformly over the nucleus and cytoplasm in Rh30 cells (23). To analyze the subcellular localization of DRAL/FHL2 more precisely, we screened several human and rodent cells or cell lines for the presence of endogenous DRAL/FHL2. Immunoblotting with an antiserum raised against the recombinant GST fusion protein (23) showed the presence of DRAL/FHL2 in the cell lysates of normal human fibroblasts, Wi26 human fibroblasts, mouse NIH 3T3, and DRAL/FHL2-transfected HEK293 cells as a prominent band migrating under reducing (Fig. 7A) or non-reducing (not shown) conditions at the position expected for an ϳ31-kDa polypeptide, in agreement with previous data (23). In cell lysates from a panel of transformed or tumor cell lines of epithelial or neuronal origin, the DRAL/FHL2 band was fainter or barely seen and had a slightly different mobility (Fig. 7A), which could indicate different post-translational modifications. Human skin fibroblasts expressed most DRAL/FHL2 and were chosen to study the subcellular localization of naturally occurring protein. After fractionation by differential centrifugation, DRAL/FHL2 was mainly found in the nuclear and cytosolic fractions with only small amounts in the cell membrane fraction (Fig. 7B). Endogenous DRAL/FHL2 Can Be Recruited to Cell Adhesion Complexes-As integrins function at the cell surface in adhesion complexes, we used immunofluorescence microscopy to establish whether DRAL/FHL2 is localized at the periphery of spread cells, where it would be expected to interact with the cytoplasmic tail of integrin subunits. Normal human skin fibroblasts were the most appropriate for the study since they express more DRAL/FHL2 than most of the established cell lines (Fig. 7). Moreover, the polyclonal antibody raised against GST-DRAL/FHL2 recognized a single major band in immunoblots of normal human skin fibroblast lysates (Fig. 7). Immunofluorescence staining with the polyclonal antiserum against DRAL/FHL2 showed that naturally expressed DRAL/FHL2 was associated with fibril-like structures within the cell body and with clusters at the cell periphery (Fig. 8, A, D, and G). Double labeling with antibodies against DRAL/FHL2 (Fig. 8, A,  D, and G) and with monoclonal antibodies against the integrin ␤ 1 subunit (Fig. 8B) or vinculin (Fig. 8H) showed overlap for FIG. 5. Association of DRAL/FHL2 with itself and with AIBP80 in human cells. After transient transfection with the indicated cDNA constructs, lysates of HEK293 cells were precipitated (P) as described in Fig. 2. Co-precipitating proteins were identified by immunoblotting (IB) with the indicated antibodies. Myctagged DRAL/FHL2 co-migrates with the IgG heavy chain and could therefore not be visualized with the antibodies against Myc.
FIG. 6. Identification of sites on DRAL/FHL2 that bind to integrin subunits and to AIBP80 or are required for DRAL/FHL2 self-interaction. A, yeast Y190 cells were co-transformed with one of the cDNA fragments encoding DRAL/FHL2 deletion mutants inserted into the pACT2 vector and with the cDNA fragments encoding one of the ligand proteins inserted into the pAS2-1 vector. The numbers in parentheses indicate DRAL/FHL2 amino acids encoded by the corresponding constructs. The interactions were detected by growth on His Ϫ medium and in a ␤-galactosidase filter assay. Scoring was as described in the legend to Table I. B, schematic representation of principal binding sites on DRAL/FHL2 identified in A. many but not all of the clusters (Fig. 8, superimposed photographs C and I). Furthermore, visualization of fibrillar actin by fluorescein isothiocyanate-conjugated phalloidin in cells stained with the antiserum against DRAL/FHL2 revealed that DRAL/FHL2 is clustered at the ends of actin bundles (superimposed photograph F). Again, not all DRAL/FHL2-positive clusters were located at the ends of the actin fibers and, vice versa, DRAL/FHL2-positive clusters were not always present at the termini of actin fibers. In contrast, the antiserum against DRAL/FHL2 did not stain any of the vinculin-positive peripheral clusters in RD-9 cells (not shown), which agrees with previous results showing that DRAL/FHL2 is down-regulated in rhabdomyosarcoma cells (23).
Cell adhesion complexes are dynamic multimolecular assemblies of kinases and adaptor proteins that are structurally or transiently associated to integrin clusters (4,5). To confirm the association of DRAL/FHL2 with cell adhesion complexes, footprints were prepared by submitting normal human fibroblasts to osmotic shock in the absence or in the presence of DSP, a membrane-permeable cross-linker. Immunofluorescence labeling showed that both integrin ␤ 1 subunit and vinculin, but not DRAL/FHL2, were present in non-cross-linked cell remnants (Fig. 9). However, after cross-linking, DRAL/FHL2 was detected in the footprints of normal human fibroblasts (Fig. 9).
Finally, mouse 3T3 fibroblasts were transiently transfected with Myc-tagged full-length or truncated DRAL/FHL2 representing the N-and C-terminal part of the molecule. Immunofluorescence detection of the tagged protein with an antibody against Myc showed that full-length DRAL/FHL2 localized to focal adhesion clusters (Fig. 10A). In contrast, the immunofluorescence staining of mouse 3T3 fibroblasts transfected with constructs representing the N-or C-terminal half of DRAL/ FHL2 was only nuclear (Fig. 10, B and C), indicating that only full-length DRAL/FHL2 can be targeted to focal adhesion clusters. DISCUSSION In this report we show that the LIM-only protein DRAL/ FHL2 is a novel integrin-binding protein. After the identification in a yeast two-hybrid screen as interacting with the cytoplasmic domain of the integrin ␣ 3A subunit, we provide evidence that DRAL/FHL2 has the capacity to interact, both in yeast and in mammalian cells, with itself, with the cytoplasmic domain of integrin ␣ 3A , ␣ 3B , ␣ 7A , and several ␤ subunits, and with integrin-binding proteins. Furthermore, studies with mutant forms of DRAL/FHL2 demonstrate that different LIM domains are responsible for those interactions. Finally, we show that DRAL/FHL2, previously described as a nuclear protein, is targeted to cell adhesion complexes. Together, these results suggest that DRAL/FHL2 may act as an adaptor protein, regulating integrin trafficking, function, or signaling.
A molecular dissection of the DRAL/FHL2 integrin ␣ subunit-binding sites showed the importance of 12 amino acid residues immediately following the conserved membrane-proximal region of the integrin ␣ 3A subunit. Several cytoplasmic proteins, including calreticulin, calcium-, and integrin-binding protein, BIN1, Mss4, AIBP63, and AIBP80 interact with the cytoplasmic tail of integrin ␣ subunits and require the highly conserved KXGFFKR sequence for optimal binding (16,17,50). DRAL/FHL2, therefore, is the first protein shown to bind the non-conserved region of integrin ␣ subunit cytoplasmic domains. Furthermore, DRAL/FHL2 has a specificity restricted to the integrin ␣ 3A , ␣ 3B , and ␣ 7A chains. These subunits have divergent sequences after the conserved KXGFFKR motif, and it was unexpected that DRAL/FHL2 interacts with all three. A deletion mutation analysis of DRAL/FHL2 itself solved the apparent contradiction, as it was revealed that different LIM domains interact with the ␣ 7A peptide and the ␣ 3 variants.
In integrin ␤ subunits the C-terminal part of the cytoplasmic domain, which includes the cyto-3 NXXY motif, is needed for binding to DRAL/FHL2. Except for the integrin ␤ 4 and ␤ 8 subunits, the cytoplasmic domains of the ␤ chains share the functionally important cyto-1, cyto-2, and cyto-3 regions (46 -48). Cyto-2 and -3 contain the NPXY and NXXY motif, respectively. NPXY motifs are recognition sites for phosphotyrosinebinding proteins such as those containing SH2 domains (51). It is believed that the integrin cyto-2 region folds in a ␤-turn so that the C-terminal part of the chain, containing the NXXY motif of cyto-3, can be brought in the vicinity of the membrane proximal conserved region (52). Mutations in the NPXY sequence, truncation or deletion of cyto-2, or peptides representing this region impair talin, filamin, and ␣-actinin binding to the ␤ 1 cytoplasmic tail as well as cell adhesion, spreading, and formation of focal adhesions (47,(53)(54)(55)(56)(57). The NXXY motif of cyto-3 is required for ICAP-1 or ␤ 3 -endonexin binding to the cytoplasmic domain of integrin ␤ 1 or ␤ 3 subunit, respectively (12,13,58). Our data show that DRAL/FHL2 binds all tested integrin ␤ subunits. Results obtained with sequential deletion mutants of the ␤ 1A subunit and further deletion mutants of several other ␤ subunits indicated that the NXXY cyto-3 motif is within the critical binding site. Thus, in contrast to ICAP-1 and ␤ 3 -endonexin, DRAL/FHL2 is a binding partner common to all integrin ␤ subunits that possess the NXXY motif. One hypothesis is that the function of ␤ subunits in integrin activation is regulated through a conformational change involv- ing folding of the C-terminal region containing cyto-3 over the N-terminal region (52). Such a change could be under the control of cytoplasmic factors. Binding of DRAL/FHL2 to the cyto-3 region may regulate the conformational changes of the integrin ␤ 1 cytoplasmic domain by either inducing or inhibiting the folding or, alternatively, by transiently competing with other proteins for binding to this domain.
Our observation that DRAL/FHL2 binds numerous proteins is not surprising because it contains several LIM domains and those are thought to function as protein interaction modules. In addition, DRAL/FHL2 has the property to bind both ␣ and ␤ integrin subunits through different LIM domains. Single LIM domains are involved in binding to the integrin ␣ 7A subunit or in DRAL/FHL2 homodimerization, while for binding to other proteins the coordinated action of several LIM domains was needed. The most striking result was that deletion of any LIM domain prevented the interaction with the integrin ␤ 1A subunit, suggesting that the three-dimensional structure of DRAL/ FHL2 is required for binding. At this point we do not know if the different interactions described here for DRAL/FHL2 can take place simultaneously or whether homodimerization of DRAL/FHL2 influences the interactions. LIM domains fold independently and are held together by a linker region (26,59). Furthermore, a single LIM domain, or even a single zinc finger module of a LIM domain, can function as protein-binding interface so that single LIM domain could be functionally bipartite (27). For example, although the LIM-only protein PINCH consists of five LIM domains, it binds integrin-linked kinase by only the most N-terminal one (60). Similarly, zyxin, a LIM-plus protein with a tandem of three LIM domains, binds to another FIG. 8. Localization of DRAL/FHL2 to adhesion complexes in normal human skin fibroblasts. The cells were cultured overnight in DMEM containing 10% fetal calf serum, fixed and doublestained with rabbit polyclonal antiserum against DRAL/FHL2 together with the mouse monoclonal antibody TS2/16 (A) against the integrin ␤ 1 subunit or F-VII (C) against vinculin, followed by the appropriate fluorescence-labeled second antibodies or by fluorescein isothiocyanateconjugated phalloidin (E). Specimens were examined and photographed under epifluorescence microscopy using separate detection channels (A, B, D, E, G, and  H). Superimposed photographs show that the integrin ␤ 1 subunits (C) and vinculin (I) are, but not always, co-localized with DRAL/FHL2. Superimposition of actin and DRAL/FHL2 images (F) reveals that DRAL/FHL2 is localized at the far ends of actin stress fibers that usually terminate in cell adhesion complexes.
LIM-containing protein, CRP1, also by a single LIM domain (27). These data, together with the fact that DRAL/FHL2 forms dimers, suggest that this novel integrin-binding protein has great potential for forming multimeric protein complexes. Moreover, sequence comparison of DRAL/FHL2-interacting proteins did not reveal any homology, except for the integrin ␤ subunits that share the cyto-1, -2, and -3 motifs, so that due to its modular structure DRAL/FHL2 can potentially be involved in many different interactions. Interestingly, further interactions of DRAL/FHL2 with the the androgen receptor (61) or with hCDC47 (62) were recently described. The interaction with the androgen receptor requires both the N-and C-terminal domains of DRAL/FHL2, and the interaction with hCDC47 involves the LIM 2 and 3 domains together with the first half LIM motif of FHL2/DRAL, respectively. Taken together, this suggests that by using different sets or combinations of its LIM domains this adaptor protein could be involved in the organization or the regulation of very diverse multimolecular complexes including transcriptional complexes.
Although DRAL/FHL2 is mainly localized in the cell nucleus and cytosol (23; this report), it was recruited to cell adhesion complexes in several cell types, including normal skin fibroblasts. There, it was clustered together with integrins and vinculin at the ends of actin stress fibers. Only full-length DRAL/FHL2 was targeted to cell adhesion complexes in transfected mouse 3T3 fibroblasts while truncated versions of the protein were not. This agrees well with the results observed in yeast interaction assays showing that binding to the integrin ␤ 1A subunit requires full-length DRAL/FHL2. In this aspect, the requirement for binding is similar to that reported for the androgen receptor (61). Several proteins from the LIM-plus family, like paxillin, zyxin, and abLIM, are adaptors involved in scaffolding of focal adhesion complexes or of the cytoskeleton (25,(27)(28)(29). The Caenorhabditis elegans LIM-only protein UNC-97, that also has both nuclear and extranuclear distribution, co-localizes in muscle with ␤ integrin in focal adhesionlike structures (63). The other LIM-only proteins PINCH and CRP1 can also be recruited to integrin signaling complexes through interactions with integrin-linked kinase (60) or via zyxin and ␣-actinin (25), respectively.
Our cell fractionation experiments showed that only a small proportion of DRAL/FHL2 is present in the membrane fraction of human fibroblasts and that DRAL/FHL2 is more abundant in the cell nucleus and the cytoplasm. This suggests that DRAL/FHL2 may also have other functions within the cell such as that recently described in the formation of transcriptional complexes (61). That only a small amount of DRAL/FHL2 was found to be membrane-associated can explain why we could not precipitate the endogenous protein with antibodies against integrin chains. Only when recombinantly overexpressed in mammalian cells, an interaction of DRAL/FHL2 and integrin cytoplasmic domains or full-length integrin ␣ 3A (this report) or ␣7A 2 could be observed. Alternatively, in mammalian cells interaction of DRAL/FHL2 with integrins may be favored by a certain conformation or activation states of the integrins such as those induced by extracellular ligand binding and which are lost under the experimental conditions required for immunoprecipitation. Remarkably, binding of DRAL/FHL2 to the androgen receptor is strictly agonist-dependent in mammalian cells (61). Moreover, in fibroblasts co-localization of DRAL/ FHL2 and vinculin, a marker of focal adhesion complexes, or integrin ␤ 1 subunit was not always seen. Furthermore, association of naturally expressed DRAL/FHL2 within adhesion complexes of fibroblasts is weak, at least weaker than that of vinculin, since it was retained in cell footprints only after cross-linking. This finding argues for a transient or regulatory role of DRAL/FHL2, a hallmark expected of proteins involved in the signal transduction cascade initiated in adhesion com-2 V. Wixler and K. von der Mark, unpublished observations. FIG. 9. Immunofluorescence staining of human skin fibroblast footprints. Fibroblasts were subjected to osmotic shock either without (a, c, and e) or with (b, d, and f) previous cross-linking of proteins with DSP and further processed for immunofluorescence staining with the mouse monoclonal antibody K20 (a and b) against integrin ␤ 1 subunit or F-VII (c and d) against vinculin, or the rabbit polyclonal antiserum against DRAL/FHL2 (e and f).  Fig. 6A. The cells were cultured on fibronectin-coated coverslips in DMEM containing 10% fetal calf serum for 60 min. After fixation, they were stained with mouse monoclonal antibody 9E10 against the Myc tag followed by Cy3-conjugated second antibodies.
plexes. The migration velocity in in vitro wound closure assays as well as the adhesion to several extracellular matrix proteins, including collagens, fibronectin, and laminins, were not changed after overexpression of DRAL/FHL2. Moreover, in the RD-9 rhabdomyosarcoma cells vinculin-positive adhesion complexes were present despite the absence of DRAL/FHL2. Thus, while DRAL/FHL2 obviously participates in protein clusters formed by integrins in fibroblasts, its presence does not appear to be essential for adhesion complexes. In this regard it is similar to LIM proteins like PINCH (60) or Hic-5 (31).
A function as a nuclear-cytoplasmic shuttling protein has been proposed for zyxin, a member of the LIM-plus family (32). The LIM domain structure has been resolved for a number of proteins, and they all show a striking similarity to the DNAinteracting CCCC module of the transcription factor GATA-1 (64 -66), suggesting that they can bind DNA (33,42). Indeed, two recent reports describe interactions of FHL2/DRAL with the androgen receptor (61) or with hCDC47 (62) and its involvement in transcriptional complexes. Finally, one of the other four and a half LIM domain containing proteins, ACT (activator of CREM in testis), stimulates the transcriptional activity of CREM and CREB (35). In conclusion, DRAL/FHL2 appears as an excellent candidate to facilitate the processing of integrin signals at the cell membrane into the cell program since such proteins are likely to be required for the transfer of information from cell adhesion complexes into the nucleus.