Class- and Splice Variant-specific Association of CD98 with Integrin b Cytoplasmic Domains*

CD98 is a type II transmembrane protein involved in neutral and basic amino acid transport and in cell fusion events. CD98 was implicated in the function of integrin adhesion receptors by its capacity to reverse suppression of integrin activation by isolated integrin b 1A domains. Here we report that CD98 associates with integrin b cytoplasmic domains with a unique integrin class and splice variant specificity. In particular, CD98 interacted with the ubiquitous b 1A but not the muscle- specific splice variant, b 1D , or leukocyte-specific b 7 cy- toplasmic domains. The ability of CD98 to associate with integrin cytoplasmic domains correlated with its capacity to reverse suppression of integrin activation. The association of CD98 with integrin b 1A cytoplasmic do- mains may regulate the function and localization of these membrane proteins. The development and function of multicellular animals re-quires integrin adhesion Integrin-dependent cell is regulated, in part, by ligand (“ac-tivation”) is migration extracellular morphogenesis Integrin activation is energy-dependent and is mediated by cell type specific

The development and function of multicellular animals requires integrin adhesion receptors (1). Integrin-dependent cell adhesion is regulated, in part, by ligand binding affinity ("activation") changes controlled by cellular signaling cascades (1)(2)(3). Regulation of integrin affinity is important in cell migration (4 -6), extracellular matrix assembly (7), and morphogenesis (8). Integrin activation is energy-dependent and is mediated by cell type specific signals operating through integrin cytoplasmic domains (9).
Complementation of dominant suppression (CODS) 1 is an expression cloning scheme used to identify proteins that modulate integrin affinity (10). CODS depends on the ability of an isolated integrin ␤ 1A cytoplasmic domain, in the form of a chimera with the ␣ subunit of the interleukin-2 receptor, to block integrin activation (dominant suppression). Proteins involved in integrin activation are isolated by their ability to complement dominant suppression. CD98, a type II transmem-brane protein first discovered as a T-cell activation antigen (11), was identified utilizing CODS. CD98, although widely expressed on proliferating cells, is generally down-regulated in quiescent cells (12). CD98 forms disulfide-bonded heterodimers with several light chains that strongly resemble permeases (13)(14)(15)(16)(17)(18)(19)(20). CD98 regulates the transport of neutral and positively charge amino acids through these light chains (14,15,17,18). Thus, CODS has identified an unexpected connection between cell adhesion and certain amino acid transporters.
The mechanism by which CD98 influences integrin function is not yet clear. CODS was predicated on the idea that it would identify integrin ␤ cytoplasmic domain binding proteins (10). Many ␤ cytoplasmic domains manifest overall sequence similarity (1,2); however, the cytoskeletal protein, talin, binds to the muscle-specific splice variant, ␤ 1D , more tightly than to ␤ 1A . In addition, the leukocyte-specific ␤ 7 cytoplasmic domain binds to filamin more tightly than to ␤ 1A (21). We have now examined interactions between CD98 and recombinant parallel-dimerized integrin ␤ 1A , ␤ 1D , and ␤ 7 cytoplasmic domains by affinity chromatography (21). Here we report that CD98 interacts with the ␤ 1A but not ␤ 1D or ␤ 7 integrin cytoplasmic domains. Furthermore, the CD98 interaction is insensitive to ␤ cytoplasmic domain mutations that abolish the binding of talin and filamin. The capacity of CD98 to complement dominant suppression correlates with its capacity to bind to the suppressive ␤ cytoplasmic domains. The interaction of the integrin ␤ 1A cytoplasmic domain with CD98 may thus serve to regulate the localization and the function of these membrane proteins.
DNA Constructs and Recombinant Proteins-cDNA encoding the expressed integrin cytoplasmic domains joined to 4 heptad repeats ( Fig.  1) were cloned into the modified pET-15 vector as described previously (21). Point mutations in ␤ 1D and ␤ 7 (Fig. 1) were performed utilizing the Quickchange kit (Stratagene). Recombinant expression in BL21 (DE3)pLysS cells (Novagen) and purification of the recombinant products were made in accordance with the manufacturers instructions (Novagen), with an additional final purification step on a reverse phase C18 high performance liquid chromatography column (Vydac). Polypeptide masses were confirmed by electrospray ionization mass spectrometry on an API-III quadrupole spectrometer (Sciex, Toronto, Ontario, Canada) and varied by less than 4 daltons from those predicted by the desired sequence.
Cell Lysates-Jurkat cells were washed twice in phosphate-buffered saline and surface-biotinylated using Sulfo-Biotin N-hydroxysuccinimide in phosphate-buffered saline according to the manufacturer's instructions (Pierce). They were then washed twice with Tris-buffered saline and lysed by sonication on ice in buffer A (1 mM Na 3 VO 4 , 50 mM NaF, 40 mM sodium pyrophosphate, 10 mM Pipes, 50 mM NaCl, 150 mM sucrose, pH 6.8) containing 1% Triton X-100, 0.5% sodium deoxycholate, 1 mM EDTA, and protease inhibitors (aprotinin, 5 g/ml leupeptin, and 1 mM phenylmethylsulfonyl fluoride). Platelet lysates were prepared as described previously (21). Subcellular fractionation of Jurkat cells was performed after surface biotinylation. The cells were washed three times in Hepes-saline (200 mM Hepes, 12 mM CaCl 2 ⅐2H 2 O, 16 mM MgSO 4 , pH 7.3-7.4), suspended in 20 mM Hepes, and homogenized with a Dounce homogenizer. An equal quantity of buffer B (20 mM Hepes, 0.5 M sucrose, 10 mM MgCl 2 , 0.1 M KCl, 2 mM CaCl 2 ⅐H 2 O with protease inhibitors) was added to the homogenate, and the mixture was centrifuged at 500 ϫ g at 4°C for 15 min. The supernatant was collected and centrifuged at 100 000 ϫ g for 30 min in a Beckman model L7-65 centrifuge. The cytoplasmic fraction (supernatant) was removed and the membrane fraction (pellet) washed in a 1:1 mixture of 20 mM Hepes and buffer B. The membrane fraction was resuspended in buffer A, 1 mM EDTA, and protease inhibitors and centrifuged at 30,000 ϫ g for 20 min.
Affinity Chromatography Experiments-Recombinant proteins were expressed in BL21(DE3)pLysS cells (Novagen) and bound to His-bind resin (Novagen) through their N-terminal His tag in a ratio of 1 ml of beads/liter of culture. Coated beads were washed with PN (20 mM Pipes, 50 mM NaCl, pH 6.8) and stored at 4°C in an equal volume of PN containing 0.1% NaN 3 . Beads were added to cell lysates diluted in buffer A, (0.05% Triton X-100, 3 mM MgCl 2 , and protease inhibitors) and incubated overnight at 4°C and then washed five times with buffer A. 100 l of SDS-sample buffer was added to the beads and the mixture was heated at 100°C for 5 min. After 10,000 rpm centrifugation in a microcentrifuge, the supernatant was fractionated by SDS-PAGE and analyzed by Western blotting. In some experiments, proteins were eluted off the beads with 100 l of elution buffer (1 M imidazole, 500 mM NaCl, 20 mM Tris-HCl, pH 7.9) and 1 ml of immunoprecipitation buffer (20 mM Tris-HCl, 150 mM NaCl, 10 mM benzamidine HCl, 1% Triton X-100, 0.05% Tween 20, and protease inhibitors) was then added. The eluted proteins were immunoprecipitated overnight at 4°C with an 4F2 antibody pre-bound to protein A-Sepharose beads (Amersham Pharma-cia Biotech). The following day, the beads were washed three times with the immunoprecipitation buffer and heated in reducing sample buffer for SDS-PAGE under reducing conditions. Samples were separated on 4 -20% SDS-polyacrylamide gels (Novex) and transferred to nitrocellulose membranes. Membranes were blocked with Tris-buffered saline, 5% nonfat milk powder and stained with streptavidin-peroxidase or with specific antibodies and appropriate peroxidase conjugates. Bound peroxidase was detected with an enhanced chemiluminescence kit (Amersham Pharmacia Biotech). Equal loading of Ni 2ϩ beads with recombinant proteins were verified by Coomassie Blue staining of SDS-PAGE profiles of SDS eluted proteins.
Flow Cytometry-Analytical two-color flow cytometry was performed as described previously (9). PAC1 binding was assessed in a subset of transiently transfected ␣␤py cells (cells positive for co-transfected Tac-␣5 as measured by 7G7B6 binding). Integrin activation was quantified as an activation index (AI) defined as (FϪ F o )/(F LIBS6 Ϫ F o ), in which F is the median fluorescence intensity of PAC1 binding, F o is the median fluorescence intensity of PAC1 binding in the presence of competitive inhibitor (Ro43-5054, 1 M), and F LIBS6 is the maximal median fluorescence intensity of PAC1 binding in the presence of the integrin activating antibody anti-LIBS6 (2 M). Percentage of reversal is calculated as (AI (␤x ϩ CD98) Ϫ AI ␤x )/(AI ␣5 Ϫ AI ␤x ). AI ␤x is the activation index of cells transfected with Tac-␤ x chimeras, AI (␤x ϩ CD98) is the AI of cells co-transfected with CD98 and Tac ␤ x chimeras, and AI ␣5 is the AI of cells transfected with Tac-␣5. The x of ␤ x can have values of 1A, 1D, and 7 for the Tac-␤ 1A , Tac-␤ 1D , and Tac-␤ 7 chimeras, respectively.

CD98 Binds to the ␤ 1A Integrin Cytoplasmic Domain-CD98
can block reduced integrin affinity caused by overexpression of free ␤ 1A cytoplasmic domains, suggesting a physical interaction between ␤ 1A and CD98 (10). To assess this potential interaction, we examined the binding of solubilized membrane proteins to the ␤ 1A cytoplasmic domain. For affinity matrices, we used model proteins in which the integrin cytoplasmic domain was joined to four heptad repeats (21). The repeats form parallel coiled-coil dimers so that the tails are dimerized and parallel. When a Jurkat cell lysate was exposed to such an affinity matrix, a cell surface polypeptide of 88 kDa bound to the ␤ 1A but not to the ␣ IIb tail ( Fig. 2A). This polypeptide was immunoprecipitated by the anti-CD98 antibody, 4F2 (Fig. 2B). Based on its mass and reactivity with anti-CD98 antibody, the ␤ 1A tail binding polypeptide was identified as CD98.
To assess the specificity of CD98 binding to ␤ integrin tails, affinity chromatography was performed with ␤ 1D , ␤ 3 , and ␤ 7 cytoplasmic domains. CD98 did not bind to ␤ 7 and binding to ␤ 1D was weak and variable (Fig. 3A). In contrast, talin and filamin (Fig. 3A) bound strongly to ␤ 1D and ␤ 7 tails, respectively, as reported (21). CD98 also bound to ␤ 3 , and binding was not altered by the presence of the ␣ IIb cytoplasmic domain (Fig.  3B). Thus, CD98 binding to integrin tails is integrin class-and splice-variant-specific.
Differential CD98 Binding to ␤ Integrin Tails Is Independent of Filamin and Talin Binding-CD98 binds well to the ␤ 1A integrin cytoplasmic domain but not to those of ␤ 1D or ␤ 7 . The binding assays were performed using talin-and filamin-1-containing cell extracts. Thus, these CD98 binding differences could be due to competition for CD98 binding by filamin-1 or talin, which bind preferentially to ␤ 7 or ␤ 1D , respectively (21). To test this possibility, we used filamin-1-deficient human melanoma cells (M2) and reconstituted cells (A7) (27) to examine the role of filamin-1 in CD98 binding. CD98 bound to the ␤ 1A tail, but not ␤ 7 , when lysates of M2 cells were used (Fig. 4A), showing that filamin-1 is not required for CD98 binding to ␤ 1A . CD98 binding to ␤ 7 was not observed in the filamin-1 null (M2) cells. Consequently, competition with filamin-1 does not account for the failure of ␤ 7 to bind CD98.
To examine the role of talin, we used cell membrane preparations with a greatly reduced talin content (Fig. 5A). CD98 extracted from these membranes bound ␤ 1A but not ␤ 1D cytoplasmic domains (Fig. 5B). Thus, talin does not prevent CD98 Depicted is an alignment of the integrin cytoplasmic domains used in this study. The underlined tyrosine (Y) was mutated to an alanine (A) to form the YA mutants. All integrin sequences with the exception of ␤ 7 correspond to those human sequences published in the Swiss-Protein data base as of May 15, 1999. In ␤ 7 , the amino-terminal Arg was changed to Lys in order to introduce a HindIII restriction site.

CD98 Binds Integrin ␤ Tails
binding to ␤ 1D , nor is it required for CD98 binding to ␤ 1A .
The Y788A mutation of ␤ 1A (Fig. 1) disrupts filamin (Fig. 4B) and talin (Fig. 5C) binding (21). Similar Tyr to Ala mutations in ␤ 7 and ␤ 1D tails, corresponding to the Y788A mutation in ␤ 1A (Fig. 1), also disrupted filamin (Fig. 4B) and talin (Fig. 5C) binding. CD98 binding to ␤ integrin tails was not affected by Tyr to Ala mutations (Figs. 4B and 5C). The Tyr to Ala mutation introduced into ␤ 1D or ␤ 7 did not increase CD98 binding, nor was CD98 binding reduced in the ␤ 1A (Y788A) mutant. These results confirm that talin or filamin competition does not account for the lack of CD98 binding to ␤ 1D and ␤ 7 and that talin or filamin binding is not required for CD98 binding to the ␤ 1A cytoplasmic domain.
CD98 Binding to Integrin Cytoplasmic Domains Correlates

FIG. 2. ␤ 1A cytoplasmic domains bind CD98. Jurkat human T cells were surface-labeled with Sulfo-Biotin N-hydroxysuccinimide, and the cells were lysed in buffer A (see "Experimental Procedures"). Panel
A depicts a reduced SDS-PAGE analysis of the biotinylated proteins that bound to Ni 2ϩ beads, coated with model proteins containing ␤ 1A (␤1A) or ␣ IIb (␣IIb) cytoplasmic tails. Adjacent lanes show the surface proteins present in the lysate (lysate) or the ones that bound to uncoated Ni 2ϩ beads (0). In panel B, the biotinylated surface proteins that bound to the ␤ 1A (␤1A) or ␣ IIb (␣IIb) tails or uncoated beads (0) were immunoprecipitated with CD98 antibody (IP) or a control IgG (IgG). The immunoprecipitates were fractionated by reduced SDS-PAGE, and biotinylated proteins were detected by streptavidin-peroxidase-generated chemiluminescence.
FIG. 3. CD98 binds to the ␤ 1A and ␤ 3 cytoplasmic domain rather than that of ␤ 1D or ␤ 7 . A, in the upper panel, surface-biotinylated Jurkat cell lysates were allowed to bind to model proteins containing the ␤ 1A , ␤ 7 , ␤ 1D , or ␣ IIb integrin tails. The bound fractions were immunoprecipitated with CD98 antibody and analyzed by SDS-PAGE, as described under "Experimental Procedures." In the lower two panels, human platelet lysates were incubated with the same tail constructs, and bound proteins were fractionated by reduced SDS-PAGE and immunoblotted with antibodies to talin or to filamin. The loading of each tail was verified by Coomassie Blue staining of the model proteins eluted from the beads and fractionated by SDS-PAGE (data not shown). B, the surface-labeled Jurkat T cell lysate used in panel A was allowed to bind to model proteins containing a heterodimer of the ␣ IIb and ␤ 3 tails, or to model proteins containing only the individual tails. Bound fractions were immunoprecipitated with CD98 antibody and analyzed by SDS-PAGE, as described under "Experimental Procedures." The membrane fraction and whole cell lysate were incubated with ␣ IIb , ␤ 1A , or ␤ 1D integrin tails and bound CD98 was detected by immunoprecipitation as described under "Experimental Procedures" (panel B). In panel C, lysates of Jurkat cells (upper) and platelets (lower) were analyzed for binding of CD98 and talin to ␤ 1A and ␤ 1D tails, and their corresponding YA (␤ 1 YA, ␤ 1D YA) mutants as described in Fig. 4. Loading of integrin tails was equal as verified by Coomassie Blue staining (data not shown).
with Complementation of Dominant Suppression-Overexpression of isolated integrin ␤ 1A cytoplasmic domains, in the form of a Tac-␤ 1A chimera, results in suppression of integrin activation. Dominant suppression is reversed by overexpression of CD98 (10). Tac-␤ 1A , Tac-␤ 1D , and Tac-␤ 7 induced dominant suppression of integrin activation (Fig. 6A). As noted above (Fig. 3), CD98 bound poorly to ␤ 1D and ␤ 7 tails, showing that CD98 binding is not required for dominant suppression. However, CD98 was much less effective at reversing the suppression induced by Tac-␤ 1D and Tac-␤ 7 (Fig. 6B). Thus the capacity of CD98 to rescue suppression correlates with its binding to the suppressive ␤ cytoplasmic domain.
CD98 Binding Is Not Sufficient to Induce Dominant Suppression-As noted above, ␤ 1A tails suppress integrin activation and bind CD98. To assess whether CD98 binding alone is sufficient to induce dominant suppression, we first examined CD98 binding to a series of ␤ 1A truncation mutants (Fig. 1). CD98 binding was lost when the C-terminal seven residues were deleted (␤ 1A C797X)) but not when the last three amino acids were eliminated (␤ 1A (801X)) (Fig. 7A). Despite maintaining its capacity to bind to CD98, the Tac-␤ 1A (801X) mutant was a poor suppressor of integrin activation (Fig. 7B), and this was not due to a quantitative reduction in the association of CD98 with ␤ 1A (801X) (Fig. 7C). Furthermore, the ␤ 1A (Y788A) mutant, which also bound CD98 (Figs. 4 and 5), failed to suppress integrin activation (Fig. 7B). Consequently, integrin ␤ cytoplasmic domain binding to CD98 is not sufficient to induce dominant suppression. DISCUSSION CD98 is implicated in several cellular functions, including amino acid transport, cell fusion events, and integrin activation (12). We previously found that CD98 reverses dominant suppression of integrin function (10). We now report that: 1) CD98 associates with the ␤ 1A integrin cytoplasmic domain; 2) CD98 interacts differentially with ␤ cytoplasmic tails in a class-and splice variant-specific manner, which is independent of the capacity of the tails to bind the cytoskeletal proteins talin and filamin; 3) CD98's capacity to associate with integrin tails correlates with its ability to overcome dominant suppression of integrin activation; 4) CD98 association with integrin tails is neither necessary nor sufficient for dominant suppression of integrin activation. Thus, the association of CD98 with integrin cytoplasmic domains may regulate the function and localization of these membrane proteins.
CD98 physically associates with ␤ 1A integrin cytoplasmic domains. This association was observed utilizing model protein mimics of dimerized integrin cytoplasmic tails, and it may account for the physical association of certain ␤ 1 integrins with CD98. 2 The specificity of the interaction was confirmed by the lack of binding to mimics containing cytoplasmic domains from ␣ IIb or several other ␤ subunits. CD98 was added to the tails in the presence of other cellular proteins, so it remains possible that an intermediary protein is required for this interaction. However, CD98 was the only surface protein observed binding to the ␤ 1A tail (Fig. 2). Moreover, we observed CD98 binding in the absence of two known integrin binding proteins, talin and filamin (Figs. 3 and 4). CD98 failed to bind to ␤ 1D and ␤ 7 cytoplasmic domains, even though these tails bind many of the same polypeptides as ␤ 1A (21). Thus, we conclude that CD98 associates with the ␤ 1A tail and that the interaction is potentially direct.
CD98 binds to integrin ␤ cytoplasmic domains with unique splice variant and class specificity. CD98 bound well to the ␤ 1A tail and the ␤ 3 tail. Binding to the ␤ 1D and ␤ 7 tails was negligible. The specificity of CD98 binding differs markedly from the specificity of talin and filamin binding, since talin binds preferentially to the ␤ 1D tail and filamin to the ␤ 7 tail (21). Moreover, the binding of both cytoskeletal proteins is sensitive to the Tyr substitution with Ala in the first "NPXY" (21) in ␤ 1A and, as shown here, in ␤ 7 and ␤ 1D . Strikingly, CD98 binding was insensitive to this mutation. Finally, although the last three residues of ␤ 1A were dispensable, the last seven residues were required for binding. Thus, the features of the ␤ tail defined here for CD98 binding identifies a novel structural specificity for integrin ␤ tail function.
CD98 binding to ␤ tails correlates with its capacity to com-2 M. Hemler, personal communication.  Tac chimera is shown. B, CD98 binding to ␤ tails correlates with its ability to reverse dominant suppression. ␣␤py cells were transfected with each of the Tac chimeras in the presence or absence of 4 g of cDNA encoding full-length CD98. 24 h after transfection, cells were collected and the Tac-positive subset of cells were analyzed for the ability to bind to the PAC1 antibody. Data are expressed as percentage reversal, which is calculated as (AI ␤x ϩ CD98 Ϫ AI ␤x )/(AI ␣5 Ϫ AI ␤x ). AI is the activation index, AI ␤x is the AI of cells transfected with Tac ␤ chimeras, AI ␤x ϩ CD98 the AI of cells transfected with CD98 and Tac ␤ x chimeras, and AI ␣5 is the AI of cells transfected with the Tac-␣5. The x of ␤ x can have values of 1A, 1D, and 7 for the Tac-␤ 1A , Tac-␤ 1D , and Tac-␤ 7 chimeras, respectively. The expression of the Tac-␤ 1A and Tac-␤ 1D chimeras were similar (mean fluorescence intensity ϭ 340 Ϯ 20 and 370 Ϯ 50 units, respectively), while Tac-␤ 7 was better expressed (mean fluorescence intensity ϭ 530 Ϯ 90 units). In the absence of CD98, Tac-␤ 7 (55 Ϯ 6% suppression) inhibited activation less than Tac-␤ 1A or Tac-␤ 1D (77 Ϯ 4% and 82 Ϯ 7% suppression, respectively). plement dominant suppression. CD98 was implicated in integrin activation by its capacity to reverse the suppression of integrin activation caused by an isolated ␤ 1A cytoplasmic domain (10). In the present work, we found that CD98 binds to the ␤ 1A cytoplasmic domain, but fails to bind well to the ␤ 7 or ␤ 1D cytoplasmic domain. Strikingly, CD98 failed to complement dominant suppression initiated by either ␤ 7 or ␤ 1D cytoplasmic domains. Consequently, the mechanism of CODS appears to involve CD98 binding to the suppressive ␤ tail. Furthermore, cross-linking of CD98 stimulates integrin ␣ 3 ␤ 1dependent adhesion in small cell lung cancer cells (10) and in certain breast cancer cell lines (28) and ␤ 1 integrin-dependent cell fusion events (29 -36). Thus, our finding that CD98-␤ 1 cytoplasmic domain interactions correlate with effects on integrin function is relevant to integrin-dependent events involved in mulinucleate giant cell formation, virally induced cell fusion, and regulation of cell adhesion.
The physical interaction of CD98 with integrin cytoplasmic domains may be involved in modulating amino acid transport regulation. CD98 is known to regulate yϩL and L type amino acid transport (14,15,17,18). This regulation is probably due to disulfide-bonded heterodimer formation with a variety of light chains, that resemble permease amino acid transporters (13)(14)(15)(16)(17)(18)(19)(20). In fact, mutations in one of these light chains (15) are a likely cause of lysinuric protein intolerance (37). CD98 may function to regulate both the expression and localization of its light chains (18). In certain cells CD98 has a basolateral localization (38). ␤ 1A integrins also manifest basolateral polarization in many cells (39,40), probably due to interactions with underlying matrix components (41) or recruitment to lateral cell contacts (42). It is noteworthy that ␤ 7 integrins are primarily involved in lymphocyte homing and ␤ 1D integrins primarily form mechanical linkages in striated and cardiac muscle (43,44). Thus, the failure of these cytoplasmic domains to bind to CD98 correlates well with their lack of a role in establishing polarity in epithelial or mesenchymal cells. Consequently, the physical association of CD98 with ␤ 1A integrin cytoplasmic domains may participate in the polarization and regulation of amino acid transporters and to modulate the function of certain integrins.