Integrin αIIbβ3 Inside-out Activation

Background: The transmembrane domain of integrins plays a critical role in mediating receptor inside-out activation. Results: Inside-out activation triggers the repartitioning of the intracellular border of αIIb but not the β3 transmembrane domain into the lipid bilayer in living cells. Conclusion: Complex conformational changes occur in the transmembrane domain upon integrin αIIbβ3 inside-out activation. Significance: Our findings represent a new mechanism for integrin inside-out activation. Integrins are a family of heterodimeric adhesion receptors that transmit signals bi-directionally across the plasma membranes. The transmembrane domain (TM) of integrin plays a critical role in mediating transition of the receptor from the default inactive to the active state on the cell surfaces. In this study, we successfully applied the substituted cysteine scanning accessibility method to determine the intracellular border of the integrin αIIbβ3 TM in the inactive and active states in living cells. We examined the aqueous accessibility of 75 substituted cysteines comprising the C terminus of both αIIb and β3 TMs, the intracellular membrane-proximal regions, and the whole cytoplasmic tails, to the labeling of a membrane-permeable, cysteine-specific chemical biotin maleimide (BM). The active state of integrin αIIbβ3 heterodimer was generated by co-expression of activating partners with the cysteine-substituted constructs. Our data revealed that, in the inactive state, the intracellular lipid/aqueous border of αIIb TM was at Lys994 and β3 TM was at Phe727 respectively; in the active state, the border of αIIb TM shifted to Pro998, whereas the border of β3 TM remained unchanged, suggesting that complex conformational changes occurred in the TMs upon αIIbβ3 inside-out activation. On the basis of the results, we propose a new inside-out activation mechanism for integrin αIIbβ3 and by inference, all of the integrins in their native cellular environment.

Integrins are a large family of adhesion receptors on the cell surfaces that mediate cell adhesion, migration, and extracellular matrix assembly (1,2). 18 integrin ␣ subunits and 8 ␤ subunits form 24 heterodimers, of which each subunit consists of a large extracellular domain, a single transmembrane domain (TM), 3 and a short unstructured cytoplasmic tail. Integrins exist in a default low affinity state on the cell surfaces. Upon stimulation by intracellular signals, integrins convert to the active state that permits extracellular ligands binding (inside-out activation), which in turn promotes interactions of intracellular proteins with the cytoplasmic tails (outside-in signaling) (3,4). In this way, integrins transmit signals bi-directionally across plasma membranes.
Biochemical analyses (5,6), electron microscopy (7,8), and fluorescent energy transfer studies (9) have established that a constraint is present in the intracellular membrane-proximal (MP) region that restrains integrin ␣ IIb ␤ 3 in the inactive state. Intracellular signals interact with the ␤ 3 cytoplasmic tail that breaks the constraint to initiate a conformational change that traverses the TMs, which ultimately triggers activation of the extracellular domain. Two intracellular proteins, talin and kindlin, are reported to interact with the ␤ 3 cytoplasmic tail and work synergistically to activate ␣ IIb ␤ 3 (10 -12). The extracellular domain of integrin ␣ IIb ␤ 3 was crystallized (13), and a low resolution electron cryomicroscopy structure (8) of the full heterodimer was also reported; however, attempts to obtain a crystal structure of the full-length receptor were not successful.
Two stretches of highly conserved amino acids in the intracellular MP region of integrin ␣ IIb ␤ 3 play pivotal roles in controlling the receptor inside-out activation ( 991 GFFKR 995 in ␣ IIb ; 717 LLITIHD 723 in ␤ 3 ) (5, 6, 14) (Fig. 1A). A salt bridge between ␣ IIb Arg 995 and ␤ 3 Asp 723 was proposed to form a clasp constraining the receptor in the inactive state (15). Intense efforts including site mutagenesis scanning (15,16), transmission electron microscopy (7), nuclear magnetic resonance (NMR) (17)(18)(19)(20), chimeric swapping (21), and computational modeling (16,22) have been invested to determine how the intracellular clasp is formed, which gave conflicting results. Recent studies using NMR on the structure of associated ␣ IIb ␤ 3 TM peptides in the lipid bicelles (20), ␣ IIb ␤ 3 TM and cytoplasmic tail complex in CD 3 CN/H 2 O (1:1) mixture (17), and by using disulfide constraints coupled with structural modeling (16) have yielded three different views on the intracellular clasp of ␣ IIb ␤ 3 : the first proposed that the clasp is formed by the interaction between residues ␣ IIb Phe 992 /Phe 993 and ␤ 3 Tyr 715 , the second proposed that the clasp is formed by an inhibitory ligand binding in the MP region, and the third proposed that the clasp is formed between ␤ 3 Lys 716 side chain and the peptide backbone of ␣ IIb 991 GFFKR 995 motif. All of the aforementioned studies have greatly improved our knowledge on the conformation of the TMs and the intracellu-lar clasp in the MP region of integrin ␣ IIb ␤ 3 ; however, questions remain. It is known that the regulated association of ␣ IIb ␤ 3 TMs is driven by domains outside of the plasma membranes, whether the NMR structures of fragmented ␣ IIb ␤ 3 TM peptides correspond to any physiological states of the receptor is obscure. Questions such as how the TMs of integrin ␣ IIb ␤ 3 undergo spatial and conformational changes upon activation and how intracellular proteins interact with the cytoplasmic tails of ␣ IIb ␤ 3 to initiate inside-out activation can only be genuinely answered by the analyses of intact integrin protein in the cellular environment.
Here, we present a successful analysis of the cellular location of integrin ␣ IIb ␤ 3 TM intracellular borders in the receptor inactive and active states in the living cells by using the substituted cysteine scanning accessibility method (SCSAM). Our data revealed a new mechanism for integrin ␣ IIb ␤ 3 inside-out activation in cell membranes.

EXPERIMENTAL PROCEDURES
Materials-Site-directed mutagenesis kits were from Strategene. Biotin maleimide, DMEM, and all cell culture reagents were from Invitrogen. Anti-integrin ␣ IIb or ␤ 3 rabbit or mouse antibodies were from Santa Cruz Biotechnology. Protein Aand G-Sepharose, streptavidin/biotinylated-horseradish peroxidase complex (streptavidin-HRP), and goat anti-rabbit IgGconjugated horseradish peroxidase were from GE Healthcare. Igepal was from Sigma. PVDF membrane was from Millipore.
Site-directed Mutagenesis-A wild-type human ␣ IIb and a ␤ 3 cDNA were used as the templates for site-directed mutagenesis. Amino acids at the position of Leu 985 to Glu 1008 in ␣ IIb and Leu 712 to Thr 762 in ␤ 3 were individually replaced with cysteines. Mutagenesis was performed using the Strategene site-directed mutagenesis kit following the manufacturer's instructions. The complete cDNA sequence of each mutant was verified by DNA sequencing.
Protein Expression-Cysteine-substituted ␣ IIb ␤ 3 was transiently expressed in the human embryonic kidney 293 cells (HEK 293) by using Lipofectamine 2000 (from Invitrogen) transfection following the manufacturer's instructions. Chinese hamster ovary (CHO) cells were also used in some experiments. Cells were grown at 37°C in a 5% CO 2 atmosphere and harvested 48 h after transfection.
Flow Cytometry-Flow cytometry assays for assessing the effects of cysteine substitutions on ␣ IIb ␤ 3 activation were performed as described previously (23). In brief, cysteine-substituted ␣ IIb or ␤ 3 constructs were co-expressed with the wildtype partners in the CHO cells. 24 h after transfection, cells were stained with antibody D57, which measures receptor surface expression, and PAC1, which detects the active state of ␣ IIb ␤ 3 , in the presence and absence of Ro43-5054 or anti-LIBS6, and then subjected to FACS scan. Ro43-5054 is a competitive antagonist of ␣ IIb ␤ 3 , which was used to estimate nonspecific PAC1 binding (F 0 ), and anti-LIBS6 is an ␣ IIb ␤ 3 -activating anti-FIGURE 1. A, amino acid sequences of the TMs and the cytoplasmic tails of integrin ␣ IIb ␤ 3 . The proposed TMs are depicted in a box with broken lines, MP regions are colored in gray, and cysteine-substituted residues are highlighted in bold. The potential ionic interaction between ␣ IIb Arg 995 and ␤ 3 Asp 723 is indicated as circled positive and negative symbols. B, BM labeling of wild-type and cysteine-substituted integrin ␣ IIb ␤ 3 . HEK 293 cells transfected with cloning vector pcDNA3, wild-type ␣ IIb ␤ 3 , and mutant ␣ IIb E1008C/␤ 3 T762C were collected and labeled with BM at room temperature for 20 min. Cells were lysed, and integrin ␣ IIb ␤ 3 heterodimer was immunoprecipitated, resolved on 10% SDS-PAGE, and transferred to a PVDF membrane. Incorporated biotin was detected by HRP-streptavidin and ECL. The blot was stripped and probed with an anti-␣ IIb light chain and an anti-␤ 3 antibody simultaneously to detect the amount of protein in each sample.
body that was used to estimate maximal PAC1 binding (F max ). The activation index was calculated as 100 ϫ (F Ϫ F 0 )/(F max Ϫ F 0 ), where F ϭ mean fluorescence intensity of PAC1 staining under the test condition. For detailed methods, see Ref. 23.
Biotin Maleimide Labeling and Immunoprecipitation-Whole cell labeling with BM was performed as described previously (24). Briefly, transfected HEK 293 cells were collected and resuspended in PBSCM (PBS containing 0.1 mM CaCl 2 and 1 mM MgCl 2 , pH 7.0) solution and subsequently labeled with BM (0.2 mM final) at room temperature for 20 min. Reactions were stopped by adding 5-fold glutathione in molar ratio. Cells were then washed with PBSCM and lysed in IPB buffer (150 mM NaCl, 1% (v/v) Igepal, 0.5% (w/v) sodium deoxycholate, 10 mM Tris-HCl, pH 7.5) containing 0.2% (w/v) BSA and protease inhibitors (from Roche Applied Science) on ice for 10 min. ␣ IIb ␤ 3 proteins were immunoprecipitated by a mouse anti-human ␣ IIb ␤ 3 monoclonal antibody (sc-21783 from Santa Cruz Biotechnology) and protein G beads for 4 h at 4°C.
Membrane Isolation and Na 2 CO 3 Treatment-Membrane treatment with Na 2 CO 3 was performed as described previously (25). Briefly, transfected cells (10-cm plate) were collected, washed (TBS buffer: 140 mM NaCl, 10 mM Tris, pH 7.4), and incubated with the homogenization buffer (10 mM Tris, pH 7.4, with Roche protease inhibitors) for 30 min on ice. Cells were then lysed by Dounce homogenization. Cell debris were removed by low speed centrifugation (4,000 ϫ g, 5 min, 4°C), and the membrane fractions were collected by high speed centrifugation (35,000 ϫ g, 30 min, 4°C). Membrane pellets were first resuspended in 100 l of 0.3 M sucrose and then mixed with 2 ml of ice-cold 0.1 M or 3 M Na 2 CO 3 , pH 11.5, and incubated for 30 min on a rotating shaker at 4°C. Membranes were then collected by centrifugation, washed with PBSCM, and resuspended in 1.0 ml of PBSCM followed by BM labeling. In some experiments, the protein levels of the stripped and unstripped membranes were determined by the BCA method (from Thermo Scientific). Equal amounts of protein samples were resolved on SDS-PAGE and stained with Coomassie Blue.
Image and Data Analysis-Films from immunoblots and biotinylation blots were scanned with a Hewlett-Packard Scanjet 5590. Scanned images were quantified with UN-SCAN-IT gel TM version 6.1 software. Biotinylation levels were calculated according to Ref. 24.
Statistical Analysis-Means Ϯ S.E. were calculated with SigmaPlot 10 software. Statistical analysis was performed using SigmaPlot 10 software.

RESULTS
Substituted Cysteine Scanning Accessibility Method for Integrin ␣ IIb ␤ 3 Conformational Analysis-SCSAM uses sulfhydryl reactive chemical probes to determine the location of the introduced cysteines in a target membrane protein that is free of endogenous reactive cysteines (26,27). BM, a membrane-permeable sulfhydryl-specific reagent, has been used successfully for the topological analysis of transmembrane proteins (24,28). SCSAM is based on the observation that the chemical reaction only occurs in the aqueous environment. Cysteines residing in the aqueous medium can react with BM, whereas cysteines residing in the lipid bilayer, the protein interior, or forming disulfide bonds cannot. BM minimally labels the endoplasmic reticulum-retained membrane proteins (24), which is probably due to the high content of glutathione in the cytosol. Integrin ␣ IIb ␤ 3 is a cysteine-rich membrane receptor that contains 20 endogenous cysteines in ␣ IIb and 56 in the ␤ 3 subunit. The recent crystal structure of the extracellular domain of ␣ IIb ␤ 3 showed that all of the cysteines in ␣ IIb ␤ 3 form disulfide bonds (13); however, conflicting reports also suggested that the receptor may have endogenous thiol isomerase activity that exposes free cysteines (29). To determine whether free reactive cysteines are present in the wild-type ␣ IIb ␤ 3 , we tested wild-type integrin ␣ IIb ␤ 3 for labeling with BM. Wild-type ␣ IIb ␤ 3 and mutant constructs ␣ IIb -E1008C/␤ 3 -T762C were expressed in the HEK 293 cells and subjected to the whole cell labeling. Fig.  1B shows that the wild-type ␣ IIb or ␤ 3 has no detectable BM labeling, whereas the light chain of ␣ IIb with a single cysteine substitution (E1008C) at the C terminus or ␤ 3 with a cysteine substitution at Thr 762 (residue at the C terminus of ␤ 3 ) was strongly labeled, suggesting that no free endogenous cysteines are present in wild-type ␣ IIb or ␤ 3 that are available for BM labeling. Western blots verified that ␣ IIb (sc-6602, from Santa Cruz Biotechnology) and ␤ 3 (sc-6627, from Santa Cruz Biotechnology) are both well expressed. Based on the results, we individually substituted single amino acids (between Leu 985 and Glu 1008 ) with cysteines in the C-terminal region of ␣ IIb that covers the proposed C terminus of the TM, the intracellular MP region, and the whole intracellular tail, and the positions between Leu 712 and Phe 730 in ␤ 3 covering the proposed MP region of ␤ 3 TM (Fig. 1A).
Whole Cell Labeling of Cysteine-substituted Integrin ␣ IIb ␤ 3 with BM-Integrin ␣ IIb carrying substituted cysteines was coexpressed with the wild-type ␤ 3 in the HEK 293 cells using Lipofectamine 2000 transfection. 48 h after transfection, cells were collected and subjected to BM labeling. Fig. 2 shows that the positive control, ␣ IIb E1008C was strongly labeled, whereas the negative control, wild-type ␣ IIb , had no labeling. Amino acids between Leu 985 and Phe 993 were not labeled, suggesting that they are embedded in the lipid bilayer; amino acids between Asn 996 and Glu 1008 were increasingly labeled, indicating the C-terminal tail is free of any cytosolic protein interactions. Interestingly, K994C was weakly labeled, whereas the adjacent residue, R995C was unlabeled compared with the controls. These data suggest that in the intact ␣ IIb ␤ 3 complex in living cells, the intracellular border of ␣ IIb TM is at Lys 994 , which leaves the proposed salt bridge ␣ IIb Arg 995 /␤ 3 Asp 723 outside of the plasma membrane at the lipid/aqueous interface in the cytosol.
We then performed whole cell labeling on the cysteine-substituted integrin ␤ 3 constructs co-expressed with wild-type ␣ IIb in HEK 293 cells. Fig. 3, A and C, shows that none of the substituted cysteines in the region between Leu 712 and Glu 726 was labeled with BM, and the first BM labeled residue was at Phe 727 marking the intracellular border of ␤ 3 TM. The intensity of BM labeling on the cysteine substitutions from F727C to F730C was similar suggesting the peptide after Phe 727 enters the aqueous medium abruptly.
Functional Analysis of Cysteine-substituted Integrin ␣ IIb ␤ 3 -Functional effects of cysteine substitutions in the TM and the MP regions of integrin ␣ IIb ␤ 3 have recently been reported (16), in that cysteine substitution of any of the residues in the 991 GFFKRN 996 region of ␣ IIb activated the receptor to various degrees, whereas substitution in the remaining portion of the ␣ IIb intracellular tail had no functional effect. Surprisingly, in the MP region of ␤ 3 , only one cysteine substitution at amino acid Lys 716 substantially activated integrin ␣ IIb ␤ 3 . Because most of the functional studies of integrin ␣ IIb ␤ 3 were performed in the CHO cells, we expressed the cysteine-substituted constructs in the CHO cells and subjected them to FACS analysis with PAC1 antibody that recognizes the active state of integrin ␣ IIb ␤ 3 . Fig. 4 shows that, indeed, our findings were in agreement with the previous report.
Intracellular Border of Integrin ␣ IIb TM in Fully Active State-Functional results indicated that the intracellular border of integrin ␣ IIb TM represented mixed states of active/inactive receptors. Substitution of ␤ 3 Lys 716 with Cys or Pro was shown to activate integrin ␣ IIb ␤ 3 close to the full activation level (16) in the HEK 293 cells. To obtain the intracellular border of ␣ IIb TM in the active state, we co-expressed ␣ IIb cysteine-substituted constructs with ␤ 3 K716P and subjected them to whole cell labeling with BM. Fig. 5 shows that activation of integrin ␣ IIb ␤ 3 had no effect on the labeling in the region of Pro 998 to Glu 1008 ; however, the region of Lys 994 -Arg 997 was no longer labeled with BM, and the first labeled residues started with Pro 998 , which differs significantly from the previous results obtained in the mixed active/inactive states of the receptor. We further tested BM labeling on the ␣ IIb cysteine-substituted constructs co-expressed with a truncated form of ␤ 3 at Lys 716 , which also produces a highly active state of ␣ IIb ␤ 3 (14). The same results were observed (data not shown). To test whether breaking of the predicted outer membrane clasp (highly activates integrin ␣ IIb ␤ 3 ) affects conformation of the region, we whole cell labeled the ␣ IIb cysteine-substituted constructs coexpressed with ␤ 3 G708I (20). Again, we observed a similar pattern of labeling (data not shown), indicating that this conformational change in ␣ IIb is irrespective to the breaking of either the inner or the outer membrane clasp.
It was previously reported that the GFFKR region of ␣ IIb may associate with intracellular peripheral proteins such as calcium-and integrin-binding protein (30) which potentially shield the substituted cysteines from being labeled. To test this, we isolated the plasma membranes from cells expressing the ␣ IIb cysteine-substituted constructs (Phe 993 -Phe 998 ) with ␤ 3 K716P and subjected to chemical stripping with 0.1 M Na 2 CO 3 prior to BM labeling. Compared with the unstripped plasma membranes, Na 2 CO 3 treatment effectively removed multiple FIGURE 2. BM labeling of cysteine-substituted ␣ IIb co-expressed with wild-type ␤ 3 . A, representative results of BM labeling on integrin ␣ IIb -substituted cysteines. 24 amino acids in the C terminus of ␣ IIb were individually substituted with cysteines and labeled with BM as described under "Experimental Procedures." B, summary of BM labeling of integrin ␣ IIb -substituted cysteines. The level of biotin incorporation into each sample was quantified by densitometry, and the signal was normalized to the amount of integrin ␣ IIb light chain present in the sample. In each experiment, the level of biotinylation was compared with that of the ␣ IIb E1008C, whose labeling was set to 100%. Data represent mean of 3-5 experiments Ϯ S.E. (error bars).
protein bands from the membrane samples as shown on the Coomassie Blue-stained SDS-PAGE (Fig. 6A). Fig. 6, B and C, shows no differences in BM labeling were detected compared with the fully active state of ␣ IIb ␤ 3 . We then tested BM labeling on the membranes stripped with 3 M Na 2 CO 3 , a stringent condition inducing both counter-ion and ionic strength that could effectively expose a deeply buried residue in the sodium bicarbonate co-transporter 1 (25). Again, we did not observe any labeling on the substituted cysteines in Phe 993 -Arg 997 (data not shown). Interestingly, the Na 2 CO 3 -treated ␣ IIb ␤ 3 protein was no longer immunoprecipitated by the antibody recognizing the protein complex, but instead, by an antibody that recognizes ␣ IIb , suggesting that a conformational change in the extracellular domain of the receptor must have been induced by the Na 2 CO 3 treatment. Our data indicated that activation of ␣ IIb ␤ 3 triggers repartition of the ␣ IIb MP region from aqueous into the lipid bilayer.
Intracellular Border of Integrin ␤ 3 -TM in Fully Active State-The flow cytometry assay showed that none of the cysteine substitutions except K716C in the ␤ 3 TM/MP region activated ␣ IIb ␤ 3 . Therefore, the first labeled cysteine-substituted residue, F727C, marks the intracellular border of ␤ 3 TM in ␣ IIb ␤ 3 inactive state in living cells. To determine the intracellular border of ␤ 3 TM in activated ␣ IIb ␤ 3 , we co-expressed ␤ 3 cysteine-substituted constructs with ␣ IIb F992A/F993A, a construct that fully activates the receptor (16) . Fig. 3, B and C, shows that activation of ␣ IIb ␤ 3 did not expose any ␤ 3 endogenous cysteines to aqueous. Interestingly, F727C again remained the first residue labeled with BM, indicating its aqueous accessibility. We were somewhat surprised that no differences were observed in the ␤ 3 TM BM labeling between the active and the inactive state of ␣ IIb ␤ 3 , and therefore we performed additional BM labeling assays on the co-expressed ␤ 3 cysteine-substituted constructs with a truncated form of ␣ IIb at Gly 991 , which generates the constitutively active state of ␣ IIb ␤ 3 (6). Indeed, the same results were obtained (data not shown). These results suggested that ␤ 3 TM/MP regions do not repartition between lipid bilayer and aqueous medium upon the receptor activation as seen in ␣ IIb subunit.
Aqueous Accessibility of the Predicted Membrane-anchored Integrin ␤ 3 Intracellular Tail-A recent NMR analysis on the complex of fragmented ␣ IIb and ␤ 3 intracellular tails has sug-FIGURE 3. BM labeling on integrin ␤ 3 -substituted cysteines. Amino acids between Leu 712 and Phe 730 in the ␤ 3 subunit were individually substituted with cysteines and labeled with BM. A, BM labeling of cysteine-substituted ␤ 3 co-expressed with wild-type ␣ IIb . B, BM labeling of cysteine-substituted ␤ 3 co-expressed with mutant ␣ IIb F992A/F993A. Similar results were obtained with co-expression of ␣ IIb Gly 991 truncation. C, summary of BM labeling of cysteinesubstituted ␤ 3 . Biotin incorporation into each sample was quantified as described in Fig. 2 and was compared with ␤ 3 -T762C, whose labeling was set to 100%. Data represent mean of 3-6 experiments Ϯ S.E. (error bars).
gested that regions of Phe 727 -Trp 739 and Tyr 747 -Tyr 759 in the ␤ 3 tail form membrane-anchored ␣-helices in loose contact with the lipid bilayer (31). To determine whether these predicted membrane-anchored helices are accessible to BM labeling, we individually substituted amino acids from Glu 731 to Thr 762 (in wild-type ␤ 3 ) with cysteines that cover the whole intracellular tail of ␤ 3 . The cysteine-substituted constructs were co-expressed with wild-type ␣ IIb in HEK 293 cells and subjected to whole cell BM labeling. Fig. 7 shows that all of the substituted cysteines are increasingly strongly labeled. Therefore, these predicted membrane-anchored helices are accessi-ble to aqueous that differs from Leu 712 -Glu 726 region that is embedded in the lipid bilayer.

Intracellular Borders of Integrin ␣ IIb and ␤ 3 TMs in Living
Cells-Here, we present for the first time the application of SCSAM in analyzing the conformational changes of integrin ␣ IIb ␤ 3 , a membrane adhesion receptor in living cells. The intracellular MP region of integrin ␣ IIb ␤ 3 has been the target of intensive study because of its critical role in controlling the receptor inside-out activation. In this study, we have unambig-  Fig. 3. B, summary of BM labeling of integrin ␣ IIb -substituted cysteines. Results were analyzed same as described in Fig. 3. Data represent mean of 3-5 experiments Ϯ S.E. (error bars). Similar labeling results were obtained on cysteine-substituted ␣ IIb co-expressed with ␤ 3 Lys 716 truncation or ␤ 3 G708I mutation. FIGURE 6. BM labeling of cysteine-substituted ␣ IIb co-expressed with ␤ 3 -K716P after Na 2 CO 3 stripping. A, comparison of protein samples from membranes without and with Na 2 CO 3 stripping. Half-fraction of the isolated cell membranes was treated with 0.1 M Na 2 CO 3 for 30 min on ice. Membranes were then pelleted by centrifugation and washed once with PBS. Membrane pellets were lysed in IPB buffer, and protein levels were determined by the BCA method. Equal amounts of protein samples were loaded on the 4 -20% SDS-polyacrylamide gel. B, representative BM labeling of cysteine-substituted ␣ IIb Phe 993 -Pro 998 co-expressed with ␤ 3 K716P after 0.1 M Na 2 CO 3 treatment. Similar labeling results were obtained after 3 M Na 2 CO 3 treatment. C, summary of BM labeling. Results were analyzed the same as described in Fig. 3. Data represent mean of 3-6 experiments Ϯ S.E. (error bars).
uously resolved the intracellular borders of integrin ␣ IIb and ␤ 3 TM domain in the receptor active and inactive states on the surface of living cells. Our data revealed that, unlike the results from in vitro analysis, the intracellular borders of the active state integrin ␣ IIb and ␤ 3 TMs reside at amino acids Pro 998 and Phe 727 , respectively. This located the proposed MP regions of both subunits in the lipid bilayer in the active state. The conclusion is supported by two sets of BM labeling experiments by using highly active mutation (K716P and G708I) or truncation (at Lys 716 ) in the ␤ 3 subunit, and highly active mutation (F992A/F993A) or truncation (at Gly 991 ) in the ␣ IIb subunit that generates constitutive active states of integrin ␣ IIb ␤ 3 .
A puzzling pattern of BM labeling was observed on the determination of ␣ IIb TM border in the receptor inactive state. When co-expressed with wild-type ␤ 3 subunit, weak labeling was detected on ␣ IIb K994C; strong labeling on N996C, R997C, and P998C; however, no labeling on F992C, F993C, and R995C. This result is readily explained by the functional data from the current and the previous studies (16), which showed cysteine substitution of Arg 995 activated the receptor close to the maximum level in the HEK 293 cells, whereas Phe 992 , Phe 993 , Lys 994 , Asn 996 , Arg 997 , and Pro 998 were ϳ0 -40% activation, indicating that Ͼ60% of the population of the mutant receptors on the cell surface were not active. Additionally, the recent Rosetta model on the ␣ IIb MP region showed that the side chain of Lys 994 points away from the dimer interface in the receptor inactive state, suggesting its accessibility to the aqueous medium (16). Therefore, labeled K994C, N996C, R997C, and P998C represent the major population of inactive receptors. Consistent with this, no BM labeling was detected on ␣ IIb R995C co-expressed either with wild-type or activating ␤ 3 subunit because both conditions activated the receptor at a high level. Based on these observations, we concluded that the labeled K994C defines the intracellular border of ␣ IIb TM in the inactive state. One possibility is that in the active state, intracellular proteins may bind to the unclasped ␣ IIb tail that shields the substituted cysteines to BM labeling. We ruled out this possibility by stripping the isolated membranes with 0.1 M or 3 M Na 2 CO 3 that removes potential bound peripheral proteins prior to BM labeling and showed no detectable differences compared with the samples without treatment.
In contrast to ␣ IIb , determination of the ␤ 3 TM intracellular border was clear and compelling because none of the cysteine substitutions in the proposed MP region except Lys 716 activated the receptor. Our data showed that none of the cysteine substitutions prior to Phe 727 , including the highly charged 722 HDRKE 726 stretch, was accessible to BM labeling, suggesting their lipid embedment. Residues starting at Phe 727 became strongly labeled, indicating their aqueous exposure. Therefore, we assigned the intracellular border of ␤ 3 TM in the inactive state to Phe 727 . Indeed, the strong labeling of the cysteine-substituted ␤ 3 tail membrane-anchored ␣-helices further supports that Phe 727 marks the intracellular lipid/aqueous interface of ␤ 3 TM in the living cells.
Comparison with the in Vitro Determined TM Borders-Our findings differ significantly from the results obtained in the in vitro studies, including glycosylation mapping in the cell-free system (32,33) and the recent NMR structures of ␣ IIb (34) and ␤ 3 TM (35) peptides in the lipid bicelles, which both placed the intracellular borders of ␣ IIb TM at Lys 994 and ␤ 3 TM at Asp 723 . Because integrin TMs are fully separated in the active state (36), these in vitro findings may represent the TM borders of the integrin active state. Because the isolated TM fragments do not contain the intracellular clasp in the MP region and more importantly do not carry the extracellular domains, the physiologic significance is questionable. It is known that the extracellular domain of integrin ␣ IIb and ␤ 3 forms a dimer independent of the TM domains (13), and further, dimerization of the extracellular domains proceed the dimerization of the TM domains during biosynthesis. It is conceivable that the ␣ IIb ␤ 3 TM domain dimerization and lipid embedding are influenced or guided by the dimerized extracellular domain. In support of this, mutations that separated the dimerized ␣ IIb ␤ 3 TM fragment in the lipid bicelles failed to activate the intact receptor in the cells (20), confirming that the extracellular domain is involved in stabilizing the ␣ IIb ␤ 3 TM dimer and that the previous in vitro observations may not be physiological.
Our study has considerable advantages over the in vitro studies because the analyses were performed on intact integrin ␣ IIb ␤ 3 protein in living cells. Our data showed, in contrast to the in vitro findings, that Lys 994 marks the intracellular border of inactive state ␣ IIb TM. When the receptor is active, the border shifts to Pro 998 , indicating a significant peptide repartitioning from aqueous medium into the lipid bilayer. The extracellular domain of integrin was shown to switch from a bent to an extended conformation in the active state (36). We suggest that the peptide repartitioning is caused by the pulling of the extended extracellular domain rather by a spontaneous repartition of the peptide itself. Indeed, upon close examination of the lipid-embedded extracellular end of ␣ IIb TM, two amphiphilic residues (Trp 997 , Trp 998 ) immediately follow the first lipid embedded residue Ile 996 , making it possible for them to repartition into the aqueous medium as a result of the stretching forces induced by the extracellular domain extension. In support of this, when the extracellular domain was missing, the 994 KRNRP 998 stretch of ␣ IIb TM was not embedded in the lipid bilayer, indicating that spontaneous lipid repartitioning of the peptide is not favorable (32)(33)(34). This phenomenon can only be observed in intact integrin in a cellular environment.
Compared with the reported in vitro determined ␤ 3 TM intracellular border, our data are more complex: exposure to the aqueous starts at amino acid Phe 727 , leaving the stretch of charged residues 722 HDRKE 726 in the membrane. In the in vitro system, to place such a highly charged peptide in the lipid environment is energetically unfavorable. The ␤ 3 TM NMR structure showed that the TM helix terminates at Asp 723 (20, 35), whereas cysteine cross-linking analyses in the intact protein suggested that the helical structure extended to Phe 730 (16).
Following the latter observation, our findings suggest that there is an additional helix turn at the C terminus of ␤ 3 TM in the lipid bilayer, ϳ5.4 Å distant from the aqueous medium. Residue Asp 723 has been shown to face the dimer interface (16,20); therefore Arg 724 , Lys 725 , and Glu 726 would be 100 degrees apart, which potentially interacts with the negatively charged phosphate groups and the positively charged choline head of membrane phospholipids, which forms ionic interactions to stabilize the end of TM in the lipid bilayer. In agreement with this, we observed that activation of the receptor had no effect on membrane embedding of this region, suggesting that it is unlikely that the C terminus of ␤ 3 TM could move out of the lipid bilayer upon activation. Lipid embedding of heavily charged peptides has been shown in the crystallized voltage sensor of the voltagegated potassium channel (37), and potentially, these charged residues are hydrated (38).
A recent analysis of the integrin ␤ 3 TM intracellular border using a ␤ 3 TM fragment suggested that ␤ 3 Lys 716 helps determine the integrin ␤ 3 TM topography (39). Removal of the positive charge at this position (active state) induced aqueous exposure of the residues in the proposed ␤ 3 MP region. Because our experiments analyzed the intact receptor in a cellular environment, the difference between our findings and this observation is potentially due to differences in the experimental approaches used.
Implications for Integrin Inside-out Activation-The TMs of integrin have been proposed to undergo "separation" (9), "piston" (40), or "scissors" (40, 41) movement upon inside-out activation. We found that, upon activation of integrin ␣ IIb ␤ 3 in the living cells, the intracellular border of ␣ IIb TM undergoes an upward shift that repartitions into the lipid bilayer, whereas the position of ␤ 3 TM border remains unchanged. It has been shown that the ␣ IIb TM is perpendicular in the lipid bilayer whereas ␤ 3 TM tilts at a significant angle when dimerized (16,20). Our finding that the membrane embedded region of ␤ 3 TM extends to Glu 726 suggests that the TM tilts at more extended angles. Taking our in vivo findings together with the reported in vitro TM structures (16,20,23), we propose a new mechanism for integrin inside-out activation (Fig. 8). In the resting state, the MP region of ␣ IIb 991 GFFKR 995 interacts with ␤ 3 Lys 716 forming a clasp that keeps the receptor inactive. Upon insideout activation, the breakage of the intracellular clasp leads to extensive conformational changes in the extracellular domain from a bent to an extended conformation. Because the two TMs are dissociated, the extended ␣ IIb extracellular domain pulls its TM upward causing four consecutive residues in the MP region repartition into the lipid bilayer. The NMR structure shows the region after ␣ IIb Val 990 has a left reverse turn rather than a helical structure; this would minimize the energy required for the peptide repartition into the lipid bilayer. Additionally, the up shift of ␣ IIb TM would draw the negatively charged 1001 EED-DEEGE 1008 stretch close to the surface of lipid bilayer, which may form ionic interactions with the positively charged choline head of membrane phospholipids to serve as a "break" for the prevention of "whiplash" of ␣ IIb extracellular domain, to keep integrin in the correct active state conformation on the surface of cells. Indeed, the unique reverse folding structure prior to the acidic stretch introduced by two prolines (998 and 999) in the ␣ IIb tail makes the break hypothesis plausible (42). FIGURE 8. Proposed mechanism of integrin inside-out activation. A, schemes showing that in the inactive state, the ␣ IIb TM (aqua) associates with ␤ 3 TM (blue) in the lipid bilayer with a clasp formed by GFFKRNR (red line) and Lys 716 (red dot) at the intracellular lipid/aqueous interface. Upon inside-out activation, extension of the ␣ IIb extracellular domain applies stretching forces on the ␣ IIb TM, causing its upward shift that results repartition of four amino acids in the MP region (red line) into the lipid bilayer and drawing the negatively charged tail (dashed black lines) close to the inner leaflet of the plasma membrane that may form ionic interactions with the positive head groups of membrane phospholipids. B, detailed positions of the residues in the ␣ IIb MP region involving lipid bilayer repartitioning upon inside-out activation. The structure of the ␣ IIb MP region was adapted from Refs. 16 and 34.