A Conserved Ectodomain-Transmembrane Domain Linker Motif Tunes the Allosteric Regulation of Cell Surface Receptors

In many families of cell surface receptors, a single transmembrane (TM) α-helix separates ecto- and cytosolic domains. A defined coupling of ecto- and TM domains must be essential to allosteric receptor regulation but remains little understood. Here, we characterize the linker structure, dynamics, and resulting ecto-TM domain coupling of integrin αIIb in model constructs and relate it to other integrin α subunits by mutagenesis. Cellular integrin activation assays subsequently validate the findings in intact receptors. Our results indicate a flexible yet carefully tuned ecto-TM coupling that modulates the signaling threshold of integrin receptors. Interestingly, a proline at the N-terminal TM helix border, termed NBP, is critical to linker flexibility in integrins. NBP is further predicted in 21% of human single-pass TM proteins and validated in cytokine receptors by the TM domain structure of the cytokine receptor common subunit β and its P441A-substituted variant. Thus, NBP is a conserved uncoupling motif of the ecto-TM domain transition and the degree of ecto-TM domain coupling represents an important parameter in the allosteric regulation of diverse cell surface receptors.

In many families of cell surface receptors, a single transmembrane (TM) α-helix separates ecto-and cytosolic domains. A defined coupling of ecto-and TM domains must be essential to allosteric receptor regulation but remains little understood. Here, we characterize the linker structure, dynamics and resulting ecto-TM domain coupling of integrin αIIb in model constructs and relate it to other integrin α subunits by mutagenesis. Cellular integrin activation assays subsequently validate the findings in intact receptors. Our results indicate a flexible yet carefully tuned ecto-TM coupling that sets the signaling threshold of integrin receptors. Interestingly, a proline at the N-terminal TM helix border, termed NBP, is critical to linker flexibility in integrins. NBP is further predicted in 21% of human singlepass TM proteins and validated in cytokine receptors by the TM domain structure of the cytokine receptor common subunit β and its P441A-substituted variant. Thus, NBP is a conserved uncoupling motif of the ecto-TM domain transition and the degree of ecto-TM domain coupling represents an important parameter in the allosteric regulation of diverse cell surface receptors.

Cells
sense their environment through transmembrane (TM) surface receptors that transmit extracellular signals into the cell. With the exception of G-protein coupled receptors, these proteins are dimers composed of subunits that each typically exhibits an extracellular domain (ectodomain) containing the ligand-binding site, a single TM α-helix and an intracellular effector domain. Integrins, which control vital cell-cell and cell-matrix adhesions (1)(2)(3), constitute one ubiquitous family of TM cell surface receptors. Receptor tyrosine kinases (RTK) represent another prominent family that activate upon the stabilization of dimeric receptor states by bound ligand (4,5). However, rather than being simple binary switches, all these receptors have the potential for allosteric regulation (1,5,6), indicating a multi-state coupling of ecto-and TM domains. This coupling and associated allosteric parameters must depend on the structural and dynamic properties of the linker between ecto-and TM domains. However, irrespective of detailed structural information on ecto-and TM domain in RTK (4,5), structural or dynamic information on linkers remains indirect and ambiguous (7)(8)(9). Similarly, structures of integrin ectodomains, TM domains and cytosolic tails are available (10)(11)(12)(13)(14)(15)(16) but little is known about their ecto-TM domain coupling. In support of a functionally relevant linker, integrin αMβ2 spontaneously activates when the TM domain is uncoupled from the ectodomain by insertion of a flexible (GGGGS)2 linker (17).
Integrins consist of heterodimeric, noncovalently associated αβ subunits that each exhibits a large ectodomain, a single TM α-helix and a short cytosolic tail (Fig. 1A). In addition to outside-in signaling, integrins also signal in the opposite direction (inside-out signaling) in relation to their function as dynamic cell adhesion molecules (1)(2)(3). This bi-directional, multi-state signaling ability renders integrins well suited to study the structural basis of ecto-TM coupling and its impact on allosteric receptor regulation. The framework of allosteric integrin regulation is established by a coupling between the structural state of αβ ectodomains and the association state of αβ TM domains (12,18). The assembled TM complex correlates with the low-affinity, bent conformation of the ectodomains and the absence of signaling (Fig. 1A). During inside-out signaling, binding of an agonist to the cytosolic β tail disrupts the TM complex (19), thereupon destabilizing the interface of subdomains α(Calf2)-β(I-EGF4/β-tail) within the ectodomains (20,21), and permitting ligand binding by the rearranged, high-affinity ectodomains (Fig. 1A). When an extracellular agonist is able to spontaneously bind the ectodomains and dissociate the α(Calf2)-β(I-EGF4/β-tail) interface, the destabilized TM complex can dissociate giving rise to outside-in signaling (21,22).
In the present study, we use NMR spectroscopy to characterize the linker structure, dynamics and ecto-TM domain coupling of integrin αIIb in model constructs, evaluate its sequence determinants in relation to other integrin α subunits, and assess its effects on receptor activation in cellular integrin activation assays. We further compare the sequence and structural features of the ecto-TM domain coupling of integrin αIIb with other human TM cell surface receptors to achieve general insight into the principles of TM cell surface receptor signaling.

Results and Discussion
Thermodynamic Description of Ectodomain-TM Domain Coupling-To obtain a simple model of integrin activation, we decompose the free energy required to activate the receptor into two terms. Specifically, the threshold of bi-directional TM signaling relates to the free energy difference between the dissociated and associated TM complex, termed ΔG°TM, and between the inactive and active ectodomain conformations, termed ΔG°E (Fig. 1A). However, the physical linkage between the TM and ectodomains must determine their allosteric coupling and ultimately set signaling thresholds. To quantify the mutual stabilization of the TM complex and the ectodomain (12), we define a coupling factor, f, with 0 ≤ f ≤ 1. To activate the ectodomain via inside-out (IO) signaling, talin binding has to provide ΔG°IO = ΔG°TM + f⋅ΔG°E (Fig. 1C). Conversely, to disrupt the TM complex and generate an outside-in (OI) signal, an extracellular agonist must provide ΔG°OI = f⋅ΔG°TM + ΔG°E. Moreover, f sets the minimum affinity of an extracellular ligand to bind to an inside-out stimulated receptor, namely (1-f)⋅ΔG°E, and analogously (1-f)⋅ΔG°TM for an intracellular ligand in an outside-in stimulated receptor (Fig. 1C).
For the prevalent integrin β1, β2 and β3 subunits, the ecto-TM domain linker sequence is conserved and consists of only 5 residues (Fig.  1B). With integrin α subunits on the other hand, longer linker lengths and higher sequence variability were observed (Fig. 1B), suggesting that modulation of f in the integrin family takes place through the properties of α subunits. Accordingly, our study focused on integrin α subunits, in particular αIIb that exhibits an 8residue ecto-TM domain linker (Fig. 1B) and combines with β3 to form the integrin αIIbβ3 fibrinogen receptor.

A Flexible Linker Loosely Couples the Ecto-and
TM Domains in Integrin αIIb-Crystallographic studies of inactive ectodomains provided precise borders of the membrane-proximal αIIb(Calf2) and β3(β-tail) domains but were unable to obtain structural information on linkers (11,23,24). NMR studies of the αIIb and β3 TM domains provided borders for these segments (12,25,26), but without the representation of preceding αIIb(Calf2) and β3(β-tail) domains, linker properties are invariably misrepresented. Calf2 has an immunoglobulinlike, β-sandwich fold with longer and more abundant β-sheets compared to typical Ig-like domains (Fig. 1D). The elongated shape of Calf2 means that most of these secondary structures are far from the membrane (Fig. 1A) and, thus, cannot influence the linker structure. Likewise, crystal structures of Calf2 verify its folding in the absence of linker (11,23). It is therefore possible to approximate Calf2 with a smaller, highly stable domain that is structurally homologous to the membrane-proximal structure of Calf2 in order to facilitate the solution NMR-based characterization of linker properties. The third IgG-binding domain of protein G, termed GB3, ensures a transition from the terminal β-sheet to the first linker residue that is similar to Calf2 (Fig. 1D). Next to the close congruence of backbone conformations, sidechain conformations matched between Calf2 and GB3 for the final three residues preceding the linker. We note that the terminal GB3 residue E56 is hydrogen bonded to K10, providing it a stable backbone structure despite its C-terminal position (27). Thus, while all structural aspects pertaining to the study of ΔG°E would require an intact ectodomain, a meaningful characterization of f is achievable within a construct that fuses GB3 to the αIIb(linker-TM) sequence ( Fig. 2A).
The GB3-linker-TM construct was reconstituted in phospholipid bicelles ( Fig. 2A-B), a well-established membrane mimic for integrins (15,26,28). To control for any size-dependent effect of the bicelle bilayer area on protein structural properties, bicelles with different shortto-long chain lipid ratios (q=0.5 and 0.3) were examined. Backbone chemical shift assignments were carried out in both environments and the obtained secondary 13 C α shifts allowed for the straightforward assessment of secondary structure (29). Identical 13 C α shifts in q=0.5 and 0.3 bicelles (Fig. 2C) showed the absence of bicelle sizedependent effects, allowing us to proceed with q=0.3 bicelles that afford more sensitive measurements. In further validation of the GB3linker-TM construct, the GB3 structure remained intact in proximity to the bicelle and the TM structure of αIIb was unchanged by the presence of GB3 (Fig. 3A). The 8-residue linker exhibited random coil properties with the notable exception of Ile964, which was coerced into extended conformations by the sidechain constraints of succeeding Pro965 (30). In the absence of the GB3 domain, the linker exhibited helical propensity (positive secondary 13 C α shifts; Fig. 3A), confirming the requirement of a Calf2 domain representation.
To gain further insight into linker properties, linker backbone dynamics on the picoto nanosecond timescale were analyzed relative to the TM domain by interpreting 15 N relaxation parameter in terms of the general order parameter S 2 (31). This parameter describes the spatial fluctuation of the N-H bond vector and is limited by 0 ≤ S 2 ≤ 1. In case of S 2 →1, relaxation is solely described by the global motion of the protein, whereas for S 2 →0 local motions fully describe 15 N relaxation. In comparison to the well-folded TM region (S 2 ≈1), the elevated linker dynamics confirmed its random-coil nature (Fig. 3C). To express the rigidity of the GB3 domain-TM domain coupling with a single parameter, the average linker S 2 value, termed <S 2 >linker, was used. <S 2 >linker, not being directly related to the thermodynamically defined f, was 0.35 ± 0.01. Evidently, a dynamically unstructured linker uncouples the ecto-and TM domain of the integrin αIIb subunit to a substantial degree.

The C-terminal Linker Proline Governs Ecto-TM Domain Coupling-The most conspicuous linker
property is the effect of the C-terminal linker proline on its preceding residue (Fig. 3A). This proline also represents the most conserved linker residues; 11 out of 18 human integrin α subunits contain a proline at their putative TM domain border (Fig. 1B). We hypothesized this proline to be a key determinant of the linker-TM transition. To ascertain the role of αIIb(Pro965) on ecto-TM domain linker properties, we have substituted it with Glu, the corresponding residue found in the integrin α3 subunit (Fig. 1B), and expanded our NMR study to the GB3-linker(P965E)-TM construct. The glutamate substitution of the linkerterminal proline had profound effects; it led to the propagation of helical propensity from the TM helix into the linker (Fig. 3B). The P965E substitution concomitantly diminished backbone dynamics ( Fig. 3C) and, with <S 2 >linker = 0.43 ± 0.02, enhanced GB3 domain-TM domain coupling, demonstrating fwt < f α IIb(P965E). The tighter domain-domain coupling was further reflected by an increase in isotropic rotational correlation times of GB3 and TM domains upon substituting Pro965 (Table 1).
To examine any effects on ΔG°TM as a result of αIIb(P965E), isothermal titration calorimetry was applied to determine the TM complex stability of αIIb(P965E)β3. To broaden our study, we further examined the alanine substitution of αIIb(Pro965), which is found in the integrin αD subunit (Fig. 1B). The secondary 13 C α shifts of linker(P965A)-TM and linker(P965E)-TM peptides showed an even more helical linker in P965A compared to P965E (Fig. 3D). Relative to ΔG°TM = -4.84 ± 0.01 kcal/mol of wild type, TM complex stabilities of αIIb(P965E)β3 and αIIb(P965A)β3 were reduced by 0.11 ± 0.01 kcal/mol and 0.09 ± 0.01 kcal/mol, respectively ( Table 2). These small differences indicate that any functional effect of P965 substitutions will be dominated by changes in f and not ΔG°TM. In conclusion, by uncoupling the linker from the TM domain conformation, the C-terminal linker proline is a pivotal determinant of linker properties and ecto-TM domain coupling.

The N-terminal TM Border Proline Modulates the
Efficiency of Integrin αIIbβ3 Activation-In adopting a membrane-centric view, the C-terminal linker proline is referred to as the N-terminal TM border proline (NBP). To determine the functional significance of the NBP residue and the factor f, we compared the efficiency of talin-induced activation of full-length integrin for αIIbβ3, αIIb(P965A)β3 and αIIb(P965E)β3. Furthermore, we assayed the variant αIIb(E961G/A963G), which resembles the glycine content and distribution of the integrin α5 linker (Fig. 1B). Because of the intrinsic flexibility of glycine, this variant served as negative control, i.e., f α IIb(E961G/A963G) < fwt < f α IIb(P965X). Talin binds the cytosolic tail of the β3 subunit (19), which dissociates the αIIbβ3 TM complex (32,33) and destabilizes the resting ectodomain to allow ligand binding (Fig. 4A). An increasing degree of ecto-TM domain coupling (f value) increasingly aligns the dissociated TM complex with the active ectodomain conformation. In the regime where linker mutations do not spontaneously cause receptor activation, we expect increasing values of f to result in higher saturating talin concentrations to compensate for a more favorable, decreased ΔG°IO (Fig. 4A). Concomitantly, with increasing f the threshold for ligand binding is lowered (Fig.   4A) and higher levels of active receptor as judged by successful ligand binding are expected.
We used a cellular assay that correlates the concentration of the talin head domain (THD) with the levels of active integrin αIIbβ3 receptors in the plasma membrane of CHO cells (Fig. 4B). Specifically, a maximal activation index, termed PAC1max, was used to quantify the levels of PAC1 ligand-binding competent receptors at saturating THD concentration. The [THD] at which Bmax/2 was reached is termed EC50. EC50 values carried uncertainties of approximately 20%; as such, they did not allow the differentiation of linker variants in terms of EC50 (Fig. 4C). However, PAC1max values unambiguously increased with increasing values of f as expected. We further note that PAC1max was larger for αIIb(P965A) than αIIb(P965E) that is in line with the higher helical propensity of the linker for αIIb(P965A) than αIIb(P965E). i.e., f α IIb(P965E) < f α IIb(P965A) (Fig. 3D). Accordingly, we demonstrate in intact receptors (Fig. 1A) that the ecto-TM domain linker properties (f factor) govern the allosteric properties of integrin αIIbβ3 and establish NBP to be a central determinant of such properties.

An N-terminal Proline Frequently Borders the TM Helix of Bitopic Membrane Proteins-
The functional relevance and sequence prevalence of NBP in integrin α subunits suggested that it may be a recurring motif in bitopic membrane proteins. Proline is a well-established N-terminal helix cap. Its propensity to populate the helix-preceding residue (Ncap position) and the first helix residue (N1 position) is documented for membrane proteins (34)(35)(36). However, no systematic study regarding the prevalence of NBP in bitopic and polytopic membrane proteins is available. We therefore searched the human genome for proteins with a single-pass TM helix and a preceding sequence of at least 90 residues to allow for the presence of a soluble domain and intervening linker. In defining the NBP motif, we accepted both Ncap and N1 positions as both sites serve to abrogate helical propensity in the linker Cterminus. NBP residue predictions were found in 20.9% of such proteins (Table 3 and Supplemental  Table 1). Next to integrin α subunits, NBP is common in RTK, immunological and cytokine receptors (Table 3), revealing a wide relevance of NBP in TM cell surface receptors. For α chains of MHC class I molecules and some killer cell immunoglobulin-like receptors TM sequences were highly homologous (Supplemental Table 1). When considering only one representative from these families, the NBP frequency was 17.4%. In contrast, in polytopic membrane proteins in the human genome, the NBP prediction frequency was only 12.4% per TM helix. This difference in frequencies is statistically significant (P<0.001; see Experimental Procedures). It appears that bitopic membrane proteins overproportionally benefit from the NBP-conferred sharp separation of linker and TM helix conformations in accordance with an additional functional role of NBP.

The Structure of the Cytokine Receptor Common
Subunit β Validates its NBP Residue-To verify the presence of NBP in another prevalent TM cell surface receptor family, we examined our NBP prediction for Pro441 of the cytokine receptor common subunit β (βc; Table 3). The βc subunit partakes in the heterodimeric assembly of Granulocyte-macrophage colony-stimulating factor, interleukin-3 and interleukin-5 receptors (37). Similar to Calf2 of αIIb, the βc ectodomain concludes with an IgG-like domain and transitions into the putative six-residue T436-ESVL-P441 linker from a terminal β-sheet. To define the role of Pro441, we have determined the TM domain structure of βc and βc(P441A) including flanking residues in phospholipid bicelles by multidimensional, heteronuclear NMR spectroscopy. Backbone and sidechain torsion angle restraints in combination with H-N, C α -C′ and C′-N bond vector restraints defined the backbone heavy atoms to precisions of 0.42 and 0.19 Å for βc and βc(P441A), respectively ( Table  4). The TM helix of wild-type βc encompassed Met442-Tyr466 (Fig. 5A), i.e., 25 residues, which signifies a small tilt in the membrane. On the Cterminal side, a glycine-Schellmann motif terminates the helix whereas on the N-terminal side, Pro441 indeed caps the helix. The residue preceding Pro441 was dynamically unstructured at random-coil conformations (Fig. 5B), thereby uncoupling linker and TM helix structures.
The structure of βc(P441A) showed a helix extension that folds the entire linker stretch (TESVLA) into helical conformation (Fig. 5A), resulting in a relatively long helix of 31 residues (Thr436-Tyr466) in the absence of the ectodomain. However, lipid interactions remained virtually unchanged between βc and βc(P441A) (Fig. 5C). The helix extension is therefore not lipid immersed but remained in the aqueous milieu. Interestingly, the βc(P441A) helix is not fraying at the N-terminus but rather starts at Thr436 without significantly affecting preceding residues (Fig.  5B). As a result, without NBP, the ecto-and TM domain of βc would be tightly coupled (Fig. 5A).
Conclusions-The linker connecting the ecto-and TM domains of integrin αIIb is flexible. Even in the absence of NBP, the linker must be considered flexible relative to folded secondary structure (Fig.  1C). However, the degree of linker flexibility in integrin α subunits appears to be carefully controlled; 39% of subunits lack NBP, subunits exhibit conspicuous and varying numbers of glycine and proline, and variations in linker lengths (Fig. 1B). On structural and functional levels, such variations adjust the degree of ecto-TM domain coupling (Fig. 3C) and control receptor activation thresholds (Fig. 4C). Correspondingly, flexibility is key to allowing allostery in integrin signaling (Fig. 1C) and the same result is expected for other TM cell surface receptors. Our assertion is supported by a 21% incidence of NBP in human single-pass TM proteins, a statistically significant increase of NBP in single-over multi-pass TM helices, and the requirement of NBP to obtain a short, dynamic linker in the cytokine receptor βc subunit (Fig. 5). Accordingly, while NBP supports TM helix initiation and TM helix-helix loop formation in general (34)(35)(36), the abundance and uncoupling effect of NBP in TM cell surface receptors indicate that the degree of ecto-TM domain coupling plays an important role in the function and signaling mechanism of these proteins.

Experimental Procedures
Integrin αIIb and βc TM Peptide Production and NMR Sample Preparation-GB3-linker-TM constructs of integrin αIIb were derived from the previously introduced pET44-GB3-TEV-linker-TM vector (25). An XhoI site, which codes for Leu-Glu, was introduced into the linker (Fig. 1B), and subsequently GB3 with a T55L substitution was subcloned into this vector via this site and a 5' NdeI site. This procedures eliminated the tobacco etch virus (TEV) protease cleavage site as to express His-tagged GB3(T55L)-αIIb(Ala958-Pro998).

NMR Spectroscopy, Calculation of General Order parameter S 2 , and Structure Calculation-NMR data was acquired on a cryoprobe-equipped Bruker
Avance 700 spectrometer at 40 °C unless otherwise stated. HNCA, HNCACB and HNCO experiments were performed to achieve backbone assignments. HNCO-based experiments were employed for the measurement of 3 JC ′ C γ and 3 JNC γ couplings (40), and the detection of 1 JC α C ′ and 1 JC ′ N as well as 1 JC α C ′ + 1 DC α C ′ and 1 JC ′ N+ 1 DC ′ N couplings (41,42). 1 JNH and 1 JNH+ 1 DNH couplings were measured using the ARTSY scheme (43). To determine H N -H N and H N -H O NOEs, a 15 N-edited NOESY spectrum was recorded (150 ms mixing time). TROSY-type H-N detection (44) was used throughout all experiments.
In reference to previous measurement conditions (28), 15 N longitudinal and transverse relaxation rates R1 and R2, respectively, and { 1 H}- 15 N NOEs were measured at 35 °C using TROSYbased pulse sequences (45). Specifically, { 1 H}-15 N NOE measurements were performed in an interleaved manner with 5 s of presaturation preceded by a recycling delay of 4 s for the NOE experiment and by a 9 s recycle delay for the reference experiment. R2 measurements were carried out with a CPMG delay of 400 μs. Data were processed and analyzed with the NMRPipe package (46). For R1 and R2 quantifications, peak intensities were fit to exponential decay and uncertainties estimated by Monte Carlo simulations as described in the literature (47). The error in { 1 H}-15 N NOE values was estimated by assuming that the uncertainty in the peak heights in the two interleaved spectra equals the r.m.s. noise in each spectrum. 15 N relaxation parameter were analyzed with the program Tensor2 (48). An isotropic rotational diffusion tensor adequately described the relaxation data, as verified by using tensors of different symmetries to compare experimental and back-calculated relaxation parameter using χ 2 statistics. To appropriately fit the general order parameter S 2 and to determine its uncertainty, Monte Carlo simulations and Fstatistics were used (48).
The structures of the bicelle-embedded TM segments of βc and βc(P441A) were calculated by simulated annealing, starting at 3000 K using the program XPLOR-NIH (49). Backbone dihedral angle constraints were obtained from the pattern of N, H α C α , C β and C′ chemical shifts (50). 3 JC ′ C γ and 3 JNC γ coupling constants instructed χ 1 side-chain angle restraints. Aside from standard force field terms for covalent geometry (bonds, angles, and improper dihedrals) and nonbonded contacts (Van der Waals repulsion), dihedral angle and interproton distance restraints were implemented using quadratic square-well potentials. In addition, a backbone-backbone hydrogen-bonding potential and torsion angle potential of mean force were employed (51,52). A quadratic harmonic potential was used to minimize the difference between predicted and experimental H-N, C α -C′ and C′-N RDCs (Δ 1 D). The final values for the force constants of the different terms in the simulated annealing target function were as previously described (12). Table 4 reports the structural statistics for the total of 20 structures calculated for βc and βc(P441A). Ensemble coordinates for βc and βc(P441A) together with energy-minimized average structures and structural constraints have been deposited in the Protein Data Bank and Biological Magnetic Resonance Bank as entries 2na8 / 25931 and 2na9 / 25932, respectively.

Talin-dependent Integrin αIIbβ3 Activation-
CHO-K1 cells were transfected with 2.5 μg of αIIb cDNA and 2.5 μg of β3 cDNA in conjunction with either 1 μg of GFP talin head domain (GFP-THD) or 0.1 μg of GFP as control. Cells were cultured for 24 hours after transfection, trypsinized, washed and co-stained with anti-αIIbβ3 antibody D57 detecting αIIbβ3 expression in the Phycoerythrin channel and with the ligandmimetic antibody PAC1 (53,54) reporting αIIbβ3 activation in the alexa647 channel. The inability of PAC1 to activate integrin αIIbβ3 is noted (53,54). PE-positive, i.e. αIIbβ3-expressing, cells were gated and further divided into 12 regions according to the expression level of talin head domain in these cells. The geometric mean fluorescence intensity (MFI) of GFP-THD were calculated from each population and plotted against the MFI of PAC1 (Fig. 4B). PAC1max and EC50 values were extracted by non-linear curve fitting.

Isothermal
Titration Calorimetry-ITC measurements of the peptides listed in Table 2 were carried on a Microcal VP-ITC calorimeter. 10 μM of β3 peptide in the 1.425 ml sample cell was titrated with αIIb peptide by injecting 9 μl aliquots over a period of 10 s each. Measurements were carried out in 43 mM DHPC, 17 mM POPC, and 25 mM NaH2PO4/Na2HPO4 pH 7.4 at 28 °C. Prior to data analysis, the measurements were corrected for the heat of dilutions of the αIIb and β3 peptides. The αIIbβ3 TM complex stoichiometry was fixed at the experimentally verified 1:1 ratio (55) and the reaction enthalpy (ΔH o ) and KXY were calculated from the measured heat changes, δHi, as described (55).
Human Genome Analysis-Entries from the Uniprot KB database with location tag of "single pass", organism tag of "Homo sapiens (Human) [9606]" and "Reviewed" status were selected. To allow for the possibility of linker, domain and signal sequence to precede the TM helix, only entries with at least 90 residues preceding the annotated TM sequence were considered. 1597 entries fulfilled these criteria. Using the program TMHMM 2.0 (56), the annotated TM sequences were re-evaluated, which left 1557 entries as some TM annotations were discarded. When predicted TM borders were accurate, entries with Pro at positions −1 (Ncap) and +1 (N1) were sought. However, inspection of the predictions for integrin subunits and other entries suggested a maximal Nterminal TM border accuracy of ±1 residues. We therefore accepted Pro at positions −2, −1, +1, +2 but discarded entries in which a negatively charged residue succeeded proline (e.g. integrin β3; Fig. 1B) as this residue is likely to constitute the Ncap position. 326 entries fulfilled these criteria (Supplemental Table 1). For reference, Pro at positions −1, +1 produced 213 entries (13.7%). To compare single-to multi-pass entries, analogous Uniprot entries with location tag "multi pass" were selected. Out of a total of 15889 predicted TM helices in these proteins, 1971 exhibited Pro at positions −2, −1, +1, +2.