Transmembrane-4 Superfamily Proteins Associate with Activated Protein Kinase C (PKC) and Link PKC to Specific β1 Integrins*

Translocation of conventional protein kinases C (PKCs) to the plasma membrane leads to their specific association with transmembrane-4 superfamily (TM4SF; tetraspanin) proteins (CD9, CD53, CD81, CD82, and CD151), as demonstrated by reciprocal co-immunoprecipitation and covalent cross-linking experiments. Although formation and maintenance of TM4SF-PKC complexes are not dependent on integrins, TM4SF proteins can act as linker molecules, recruiting PKC into proximity with specific integrins. Previous studies showed that the extracellular large loop of TM4SF proteins determines integrin associations. In contrast, specificity for PKC association probably resides within cytoplasmic tails or the first two transmembrane domains of TM4SF proteins, as seen from studies with chimeric CD9 molecules. Consistent with a TM4SF linker function, only those integrins (α3β1, α6β1, and a chimeric “X3TC5” α3 mutant) that associated strongly with tetraspanins were found in association with PKC. We propose that PKC-TM4SF-integrin structures represent a novel type of signaling complex. The simultaneous binding of TM4SF proteins to the extracellular domains of the integrin α3 subunit and to intracellular PKC helps to explain why the integrin α3 extracellular domain is needed for both intracellular PKC recruitment and PKC-dependent phosphorylation of the α3 integrin cytoplasmic tail.

Integrin-dependent cell adhesion, through integration of cell signaling pathways and cytoskeletal reorganization, markedly influences cell growth, death, and differentiation (1)(2)(3). Signaling through many different integrins causes similar calcium fluxes, pH changes, and activation of focal adhesion kinase. However, specific integrins may also differ markedly from each other in support of cell cycle progression, cell survival, or gene induction (4 -6). Consistent with signaling differences, different integrin cytoplasmic domains may interact with a number of specific integrin-associated proteins (7).
The PKC family of phospholipid-dependent serine and threonine kinases participates in a wide spectrum of biological activities (18 -20). Activation of cytosolic PKC by phorbol ester or diacylglycerol occurs in parallel with PKC translocation to cellular membranes. Membrane association is largely attributed to specific PKC interactions with phosphatidylserine. Various PKC isoforms also associate with a number of specific binding proteins (20). However, aside from PKC interaction with the transmembrane proteoglycan syndecan-4 (21), a role for specific transmembrane proteins during PKC translocation has not previously been suggested.
Here we demonstrate that upon activation and translocation, conventional PKCs associate closely with several different TM4SF/tetraspanin proteins. Upon PKC activation, those integrins (␣ 3 ␤ 1 , ␣ 6 integrins) already constitutively associated with TM4SF proteins then become linked to PKC. Within these PKC-TM4SF-integrin complexes, integrin ␣ 3 and ␣ 6 tails are phosphorylated in a PKC-dependent manner. The presence of TM4SF linker proteins helps to explain how association of intracellular PKC may be determined by integrin extracellular domains.
Immunoprecipitation and Reimmunoprecipitation-Cells were surface-labeled with Na 125 I (PerkinElmer Life Sciences) using lactoperoxidase by an established protocol, or cells were 32 P-labeled by growth in sodium phosphate deficient medium supplemented with [ 32 P]orthophosphate (PerkinElmer Life Sciences) for 3-6 h. In all experiments involving PMA stimulation, cells were treated with 100 nM PMA for 20 -30 min at 37°C prior to lysis. Cells were lysed in immunoprecipitation buffer (1% Brij 96 or 1% Brij 99, 25 mM HEPES, pH 7.4, 150 mM NaCl, 5 mM MgCl 2 , 2 mM phenylmethylsulfonyl fluoride, 20 mg/ml aprotinin, and 10 mg/ml leupeptin) for 1 h at 4°C. For 32 Plabeled cells, immunoprecipitation buffer was supplemented with phosphatase inhibitors (1 mM sodium orthovanadate, 1 mM NaF, and 10 mM ␤-glycerophosphate). Immunoprecipitations and reimmunoprecipitations were then carried out as described (10,44). Immune complexes collected on beads were then washed three times with immunoprecipitation buffer and analyzed by SDS-PAGE under nonreducing conditions, and radiolabeled proteins were visualized by autoradiography.
Western Blot Analysis-For Western blot analysis, immunoprecipitated samples were subjected to SDS-PAGE under reducing conditions and then electrophoretically transferred to nitrocellulose membrane. After blocking with 5% nonfat milk in PBS-Tween 20 buffer at room temperature for 1 h, nitrocellulose membranes were sequentially blotted at room temperature for 1 h with specific antibody and then horseradish peroxidase-conjugated goat anti-mouse IgG. Each step was followed by four 15-min washes with PBS-Tween 20 buffer. Membranes were then developed using chemiluminescence (Renaissance; PerkinElmer Life Sciences).
Immunofluorescence-Circular glass coverslips (12 mm; Fisher) were coated with fibronectin (10 g/ml) in 0.1 M NaHCO 3 at 4°C overnight. HT1080 cells were harvested in PBS with 2 mM EDTA, washed once in serum-free Dulbecco's minimal essential medium, and plated on coverslips for 1-2 h at 37°C. Some HT1080 cells were treated with 100 nM PMA for the last 20 min prior to rinsing in PBS and fixing in PBS containing 3% paraformaldehyde for 10 min. Permeabilization was with 0.05% Triton X-100 in PBS for 2 min at room temperature. Nonspecific binding sites were blocked with 20% goat serum in PBS for 1 h at room temperature. Primary mAbs (ϳ1 g/ml final concentration) were diluted in 20% goat serum/PBS and incubated with cells for 1 h at room temperature. Coverslips were washed four times with PBS and then incubated for 30 min with rhodamine-conjugated secondary antibodies. Finally, coverslips were washed four times with PBS, mounted on glass slides in FluroSave reagent (Calbiochem), and analyzed using an Axioskop fluorescent microscope (Zeiss).
Cells, Transfectants, and Mutants-HT1080 fibrosarcoma, Jurkat T leukemia, and K562 erythroleukemia cells were cultured in RPMI medium containing 10% fetal bovine serum. K562 cells transfected with mutant and wild type integrin ␣ 2 , ␣ 3 , ␣ 4 , and ␣ 6 subunits were prepared as previously described (10,45). Mutant subunits include X3C0, in which the ␣ 3 cytoplasmic tail is deleted; X3TC5, in which the transmembrane portion and cytoplasmic tail of ␣ 3 are replaced by those of ␣ 5 ; and X2C3, in which the ␣ 2 integrin cytoplasmic tail is replaced by that of ␣ 3 (46). Transfected integrins were expressed at comparable levels (i.e. varied by less than a factor of 2) on the surface of K562 cells.
Chimeric integrin TM4SF proteins were produced by the overlapping oligonucleotide polymerase chain reaction technique. In the reciprocal CD9-il.A15 and A15-il.CD9 chimeras, the intracellular loops from CD9 (QESQC) and A15 (RGSPW) were swapped. The A15-lel34.CD9 chimera was produced by replacing, in A15, the large extracellular loop and flanking TM3 and TM4 domains with the corresponding region from CD9. The sequence (with CD9 underlined) becomes . . . GSPWM/ LGLFF . . . MILCC/FITAN . . . The entire polymerase chain reaction regions of the chimeric constructs were sequenced to confirm fidelity. Chimeric cDNA was subsequently cloned into the expression plasmid pCR3.1-uni (Invitrogen, Carlsbad, CA) and stably transfected into K562 cells via electroporation at 960 microfarads and 280 V using a gene pulser. Transfectants were selected with 1 mg/ml G418 (Life Technologies, Inc.) and subcloned by limiting dilution. Positive subclones stably expressing chimeric integrin subunits were assessed, pooled, and sorted by flow cytometry, using monoclonal antibodies specific for the large extracellular loops.

PKC Forms Complexes with Specific Tetraspanin Proteins-
During our studies of tetraspanin protein association with intracellular PI 4-K (16,17), we noticed that another enzyme, PKC, showed an even stronger tetraspanin association. From a series of K562 erythroleukemia cell transfectants, the tetraspanin protein CD81 was immunoprecipitated, and then PKC␤II (a conventional PKC isoform) was detected by immunoblotting, provided that that the K562 cells had been activated with PMA (Fig. 1A, compare upper and lower panels). Immunoprecipitates of CD98 (right lanes) yielded no associated PKC. To expand these results, we next immunoprecipitated multiple tetraspanin proteins from Jurkat T cells, and immunoblotted to detect PKC␣ (another conventional PKC isoform). As indicated, the TM4SF proteins CD9, CD81, and CD82 each showed a clear association with PKC␣ (Fig. 1B, upper panel). In contrast, PKC␣ was not present in immunoprecipitates of A15 (another TM4SF protein), CD71 (transferrin receptor), CD98 (another prominent transmembrane protein), or MHC class 1. Cell surface levels of A15, CD71, CD98, and MHC class 1 were each greater than that of CD9, and were comparable with CD81 and CD82 (not shown). Association of PKC␣ with TM4SF proteins was not observed unless Jurkat cells were pretreated with PMA (Fig. 1B, upper panel). PMA had no effect on the levels of TM4SF proteins (CD9, CD81, CD82, and A15) or control proteins (CD71, CD98, and MHC class I) directly immunoprecipitated using appropriate mAbs (not shown). The TM4SF protein CD151 also co-precipitated with PKC (see below). Compared with total cell lysate samples in Fig. 1, A and B, comparable levels of PKC were detected in CD9 and CD81 immunoprecipitates derived from 20 -30-fold more cell equivalents. Thus, ϳ3-5% of the total PKC may be associated with each of these TM4SF proteins.
To confirm results in Fig. 1B, PKC␣ was immunoprecipitated from a lysate of Jurkat cells that had been PMA-treated and 125 I-labeled. Reimmunoprecipitations were then carried out, showing that CD9, CD53, CD81, and CD82 were associated with PKC␣ (Fig. 1A, lanes a-d). In control experiments, a nontetraspanin protein (MHC-1) was not recovered from PKC␣ immunoprecipitates (lane e) although it could be directly immunoprecipitated (lane k). Furthermore, no tetraspanins or other proteins were recovered from immunoprecipitations of PI 3-K (lanes f-j) or PKA (not shown).
Another means of inducing PKC activation is through second antibody cross-linking of the CD3-T cell receptor complex on the surface of T cells (47). Antibody/second antibody-induced cross-linking of the CD3-T cell receptor on 125 I-surface-labeled Jurkat T cells triggered association of PKC␣ with CD81 and a few other unknown cell surface proteins (Fig. 2, lane b). In contrast, antibody/second antibody cross-linking of MHC class I, CD28, CD81, and CD98 (all abundant on Jurkat T cells) failed to stimulate PKC␣ association with CD81 protein com-plexes (lanes a and c-e). In another control experiment, CD81 and associated proteins were not associated with PKA following CD3 antibody cross-linking (lane f). The identity of CD81 was verified in Fig. 2B. After antibody cross-linking of CD3 on Jurkat cells, a PKC␣ immunoprecipitation was carried out, and from the resulting protein complex, CD81 could be reimmunoprecipitated (Fig. 2B, lane i). In contrast, ␣ 5 ␤ 1 integrin, CD98, and MHC-1 proteins could not be reimmunoprecipitated (lanes h, j, and k). Likewise, CD81 could not be reimmunoprecipitated from a PI 3-K immunoprecipitate after CD3 antibody crosslinking of Jurkat cells (lane l). A direct immunoprecipitation of CD81 and associated proteins is shown in lane g. The pattern of surface-labeled proteins looks remarkably similar to that immunoprecipitated with an anti-PKC␣ antibody (lane b), thus providing further evidence for the presence of PKC␣ in CD81 complexes.
From human peripheral blood mononuclear cells stimulated with either PMA or anti-CD3 antibody cross-linking, PKC␣ again formed complexes with TM4SF proteins (not shown). Another agent that stimulates PKC translocation (bryostatin 1 (48)) also induced association of PKC with TM4SF proteins (not shown). In other experiments, PKC inhibitors (chelerythrine D, Go6976, calphostin C, and staurosporine) did not inhibit PMAinduced association of PKC with TM4SF proteins (data not shown). Thus, although PKC translocation is required for TM4SF protein association, PKC activity is not required.
TM4SF-PKC Interactions Are Stabilized by Covalent Crosslinking-To characterize TM4SF-PKC interactions further, intact K562 cells were treated with dithiobis(succinimidyl propionate) (DSP), a homobifunctional cross-linking agent with a span of 12 Å. Without cross-linker, and under relatively mild detergent conditions, immunoprecipitation of CD81 from PMAstimulated K562 cells yielded associated PKC␤II that was readily visualized by Western blotting (Fig. 3A, lane a), in FIG. 1. PKC association with TM4SF proteins. A, K562 transfectants treated with or without PMA were lysed in 1% Brij 96, and then immunoprecipitates of CD81 (a representative TM4SF protein) and CD98 mAb (negative control) were prepared. After SDS-PAGE and electrophoretic transfer, PKC␤II was detected by immunoblotting. For integrin expression in the various K562 transfectants, see Fig. 7A. B, Jurkat cells treated with or without PMA were lysed in 1% Brij 99 and then immunoprecipitated (IP) with mAb to TM4 proteins (CD9, CD81, CD82, and A15) or other cell surface molecules (CD71, CD98, and MHC-1). After SDS-PAGE and electrophoretic transfer, whole cell lysate and immunoprecipitated proteins were blotted using anti-PKC␣ mAb. C, PMA-treated, 125 I-labeled Jurkat cells were lysed in 1% Brij 99. PKC␣ (lanes a-e) and PI 3-K (lanes f-j) complexes were immunoprecipitated, dissociated, and then reprecipitated with antibodies to TM4SF proteins (CD9, CD53, CD81, and CD82) or to MHC class I as indicated. Also, MHC-1 was directly immunoprecipitated (lane k). Note that mAb to PKC and PI 3-K yielded comparable amounts of appropriate target proteins from metabolically labeled lysate (not shown).

FIG. 2. Stimulation of CD3 promotes PKC-TM4SF protein complex formation.
A, 125 I-labeled Jurkat cells were incubated for 1 h with mAb to the indicated cell surface proteins, and then a secondary rabbit anti-mouse antibody was added to cross-link the primary mAb. Following cell lysis (in 1% Brij 99), PKCa (lanes a-e) or PKA (lane f) was immunoprecipitated (I.P.), and proteins were resolved by SDS-PAGE and then visualized by autoradiography. B, 125 I-labeled Jurkat cells were incubated for 1 h with mAb to CD3, and then second antibody cross-linking was carried out as in A. Following cell lysis (in 1% Brij 99), PKCa (lanes h-k) or PI 3-K (lane l) was immunoprecipitated. Complexes were then dissociated, and reprecipitations were carried out as indicated. In a control experiment, CD81 was directly immunoprecipitated (lane g) from cells untreated with antibodies. agreement with results in Fig. 1A. Without cross-linker, association with PKC␤II was lost when 0.2% SDS was included in the Brij 96 lysis buffer (Fig. 3A, lane c). However, after treatment of intact cells with DSP cross-linker, association with PKC␤II persisted despite the stringent detergent conditions (lane e). The PKC␤II protein was not present in control Ig immunoprecipitations under any conditions (lanes b, d, and f). In a second experiment (Fig. 3B), another TM4SF protein (CD151) maintained association with PKC␤II under relatively harsh detergent conditions (1% Brij 96 plus 0.2% SDS) only when cells were treated with both PMA (to activate PKC) and DSP cross-linker. Although CD71 (transferrin receptor) was present on the cell surface at a high level, it did not cross-link to PKC␤II regardless of PMA or DSP treatment.
PKC Specificity-Conventional PKCs, including PKC␣ and PKC␤II, undergo calpain-dependent proteolytic conversion to PKM, a form lacking the N-terminal C1 and C2 regulatory domains (49). The addition of calcium and the omission of the protease inhibitor leupeptin caused the partial conversion of intact PKC␤II in K562 cells (Fig. 4A, lane a) into two closely migrating fragments of ϳ50 kDa characteristic of PKM (lane b). However, CD81 immunoprecipitation from PMA-treated K562 cells yielded only the intact PKC␤II fragment and no PKM (lane g). No PKM was observed even upon overexposure of lane g, such that the PKC intensity was comparable with that in lanes a-d (not shown). Also, no PKC was co-precipitated with CD81 in the absence of PMA treatment (lane e) or with control Ig (lanes f and h). In conclusion, the PKM form of PKC␤II, lacking N-terminal regulatory domains, does not associate with CD81. In addition, we failed to recover PKC⑀, PKC, or PKC from TM4SF immunoprecipitates, despite testing multiple cell lines and 4 -6 different TM4SF proteins for each isozyme (Fig. 4B).
TM4SF Specificity-As seen in Fig. 1, PKC␣ interacted with four different TM4SF proteins (CD9, CD53, CD81, and CD82) but not another TM4SF protein (A15/Talla1). Showing a similar specificity, PKC␤II in K562 cells interacted again with CD9 but not A15 (Fig. 5). With A15 being negative for PKC association, an opportunity for a chimeric mapping approach was provided. So far, nearly all TM4SF associations and functions have mapped to the large extracellular loop (42, 50 -52). However, the A15-lel34.CD9 chimera (A15 with the large extracellular loop and flanking transmembrane domains of CD9) failed to interact with PKC␤11 in K562 cells (Fig. 5). Thus, the CD9 large extracellular loop and flanking transmembrane domains are not sufficient for PKC association. Because the short inner loop of A15 (RGSPW sequence) is quite distinct from most other known tetraspanins (53), we hypothesized that it could selectively prevent PKC association. However, replacement of the short inner loop QESQC of CD9 with the loop RGSPW from A15 (CD9-il.A15) did not abolish PKC interaction. Conversely, replacement of the A15 RGSPW with the QESQC from CD9 (A15-il.CD9) did not confer PKC interaction. For additional mutants, in which N-terminal or C-terminal domains of CD9 were replaced by corresponding A15 domains, results regarding PKC association were inconclusive (not shown). At present, we conclude that TM4SF specificity is determined by regions   FIG. 3. Biochemical cross-linking of TM4SF-PKC complexes. A, intact K562 cells were treated with PMA (100 nM at 37°C, 20 min) and then (4°C for 60 min) with or without 1 mM DSP (a membrane-permeable cross-linker; Pierce). Cells were next lysed in either 1% Brij 96 or 1% Brij 96 plus 0.2% SDS (as indicated), and then antibody to CD81 or control Ig (cIg) was used for immunoprecipitation (I.P.). Samples were then reduced (boiled in 5% ␤-mercaptoethanol), fractionated by SDS-PAGE, and immunoblotted with anti-PKC␤II antibody. B, intact K562 cells were treated with or without PMA and with or without DSP as indicated. Cells were lysed in 1% Brij 96 plus 0.2% SDS, and then CD71 (transferrin receptor) and CD151 (TM4SF protein) were immunoprecipitated, and PKC␤II was immunoblotted. other than the short inner loop, the large outer loop, or transmembrane 3 or 4.
PKC Isoforms Are Induced to Form Complexes with Specific Integrins-Because TM4SF proteins associate with specific integrins (see Introduction) and with activated PKC, we predicted that TM4SF proteins may link activated PKC to integrins. In this regard, antibodies to either CD81 (from HT1080 cells) or PKC␣ (from activated HT1080 cells) co-immunoprecipitated similar surface-labeled proteins (Fig. 6A, lanes b and c) that precisely comigrate with ␣ 3 ␤ 1 integrin (lane a). Upon longer exposure of lanes b and c (see lanes k and l), a small amount of surface-labeled CD81 was also obvious in both CD81 and PKC␣ immunoprecipitations. Notably, integrin-like proteins were not obtained using anti-PKA, anti-PLC␥, or anti-PI 3-K mAbs (lanes d-f) and were not seen if HT1080 cells were not stimulated with PMA (lanes g-j). Upon long exposure (lane m), immunoprecipitation of another conventional PKC (PKC␥) also yielded integrin-like protein, whereas control PLC␥ (lane n) did not. As seen previously (12), PMA pretreatment was not required for CD81-integrin association (lanes b and k). Anti-PKC, anti-PKA, anti-PI 3-K, and anti-PLC␥ mAbs immunoprecipitated comparable amounts of appropriate target proteins from 35 S-labeled cells (not shown). Considering that the immunoprecipitations are from 125 I-labeled HT1080 cell total lysate, the patterns of proteins associated with CD81 and activated PKC are remarkably simple. Thus, few other cell surface proteins may be present in PKC␣-CD81-␣ 3 ␤ 1 complexes under these conditions. To identify specific ␤ 1 integrins co-immunoprecipitated with PKC␣, complexes derived from 125 I-labeled HT1080 cells were dissociated and subjected to reimmunoprecipitation (Fig. 6B). As indicated, anti-integrin antibodies yielded integrin ␤ 1 , ␣ 3 , and ␣ 6 subunits (lanes f, h, and j) but not ␣ 2 or ␣ 5 subunits (lanes g and i). Each of these subunits was expressed in HT1080 cells at moderate to high levels as seen by direct immunoprecipitation (lanes a-e) and by flow cytometry (not shown). Complexes obtained using anti-PI 3-K failed to yield reprecipitated integrin subunits (lanes k-o).
To confirm co-immunoprecipitation of specific integrins with PKCs, we analyzed another conventional isoform (PKC␤II) and another cell line (K562). From cell surface 125 I-labeled K562-␣ 3 and K562-␣ 6 Brij 96 lysates, anti-PKC␤II antibody co-immunoprecipitated abundant integrins (Fig. 7A, lanes g and j). However, little if any integrin was co-precipitated from K562-␣ 2 , K562-␣ 4 , or K562 mock transfectants containing substantial endogenous ␣ 5 (lanes f, h, and i). Labeled proteins of ϳ120 kDa in these latter lanes do not resemble heterodimeric integrins and appear to be background proteins. All integrins tested were well expressed in their respective K562 transfectants (lanes a-e). No integrins were detected from normal rabbit IgG control immunoprecipitations (lanes k-o) or from PKC␤II immunoprecipitations from cells not treated with PMA (not shown).
represented in the respective K562 transfectants (Fig. 7A,  lanes a-e). Furthermore, in the absence of PMA treatment of K562 cells, no PKC␤II was detected from integrin immunoprecipitations, although it was present in the whole cell lysate (Fig. 7B, upper panel). While PMA greatly stimulated PKC␣ and PKC␤II association with integrins, PMA had no effect on the solubilization and direct immunoprecipitation of any of the integrins tested. Besides PMA, bryostatin 1 also induced PKCintegrin association, consistent with its ability to induce PKC-TM4SF association (as mentioned above). As seen for TM4SF association (see above), PKC inhibitors did not prevent integrin association (not shown). Thus again, PKC translocation was required for integrin association, but PKC catalytic activity was not needed.
As seen by immunofluorescent staining of HT1080 cells spread on fibronectin (Fig. 8), localization of integrin ␣ 3 and TM4SF protein CD81 to lamellipodia was substantially increased following PMA treatment. Additionally, PKC␣ was translocated from the cytoplasm to lamellipodia, whereas PI 3-K distribution was relatively unaffected upon PMA treatment. These results are again consistent with PKC-CD81-␣ 3 integrin complex formation.
Specificity for Intracellular PKC Determined by Integrin Extracellular Domain-Specificity for TM4SF-integrin association resides within the ectodomains of both TM4SF proteins and integrins (42). Thus, if PKC is linked to integrins via a TM4SF linker protein, an extracellular integrin site should be needed for recruitment of intracellular PKC. In this regard, deletion of the cytoplasmic tail of ␣ 3 (K562-X3C0 transfectant, Fig. 9A, lane c) or exchange of the ␣ 3 transmembrane and tail regions with those from ␣ 5 (K562-X3TC5 transfectant, lane d) did not diminish association with PKC␤II, relative to that seen for wild type ␣ 3 integrin (K562-␣ 3 , lane b). Conversely, an ␣ 2 integrin bearing an ␣ 3 tail (K562-X2C3) failed to associate with PKC␤II (lane e). Also, anti-PKC␤II antibody did not co-precipitate integrin from cells bearing predominately ␣ 5 ␤ 1 (mock-transfected K562 cells, lane f), and none of the K562 transfectants yielded integrins with normal rabbit IgG control antibody (lanes  g-h). Thus, the integrin ␣ 3 chain extracellular domain determines specificity for interaction of an intracellular protein (PKC) with integrin. These results are consistent with PKCintegrin association requiring a transmembrane linker protein such as a tetraspanin. Similar to results seen in Fig. 6A, anti-PKC␤II antibody co-immunoprecipitation of ␣ 3 integrin was remarkably devoid of other surface-labeled proteins (Fig. 9A,  lanes b-d).
Potential Relevance of PKC Recruitment to ␣ 3 and ␣ 6 Integrins-Cytoplasmic domains from integrin ␣ 3A and ␣ 6A but not ␣ 2 , ␣ 4 , or ␣ 5 are phosphorylated by a mechanism that involves activated PKC (29,31,54). Furthermore, ␣ 3A phosphorylation may regulate integrin-dependent cell motility, signaling, and cytoskeletal organization (31). If TM4SF proteins are required to link PKC to integrin, then an integrin unable to associate with TM4SF proteins should not be phosphorylated even if the correct ␣ tail is present. Consistent with this prediction, the ␣ 3 tail present within an X2C3 integrin chimera was not phosphorylated in PMA-treated K562-X2C3 cells (Fig. 9B, lane i), although the ␣ 3 tail was phosphorylated in K562-␣ 3 cells (lane k).
FIG. 8. PMA-induced co-distribution of PKC␣, ␣ 3 ␤ 1 integrin, and CD81. HT1080 cells plated on fibronectin were untreated (left panels) or treated (right panels) with PMA and then stained with primary antibodies against either integrin ␣ 3 (A3-X8), CD81 (M38), PKC␣ or PI 3-K, followed by rhodamine-conjugated secondary antibody as described under "Experimental Procedures." Also, no phosphorylation was observed if the ␣ 3 tail was deleted (lane m) or if the tail and transmembrane regions were replaced with those of ␣ 5 (lane o). In addition, no phosphorylation was observed in CD98 control immunoprecipitations or if PMA treatment was omitted (lanes a-h).

DISCUSSION
TM4SF-PKC Association-Activation and translocation of PKC promoted a relatively robust association of conventional PKC isoforms with multiple TM4SF proteins. TM4SF-PKC complexes were seen in multiple cell lines (including both adherent and nonadherent cells), were seen for several different tetraspanins (CD9, CD53, CD81, CD82, and CD151), and were promoted by multiple PKC activating stimuli (phorbol ester, bryostatin 1, or CD3-T cell receptor triggering). Analysis of TM4SF-PKC complexes in Jurkat T cells suggests a reasonable stoichiometry. Although each TM4SF protein might only associate with ϳ3-5% of the total PKC, the presence of about five or more different TM4SF proteins in a given cell type would engage a substantially greater fraction of the total PKC. Four kinds of biochemical evidence support the presence of TM4SF-PKC complexes. (i) Immunoprecipitation of TM4SF proteins yielded PKC; (ii) immunoprecipitation of PKC yielded multiple TM4SF proteins; (iii) immunoprecipitation of either TM4SF protein (i.e. CD81) or PKC yielded essentially identical protein complexes (e.g. Fig. 2); and (iv) TM4SF-PKC complexes were stabilized by covalent cross-linking. The covalent cross-linking results provide perhaps the most compelling evidence for TM4SF-PKC association. The span of the cross-linking agent (12 Å) is such that only highly proximal interactions would be captured. Furthermore, the membrane-permeable cross-linking agent was added to intact cells and thus captured the native complexes before exposure to any detergents. Finally, the cross-linked complexes were solubilized using relatively stringent detergent conditions, such that uncross-linked proteins are largely removed.
The occurrence of TM4SF-PKC complexes was specific, with respect to both TM4SF proteins and PKC. Considering the tendency of TM4SF proteins to associate with each other, it was reassuring to find that at least one prominently expressed TM4SF protein (A15/Talla1) did not associate with PKC in Jurkat cells. Among PKC isozymes, conventional PKCs ␣, ␤II, and ␥ associated with TM4SF proteins, whereas representative other PKC types (⑀, , and ) did not. Likewise, PKM (PKC␤II lacking regulatory domains (49)) failed to associate. Thus, unique features within the regulatory regions (e.g. C1 or C2 domains) of conventional PKCs are critical for TM4SF association. The C1 and C2 regulatory domains interact with membrane diacylglycerol and phosphatidylserine, respectively, and play key roles in PKC activation and translocation (19). We now suggest that C1 and/or C2 domains of PKC also could directly interact with TM4SF proteins, thus facilitating membrane targeting of conventional PKC isoforms. Alternatively, the C1 and/or C2 domains could act indirectly. They may only be required insofar as they bring activated PKC into proximity with the membrane, while other PKC domains then associate with TM4SF proteins. While the C2 domain of PKC looks promising, it remains to be demonstrated which particular PKC regulatory domains are especially important for enabling and/or mediating TM4SF protein interactions.
Thus far, nearly all tetraspanin sites tested have mapped to the large extracellular loop (42, 50 -52). The crystal structure of CD81 indicates that the large extracellular loop is also involved in protein multimerization (55). Thus, if PKC were being recruited indirectly, by means of TM4SF protein interactions with each other or with other surface proteins, then the large extracellular loop might be required. However, our CD9/A15 chimeras allowed us to map the PKC interaction site away from the large extracellular loop. This provides further evidence in support of a more direct interaction of TM4SF proteins with intracellular PKC. The intracellular large loop and transmembrane domains 3 and 4 were also ruled out, thus leaving the cytoplasmic tails and transmembrane domains 1 and 2 as remaining candidate sites for determining PKC specificity.
Compared with tetraspanin interactions with other intracellular signaling proteins (PI 4-kinase, phosphatase, and GTP binding proteins) the PKC interactions described here are quite distinct and perhaps more robust. Those other interactions have not been demonstrated using covalent cross-linking. Furthermore, tetraspanin association with PI 4-K was observed, not in Brij 96, but in the less stringent Brij 99 conditions. Also, whereas tetraspanin-PKC associations are induced, association with PI 4-K is constitutive (8,16). Finally, PI 4-K associates well with TM4SF protein A15 but not with CD82 and CD53 (17), whereas PKC associates well with CD82 and CD53, but not with A15. Thus, TM4SF proteins may have distinct sites for recruitment of these two key signaling enzymes (PI 4-K and PKC). The CD53 and CD63 tetraspanins may associate with a low level of an unidentified phosphatase activity (56).
Again, this appears to be distinct from PKC association, since CD63, compared with other tetraspanins, showed relatively little PKC association. Elsewhere, association of CD9 with unidentified GTP-binding proteins was demonstrated (57), but  -k). Co-precipitated proteins (lanes b-d) align with control ␣ 3 ␤ 1 (directly immunoprecipitated, lane a). As indicated by flow cytometry, the X3C0, X3TC5, X2C3, and ␣ 3 proteins were all expressed at comparable levels (not shown). B, K562 transfectants, with or without PMA stimulation, were labeled with 32 P, lysed in 1% Triton X-100, and then immunoprecipitated using the relevant mAb to integrin ␣ 2 or ␣ 3 or control mAb to CD98. due to the rather nonstringent conditions utilized, this association is probably part of a large complex.
Tetraspanins Link PKC to Integrins (PKC-TM4SF-Integrin Model)-Reciprocal co-immunoprecipitation experiments showed that specific integrins (␣ 3 ␤ 1 , ␣ 6 ␤ 1 ) form complexes with conventional PKCs. Our evidence suggests that tetraspanin proteins provide a key linker function between PKC and integrins. First, we found PKC, tetraspanins, and integrins all within the same complexes. For example, in many experiments (e.g. Fig. 6A and not shown), antibodies to PKC and tetraspanins both yielded the same pattern of co-immunoprecipitated integrins. Also, all experiments showing PKC-integrin complexes were carried out under conditions (1% Brij 99, 1% Brij 96) in which TM4SF-integrin complexes are maintained (12,44). Second, immunofluorescence staining revealed a similar localization pattern for ␣ 3 ␤ 1 integrin, CD81, and activated PKC␣ at the periphery of spread cells. Third, only tetraspanin proteins have been shown (by covalent cross-linking) to have high proximity to both PKC (as shown here) and relevant integrins (42). Fourth, linkage through tetraspanin proteins explains how an extracellular integrin site could determine specificity for an associated intracellular enzyme such as PKC. Importantly, it is the extracellular domains of both integrin ␣ chains and TM4SF proteins that determine specificity for TM4SF-integrin association (8,10,44), whereas it is intracellular and/or transmembrane domains of tetraspanin proteins that most likely determine tetraspanin-PKC association. In this regard, the TM4SF protein CD151 may similarly link the integrin ␣ 3 ␤ 1 extracellular domain to another intracellular enzyme, PI 4-K (8). Fifth, those integrins (e.g. ␣ 3 ␤ 1 and ␣ 6 ␤ 1 ) that associate strongly with tetraspanins were seen in association with PKC, whereas integrins not well associated with tetraspanins (␣ 2 ␤ 1 and ␣ 5 ␤ 1 ) did not associate with PKC. Perhaps most importantly, experiments with chimeric integrins showed that ␣ chain mutants lacking capability for tetraspanin association also lost PKC association, whereas ␣ chain tail and transmembrane mutants that retain tetraspanin association also retained PKC association. Together, these results strongly support a PKC-TM4SF-integrin arrangement.
An alternative model would involve integrins providing a link between PKC and tetraspanins (PKC-integrin-TM4SF model). However, this model does not explain how integrin association with PKC, an intracellular enzyme, would be specified by the extracellular domain of the integrin ␣ chain. Likewise, the model does not account for the formation of PKC-TM4SF complexes in the absence of associated integrins (e.g. as seen in K562 cells). Finally, we have not yet observed PKCintegrin complexes in the absence of TM4SF proteins. In another report describing PKC␣-␤ 1 integrin complexes (28), the potential presence of TM4SF proteins was not addressed. At present, we prefer a model (PKC-TM4SF-integrin) in which, upon activation and translocation, PKC is brought into direct association with either TM4SF proteins or preexisting TM4SFintegrin complexes in cellular membranes.
Do Other Proteins Contribute to the Linking of PKC to Integrins?-Because integrins, PKC␣, and tetraspanin proteins have all been found in organized lipid microdomains (58 -60), we considered that our PKC-TM4SF-integrin complexes may occur in the context of large, incompletely solubilized membrane aggregates. However, in direct co-immunoprecipitation experiments (such as shown in Figs. 2, 5A, and 8A), antibodies to PKC and/or tetraspanins yielded (especially in HT1080 and K562 cells) a remarkably clean pattern of 125 I-labeled surfacelabeled proteins. In some experiments, we observed no prominent surface-labeled proteins, aside from surface-labeled integrin, and a low amount of labeled TM4SF protein.
In 1% Brij 96 lysate conditions, the majority of tetraspanin complexes appeared in the dense fractions of sucrose gradients, indicating that associations with other proteins (including ␣ 3 integrin) typically do not depend on a low density lipid microdomain (60). Furthermore, in 1% Brij 96, the majority of CD81 and CD9 complexes were included well within Sepharose 6B gel filtration columns, again indicating that they are well solubilized and of a reasonable size (Ͻ2 ϫ 10 6 Da). Even in the less stringent (i.e. less hydrophobic) Brij 99 detergent, tetraspanin-integrin complexes appeared to be well solubilized and of reasonable size (60). These prior results, coupled with our co-immunoprecipitation and biochemical cross-linking results, suggest that few other proteins, besides tetraspanins, may be needed to facilitate PKC-integrin complex formation.
PKC may interact with a number of substrate proteins (20) and other intracellular proteins termed receptors for activated C kinase (RACKs) (61). However, none of these are transmembrane proteins. A protein associating with PKC␤II, termed RACK1, may bind to the integrin ␤ 1 cytoplasmic domain. Furthermore, association of integrins with intact RACK1 was promoted upon stimulation with PMA (62). However, both the RACK1 integrin specificity (␤ 1 and ␤ 2 cytoplasmic domains) and the PKC specificity (restricted to PKC␤) are distinct from the specificities seen here for integrin-TM4SF-PKC complexes. PKC␣ may also interact with syndecan-4, a transmembrane proteoglycan regulating localization of PKC to focal adhesions (21). However, syndecan complexes may be distinct, since TM4SF proteins are not usually in focal adhesions (63).
Functional Role of PKC-TM4SF-integrin Complexes-The ␣ 3 and ␣ 6 integrins undergo serine phosphorylation, dependent on both activated PKC and an unidentified serine kinase (31). Mutation of the ␣ 3 phosphorylation site caused alterations in cell morphology, ␣ 3 integrin distribution and signaling, actin distribution, and cell migration (31). Our evidence suggests that TM4SF proteins may play a role during PKC-dependent integrin phosphorylation. First, as indicated above, TM4SF proteins are closely associated with both integrins and activated PKC. Second, only those integrins (␣ 3 and ␣ 6 integrins) able to associate with TM4SF proteins become phosphorylated. In the most informative example, the chimeric X2C3 integrin failed to associate with TM4SF proteins and was not phosphorylated, although it contained the ␣ 3 cytoplasmic tail phosphorylation site and was well expressed at the cell surface. Third, the same agents that promoted integrin ␣ 3 or ␣ 6 phosphorylation (PMA, bryostatin 1) also promoted activated PKC-TM4SF complex formation.
The formation of integrin-TM4SF-PKC complexes may allow PKC localization to be closely coordinated with cell adhesion involving particular integrins. Thereby PKC may become optimally positioned to regulate a host of downstream events involving cytoskeletal organization and signaling. For example, PKC regulates the interaction of cellular membranes with several cytoskeletal proteins (including the myristoylated alaninerich C kinase substrate protein) that are also PKC substrates (20,64). Notably, the myristoylated alanine-rich C kinase substrate protein also colocalizes with TM4SF proteins at the periphery of spread cells (63). Integrin-TM4SF-PKC complexes may be particularly important during cell migration, since TM4SF proteins (13)(14)(15), TM4SF-integrin complexes (8,65), PKC (66), and PKC-integrin complexes (28) have each been linked to cell migration.
The role of PKC in the context of integrin-TM4SF-PKC complexes is distinct from previously described PKC-dependent modulation of integrin adhesion (23, 24), cell spreading (26), and focal adhesion formation (67). For example, some integrins (e.g. ␣ L ␤ 2 , ␣ 5 ␤ 1 , and ␣ 2 ␤ 1 ) that show PKC-dependent adhesion and/or spreading functions are not those typically found in integrin-TM4SF-PKC complexes under moderately stringent detergent conditions. In addition, PKC-dependent triggering of cell adhesion, for example through CD28 on T cells (68), can occur in the absence of integrin-TM4SF-PKC complex formation (not shown). Also, in contrast to cell migration, cell adhesion is typically not regulated by TM4SF proteins (8,44).
In conclusion, studies of integrin signaling can now be expanded to include not only integrin cytoplasmic domains but also membrane-proximal ␣ chain extracellular domains. These latter domains provide specificity for the formation of integrin-TM4SF-PKC signaling complexes. Results shown here contribute to an emerging paradigm whereby association and activity of intracellular signaling enzymes can be determined through integrin extracellular domains. Also, for the first time we have demonstrated a close and possibly direct association of PKC with a class of widely expressed transmembrane proteins (TM4SF proteins) that probably play a role in PKC activation, translocation, subcellular distribution, and signaling.