A novel link between integrins, transmembrane-4 superfamily proteins (CD63 and CD81), and phosphatidylinositol 4-kinase.

Enzymatic and immunochemical assays show a phosphatidylinositol 4-kinase in novel and specific complexes with proteins (CD63 and CD81) of the transmembrane 4 superfamily (TM4SF) and an integrin (α3β1). The size (55 kDa) and other properties of the phosphatidylinositol 4-kinase (PI 4-K) (stimulated by nonionic detergent, inhibited by adenosine, inhibited by monoclonal antibody 4CG5) are consistent with PI 4-K type II. Not only was PI 4-K associated with α3β1-CD63 complexes in α3-transfected K562 cells, but also it could be co-purified from CD63 in untransfected K562 cells lacking α3β1. Thus, TM4SF proteins may link PI 4-K activity to the α3β1 integrin. The α5β1 integrin, which does not associate with TM4SF proteins, was not associated with PI 4-K. Notably, α3β1-CD63-CD81-PI 4-K complexes are located in focal complexes at the cell periphery rather than in focal adhesions. The novel linkage between integrins, transmembrane 4 proteins, and phosphoinositide signaling at the cell periphery may play a key role in cell motility and provides a signaling pathway distinct from conventional integrin signaling through focal adhesion kinase.


RESULTS AND DISCUSSION
Immunofluorescent Staining of Integrins and TM4SF Proteins-To gain initial clues regarding the functional importance of integrin-TM4SF complexes, we analyzed their cellular distribution by immunofluorescence. Previous studies showed that standard fixation and permeabilization procedures removed substantial amounts of TM4SF proteins from the cell surface, thereby precluding detailed analysis of the complex distribution (11). To overcome this problem, we pretreated cells with chemical cross-linker prior to fixation and permeabilization. HT1080 cells plated on laminin in serum-free medium could assemble structures strongly resembling classical integrin focal adhesions (27,28) as indicated by staining with anti-integrin ␣ 6 ( Fig. 1a) or anti-vinculin (Fig. 1b) mAbs. In addition, integrin ␣ 6 ( Fig. 1a) and vinculin (Fig. 1b) were dis-tributed in complexes throughout the cell periphery in a pattern strongly resembling plasma membrane "focal complexes" that were recently described (29). In comparison, two TM4SF proteins (CD63 and CD81) that associate with the ␣ 6 integrin (10, 11) were detected in focal complexes but excluded from focal adhesions (Fig. 1, c and d). The ␣ 3 ␤ 1 integrin (Fig. 1e) also showed peripheral focal complex-type staining but no focal adhesion staining, whereas another prominent membrane protein, emmprin (24), showed a uniform punctate distribution and was not present in either focal adhesions or focal complexes (Fig. 1f). Notably, in ␣ 3 -transfected RD cells (18) plated on laminin-1, fibronectin, or a 40-kDa fragment of fibronectin, both ␣ 3 ␤ 1 integrin and TM4SF proteins were again detected in focal complexes and excluded from the focal adhesions (data not shown). Inability of TM4SF proteins to cluster into focal adhesions even when an appropriate integrin is present (e.g. ␣ 6 ␤ 1 in Fig. 1a) suggests that function of the integrin-TM4SF complexes may be specifically relevant to focal complexes rather than focal adhesions. In subsequent experiments we focused on the ␣ 3 ␤ 1 -CD63-CD81 complex because it is far more abundant in HT1080 cells than ␣ 6 ␤ 1 -CD63-CD81.
Integrin-TM4SF Protein Association with Phosphatidylinositol 4-Kinase-Lamellipodial and filopodial focal complexes may trigger signal(s) leading to the reorganization of actin cytoskeleton and focal adhesion assembly (29,30). Because phosphoinositides may be potent effectors of actin polymerization (31-33), we investigated whether phosphoinositide kinase activity could be co-purified with the ␣ 3 ␤ 1 -TM4SF complex. Integrin and TM4SF immunoprecipitates prepared from HT1080 cells were assayed for phosphoinositide kinase activity. The reaction products co-migrated on TLC plates with standard PIP but not PIP 2 or PIP 3 ( Fig. 2A). Incorporation of 32 P into PIP was observed with ␣ 3 , ␤ 1 , CD63, and CD81 immunoprecipitates ( Fig.  2A, lanes a and c-e) but not with ␣ 5 or negative control P3 immunoprecipitates ( Fig. 2A, lanes b and f). This result is consistent with previous results showing that even when ␣ 5 ␤ 1 is abundantly expressed (e.g. on HT1080 and K562 cells), it is not associated with TM4SF proteins (10 -12). Notably, an immunoprecipitate of NAG2, another TM4SF protein associated with ␣ 3 ␤ 1 integrin, 3 did not exhibit associated phosphatidylinositol kinase activity (data not shown). This provides additional evidence for the specificity of interaction between ␣ 3 ␤ 1 -CD63-CD81 and phosphatidylinositol kinase.
Although TLC analysis separated PIP from PIP 2 and PIP 3 , it did not discriminate between different PtdIns phosphate species, e.g. PtdIns 3-phosphate and PtdIns 4-phosphate (and Pt-dIns 5-phosphate, if such a product exists). To determine the position of phosphorylation, the [ 32 P]PtdIns phosphate product generated by the CD63 immunoprecipitate was extracted from the TLC plate, deacylated, and analyzed by HPLC. This deacylated product co-migrated identically with authentic deacylated 3 H-labeled phosphatidylinositol 4-phosphate but apart from standard phosphatidylinositol 3-phosphate (Fig. 2B). Thus, there is PI 4-K activity in the ␣ 3 ␤ 1 -TM4SF complex.
To determine the type of PI 4-K associated with the ␣ 3 ␤ 1 -TM4 complexes, lipid kinase reactions on ␣ 3 and CD63 immunoprecipitates were carried out in the presence of Triton X-100 (0.3%), adenosine (200 nM) or mAb 4C5G (5 g/ml). Previous data showed that activity of PI 4-K type II can be stimulated by nonionic detergent and inhibited by adenosine and the 4C5G mAb, whereas all three reagents have little or no effect on PI 4-K type III (25,34). Adding adenosine and 4C5G mAb to the reactions decreased the activity of the enzyme by 70 -80%, whereas Triton X-100 had a stimulatory effect (15-20-fold) (data not shown), thus indicating that ␣ 3 ␤ 1 -TM4SF complex is associated with a PI 4-kinase with type II properties. This conclusion was extended by Western blotting with an anti-PI 4-K polyclonal antibody that detected a protein of 55 kDa, characteristic of PI 4-K type II. Notably, the 55-kDa protein was present (Fig. 2C) in anti-␣ 3 (lane a) and anti-CD63 (lane c) but not in anti-␣ 5 or negative control immunoprecipitates (lanes b and d). Compared with the total lysate sample (Fig. 2C, lane e), comparable levels of 55-kDa protein were detected in CD63 and ␣ 3 lanes that were derived from 20-fold more cell equivalents. Thus, approximately 5% or more of the 55-kDa PI 4-K protein may be present in a complex with ␣ 3 integrin and/or CD63.
The amount and the activity of PI 4-K co-immunoprecipitated with anti-CD63 mAbs was consistently greater than that detected with anti-integrin or anti-CD81 mAbs (Fig. 2, A and  C), suggesting that the CD63 interaction with PI 4-K may not require ␣ 3 ␤ 1 integrin. Indeed, PI 4-K could be co-purified with CD63 protein from K562 cells, which do not express appreciable levels of ␣ 3 ␤ 1 integrin (Fig. 2D, lane b). As expected for these cells, ␤ 1 integrins (predominantly ␣ 5 ␤ 1 ) lacked associated phosphoinositide kinase activity (Fig. 2D, lane c). However, when the ␣ 3 ␤ 1 heterodimer was expressed in K562 cells (after transfection of ␣ 3 subunit cDNA), PI 4-K was then co-immunoprecipitated with ␤ 1 integrins (Fig. 2D, lane g). Together these results suggest that TM4SF proteins may link ␣ 3 ␤ 1 integrin to PI 4-K.
Adhesion-dependent stimulation of phosphatidylinositol 4,5bisphosphate production is an established biological phenomenon that may be controlled by members of the Rho family of small GTPases (35,36). The present demonstration of physical association between ␣ 3 ␤ 1 integrin and PI 4-K, an intracellular enzyme that controls the first step in biosynthesis of PIP 2 , suggests another link between integrin activation and metabolism of phosphoinositides. The ␣ 3 ␤ 1 -CD63-CD81-PI 4-Klinked complex is distinct from the conventional FAK-related pathway insofar as its specificity for a particular ␤ 1 integrin (e.g. ␣ 3 ␤ 1 but not for ␣ 5 ␤ 1 ). Moreover, triggering of the ␣ 3 ␤ 1 -CD63-CD81-PI 4-K complex with anti-TM4 mAbs (to either CD63 or CD81) failed to induce tyrosine phosphorylation of 120 -130-kDa cellular proteins (Fig. 3, lanes e and f). In contrast, tyrosine phosphorylation of 120 -130-kDa cellular proteins that probably correspond to FAK and Cas (37)(38)(39)(40) was induced by all three anti-integrin mAbs (Fig. 3, lanes a, b, and  d). Thus, we hypothesize that the fraction of ␣ 3 ␤ 1 in ␣ 3 ␤ 1 -CD63-CD81-PI 4-K complexes may be distinct from that which signals through FAK or Cas.
What could be the function of the ␣ 3 ␤ 1 -CD63-CD81-PI 4-K complex in cells? The formation of an integrin-TM4SF-PI 4-K complex is not adhesion-dependent, because it is observed in K562 cells grown in suspension (e.g.. see Fig. 2D, lane g). Rather, given its prominent clustering at the periphery of spread cells, it is possible that an ␣ 3 ␤ 1 -CD63-CD81-PI 4-K adhesion complex may direct lamellipodial and filopodial protrusions during cell migration. Indeed, some properties of the complex may be well suited for this purpose. First, the ␣ 3 ␤ 1 -CD63-CD81-PI 4-K complex can be easily extracted from the cell membrane, thus suggesting that its interaction with ECM substrate is not very strong (11). These weak and transient interactions are particularly important at the leading edge of lamellipodia because they allow a cell to sample the substrate before deciding where to move. In this regard, the presence in indicated. C, integrins or CD63 were immunopurified and probed by Western blotting with a rabbit polyclonal antibody raised against recombinant PI 4-K␣, a 97-kDa protein that shares similar enzymatic characteristics with type II PI 4-K (46). D, integrins or CD63 were immunopurified from K562 or ␣ 3 -transfected K562 (18) cells, and the presence of PI 4-K activity in the immunoprecipitates was tested as in A. The lanes marked PI 3-K show PIP, PIP 2 , and PIP 3 standards generated by a PI 3-K reaction.
FIG. 2. Specific association of PI 4-K with ␣3␤1-TM4SF complexes. A, integrins or TM4SF proteins were immunopurified, and phosphoinositide kinase activity in the immunoprecipitates was assayed as described under "Materials and Methods." Arrows indicate TLC migration positions of the phosphoinositide markers, derived from a PI 3-K immunoprecipitation. B, [ 32 P]PtdIns phosphate produced by CD63-associated phosphoinositide kinase activity (as in A) was eluted from a TLC plate, deacylated, and analyzed by HPLC as described (26).  -d), anti-TM4SF (lanes e and f), or with anti-CD109 control (lane g) mAbs and then with goat anti-mouse Ig polyclonal antibody. Clustering of the membrane proteins was induced at 37°C for 15 min, and total cellular lysates were probed with anti-pTYR (A). As a control for loading, the filter was reprobed (B) with anti-BIP antibody (47). The arrow indicates the positions of the 120 -125-kDa proteins (A). the complex of PI 4-K, an enzyme implicated in vesicular transport (41,42), and CD63, a protein that has a YXXM internalization signal (43), could help to perpetuate the process of sampling through the recycling of ␣ 3 ␤ 1 within the leading edge. Second, the magnitude of the biochemical signal (synthesis of phosphoinositides) produced by activated complex could be a decisive factor in determining the degree of actin polymerization at the leading edge and further guiding lamellipodial and filopodial protrusions. In this regard, TM4SF proteins that can interact with one another (11,44,45) and with ␣ 3 ␤ 1 integrin (11) may regulate lateral clustering of the complex, thus affecting potency of the signal.