Interaction of the Integrin β1 Cytoplasmic Domain with ICAP-1 Protein*

In a yeast two-hybrid screen, a protein named ICAP-1 (β1 integrin cytoplasmic domain associated protein) associated with the integrin β1 cytoplasmic tail but not with tails from three other integrin β subunits (β2, β3, and β5) or from seven different α subunits. Likewise in human cells, ICAP-1 associated specifically with the β1 but not β2, β3, or β5 tails. The carboxyl-terminal 14 amino acids of β1 were critical for ICAP-1 interaction. ICAP-1 is a ubiquitously expressed protein of 27 and 31 kDa, with the smaller form being preferentially solubilized by Triton X-100. Phosphorylation of both 27- and 31-kDa forms was constitutive but was increased by 1.5–2-fold upon cell spreading on fibronectin, compared with poly-l-lysine. Also, ICAP-1 contributes to β1 integrin-dependent migration because (i) ICAP-1 transfection markedly increased chemotactic migration of COS7 cells through fibronectin-coated but not vitronectin-coated porous filters, and (ii) support of β1-dependent cell migration (in Chinese hamster ovary cells transfected with various wild type and mutant β1 forms) correlated with ICAP-1 association. In summary, ICAP-1 (i) associates specifically with β1 integrins, (ii) is phosphorylated upon β1 integrin-mediated adhesion, and (iii) may regulate β1-dependent cell migration.

Integrin-dependent cell adhesion helps to control cell proliferation and apoptosis, as well as cell spreading, migration, morphogenesis, and differentiation (1)(2)(3)(4)(5). Upon cell adhesion, integrin engagement leads to downstream activation of focal adhesion kinase, mitogen-activated protein kinase, and many other key signaling molecules (6). At the same time, integrindependent reorganization of cytoskeletal proteins and signaling complexes facilitates growth factor signaling (7,8). A distinctive property of integrins is that they not only deliver "outside-in" signals upon engagement with ligand but also their function is regulated by "inside-out" signals (9 -12). In this regard, integrin function can be strongly modulated upon overexpression of various oncogenes (13)(14)(15) or upon engagement of various cell-surface receptors with ligands or antibodies (10,16,17).
We also have undertaken a yeast two-hybrid screen to identify ␤ 1 tail-associated proteins. In the yeast, we initially identified two candidate ␤ 1 tail-interacting proteins. These were (i) a fragment of RACK1 and (ii) a protein called ICAP-1 (integrin cytoplasmic tail associated protein). Additional yeast two-hybrid studies suggested that the RACK1 interaction was nonspecific. However, the ICAP-1 protein did show specific interaction with the ␤ 1 tail, both in yeast and in human cell lines. Furthermore, the site of ICAP-1 association was mapped to the 14 carboxyl-terminal amino acids of ␤ 1 (which includes an NPXY motif); ICAP-1 phosphorylation was found to be regulated upon cell spreading on fibronectin, and ICAP-1 appeared to play a role in ␤1-dependent cell migration.
Yeast Two-hybrid Screening-Yeast genetic screening for proteins that bind to the integrin ␤ 1 cytoplasmic tail was carried out essentially as described (49,50). Integrin ␤ 1 subunit cDNA encoding for carboxylterminal amino acid residues 750 -798 (see Table I) was amplified from full-length ␤ 1 cDNA by polymerase chain reaction (PCR). This PCR product was ligated into plasmid pEG202N to generate the LexAintegrin fusion "bait" plasmid pEG202-␤1. Host yeast strain EGY48 (MAT␣, his3, trp1, ura3-52, leu2::pLEU 2-LexAop6, constructed by E. Golemis, Massachusetts General Hospital, Boston) was cotransformed with pEG202N bait and pSH18-34 reporter plasmids to verify that the bait plasmid is itself transcriptionally inert. Also, by using the pJK101 reporter plasmid we confirmed that baits (LexA-integrin cytoplasmic tail fusions) could be expressed inside the nucleus of EGY48 yeast cells and bind to LexA operator.
A yeast expression library with a complexity of 10 6 was generated from oligo(dT)-primed cDNA from HeLa (human cervical carcinoma cell line) mRNA. The cDNA was cloned unidirectionally into the EcoRI/XhoI sites of galactose-inducible, TRPϩ yeast expression vector pJG4-5 (constructed by J. Gyuris, Massachusetts General Hospital, Boston). For genetic screening, yeast strain EGY48 was transformed sequentially with pEG202-␤1, pSH18-34, and pJG4-5 using the lithium acetate method (51). Approximately 2 ϫ 10 6 yeast transformants were pooled and subjected to selections as described (49,50). Positive interaction is defined as (i) growth on leucine-deficient, galactose-conditioned medium but not on leucine-deficient, glucose-conditioned medium, and (ii) forming blue colonies on galactose X-gal plates but not on glucose X-gal plates. Plasmid DNAs from positive colonies were rescued using Escherichia coli KC8. Retransformation of EGY48 with prey plasmid DNA, pSH18-34, and pEG202N-␤1cyto plasmid DNA was done to confirm the interaction. As described above for pEG202-␤1, other bait plasmids were constructed to encode for the various integrin cytoplasmic domains listed in Table I. Also constructed were bait plasmids containing chimeric ␤ 1 /␤ 5 integrin cytoplasmic domains (bottom of Table I).
Cloning of Full-length ICAP-1 cDNA-A human HeLa S3 5Ј-stretch plus cDNA library in bacterial phage lambda gt11 vector (CLONTECH, Palo Alto, CA) was plated and transferred to Colony/Plaque Screen membrane (NEN Life Science Products) according to the manufacturer's protocol. The library was screened by in situ hybridization (52) with 32 P-labeled insert as a probe (obtained from Clone No. 4, Fig. 1). After an additional two rounds of purification, eight positive lambda phage clones were obtained, and the cDNA inserts were sequenced using a double-stranded DNA cycle sequencing kit (Life Technologies, Inc.).
For stable eukaryotic expression, we used the pECE vector containing a full-length ␤ 1 insert (54) which is designated here as pECE-␤ 1 cyto1.1. The pECE-␤ 1 cyto5.5 construct codes for wild type ␤ 1 extracellular and transmembrane domains fused to the ␤ 5 cytoplasmic domain (25). Additional ␤ 1 /␤ 5 cytoplasmic tail exchange mutants, in the pECE vector, were generated by sequential PCR (52). These contained ␤ 1 extracellular and transmembrane domains, fused to ␤ 1 , ␤ 5 , or ␤ 1/5 chimeric cytoplasmic tails listed at the bottom of Table I. CHO cells negative for dihydrofolate reductase gene (dhfrϪ) were grown in MEM ␣ϩ medium with 10% FCS, and then switched to MEM ␣Ϫ with 10% dialyzed fetal calf serum (JRH Biosciences, Lenexa, KS) after transfection. The dhfrϩ p901 vector (55) was provided by Dr. M. Rosa (Biogen Co., Cambridge, MA). For transfection, CHO cells were electroporated with a mixture of p901 dhfrϩ plasmid DNA and pECE-␤ 1 or -␤ 1 /␤ 5 mutant plasmid DNA at a ratio of 1 to 10, using a gene pulser (Bio-Rad) set at 280 V and 960 microfarads. Growth of transfected CHO cells in MEM ␣Ϫ medium and selection of positive clones by flow cytometry were carried out as described (25). The CHO-␤ 1 and -␤ 1 /␤ 5 mutants were selected to have comparable surface expression levels as measured by flow cytometry using anti-␤ 1 mAb A-1A5.
For eukaryotic transient expression, full-length ICAP-1 cDNA was ligated in frame into a modified pMT2HA vector (56) to form pMT2HA-ICAP-1, which encodes for the influenza hemagglutinin (HA) antigen epitope just upstream of ICAP-1. Calcium phosphate (52) was used to transiently transfect pMT2HA-ICAP-1 into HEK293 or COS7 cells, and cells were analyzed after 48 h. Immunofluorescence analysis revealed that ICAP was typically expressed in 20 -40% of HEK293 cells, and transfection into COS7 was at least as efficient.
In Vitro Immunoprecipitation and GST Fusion Protein Assays-The HEK293 cell line was labeled with ϳ0.15 mCi [ 32 P]orthophosphate (10 mCi/ml) in sodium phosphate-deficient DMEM supplemented with 10% dialyzed FCS and antibiotics. Labeling was typically begun at 48 h after HEK293 cell transfection and continued for 3 h unless otherwise indicated. For immunoprecipitation, cells were lysed either in Triton X-100 buffer (1% Triton X-100, 25 mM HEPES, pH 7.4, 150 mM NaCl, 5 mM MgCl 2 , 2 mM phenylmethylsulfonyl fluoride, 20 g/ml aprotinin, and 10 g/ml leupeptin) or in RIPA buffer (Triton X-100 buffer supplemented with 0.2% SDS and 1% deoxycholate) at 4°C for 1 h. After centrifugation at 12,000 rpm for 10 min, soluble material was precleared by incubation with normal rabbit serum immobilized on protein A-Sepharose 4B beads (Amersham Pharmacia Biotech) at 4°C for 1 h. Next, immune complexes were collected on protein A beads already pre-bound with rabbit anti-ICAP-1 antibodies and washed four times with cell lysis buffer. Immunoprecipitated proteins were separated on SDS-polyacrylamide gel electrophoresis, and then dried gels were exposed with Tail sequences from each integrin subunit include residues from the end of the transmembrane domain to the carboxyl terminus. Amino acid sequence derived from the ␤ 1 tail is shown in bold.
b These peptides were also incorporated into GST fusion proteins. c Besides being tested in yeast, the listed carboxyl-terminal domains, together with ␤ 1 extracellular and transmembrane domains, were tested in the context of intact integrin (Figs. 6 and 8C).
For GST fusion protein assays, HEK293 and CHO cell lysates (prepared as above) were incubated with glutathione-conjugated Sepharose beads (Amersham Pharmacia Biotech) for 1 h to remove background binding material. Lysates were then incubated overnight at 4°C with GST or GST fusion proteins pre-bound to glutathione-conjugated Sepharose beads. Beads were then washed three times with lysis buffer, and bound proteins were eluted in Laemmli sample buffer and subjected to SDS-polyacrylamide gel electrophoresis under reducing conditions. Separated proteins were electrophoretically transferred to nitrocellulose membranes (Schleicher & Schuell) at 4°C overnight. Membranes were blocked with 5% fat-free dried milk in PBS/Tween 20 buffer at 25°C for 1 h and then sequentially blotted with specific mAb and horseradish peroxidase-conjugated goat anti-mouse IgG antibody, followed by four washes (15 min each) with PBS/Tween 20 buffer after each blot. Proteins were visualized using Renaissance chemiluminescent assay (NEN Life Science Products).
Cell Migration Assay-Migration assays were performed essentially as described (25), using 96-well chambers and framed polycarbonate filters with 8-m pores (Neuroprobe, Cabin John, MD). Filters were spotted with fibronectin, vitronectin, or poly-L-lysine diluted in 0.1 M NaHCO 3 , allow to dry, rinsed with PBS, and assembled with matrixside down in the chamber. Lower wells of the chamber contained 33 l of MEM ␣ medium (for CHO cells) or DMEM (for COS7 cells) with 10% FCS, unless indicated otherwise. Cells harvested in PBS with 2 mM EDTA were labeled using BCECF-AM (2Ј,6Ј-bis(2-carboxyethyl)-5(6)carboxyfluorescein acetoxymethyl ester; Molecular Probes, Eugene, OR) for 30 min, pelleted, and resuspended at 3 ϫ 10 5 cells/ml in 1% FCS (for CHO cells) or 0.1% FCS (for COS7 cells). After no preincubation (COS7 cells) or with anti-hamster ␣ 5 ␤ 1 mAb PB1 for 30 min on ice (CHO cells), cells (suspended in 100 l) were added to upper wells of the chamber and allowed to migrate at 37°C for 4 h. After migration, cells attached to the upper side of the filter were mechanically removed by scraping, and cells on the lower side were quantitated using a Cytofluor 2300 fluorescence measurement system (Millipore Corp., Bedford, MA). Percent cell migration equals: (cell fluorescence on filter with matrix coating Ϫ control cell fluorescence on filter without matrix)/(total fluorescence of input cells) ϫ 100.

Yeast 2-Hybrid Selection and Cloning of ␤ 1 Integrin Tailbinding
Proteins-A HeLa cell library was expressed in 2 ϫ 10 6 yeast transformants, selection for interaction with the integrin ␤ 1 cytoplasmic tail protein was carried out, and 25 positive clones were obtained. Among these clones, 9 coded for protein fragments that included the carboxyl-terminal half of the receptor of activated protein kinase C (RACK1) protein. The carboxyl-terminal half of RACK1 interacted strongly with ␤ 1 , weakly with ␤ 5 , and not at all with ␤ 2 or ␤ 3 integrin tail bait proteins. However, yeast 2-hybrid analyses also revealed interactions between the RACK1 carboxyl-terminal fragment and integrin ␣ V and ␣ 4 cytoplasmic tail bait proteins (Table II). Because the ␣ V and ␣ 4 tail sequences show no obvious similarity to the ␤ 1 tail (Table I), the RACK1 interactions appeared to be nonspecific and were not pursued further. Another 7 of the initial 25 positive clones coded for a related group of polypeptides, with identical carboxyl termini but variable amino termini (Fig. 1). These results suggest that regions essential for interaction with the integrin ␤ 1 tail reside within the 162 residues present in the shortest clones (clones 4 and 5). Two of the polypeptide sequences contained divergent amino termini (clones 6 and 7), which did not appear in full-length clones (as obtained below), and thus may be cloning artifacts.
Open reading frames coding for the longest polypeptides did not include a methionine start site. Thus, to obtain a full-length sequence for the ICAP-1 protein, we used a cDNA probe corresponding to clone 4 to screen a bacteriophage lambda gt11-HeLa cell cDNA library. The resulting sequence contained a putative ATG start codon, just upstream of the sequence represented in clone 1. This methionine is present in a near consensus translation initiation sequence (57) and is located downstream of stop codons in all three frames, suggestive of an authentic start codon (Fig. 2).
The full-length ICAP-1 consists of 200 amino acids and is rich in serine (16%), with the amino-terminal 50 amino acids containing 19 serine residues. There are three possible protein kinase C phosphorylation sites at Ser-20 and Ser-46 and Ser-197 (58), and one cAMP or cGMP-dependent protein kinase phosphorylation site at Ser-10 (59). No signaling motifs or domains, such as SH2 or SH3, were found in ICAP-1. Subsequent to our isolation of ICAP-1, an identical protein was described and named "ICAP-1" (44). Also, an unpublished sequence coding for the mouse homologue of ICAP-1 has appeared in GenBank TM (accession number AJ001373).
Interaction of ICAP-1 with the Integrin ␤ 1 Cytoplasmic Tail Is Highly Specific-In the context of the yeast two-hybrid system, ICAP-1 (as a pJG4.5-ICAP-1 prey construct) interacted strongly with the ␤ 1 tail but failed to interact with the integrin ␤ 2 , ␤ 3 , or ␤ 5 tails (Table II). Also ICAP-1 did not associate with 7 different integrin ␣ chain tails ( Table II). All of the pEG202encoded bait proteins containing integrin ␣ or ␤ chain cytoplasmic domains were able to bind to LexA but by themselves were transcriptionally inert, thus they meet the criteria for bait constructs suitable for study in the two-hybrid system. In other yeast two-hybrid experiments, ICAP-1 failed to interact with additional bait proteins including phosphatidylinositol 3-kinase 85-kDa subunit, Max, v-Myc, p300 CH3 domain, CD2 cytoplasmic domain, and the LAR phosphatase cytoplasmic domain.
To determine the subregion of the ␤ 1 cytoplasmic domain that is critical for ICAP-1 association, we utilized bait plasmid pEG202 to synthesize chimeric ␤ 1 /␤ 5 cytoplasmic tail mutants (listed in Table I, bottom). In yeast, both the wild type ␤ 1 tail (cyto.11) and the cyto.51 chimera showed strong interaction with ICAP-1, whereas cyto.15 and cyto.55 did not.
Tissue Expression and Biochemical Features of ICAP-1-Northern blotting showed that ICAP-1 mRNA is present in nearly all human tissues (Fig. 3). It was highly expressed in heart, colon (mucosal lining), skeletal muscle, and small intestine, barely detectable in liver, and present at intermediate levels in all other tissues. The major ICAP-1 transcript was 1.2 kilobase pairs, with variable amounts of another form at ϳ1.8 kilobase pairs.
Anti-ICAP-1 antiserum immunoprecipitated a protein of The clone B1-7 encoding RACK1 141-317 residues was used as a prey.
b Positive interaction denotes growth of blue colonies only on X-gal indicator plates lacking leucine. c ␤-Galactosidase (␤-Gal) activity is the mean of four independent measurements.
FIG. 3. Tissue expression of ICAP-1 mRNA. Filters containing mRNA from multiple human tissues (CLONTECH) were used for Northern blotting according to manufacturer's instructions. ICAP-1 cDNA probe was prepared by EcoRI/XhoI digestion from pJG4.5 vector and labeling with [␣-32 P]dCTP using RadPrime DNA kit (Life Technologies, Inc.). After stripping of the ICAP-1 probe, filters were rehybridized with 32 P-labeled human actin cDNA. kb, kilobase pair; PBL, peripheral blood lymphocyte.
ϳ27 kDa from Triton X-100 lysate of 35 S-labeled ICAP-1-transfected HEK293 cells that was not obtained from mock-transfected cells and was not seen using preimmune rabbit serum (not shown). A protein of ϳ27 kDa was also obtained upon anti-HA Western blotting of HA-tagged ICAP-1 from Triton X-100 lysate of ICAP-1-transfected COS7 cells (data not shown). From A431 cells, HeLa cells, Jurkat cells, human endothelial cells, and human fibroblasts lysed in RIPA buffer, anti-ICAP-1 antiserum blotted endogenously expressed ϳ27and ϳ31-kDa proteins, with the latter being particularly prominent in endothelial cells and fibroblasts (not shown). Whereas the ϳ27-kDa protein was routinely observed when using mild detergent extraction (e.g. Fig. 4, lanes 3 and 4), visualization of the ϳ31-kDa protein was enhanced by use of stringent detergent lysis conditions (Fig. 4, lanes 7 and 8). Analysis of the Triton X-100 pellet revealed that the ϳ31-kDa protein was indeed retained in the Triton-insoluble fraction (compare lanes 9 and 10). In contrast, analysis of the RIPA-insoluble fraction (lane 12) indicated that the majority of both 27-and 31-kDa proteins had already been extracted (lane 11). Adhesion to fibronectin (in comparison to cell suspension) did not alter the appearance of either form of ICAP-1 protein (Fig. 4, compare  lanes 3 and 4 and 7 and 8).
In a reciprocal experiment we next analyzed binding of solubilized ␤ 1 integrin to immobilized GST-ICAP-1. First, CHO cells were transfected to stably express wild type or mutant human ␤ 1 subunits. In each case, the ␤ 1 extracellular and transmembrane domains were present, whereas the cytoplasmic tail was either unaltered or fully or partly exchanged with regions of the ␤ 5 tail (See Table I, bottom, for sequences). Upon incubation with GST-ICAP-1 fusion protein, selective binding of wild type ␤ 1 (␤ 1 cyto.11) and ␤ 1 cyto.51, but not ␤ 1 cyto.55 or ␤ 1 cyto.15, was observed (Fig. 5B), as detected by Western blotting with anti-human ␤ 1 mAb A-1A5. No ␤ 1 was found to associate with immobilized GST control protein (not shown). Wild type human ␤ 1 and various tail mutants were present in CHO cells at comparable levels as seen by cell-surface flow cytometry (Fig. 6) and also as indicated by blotting with antihuman ␤ 1 mAb A-1A5 (not shown).
Regulation of ICAP-1 Phosphorylation-Because of the high serine composition and putative protein kinase C phosphorylation sites, we tested whether ICAP-1 might be phosphorylated. First, ICAP-1-transfected HEK293 cells were incubated with [ 32 P]orthophosphate for 2 h while in suspension, and for another 1 h while spreading, prior to lysis using RIPA buffer. Then, anti-ICAP-1 antibody was used to immunoprecipitate phosphorylated proteins of ϳ27 and 31 kDa from 32 P-labeled  1-8). Alternatively, HA-ICAP-1-HEK293 cells were lysed (with Triton or RIPA) in suspension at 4°C for 30 min (lanes 9 and 11), and insoluble materials were further solubilized in Laemmli sample buffer (lanes 10 and 12). After separation by SDS-polyacrylamide gel electrophoresis, Western blotting was carried out using anti-HA mAb 12CA5. HEK293 cells (Fig. 7A, lanes 3 and 4). These proteins were not precipitated using preimmune serum or from mock-transfected cells (lanes 1, 2, and 5-8). Notably, phosphorylation was enhanced by ϳ2-fold for the 27-kDa protein, and 1.5-fold for the 31-kDa protein when ICAP-1-transfected HEK293 cells were spread on FN (lane 4) compared with poly-L-lysine (lane 2). In contrast, the level of phosphorylation of background proteins was unchanged as determined by comparison of protein band densities (FN/PLL ratios ϭ 1.0). A long exposure of Fig. 7A confirmed that none of the many phosphorylated non-ICAP-1 proteins (including Control Band 1 and Control Band 2) were altered. In a separate experiment (not shown), phosphorylation of 27-and 31-kDa ICAP-1 proteins was again increased (by 1.8and 2.0-fold), respectively, upon adhesion to fibronectin compared with PLL. Again, phosphorylation of all other (non-ICAP-1) proteins was unchanged.
A report elsewhere (44) has suggested that the more slowly migrating form of ICAP-1 may represent a phosphorylated form of the protein that may appear at elevated levels upon cell adhesion and spreading on fibronectin for 15 or 30 min (44). Thus, to supplement our results obtained upon adhesion to fibronectin for 1 h (Fig. 7A), we analyzed additional time points (Fig. 7B). At no time point from 15 to 120 min did we observe that the slowly migrating form of ICAP-1 (ϳ31 kDa) was highly phosphorylated relative to the 27-kDa protein, even though the 31-kDa protein was well represented (e.g. see Fig. 4). Indeed, under identical extraction conditions, the 31/27-kDa ratio was 0.42 in terms of total protein but only 0.13 in terms of phos-phorylated protein.
ICAP-1 May Contribute to Cell Migration-In further experiments, COS7 cells transiently transfected with ICAP-1 were found to undergo increased transwell migration, when the FCS chemoattractant was held constant at 10% and different FN levels were coated onto the underside of the filter (Fig. 8A, left  panel). Also, ICAP-COS7 cells showed preferential migration compared with Mock-COS7 cells when FN coating was held constant at 10 g/ml and different FCS chemoattractant levels were used (Fig. 8A, right panel). Although ICAP-1 caused an elevation of ␤ 1 -dependent migration on fibronectin, it did not alter ␤ 1 -independent migration on vitronectin, as seen in two separate experiments (Fig. 8B, right and left panels). Because COS cells express moderate to high amounts of ␣ V and ␤ 5 , but little ␤ 3 , we suspect that vitronectin-dependent migration is largely mediated by ␣ V ␤ 5 .
CHO transfectants stably expressing comparable surface levels of human wild type or chimeric ␤ 1 (see Fig. 6) were also tested for migration. The assay was performed in the presence of anti-hamster ␣ 5 ␤ 1 mAb PB1 to block the contribution of endogenous hamster ␣ 5 ␤ 1 (Fig. 8C). The CHO-␤ 1 .cyto11 and -␤ 1 .cyto51 transfectants showed substantially more migration than either the CHO-␤1.cyto55 or -␤1cyto1.5 transfectants. This differential migratory behavior precisely coincides with the differential abilities of these mutants to bind to ICAP-1 (e.g. as seen in Table I and Fig. 5B). In the absence of 10% FCS as a chemoattractant, none of the cells showed very much migration (not shown).

DISCUSSION
Specific Association of ICAP-1 with Integrin ␤ 1 Tail-Here we have identified and characterized ICAP-1, a 200 amino acid phosphoprotein specifically associating with the ␤ 1 integrin tail. Interaction seen in a yeast two-hybrid assay was confirmed in reciprocal experiments using ICAP-1-and ␤ 1 integrin-transfected human and hamster cell lysates. In both systems the association was highly specific. In both mammalian cell lysates and in yeast, replacement of the carboxyl-terminal 14 amino acids of ␤ 1 with the carboxyl-terminal 24 amino acids of ␤ 5 resulted in loss of ICAP-1 association. Conversely, the reciprocal exchange (␤ 5 tail with terminal 14 residues from ␤ 1 ) allowed strong association. Thus the carboxyl-terminal "SAVT-TVVNPKYEGK" sequence in ␤ 1 is required for ICAP-1 interaction. While this work was in progress, another group de-scribed ICAP-1 as a protein that associated selectively with the integrin ␤ 1 tail (44). Consistent with results shown here, residues critical for ICAP-1 association resided within the carboxyl-terminal 13 residues of the ␤ 1 tail (44).
Association of ␤ 1 Tail with RACK1?-In another report, the carboxyl-terminal portion of RACK1 was isolated by a yeast two-hybrid approach and suggested to interact specifically with integrin ␤ 1 , ␤ 2 , and ␤ 5 tails (43). We also obtained a carboxylterminal fragment of RACK1 upon yeast two-hybrid screening but did not study it further due to an apparent lack of interaction specificity. Although it is still possible that RACK1 could specifically participate in integrin functions, future studies will need to explain its ability to bind to multiple peptide sequences that are seemingly unrelated.
Distribution and Size of ICAP-1-Northern blotting showed FIG. 7. Effects of cell adhesion on ICAP-1 phosphorylation. A, ICAP-1-HEK293 and mock-HEK293 cells were incubated with [ 32 P]orthophosphate while in suspension for 2 h. Then these cells were plated on plastic surfaces that had been pre-coated with either 10 g/ml fibronectin (Life Technologies, Inc.) or 10 g/ml PLL (Sigma) and blocked with 0.1% heat-inactivated bovine serum albumin (Sigma). An additional incubation with [ 32 P]orthophosphate (in phosphate-free DMEM) was then carried out for 1 h at 37°C while cells were adhering and spreading. Cells were then washed, lysed in RIPA buffer, and immunoprecipitated as described under "Experimental Procedures," using the indicated antibodies. The indicated ratios (FN/PLL) were determined using integrated density values (AlphaImager 2000 Documentation & Analysis System, Alpha Innotech Co., San Leandro, CA) for phosphorylated protein bands obtained from cells on fibronectin and poly-L-lysine. B, ICAP-1-transfected HEK293 cells were incubated as in A, except that adhesion to fibronectin was carried out for various periods.
that the ICAP-1 protein is widely expressed in many human tissues, as previously shown for the ␤ 1 integrin subunit. Also by Western blotting, ICAP-1 was ubiquitously expressed in most cultured cell lines. It is not yet clear whether the appearance of RNA of two different sizes (1.8 and 1.2 kilobase pairs) represents alternative splicing or different polyadenylation sites as previously suggested (44). Chang et al. (44) described an apparent alternatively spliced 16-kDa form of ICAP-1 (ICAP-1␤) that lacks amino acids 128 -177 and does not interact with integrin ␤ 1 cytoplasmic domain. We did not observe such a form while screening for full-length ICAP-1, possibly because we screened a different cDNA library.
In analyses of several cell lines, and cells transfected with ICAP-1 cDNA, we detected major (ϳ27 kDa) and minor (ϳ31 kDa) ICAP-1 proteins, with the latter only being seen using stringent detergent conditions. Both the ϳ27 and ϳ31-kDa proteins incorporated 32 P label, with phosphorylation of the more rapidly migrating ϳ27-kDa form being particularly prominent. Elsewhere, it was suggested that the more slowly migrating form of ICAP-1 may be preferentially phosphorylated, because it disappeared upon incubation of lysate in the absence of phosphatase inhibitors (44). Our direct phosphorylation results contradict that conclusion. We cannot explain why phosphatase inhibitors may have facilitated the maintenance of the more slowly migrating form, except to suggest that this effect may be indirect and possibly involve other components in the cell lysate. At present, the biochemical basis for the larger size of the 31-kDa ICAP-1 protein and its relative resistance to detergent extraction (compared with the 27-kDa protein) are not clear.
Elsewhere it was also shown that appearance of the larger ICAP protein form was favored upon cell adhesion to fibronectin, whereas it was greatly diminished when cell matrix interaction was disrupted (44). We did not observe an adhesion-dependent change in levels of either the 27-or 31-kDa form of ICAP-1 (e.g. see Fig. 4). This disparity possibly could be explained by our use of a human embryonic kidney cell line (HEK293) instead of the osteosarcoma cell line (UTA-6) used in the other study.
Functional Relevance of ␤ 1 Tail Association with ICAP-1-Association of the ␤ 1 tail with ICAP-1 may be relevant for multiple reasons. First, phosphorylation of both 27-and 31-kDa forms of ICAP-1 was selectively promoted upon cell adhesion and spreading on fibronectin but not on poly-L-lysine. Thus, ICAP-1 phosphorylation appears to be regulated during the outside-in signaling that occurs upon integrin engagement with ligand. In future studies, it will be important to place ICAP-1 phosphorylation into the context of established integrindependent signaling events, such as the phosphorylation of focal adhesion kinase, paxillin, and other downstream targets (6). A previous report suggested that constitutively activated RhoA might down-regulate ICAP-1 phosphorylation (44). In contrast, we found that RhoA.V14 transfection into NIH3T3 cells caused no elevation in ICAP-1 phosphorylation (not shown). This discrepancy is perhaps easily explained, considering (as discussed above) that Chang et al. (44) appear not to have been actually measuring ICAP-1 phosphorylation.
Second, ICAP-1 interactions with the integrin ␤ 1 tail may support cell migration. In one set of experiments, expression of ICAP-1 in COS7 cells was associated with increased ␤ 1 -dependent cell migration on fibronectin but not ␤ 1 -independent migration on vitronectin. In another set of experiments, the carboxylterminal amino acids within the ␤ 1 tail that were needed for ICAP-1 association were also required for enhanced cell migration. The carboxyl terminus of ␤ 5 could not substitute for ␤ 1 to support migration. These results may help to explain previously noted differences between the integrin ␤ 1 and ␤ 5 tails in terms of supporting cell migration (25).
Other functions known to require the carboxyl-terminal 14 amino acids of ␤ 1 could potentially also involve ICAP-1. For example, the carboxyl-terminal "SAVTTVVNPKYEGK" sequence in the ␤ 1 tail includes amino acids (Thr-788, Thr-789, Asn-792, and Tyr-795 in human ␤ 1 ) that help to regulate integrin affinity for ligand (33) and amino acids (Asn-792 and FIG. 8. ICAP-1 effects on cell migration. Migration that was both chemotactic and haptotactic was carried out as described under "Experimental Procedures." A, migration of ICAP-transfected COS7 cells was determined using porous filters coated on the underside with various doses of fibronectin with 10% FCS in the lower well (left panel), or using 10 g/ml fibronectin with different doses of FCS in the lower well (right panel). B, filters were coated with either 10 g/ml fibronectin or 10 g/ml vitronectin, with 10% FCS in the lower well. Two separate experiments were carried out on different days, using cells derived from different transfections. In each experiment, a common pool of transiently transfected COS7 cells was divided for testing on the two different substrates. In each of the seven experiments on fibronectin (A and B), ICAP-COS7 migration was significantly greater than Mock-COS7 migration (two-tailed p value Ͻ 0.008). Migration on vitronectin was not significantly different. C, migration of CHO-␤ 1 and -␤ 1/5 transfectants was determined using 10 g/ml fibronectin coating and 10% FCS in the lower well, with 1% FCS in the upper well. For all migration experiments (A-C) each data point represents the mean Ϯ S.D. from six replicates.
Tyr-795) required for cytoskeletal association (32). Deletion of the carboxyl-terminal 13 amino acids results in loss of integrin co-localization with talin, ␣-actinin, and focal adhesion kinase (60). Furthermore, alternatively spliced ␤ 1 B (21) and ␤ 1 C (61) tails lack the critical carboxyl-terminal residues present in ␤ 1 A and thus should not bind to ICAP-1. This could at least partially explain why functions of ␤ 1 A are markedly different from the functions of ␤ 1 B or ␤ 1 C (21,23). In this regard, expression of ␤ 1 B in CHO cells resulted in a severe reduction of cell motility on fibronectin (62), analogous to the diminished motility of our non-ICAP-1 binding ␤1cyto.15 mutant.
Other ␤ 1 Tail-associated Proteins-At present, the integrin ␤ 1 tail has been suggested to interact directly with cytoskeletal proteins (␣-actinin (38), talin (39,40), filamin (40), and paxillin (37,41)), protein kinases (focal adhesion kinase (37), integrinlinked kinase (42)), and other proteins (RACK1 (43) and ICAP-1 (44)). Interestingly, the ICAP-1 protein shows no similarity to any of these other proteins, and as far as we are aware, the ICAP-1 interaction is the only one to map to the carboxyl-terminal 14 amino acids of the ␤ 1 tail. In addition, the ICAP-1 protein is completely distinct from proteins that may interact with other integrin ␤ tails, such as cytohesin-1 (63) and ␤ 3 -endonexin (64). Despite the growing list of integrin cytoplasmic tail-associated proteins detected by yeast two-hybrid screening (65), few if any of these associations have yet been independently established by more than one laboratory. A strength of the current report is that now the ICAP-1 interaction with the ␤ 1 tail has been independently determined, by at least two distinct laboratories, upon screening of two different cDNA libraries.
In conclusion, we have demonstrated a direct and specific interaction between the widely expressed ICAP-1 protein and the widely expressed integrin ␤ 1 A cytoplasmic tail. Furthermore, we provide evidence that the integrin-ICAP-1 interaction may be relevant to cell migration and adhesion and spreading functions carried out by ␤ 1 integrins, and we suggest that phosphorylation of ICAP-1 could play a role in these events.