Breaking the Integrin Hinge

Integrins are heterodimeric (a, b) cell adhesion receptors. We demonstrate that point mutations in the cytoplasmic domains of both the a and b subunits promote constitutive signaling by the integrin aIIbb3. By generating charge reversal mutations, we show these “activating” mutations may act by disrupting a potential salt bridge between the membrane-proximal portions of the a and b subunit cytoplasmic domains. Thus, the modulation of specific interactions between the a and b subunit cytoplasmic domains may regulate transmembrane signaling through integrins. In addition, these activating mutations induce dominant alterations in cellular behavior, such as the assembly of the extracellular matrix. Consequently, somatic mutations in integrin cytoplasmic domains could have profound effects in vivo on integrin-dependent functions such as matrix assembly, cell migration, and anchorage-dependent cell growth and survival.

Integrins are heterodimeric (␣, ␤) cell adhesion receptors. We demonstrate that point mutations in the cytoplasmic domains of both the ␣ and ␤ subunits promote constitutive signaling by the integrin ␣ IIb ␤ 3 . By generating charge reversal mutations, we show these "activating" mutations may act by disrupting a potential salt bridge between the membrane-proximal portions of the ␣ and ␤ subunit cytoplasmic domains. Thus, the modulation of specific interactions between the ␣ and ␤ subunit cytoplasmic domains may regulate transmembrane signaling through integrins. In addition, these activating mutations induce dominant alterations in cellular behavior, such as the assembly of the extracellular matrix. Consequently, somatic mutations in integrin cytoplasmic domains could have profound effects in vivo on integrin-dependent functions such as matrix assembly, cell migration, and anchorage-dependent cell growth and survival.
The integrin family of cell adhesion receptors are heterodimers of ␣ and ␤ transmembrane subunits that play key roles in important biological processes such as inflammation, wound healing, and cell growth and survival. Integrins modulate their affinity for ligands via a process termed "activation" or "inside-out" signaling (1,2). Furthermore, ligand binding to integrins changes the activities of cytoplasmic kinases, GTPases, and phospholipases ("outside-in" signaling) (2)(3)(4)(5). Thus, integrins are bidirectional signaling receptors conducting information both into and out of the cell.
Inside-out signaling may involve the propagation of a conformational change from the integrin cytoplasmic domains to the extracellular domains resulting in high affinity ligand binding (6,7). Integrin ␣ and ␤ subunit cytoplasmic domains share a similar membrane-proximal organization with apolar and polar sequences following sequentially after the mem-brane-cytoplasm interface (Fig. 1A). The conserved sequences for the ␣ and ␤ subunits are -GFFKR and LLv-iHDR (highly conserved residues are uppercase, less conserved residues are lowercase, and dashes represent nonconserved residues). Deletion of these sequences in either ␣ or ␤ subunit cytoplasmic domain "activates" integrins, locking them in the high affinity state (8 -11). Consequently, we termed this region the integrin "hinge." Hence, the capacity of the conserved membrane-proximal motifs to regulate the affinity state of integrins may depend on an interaction between them that constrains integrins to a low affinity state. Furthermore, studies on integrin assembly suggest that there may an interaction between the membrane-proximal portions of the ␣ and ␤ subunit cytoplasmic domains (12).
In this paper, we describe point mutations in this hinge region that activate ␣ IIb ␤ 3 . Moreover, by generating complementary charge reversal mutations, we show these activating mutations may act by disrupting a potential salt bridge between the membrane-proximal portions of the ␣ and ␤ subunit cytoplasmic domains. In addition, we demonstrate that these active mutations can induce the constitutive outside-in signaling as assayed by the phosphorylation of pp125 FAK and the ligand independent recruitment of ␣ IIb ␤ 3 to focal adhesions.
cDNA Constructs-pCDM8 expression constructs encoding wild-type ␣ IIb and ␤ 3 were constructed as described (9). The generation of expression constructs in pCDM8 encoding point mutation ␣ IIb and ␤ 3 cDNAs were undertaken using PCR 1 mutagenesis (15). CDM8 expression vectors encoding ␤ 3 mutants were constructed by cloning a 0.9-kb MluI and PstI cut PCR fragment encompassing the mutations into MluI and PstI cut pCDM8. This construct was then cut with AflII and DraIII, and the 3-kb fragment ligated with the 3.5-kb AflII-DraIII fragment of pCD␤ 3 . pCDM8 expression vectors encoding ␣ IIb mutants were constructed by tripartite ligation of a 0.6-kb XbaI-BamHI cut PCR fragment encompassing the mutations, the 4.1-kb BamHI-DraIII fragment of pCD␣ IIb , and the 2.6-kb DraIII-XbaI fragment of pCDM8. All constructs were verified by DNA sequencing and purified by CsCl centrifugation before transfection. Mutagenic oligonucleotides were synthesized on a model 391 DNA synthesizer.
PAC1 Binding-The cDNAs expressing ␣ IIb ␤ 3 variants were transiently transfected into Chinese hamster ovary (CHO-K1) as described (16). PAC1 binding was then analyzed by two-color flow cytometry as described (9,10). Briefly, FITC-PAC1 binding was analyzed only on a gated subset of cells positive for ␣ IIb ␤ 3 expression detected with a biotinylated non-function blocking antibody to ␣ IIb ␤ 3 (D57) and phycoerythrin/strepavidin. To define affinity state, histograms depicting PAC1 staining in the absence or presence of the competitive inhibitor Ro 43-5054 (32) were compared. The activation index (AI) was defined as 100 where F o is the median fluorescence intensity of PAC1 binding and F r is the median fluorescence intensity of PAC1 binding in the presence of competitive inhibitor (Ro 43-5054, 1 M). F o LIBS6 is the median fluorescence intensity of PAC1 binding in the presence of 2 M anti-LIBS6, and F r LIBS6 is the median fluorescence intensity of PAC1 binding in the presence of 2 M anti-LIBS6 and competitive inhibitor.
Analysis of pp125 FAK Phosphorylation-100-mm tissue culture plates were coated with 5 mg/ml BSA or 100 g/ml fibrinogen and blocked with 5 mg/ml BSA. The cells were harvested and resuspended in incubation buffer (137 mM NaCl, 2.7 mM MgCl 2 , 5.6 mM glucose, 3.3 mM NaH 2 PO 4 , and 20 mM HEPES, pH 7.4). 1 ϫ 10 7 cells were added to each coated dish and incubated at 37°C for 90 min. Cells were then washed in PBS and lysed with RIPA buffer containing 1 mM phenylmethylsulfonyl fluoride, 1 mM Na 3 VO 4 , and 100 kallikrein inactivating units/ml aprotinin. Lysates were clarified by centrifugation, and the protein content was determined with the BCA reagent. 300 g of protein from each lysate were incubated overnight at 4°C with the anti-pp125 FAK rabbit polyclonal antibody BC3 (a generous gift from J. Thomas Parsons, Charlottesville, VA). Immune complexes were precipitated at 4°C with protein A-Sepharose and washed extensively in ice-cold RIPA buffer containing 1 mM Na 3 VO 4 . The immune complexes were extracted into Laemmli sample buffer containing 10% ␤-mercaptoethanol, subjected to SDS-polyacrylamide gel electrophoresis on 7.5% polyacrylamide gels, and electrotransferred. Western blots were prepared and analyzed for phosphotyrosine-containing proteins as described (17) using a mixture of the anti-phosphotyrosine antibodies PY20 and PY72 (a generous gift from Bart Sefton, La Jolla, CA).
Immunofluorescence-CHO cells were transiently transfected as described (16) and, after 48 h, were cultured on fibronectin-coated coverslips for 2 h at 37°C. The cells were fixed with 3.7% paraformaldehyde (methanol-free) and permeabilized with 0.2% Triton X-100. The coverslips were blocked in 10% normal goat serum, incubated with the rabbit polyclonal antibody 2308 (anti-human ␣ IIb ) 2 and the anti-hamster ␤ 1 mAb 7E2 (18) in 10% normal goat serum for 60 min, washed with PBS, and then incubated with the secondary antibodies, FITC-conjugated goat anti-mouse and FITC-conjugated goat anti-rabbit in 10% normal goat serum for an additional 30 min. The coverslips were washed in PBS and mounted using FITC-guard mounting medium. The slides were examined with a Leitz Orthoplan microscope with a 100ϫ oil immersion objective. Photographs were taken on Kodak Tmax 400 film.
Fibronectin Matrix Assembly-Cells were suspended in ␣-minimal essential medium containing 10% fetal calf serum depleted of fibronectin by gelatin affinity adsorption and supplemented with purified human plasma fibronectin. All reagents were dialyzed against three changes of 100 volumes of ␣-minimal essential medium and passed through a 0.2-m filter prior to addition to the culture medium. Cells were plated in 12-well HTC slides at a final density of 2.3 ϫ 10 5 cells/ml and cultured for 41 h. The cells were then fixed with 3.7% paraformaldehyde and stained with polyclonal rabbit anti-human plasma Fn antibody MC54 (32 g/ml) and Cy3-conjugated goat anti-rabbit IgG antibodies (7.5 g/ml). Stained cell monolayers were observed using a Nikon FXA epifluorescence microscope, and representative fields were photographed using Kodak T-Max 400 or Ektachrome 1600 direct positive slide film.

RESULTS AND DISCUSSION
To define those membrane-proximal residues of the ␣ subunit important for affinity modulation, we substituted specific residues in ␣ IIb with Ala. These variants were then co-expressed in Chinese hamster ovary (CHO) cells with a wild-type ␤ 3 and the binding of PAC1, an antibody specific for the active conformation of ␣ IIb ␤ 3 was used to define affinity states (14) (Fig. 1B). The Ala substitution of ␣ IIb (F992), ␣ IIb (F993), and ␣ IIb (R995) activated ␣ IIb ␤ 3 as determined by high affinity PAC1 binding (Fig. 1, B and C). In contrast, the Ala substitution of ␣ IIb (G991) and ␣ IIb (K994) had minimal effect (Fig. 1B). Therefore, Ala substitution of specific residues in the conserved membrane-proximal GFFKR motif of the ␣ subunit can activate an integrin.
FIG. 1. The Ala substitution of specific residues in the membrane-proximal regions of the ␣ IIb and ␤ 3 cytoplasmic domains activates ␣ IIb ␤ 3 . A, schematic representation of the topology of the transmembrane and cytoplasmic domains of ␣ IIb ␤ 3 . The conserved membrane-proximal sequences of the ␣ IIb and ␤ 3 are illustrated (highly conserved residues are uppercase, less conserved residues are lowercase). Integrin ␣ and ␤ cytoplasmic domains share a similar membraneproximal organization with apolar and polar sequences following on sequentially after the W-K membrane-cytoplasm interface. The conserved sequences for the ␣ and ␤ subunits are -GFFKR and LLv-iHDR (dashes represent unconserved residues), respectively. Deletion of these sequences lock integrins in the high affinity state. The ␤ 4 and ␤ 8 subunits lack these conserved membrane-proximal sequences which suggests that they may signal via a different mechanism to other integrins. B, flow cytometry histograms illustrating PAC1 binding to ␣ IIb ␤ 3 and ␣ IIb (F992A)␤ 3 . Depicted are flow cytometry histograms illustrating PAC1 binding in the presence (open histogram) or absence (filled histogram) of competitive inhibitor, Ro 43-5054 to CHO cells expressing wild type ␣ IIb ␤ 3 and ␣ IIb (F992A)␤ 3 .The peptidomimetic Ro 43-5054, is a selective inhibitor of ligand binding to ␣ IIb ␤ 3 (32). The ␣ IIb (F992A)␤ 3 transfectants specifically bind PAC1, illustrating that this mutation activates ␣ IIb ␤ 3 . In contrast, cells expressing wild-type ␣ IIb ␤ 3 , which is in the low affinity state, can only bind PAC1 in the presence of an activating antibody (anti-LIBS6) that binds to the extracellular domain of ␤ 3 (13). C, activation indices of the ␣ IIb ␤ 3 mutants.
␣ IIb (R995A) may activate ␣ IIb ␤ 3 by the disruption of an interaction between the ␣ IIb and ␤ 3 cytoplasmic domains. The ␤ 3 cytoplasmic domain contains a highly conserved Asp residue that is at a similar displacement from the proposed cytoplasmmembrane interface as the highly conserved Arg-995 of ␣ IIb . This raises the possibility that the membrane-proximal regions may interact via a salt bridge formed between ␣ IIb (R995) and ␤ 3 (D723). To test this idea, we expressed ␣ IIb ␤ 3 (D723A); this integrin bound PAC1 with high affinity (Fig. 1C). To further test the proposed salt bridge we constructed the "chargereversal" mutants, ␣ IIb (R995D) and ␤ 3 (D723R). Both the single mutations, ␣ IIb (R995D)␤ 3 and ␣ IIb ␤ 3 (D723R), were in the high affinity state and exhibited spontaneous PAC1 binding (Fig. 2). However, the double charge reversal mutant ␣ IIb (R995D)␤ 3 (D723R) complemented the activating effect of the individual mutations. We suggest that this double mutation may restore the potential salt bridge between the ␣ and ␤ subunits, reforming the structural constraint which prevents the activation of the integrin. As a control, we examined the affinity state of double mutants ␣ IIb (R995A)␤ 3 (D723A), ␣ IIb (R995A)␤ 3 (D723R), and ␣ IIb (R995D)␤ 3 (D723A). All of these variants bound PAC1 spontaneously (data not shown).
The association of integrins with specialized cytoskeletal structures termed focal adhesions, is regulated by ligand binding to their extracellular domains (21). As another assay for outside-in signaling, we analyzed the effect of the activating mutations on integrin targeting to focal adhesions. To ensure that the targeting of the ␣ IIb ␤ 3 variants to focal adhesions was independent of ligand binding, each was expressed with a ligand binding-deficient ␤ 3 mutant, ␤ 3 (D119Y) (22). When expressed in CHO cells plated on fibronectin, both ␣ IIb (R995D)␤ 3 (D119Y) and ␣ IIb ␤ 3 (D723R)(D119Y) were spontaneously recruited to focal adhesions formed by endogenous hamster integrins (Fig. 3b). In contrast, the distribution of ␣ IIb (R995D)␤ 3 (D723R)(D119Y) was diffuse, similar to that of ␣ IIb ␤ 3 (D119Y), and it was not recruited to focal adhesions formed by the endogenous integrins (Fig. 3b). Therefore, we conclude that these activating point mutations allow a spontaneous association of the integrin with the cytoskeleton in the absence of ligand binding. Thus, activating membrane-proximal point mutations in both integrin ␣ and ␤ subunits can induce constitutive bidirectional transmembrane signaling. The cells were in suspension on BSA-coated plates, but adhered to and spread on the fibrinogen-coated plates. The cells were then processed for analysis of pp125 FAK tyrosine phosphorylation (17). In contrast to cells expressing wild-type ␣ IIb ␤ 3 , those expressing ␣ IIb (F992A)␤ 3 and ␣ IIb ␤ 3 (D723A) exhibited tyrosine phosphorylation of pp125 FAK when incubated in suspension over BSA. As expected, cells expressing ␣ IIb ␤ 3 phosphorylated pp125 FAK when plated on fibrinogencoated plates, as did cells expressing ␣ IIb (F992A)␤ 3 or ␣ IIb ␤ 3 (D723A) (data not shown). Untransfected CHO cells did not phosphorylate pp125 FAK when plated on BSA or fibrinogen-coated plates. b, recruitment of ␣ IIb (R995D)␤ 3 and ␣ IIb ␤ 3 (D723R) to focal adhesions in a ligand independent manner. CHO cells were transiently transfected and, after 48 h, were cultured on fibronectin for 2 h. The cells were then stained with a mixture of anti-human ␣ IIb (panels A, B, C, and D) and antihamster ␤ 1 (mouse monoclonal 7E2). In all experiments, ␤ 1 was localized in punctuate structures along the cell edge and along the ventral surface, characteristic of focal adhesions (data not shown). ␣ IIb ␤ 3 (D119Y) had a uniform cell surface distribution when plated on fibronectin (panel A) and was not localized to focal adhesions. In contrast, ␣ IIb (R995D)␤ 3 (D119Y) (panel B) and ␣ IIb ␤ 3 (D723R)(D119Y) (panel C) were recruited to the focal adhesions. ␣ IIb (R995D)␤ 3 (D723R)(D119Y) (panel D) behaved similarly to ␣ IIb ␤ 3 (D119Y) and was not present in focal adhesions.
There presently exists no three-dimensional structure of the native cytoplasmic domains of ␣ IIb ␤ 3 . Furthermore, a high resolution structure of this transmembrane protein may be difficult to acquire. Consequently, a mutational analysis, similar to those conducted in bacterial chemoattractant receptors (24) and G protein-coupled receptors (23,26), can provide a viable alternative to develop a structural hypothesis of transmembrane signaling. The approach described here has led us to propose a plausible and testable mechanism for integrin signaling. Indeed, the present studies may provide insight into a general mechansim of signaling mediated by a variety of transmembrane receptors. Point mutations can constitutively activate such structurally diverse receptors such as G proteincoupled receptors, growth factor receptors, and bacterial chemoattractant receptors (23)(24)(25)(26). However, only in the ␤ 2adrenergic receptor has constitutive bidirectional signaling been reported (26). In common with integrins, the ability of specific mutations to activate these receptors has been ascribed to the release of a "constraint" that maintains the receptor in an off state. As the topography of integrins is comparatively simple, having two parallel membrane-spanning subunits, we have been able to identify such a constraint. Utilizing charge reversal mutations, we provide direct mutational evidence for a salt bridge constraining integrins into a nonsignaling state.
As reported here, membrane-proximal point mutations in both integrin ␣ and ␤ subunits can cause constitutive bidirectional transmembrane signaling. Signals from integrins can influence cell growth and death, and the assembly of the extracellular matrix (27)(28)(29). This raises the intriguing possibil-ity that activating integrin mutations may produce dominant phenotypes in vivo. The assembly of a fibronectin matrix, a process important in wound healing and cell migration during development, is regulated by integrin affinity state (29,30). Therefore, fibronectin matrix assembly could be perturbed by activating integrin mutations. To test this idea, we used CHO B2 cells that are unable to assemble a fibronectin matrix due to a lack of the appropriate integrins (31). Transfection of these cells with the constitutively active mutant ␣ IIb ␤ 3 (D723R) enabled them to assemble a fibronectin matrix (Fig. 4). In contrast, CHO B2 cells expressing wild-type ␣ IIb ␤ 3 failed to make a fibronectin matrix (Fig. 4). Thus, activating point mutations in the integrin cytoplasmic domains can influence the assembly of the extracellular matrix. It will be interesting to determine if such mutations could account for some of the increased deposition of extracellular matrix that characterizes certain pathological states.