Structure of the Integrin αIIb Transmembrane Segment*

Integrin cell-adhesion receptors transduce signals bidirectionally across the plasma membrane via the single-pass transmembrane segments of each α and β subunit. While the β3 transmembrane segment consists of a linear 29-residue α-helix, the structure of the αIIb transmembrane segment reveals a linear 24-residue α-helix (Ile-966 -Lys-989) followed by a backbone reversal that packs Phe-992-Phe-993 against the transmembrane helix. The length of the αIIb transmembrane helix implies the absence of a significant transmembrane helix tilt in contrast to its partnering β3 subunit. Sequence alignment shows Gly-991-Phe-993 to be fully conserved among all 18 human integrin α subunits, suggesting that their unusual structural motif is prototypical for integrin α subunits. The αIIb transmembrane structure demonstrates a level of complexity within the membrane that is beyond simple transmembrane helices and forms the structural basis for assessing the extent of structural and topological rearrangements upon αIIb-β3 association, i.e. integrin transmembrane signaling.

Integrin cell-adhesion receptors transduce signals bidirectionally across the plasma membrane via the single-pass transmembrane segments of each ␣ and ␤ subunit. While the ␤3 transmembrane segment consists of a linear 29-residue ␣-helix, the structure of the ␣IIb transmembrane segment reveals a linear 24-residue ␣-helix (Ile-966 -Lys-989) followed by a backbone reversal that packs Phe-992-Phe-993 against the transmembrane helix. The length of the ␣IIb transmembrane helix implies the absence of a significant transmembrane helix tilt in contrast to its partnering ␤3 subunit. Sequence alignment shows Gly-991-Phe-993 to be fully conserved among all 18 human integrin ␣ subunits, suggesting that their unusual structural motif is prototypical for integrin ␣ subunits. The ␣IIb transmembrane structure demonstrates a level of complexity within the membrane that is beyond simple transmembrane helices and forms the structural basis for assessing the extent of structural and topological rearrangements upon ␣IIb-␤3 association, i.e. integrin transmembrane signaling.
Integrins are a major class of adhesion receptors that mediate cell migration and extracellular matrix assembly, as well as inflammation, thrombosis, and tumor metastasis (1). In addition to responding to signals that originate inside the cell and regulate receptor affinity, integrins can respond to the binding of extracellular ligands by activating intracellular signaling pathways (1). Integrins are noncovalent heterodimers with large extracellular and small intracellular domains that are connected by single-pass transmembrane (TM) 3 segments. An extensive set of experiments provides evidence that signals are transduced bidirectionally across the plasma membrane via the TM segments of the ␣ and ␤ subunits (2)(3)(4). Specifically, the inactive receptor state is stabilized by the hydrophobic packing of the TM helices and adjacent electrostatic ␣IIb-␤3 interactions, whereas integrin activation ensues from the separation of the TM segments (2, 4 -8). However, the precise structural and topological processes underlying integrin inside-out and outside-in signaling events are currently unknown.
Single-pass TM segments likely traverse the membrane as ␣-helices. The recently determined structure of the integrin ␤3 TM segment reconstituted in small bicelles fulfills this expectation, but the TM helix is not restricted to residues between the first charged residues, Asp-692 and Lys-716, on the extraand intracellular sides, respectively (9). Rather, the five mostly hydrophobic residues following Lys-716 are part of the 29-residue TM helix, encoding a substantial helix tilt relative to the membrane. Apparently, this allows Lys-716 to snorkel its side chain out of the hydrophobic lipid core. Integrin ␣ subunits exhibit a similar pattern of hydrophobicity; a lysine residue constitutes the first charged residue on the intracellular side and is followed by four mostly apolar residues (see Fig. 1). Nevertheless, the structure of the ␣IIb TM segment reported herein shows that the TM helix ends at Lys-989 and that the following four residues loop back to immerse Phe-992-Phe-993 in the hydrophobic lipid core. Moreover, the 24-residue TM helix formed can immerse in a typical bilayer without a significant helix tilt. The ␣IIb TM structure reveals an unexpected level of complexity within the integrin TM region that appears crucial for understanding the association with its partnering ␤3 TM segment. Furthermore, the structure highlights that single-pass TM segments are not restricted merely to helical conformations.

EXPERIMENTAL PROCEDURES
Integrin ␣IIb Expression and Purification-Synthetic oligonucleotides coding for human integrin ␣IIb residues Ala-958 -Pro-998 were assembled by PCR and subcloned into the pET-44 expression vector (Novagen, Inc.) with the third IgGbinding domain of protein G (GB3) as N-terminal fusion protein and an intervening tobacco etch virus protease cleavage site. Expression and purification proceeded as described for the ␤3 TM peptide (9) except that the ␣IIb peptide was cleaved from the fusion protein using tobacco etch virus protease at a molar ratio of 1:50 overnight at room temperature in 50 mM Tris⅐HCl, pH 8.0, 0.5 mM EDTA, 1 mM dithiothreitol, and 0.5 M urea, leaving Gly as N-terminal residue. To introduce a high affinity Cu 2ϩ -binding site, the Gly-Gly-His ATCUN motif (10,11) was introduced by PCR for residues corresponding to ␣IIb(957-959). GGH-␣IIb peptide expression and purification proceeded in the same manner as with ␣IIb. Isotope labeling * This work was supported by an award from the American Heart Association (to T. S. U.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. was achieved by supplying 15 N-NH 4 Cl and 13 C-D-glucose and culturing in 99 and 50% D 2 O for highly and moderately deuterated peptide, respectively. NMR Sample Preparation-␣IIb peptide concentration was measured in acetonitrile-water solution by UV spectroscopy (⑀ 280 nm ϭ 16,500 M Ϫ1 cm Ϫ1 ), and defined amounts of peptide were freeze-dried. Peptide was taken up in 320 l of bicelle solution to give an ␣IIb peptide concentration of ϳ0.6 mM. The bicelle solution was comprised of 105 mM long-chain lipids and 350 mM 1,2-dihexanoyl-sn-glycero-3-phosphocholine (DHPC), 25 mM HEPES⅐NaOH, pH 7.4, 6% D 2 O, 0.02% w/v NaN 3 . As long-chain lipids, 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) and 1,2-dimyristoyl-sn-glycero-3-phosphocholine) were used, respectively. No significant H-N ␣IIb chemical shift changes were detected between the two lipid environments (supplemental Fig. 1), showing that equivalent structures were adopted. Supplemental Fig. 2 illustrates the spectral quality obtained.
To align the ␣IIb-bicelle complex relative to the magnetic field stretched, negatively charged polyacrylamide gels were used (12,13). A 320-l gel was polymerized in 150 mM Tris⅐HCl, pH 8.0, from a 4.2% w/v solution of acrylamide, 2-acrylamido-2-methyl-1-propanesulfonate (AMPS), and bisacrylamide with a monomer-to-cross-linker ratio of 49:1 (w/w) and a molar ratio of 94:6 of acrylamide to AMPS. For easier comparisons between gels, AMPS was counted as acrylamide in the given weight ratios (14). The gel was dialyzed overnight in 50 ml of 100 mM NaH 2 PO 4 /Na 2 HPO 4 , pH 6.8, followed by H 2 O for 10 h and another H 2 O dialysis overnight. The gel was dried completely and soaked in 320 l of ␣IIb-bicelle (POPC:DHPC) solution for 24 h and transferred into an open-ended NMR tube, as described (15). A 2 H splitting of 1.5 Hz was observed.
NMR Spectroscopy-NMR experiments were carried out on a cryoprobe-equipped Bruker Avance 700 spectrometer at 35°C except for the gel-aligned sample, which was measured at 40°C. Data were processed and analyzed with the nmrPipe package (16) and CARA. Throughout all experiments, the solvent was kept at ϩI z whenever possible (17,18), and TROSY-type H-N detection was used (19).
Structure Calculations-Structures of the well folded Ile-966 -Lys-994 residues (see Fig. 4A) were calculated by simulated annealing, starting at 3000 K using the program XPLOR-NIH (27). The peptide termini were represented by random-coil conformations. Besides standard force field terms for covalent geometry (bonds, angles, and improper torsions) and nonbonded contacts (van der Waals repulsion), the following experimental and empirical potentials were included. The difference between predicted and experimental residual dipolar couplings was described by a quadratic harmonic potential. Dihedral angle restraints were implemented using a quadratic square-well potential and merely aided in the convergence of the residual dipolar coupling restraints. , backbone dihedral angle constraints were extracted from N, C ␣ , C ␤ , and CЈ chemical shifts using the program TALOS (28). 1 side-chain angle restraints were derived from the 3 J CЈC␥ and 3 J NC␥ coupling constants (supplemental Table 2). NOE interproton distance restraints were referenced to the linear ␣-helical structure of Ile-966 -Lys-989 and were incorporated using a quadratic square-well potential. A backbone- backbone hydrogen-bonding potential (29) was used. A torsion angle potential of mean force (30), which was modified to implement the preferred rotamers in transmembrane helices (31,32), was employed. Moreover, the higher side-chain rotamers ( 2 -4 ) of Lys-989 were selected to extend toward the C terminus, i.e. snorkeling its side chain out of the membrane core. A total of 20 structures were calculated. The structural statistics are summarized in supplemental Table  1. The structures have been deposited in the Protein Data Bank (accession number 2k1a) together with the energyminimized average structure.

RESULTS AND DISCUSSION
Structure Determination and Description-The structure of the ␣IIb TM segment reconstituted in small bicelles was determined by multidimensional heteronuclear NMR spectroscopy, making use of 2 H/ 13 C/ 15 N-labeled peptide encompassing residues Ala-958 -Pro-998 (Fig. 1). Bicelles provide a small, symmetric lipid bilayer whose rim is stabilized by short-chain lipids (33,34), and they appear well suited for the study of single-pass TM segments (9). Under the experimental conditions employed, the presence of a paramagnetic tag on ␣IIb was not sensed by untagged ␣IIb peptides, and no significant concentration-dependent chemical shift changes were observed (supplemental Fig. 3), clearly showing that the ␣IIb structure is not affected by any homo-oligomerization tendency. To validate the choice of bicelle lipids, the effects of lipid headgroup and tail length on ␣IIb spectral parameters were examined. Between glycerophosphatidylcholine-based lipids with 16/18 and 14/14 hydrocarbon tails, no significant 1 H N and 15 N chemical shift changes were detected (supplemental Fig. 1). Moreover, the introduction of lipids carrying phosphatidylserine headgroups had no effects on the ␣IIb backbone structure, as evidenced by the invariance of 13 C ␣ secondary shifts ( Fig. 2A), which correlate with backbone torsion angles (35). The newly introduced lipids were sensed predominantly by ␣IIb residues on the extracellular face (Fig. 2, B and C). Thus, the detected ␣IIb structure is stable in lipid environments that resemble the composition of its native platelet membrane (36).
The ␣IIb TM structure was calculated from an extensive set of bond vector orientation, internuclear distance, and torsion angle constraints (supplemental Tables 1 and 2). For the ensemble of 20 calculated simulated annealing structures (Fig.  3A), coordinate precisions of 0.22 Å were obtained for the backbone-heavy atoms of the structured ␣IIb residues (Ile-966 -Lys-994). The structure is characterized by a linear ␣-helix for Ile-966 -Lys-989 followed by a reversal of backbone direction that conversely immerses Phe-992-Phe-993 toward the lipid hydrocarbon core of the intracellular membrane face (Fig. 3, D  and E). By exhibiting increased protection from the paramagnetic Mn 2ϩ EDDA 2Ϫ agent present in the aqueous phase as well as from H N -H 2 O exchange, the increased Phe-992-Phe-993 immersion was directly detectable within the bicelle membrane model (Fig. 4, B and C). Phe-993 packs against the TM helix, as evidenced by close interproton distances between its aromatic ring and nuclei from Trp-988, Leu-985, and Val-984 and its well defined 1 rotameric state (supplemental Table 2). The packing of Phe-992 is characterized by hydrophobic contacts to Val-990 FIGURE 2. Effect of bicelle lipid composition on ␣IIb chemical shifts. The substitution of one-third of bicellar POPC lipids with 1-palmitoyl-2-oleoyl-snglycero-3-[phospho-L-serine] (POPS) lipids has no effects on the 13 C ␣ secondary shifts (A) and, thus, ␣IIb backbone structure. Secondary 13 C ␣ chemical shifts, defined as the difference between the observed and tabulated random-coil 13 C ␣ shift of a residue, correlate with the underlying backbone conformation (35). Pro-965 and Pro-998 limit the backbone conformation of their preceding residues to extended conformations (42), as evidenced by their large negative 13 C ␣ secondary shifts. 13  and Met-987; however, overall packing appears less defined, leading to an averaged 1 rotameric state (supplemental Table  2). C-terminal to Lys-994, fast time scale backbone dynamics increase significantly (Fig. 4A), leading to dynamically unstructured conformations, with Arg-997 forced into extended backbone conformations by Pro-998 ( Fig. 2A). Incidentally, the extracellular side of the TM segment is also strongly affected by a proline residue (Fig. 3, B and C); Pro-965 restricts Ile-964 to extended conformations, as evidenced by its large negative 13 C ␣ secondary shift ( Fig. 2A), which may benefit dynamic interactions with the hydrophobic lipid tails (Fig. 3, B and C). Backbone dynamics increase steadily from the start of the TM helix at Ile-966 to the N terminus (Fig. 4A).
Structural similarities to the integrin ␣IIb TM segment were found in the Protein Data Bank for a ␤ peptide (1 nkz, chain F) of the light-harvesting complex II from Rhodopseudomonas acidophila (37) (supplemental Fig. 5A). Somewhat more remotely related is a TM helix of cytochrome c oxidase from Thermus thermophilus (1xme, chain A) (38) (supplemental Fig. 5B).

Membrane Embedding and Relationship to Other Integrin ␣
Subunits-Sequence alignment of all 18 human integrin ␣ subunits ( Fig. 1 and supplemental Fig. 4) suggests that the ␣IIb extra-and intracellular transitions to the TM helix are characteristic of ␣ subunits. At the extracellular helix border, half of all ␣ subunits exhibit a proline residue preceded by one that is hydrophobic. The Gly-Phe-Phe motif on the intracellular side is the only fully conserved tri-residue sequence element among integrin ␣ TM segments, suggesting that the observed backbone reversal and Phe-Phe membrane immersion are general structural features of integrin ␣ subunits. These transitions are fundamentally different from the ␤3 TM segment, where irregular structure and helix fraying characterized the transitions to the extra-and intracellular domains, respectively (9). Moreover, in contrast to the 29-residue ␤3 TM helix, the 5-residueshorter ␣IIb TM helix does not require any tilt within a typical lipid bilayer (Fig. 5). The structural complexity of the Gly-991-Phe-993 motif and ensuing complicated Mn 2ϩ EDDA 2Ϫ protection make it difficult to determine the TM helix center  (Fig. 4A), and their depicted conformation is but one of many possible conformations. D and E, views of the intracellular face of the energy-minimized, average structure. Residues C-terminal to Lys-994 become dynamically unstructured. To clearly depict the side-chain orientations, helix periodicity is not maintained between the extra-and intracellular views.
within the bicelle (Fig. 4B); in the absence of this information, the helix is assumed to be centered in the membrane (Fig. 5).
The lipid tail-to-headgroup transition on the extracellular side is estimated to take place before the TM helix commences, i.e. between Pro-965-Ile-966 (Fig. 1). In this region, protection from paramagnetic Mn 2ϩ EDDA 2Ϫ is significantly reduced (Fig.  4B). This transition is also in agreement with the Pro-998 -Leu-999 transition detected for the ␣5 subunit (39), which is similar to ␣IIb (Fig. 1). It is also noted that, at the extracellular helix border, no sharp hydrophobic-hydrophilic transition takes place (Fig. 1), which may result in a border variation between ␣ subunits (39). The membrane borders depicted in Fig. 1 should therefore be considered as minimal for monomeric ␣ subunits. On the intracellular side, the ␣IIb TM segment is expected to leave the lipid core after Phe-993 at the earliest. Lys-994 dynamics and protection are still somewhat high (Fig. 4), but, in conjunction with glycosylation mapping studies of integrin ␣2 and ␣5 subunits (40), the membrane border is estimated to lie between Phe-993-Lys-994. By virtue of a higher sequence homology on the intracellular TM face ( Fig. 1 and supplemental Fig. 4), this transition appears to be more uniform among integrin ␣ subunits. Finally, it should be noted that, on the basis of the Mn 2ϩ EDDA 2Ϫ protection data of the ␣IIb and ␤3 TM segments (9) (Fig. 4B), immersion of the ␣IIb TM helix in the membrane is nevertheless slightly deeper than is the case for the ␤3 TM helix, as depicted in Fig. 5.
A stretch of continuous ␣-helix for residues Val-990 -Phe-993 appears less preferable than the observed conformation for three reasons. First, the two Phe-992-Phe-993 aromatic rings prefer to immerse in the hydrophobic lipid tail environment instead of in the electrostatically complex headgroup interface (41). Second, any TM helix extension would necessitate a helix tilt, which will require Phe-992-Phe-993 to reposition further outward into the headgroup region because they would come to lie on the same helix side as snorkeling Lys-989. In contrast, the ␤3 subunit has Ile-719 and Thr-720 following Lys-716 as the third and fourth residues (supplemental Fig. 4). Third, Gly-991 may destabilize helical conformation. Thus, as a paradigm for integrin ␣ and ␤ subunits, the four residues following ␣IIb(Lys-989) dispose its TM helix toward a short helix and the absence of a tilt, whereas the five residues following ␤3(Lys-716) create a bias for a long TM helix and a substantial tilt (Fig. 5).
The use of the Gly-Phe-Phe sequence motif is remarkable since, for example, the introduction of a second charged residue after ␣IIb(Lys-989) would equally well cause a short ␣IIb TM helix (39,40). It appears likely that the highly conserved Gly-Phe-Phe motif has additional functional meaning.
Implications for Integrin ␣IIb ␤3 Association-Integrin bidirectional signaling proceeds with changes in ␣-␤ TM helix association (2, 4 -8). The individual ␣IIb and ␤3 TM structures, which correspond to activated integrin conformations (2)(3)(4), therefore lay the foundation for evaluating the extent of structural and topological rearrangements upon ␣IIb-␤3 association, i.e. integrin TM signaling. Moreover, the structures show that, in the activated receptor, the ␣-␤ extracellular domains connect to their respective transmembrane segments at different membrane crossing angles (Fig. 5). Alanine substitution of Phe-992 or Phe-993 leads to receptor activation (5), showing that interactions of these residues with the ␤3 subunit are required for maintaining the inactive state. By either interacting with the ␤3 subunit in the conformation of the monomeric ␣IIb TM segment or shifting Val-990 -Phe-993 to another, for example helical, conformation, the Gly-991-Phe-993 structural motif is central to determining the ␣IIb-␤3 relative orientation and topology not only in the activated but also in the resting receptor states (Fig. 5). Interestingly, in the reported cytoplasmic ␣IIb-␤3 complex structure, which extends up to ␣IIb(Lys-989), Phe-992-Arg-997 are in helical conformation (7). Cysteine scanning mutagenesis of resting integrin ␣IIb-␤3 has identified close interhelical contacts (2), whereas leucine scanning has revealed additional dimerizationsensitive residues (8). These interhelical contacts are mainly clustered near the extracellular helix sides, and the association of the individual ␣IIb and ␤3 structures appears able to fulfill these contacts (Fig. 5). Likewise, the salt bridge detected between ␣IIb(Arg-995) and ␤3(Asp-723) (5) is compatible with such a structure, although this is less significant since ␣IIb(Arg-995) is highly dynamic in the monomeric ␣IIb TM structure (Fig. 3). Clearly, additional structural information is required to define the associated ␣IIb-␤3 structure. In conclusion, the unusual Gly-991-Phe-993 structural motif of ␣IIb, which appears prototypical for integrin ␣ subunits, reveals a high degree of structural complexity at the TM level and lays the foundation for understanding the transition from activated to resting receptor states.