Insight into pathological integrin αIIbβ3 activation from safeguarding the inactive state

Abstract

The inhibition of physiological activation pathways of the platelet adhesion receptor integrin αIIbβ3 may fail to prevent fatal thrombosis, suggesting that the receptor is at risk of activation by yet an unidentified pathway. Here, we report the discovery and characterization of a structural motif that safeguards the receptor by selectively destabilizing its inactive state. At the extracellular membrane border, an overpacked αIIb(W968)-β3(I693) contact prevents αIIb(Gly972) from optimally assembling the αIIbβ3 transmembrane complex, which maintains the inactive state. This destabilization of approximately 1.0 kcal/mol could be mitigated by hydrodynamic forces but not physiological agonists, thereby identifying hydrodynamic forces as pathological activation stimulus. As reproductive life spans are not generally limited by cardiovascular disease, it appears that the evolution of the safeguard was driven by fatal, hydrodynamic force-mediated integrin αIIbβ3 activation in the healthy cardiovascular system. The triggering of the safeguard solely by pathological stimuli achieves an effective increase of the free energy barrier between inactive and active receptor states without incurring an increased risk of bleeding. Thus, integrin αIIbβ3 has evolved an effective way to protect receptor functional states that indicates the availability of a mechanical activation pathway when hydrodynamic forces exceed physiological margins.

Introduction

The cardiovascular system is the first organ to become functional during human embryogenesis and, pointedly, diseases of this system constitute the leading cause of human demise.¹ Death generally arises from unwarranted platelet aggregation and ensuing occlusive thrombus formation in the lumen of a vital blood vessel.^2,3 The adhesive properties of platelets correlate with the activation state of their dominant adhesion receptor, integrin αIIbβ3.^4,5 Inactive integrin αIIbβ3 is expressed to levels of approximately 80,000 copies per platelet and comprises 80% of total surface proteins.⁶ Physiologically, the receptor is activated from the inside (Figure 1a), rendering activation indirect and dependent on more than one activating signal.^4,5,7 Integrin αIIbβ3 has been postulated to exhibit the most tightly regulated switch between inactive and active states in the integrin family,⁸ which would make the free energy difference between inactive and active states, termed ΔG°_A (Figure 1a–b), largest among the 24 human integrins. These properties likely arose from considerable evolutionary pressure to prevent unwarranted integrin αIIbβ3 activation in relation to the immediate and systemic danger of pathological thrombosis. Nonetheless, the continuing high fatalities from seemingly unwarranted thrombosis¹ highlight that these properties do not eliminate pathological integrin αIIbβ3 activation, questioning whether pathological activation proceeds through physiological pathways.

Figure 1. Structural and sequence context of the integrin αIIbβ3 destabilizing motif.

(a) Integrin αIIbβ3 activation (inside-out signaling) requires the dissociation of αIIb-β3 TM complex by talin.⁹ This event removes the mutual stabilization of the TM complex and bent ectodomains,¹⁰ allowing their rearrangement to an extended open conformation that binds ligand with high affinity. Structural models are based on PDB entries 3fcs, 2k9j, 2vdl, 2k1a and 2rmz. (b) The safeguarding motif enlarges ΔG°_A by ΔG°_B for pathological but not physiological activation stimuli. The integrin αIIbβ3 activation state is identified by the association state of its TM complex, either αIIb·β3 or αIIb+β3. (c) Structure of the integrin αIIbβ3 TM complex (PDB entry 2k9j). The αIIb(W968)β3(I693) interaction localizes to the extracellular membrane border. OMC and IMC denote the outer and inner membrane clasps, respectively.¹⁰ To illustrate side chain geometries, the protein surface is shown at distances corresponding to less than the van der Waals radii of the underlying atoms. (d) Sequence alignment of αIIb and β3 TM sequences from diverse vertebrates. Amino acid numbering follows the human subunits. Conserved amino acids are colored by the Jalview multiple alignment editor⁶⁹ using the ClustalX color scheme.

The binding of talin to the cytosolic β3 tail constitutes the final step of integrin αIIbβ3 activation.⁹ This event dissociates the transmembrane (TM) complex that maintains the inactive state to destabilize the interface of αIIb(Calf2)-β3(I-EGF4/β-tail) subdomains within the ectodomains.^9–12 The loss of these intersubunit contacts and associated entropic stabilization allows ectodomain rearrangement to expose the high affinity ligand-binding site to bind multivalent fibrin with high affinity (Figure 1a), resulting in the cross-linking and aggregation of platelets. In a cardiovascular system that is compromised by disease, it appears that ΔG°_A is too small to prevent integrin αIIbβ3 activation. However, disfavoring thrombosis invariably promotes bleeding, which is equally life threatening. These opposing interests must have influenced the evolution of integrin αIIbβ3 and the value of ΔG°_A. Here, we report the discovery of a safeguarding motif that effectively enlarges ΔG°_A by buffering the inactive state against pathological stimuli but not physiological signals required to arrest bleeding. The nature of the motif identifies a general mechanism to safeguard receptor functional states and suggests an alternative, pathological pathway for integrin αIIbβ3 activation.

Results

The inactive state of integrin αIIbβ3 is intrinsically destabilized in most vertebrates

A receptor functional state may be buffered by incorporating a destabilizing motif in this state that is removed by pathological but not physiological stimuli. In other words, physiological stimuli must provide ΔG°_A to activate the receptor whereas pathological stimuli must supply ΔG°_A + ΔG°_B where ΔG°_B denotes the free energy difference between the presence and absence of the destabilizing motif in the buffered state (Figure 1b). Accordingly, we examined integrin αIIbβ3 for a destabilizing interaction, i.e., an interaction that would stabilize the inactive state when removed. To narrow down the structural region that may contain such a motif, we considered that ΔG°_A depends on the free energy differences between the inactive and active ectodomain conformations, termed ΔG°_E, and the dissociated and associated TM complex, termed ΔG°_TM (Figure 1a).¹³ For the endoglycosidase H-treated integrin α5β1 ectodomain, ΔG°_E was reported as −2.0 ± 0.2 kcal/mol.¹⁴ This value is smaller than ΔG°_TM measured for integrin αIIbβ3 of −4.84 ± 0.01 kcal/mol,¹⁵ suggesting that it would be most effective to introduce a destabilizing motif in the TM complex. Within the inner membrane clasp (IMC), the TM complex structure is optimized to receive activating signals from cytosolic proteins (Figure 1a,c)^10,16 and appears to be free of conformational constraints with all known point mutations destabilizing the TM complex.^17,18 In the outer membrane clasp (OMC), αIIb(Gly972), αIIb(Gly976) and β3(G708) mediate helix-helix contacts by “knobs-in-holes” packing (Figure 1c). All known point mutations of these contacts are again destabilizing.^17–19 Remaining as candidate for a destabilizing TM interaction is the first helix-helix contact at the extracellular membrane border, αIIb(W968)-β3(I693).

We assessed the contribution of αIIb(W968)-β3(I693) to TM complex stability in phospholipid bicelles by examining their mutation using isothermal titration calorimetry.^15,20 Ala and Val substitution of αIIb(W968) stabilized the TM complex by −0.90 ± 0.04 and −0.84 ± 0.03 kcal/mol, respectively (Table 1). Substituting the in relation to Trp smaller side chain of β3(I693) for Ala led to a complex stabilization of −0.65 ± 0.03 kcal/mol and dual I693A/W968V substitutions improved complex stability by −1.04 ± 0.02 kcal/mol (Table 1). Thus, integrin αIIbβ3 may employ a buffer of up to ΔG°_B = 1.04 ± 0.02 kcal/mol to safeguard its inactive state. The physiological significance of this added stabilization can be judged by comparing it to integrin the αIIbβ3(A711P) variant that rearranges the IMC to attain a TM complex stabilization of −0.82 ± 0.01 kcal/mol.^21,22 The β3(A711P) stabilization prevents integrin activation by talin,^23,24 indicating that the full removal of αIIb(W968)-β3(I693) strain could even block physiological integrin activation. In all 18 human α subunits, only αIIb possesses Trp at this position whereas other subunits utilize Val/Ile/Leu (Figure S1). Nonetheless, αIIb(W968)-β3(I693) is highly conserved across a diverse range of vertebrates (Figure 1d) identifying this interaction as important to the cardiovascular system of vertebrates.

Table 1.

Thermodynamic stability of mutant αIIbβ3 TM complexes

Peptides	KXYa	ΔH° [kcal/mol]	TΔS° [kcal/mol]	ΔG° [kcal/mol]	ΔΔG°b [kcal/mol]	ΔΔG°,′ c [kcal/mol]
αIIb + β3d	3250 ± 60	−16.0 ± 0.1	−11.1 ± 0.1	−4.84 ± 0.01	–	–
αIIb(W968V) + β3	13200 ± 600	−14.3 ± 0.2	−8.6 ± 0.2	−5.68 ± 0.03	−0.84 ± 0.03	–
αIIb(W968A) + β3	14000 ± 1000	−14.1 ± 0.3	−8.4 ± 0.3	−5.74 ± 0.04	−0.90 ± 0.04	–
αIIb + β3(I693A)	9500 ± 500	−13.9 ± 0.3	−8.4 ± 0.3	−5.49 ± 0.03	−0.65 ± 0.03	–
αIIb(W968V) + β3(I693A)	18600 ± 800	−16.0 ± 0.2	−10.2 ± 0.2	−5.88 ± 0.02	−1.04 ± 0.02	–
αIIb(P965A) + β3d	2790 ± 50	−12.9 ± 0.1	−8.1 ± 0.1	−4.75 ± 0.01	0.09 ± 0.01	–
αIIb(P965A/W968V) + β3	9700 ± 300	−14.6 ± 0.1	−9.1 ± 0.1	−5.50 ± 0.02	0.18 ± 0.04	−0.09 ± 0.04
αIIb(G972A) + β3	1080 ± 30	−14.2 ± 0.2	−10.1 ± 0.2	−4.18 ± 0.01	0.66 ± 0.01	–
αIIb(W968V/G972A) + β3	1570 ± 50	−14.4 ± 0.2	−10.0 ± 0.2	−4.40 ±0.02	1.28 ± 0.04	−0.62 ± 0.04
αIIb(G972S) + β3	810 ± 20	−14.8 ± 0.2	−10.8 ± 0.2	−4.01 ± 0.01	0.83 ± 0.01	–
αIIb(W968V/G972S) + β3	1300 ± 50	−12.5 ± 0.3	−8.2 ± 0.3	−4.29 ± 0.02	1.39 ± 0.04	−0.56 ± 0.04

^aMeasurements performed in 43 mM DHPC, 17 mM POPC, 25 mM NaH₂PO₄/Na₂HPO₄ pH 7.4 solution at 28 °C.

^bΔΔG° = ΔG°αIIbβ3,mutant −ΔG°αIIbβ3 or ΔG°αIIb(W968V)β3,mutant −ΔG°αIIb(W968V)β3 for double αIIb substitutions.

^cΔΔG°^,′ = (ΔG° αIIbβ3,mutant −ΔG°αIIbβ3) − (ΔG°αIIb(W968V)β3,mutant −ΔG°αIIb(W968V)β3)

^dValues taken from ref.^13,15.

Trp968 acts as a local wedge that prevents optimal αIIb(G972)-β3(L697) contacts

To understand the structural basis of αIIb(W968)-β3(I693) destabilization, we determined the αIIb(W968V)β3 TM complex structure by multidimensional, heteronuclear NMR spectroscopy (Table S1 and Figure S2). Based on the replacement of Trp with Val/Ile/Leu in all other human integrins (Figure S1), this structure is also suitable for modeling additional TM complex structures. We applied the structure calculation protocol introduced for the αIIbβ3(A711P) TM complex; it improved the number and quality of structural restraints over the wild-type TM complex structure by combining perdeuterated with protonated subunits.^10,21 To discuss relatively small structural changes, it was most appropriate to compare the TM complex structures of αIIb(W968V)β3 and αIIbβ3(A711P). Both mutations induce structural changes that do not overlap (Figure 2a).

Figure 2. Structure of the integrin αIIb(W968V)β3 TM complex.

(a) Chemical shift differences of β3 TM backbone ¹⁵N nuclei between non-covalent associations with αIIb and αIIb(W968V) TM peptides illustrate the propagation of structural changes in the β3 TM helix for the W968V substitution. For comparison, the corresponding αIIb shift differences between β3 and β3(A711P) associations are also shown. The panels align β3(G708) with αIIb(L980). (b) Structure of the αIIb(W968V)β3 TM complex in comparison to the αIIbβ3(A711P) variant (PDB entry 2n9y). The average structures were superimposed on the N^H/C^α/C′ coordinates of αIIb(698–704) and β3(973–979). (c) Comparison of the 3D NOESY-TROSY strips of β3(L697) of disulfide-linked αIIb(A963C)–²H/¹⁵N-β3(G690C/A711P) and disulfide-linked αIIb(A963CG/W968V)–²H/¹⁵N-β3(G690CG). NOEs to (protonated) lipids are indicated by green lines. Spectra were recorded at 40 °C and 700 MHz.

The smaller size of Val relative to Trp at position 968 allowed β3(Ile693) to adopt a relaxed side chain orientation that indicates the absence of steric clashes (Figure 2b). For the two N-terminal turns of the TM helices, interhelical distances became closer in the presence of αIIb(W968V) that allowed an overall subtle repacking of αIIb(G972)-β3(L697) (Figure 2b). Specifically, packing at the most favorable van der Waals distances now seems achieved to which adjustments in helical curvature and rise contributed as well (Figure 2b and Figure S2c). Next to close interproton distances (NOE signals) between αIIb(G972/H^N) and the side chain of β3(L697), such proximities were also detected between β3(L697/H^N) and αIIb(W968V) (Figure 2c). However, for αIIb(W968) substitutions to Val and the smaller Ala, improvements in TM complex stabilities were similar (Table 1), indicating that no favorable interaction arose directly from the presence of Val.

Further insight into repacking was obtained from β3 chemical shifts differences between αIIb and αIIb(W968V)-bound states. Chemical shifts represent highly sensitive probes of the structural environment of atom nuclei with backbone H^N nuclei particularly sensitive to hydrogen bonding and backbone ¹⁵N nuclei to side chain torsion angles.²⁵ Leu697 experienced the largest ¹⁵N shift change followed by Ile693 (Figure 2a) further supporting the repacking of both sites. At β3(Met701), which is the helix-helix contact subsequent to Leu697, shift changes were already much smaller and, after one additional turn, essentially neutral. In contrast, H^N shift changes were still prominent for Met701, showing that optimal repacking led to slightly longer-range adaptations of helix curvature (Figure S2c). Nonetheless, Trp968 indeed acted as a local wedge and overpacked its contact with β3(Ile693) and caused the underpacking of the αIIb(G972)-β3(L697) interaction.

The safeguarding motif encompasses the WxxxG sequence

To assess the contributions of αIIb(W968)-β3(I693) over- and αIIb(G972)-β3(L697) under-packing to W968V-conferred TM complex stabilization, we quantified the free energy difference between the effect of a mutation on αIIbβ3 and αIIb(W968V)β3 TM complex stabilities in bicelles, termed ΔΔG°^,′. First, we measured this parameter for αIIb(P965A) that is not part of the TM complex interface (Figure 1c). With ΔΔG°^,′= −0.09 ± 0.04 kcal/mol (Table 1), the role of Pro965 was essentially unchanged in both TM complex structures, making the observed value indicative of the measurement accuracy. For αIIb(G972A) and αIIb(G972S), ΔΔG°^,′ values were approximately −0.6 kcal/mol (Table 1). With a total αIIb(W968V)-mediated TM complex stabilization of −0.84 ± 0.03 kcal/mol, repacking of αIIb(G972) accounted for most of this improvement, whereas the removal of αIIb(W968)-β3(I693) strain was secondary in effect. This implies that Trp indeed wedged the αIIb-β3 TM helices apart but TM complex destabilization mainly results from Gly underpacking. The diminished buffering capacity ΔG°_B = 0.22 ± 0.02 kcal/mol with Ala in place of Gly972 (Table 1) further illustrates this view and represents the mere removal of αIIb(W968)-β3(I693) strain. As such, the stability gain obtained from removing αIIb(W968)-β3(I693) overpacking largely arose from alleviating αIIb(G972) underpacking, which identifies the WxxxG sequence as hallmark of the integrin αIIbβ3 safeguarding motif.

Integrin αIIbβ3 inside-out activation does not trigger the safeguarding motif

To verify the presence of ΔG°_B in the full-length receptor in its native membrane environment and to test the response of the safeguarding motif to inside-out signaling, we measured activation levels of wild-type and mutant integrin αIIbβ3 in Chinese hamster ovary (CHO) cells as a function of the cellular talin concentration.²⁶ The talin head domain (THD) activates integrin αIIbβ3 by breaking the αIIb(R995)-β3(D723) interaction of the IMC and reorienting the β3 TM helix to dissuade αIIb partnering (Figure 1a,c).^9,16 To quantify integrin activation, we measured the cellular THD concentration at 50% activation, termed EC₅₀, and the percentage of activated receptors obtained at saturating THD concentrations, termed AI_THD (Figure S3). EC₅₀ values carried relatively large uncertainties and did not allow the differentiation of most receptor variants (Figure 3b). However, the αIIb(W968A/V)β3 receptors unambiguously resulted in lower numbers of activated receptors (AI_THD values) than wild type (Figure 3a) in confirmation of the presence of the destabilizing motif in the native receptor.

Figure 3. Talin-mediated inside-out activation of integrin αIIbβ3 and variants.

(a) The percentage of activated receptors obtained in CHO cells at saturating THD concentrations, termed AI_THD, and (b) the cellular THD concentration at 50% activation, termed EC₅₀, are correlated to the thermodynamic TM complex stabilities (ΔΔG°) in bicelles (Table 1). To guide the eye, linear fits are shown in red. The squared correlation coefficient (R²) is indicated. Error bars represent the standard error. The significance level between the AI_THD differences of W968V/A and wild type is 0.05 (two-tailed Welch’s test).

The cooperative collapse of the dimer interface by THD will remove αIIb(Val968)-β3(Ile693) strain, however, at this point, it is too late to cause a net stabilization of the TM complex. To verify this statement experimentally, we also examined the αIIb(G972A) and αIIb(W968V/G972A) mutants in CHO cells to allow for a better characterization of the AI_THD-ΔΔG° correlation. First, despite some differences between membrane and bicelle lipid environments,^27,28 we note a relatively high quality of the AI_THD-ΔΔG° correlation (Figure 3a). Evidently, bicelles adequately represented the cellular lipid environment pertaining to the αIIb(W968)-β3(I693) interaction. Second, when assuming a linear AI_THD-ΔΔG° relationship, it appears that the difference in AI_THD values between the W968X-substituted and wild-type receptor is slightly larger than predicted based on their corresponding ΔΔG° difference (Figure 3a). For THD-triggered safeguarding the opposite behavior is expected. This increase may stem from an increased safeguarding capacity in the cell membrane compared to the bicelle environment. Third, W968V/A and W968V/G972A have their safeguarding motif removed. If THD-triggered safeguarding would take place, AI_THD of wild type would be smaller than AI_THD of W968V/G972A but this is not the case. Wild type is indistinguishable from W968V/G972A and far from AI_THD of W968X, which is close to the limit of safeguarding (Figure 3a). In sum, the safeguarding motif is present in the native receptor and not triggered by inside-out signaling.

A force acting on the TM complex may trigger the safeguarding motif

To act as a safeguarding motif, overpacking of αIIb(W968)-β3(I693) must be avoidable to permit improved packing of αIIb(G972). To understand the correlation of Trp968 over- with Gly972 under-packing in more detail, we compared the packing of Gly972 between wild-type and W968V-substituted TM complexes in all-atom molecular dynamics simulations of 2 μs duration. Gly972 packing was quantified by the closest contacts made by its H^α1 and H^α2 atoms to any β3 atom (Figure 4b,d). Accordingly, the sum of the minimal Gly972/H^α1 and H^α2 contact distances, termed d_min(H^α1+H^α2), was computed for each simulation step. This trajectory showed not only a decrease in Gly972 baseline packing for W968V but also diminished amplitudes of Gly972 contact losses (Figure 4a). Overall, mean d_min(H^α1+H^α2) decreased by only 0.15 Å for W968V (P< 0.001) illustrating the high sensitivity (~0.6 kcal/mol; Table 1) of this site to packing defects. In contrast, C^α distances between Trp/Val968-Ile693 decreased by 0.52 Å for W968V (Figure S4a) but improved TM complex stability by only ~0.2 kcal/mol (Table 1). Thus, an activation of the safeguarding motif may adjust a relatively plastic Trp968-Ile693 contact to lower d_min(H^α1+H^α2) by reducing baseline packing and/or diminishing fluctuation amplitudes.

Figure 4. Molecular dynamics simulations of integrin αIIbβ3 and αIIb(W968V)β3 TM complexes.

(a) Comparison of the trajectories of minimal αIIb(G972/H^α1+H^α2) distances to the partnering β3 helix, termed d_min(G972/H^α1+H^α2), between αIIbβ3 and αIIb(W968V)β3 TM complexes. Mean d_min values were 5.11 ± 0.46 Å and 4.95 ± 0.38 Å for αIIbβ3 and αIIb(W968V)β3, respectively. For all simulations, mean d_min values were calculated for simulation times ≥0.1 μs to ensure initial equilibration of the system. Despite wide distance fluctuations, the large number of observed states renders differences of 0.01 Å statistically significant (P< 0.001). Eliminating d_min values above 5.495 Å (green line) for αIIbβ3 matches mean d_min(G972/H^α1+H^α2) distances between αIIbβ3 and αIIb(W968V)β3. (b) Comparison of d_min trajectories of Gly972 H^α1 and H^α2 in the αIIbβ3 TM complex. Mean d_min values are indicated. For reference, mean d_min for H^α1 and H^α2 of W968V were 2.47 ± 0.24 Å and 2.48 ± 0.28 Å, respectively. c Comparison of the d_min trajectories of G972 H^α1 and H^α2 in the αIIbβ3 TM complex with force applied. Mean d_min values are indicated. (d-e) Illustration of αIIbβ3 TM complex structure with applied forces. The black arrow originating on αIIb(G966/C^α) indicates the origin and direction of the pulling force. The force acting on β3(I721/C^α) constrained it to its initial position. Forces acted only along the directions indicated (x dimension) whereas coordinates orthogonal to these directions (y and z dimensions) were not restricted. Panel d depicts the forces at the start of constant velocity pulling (0 μs). Panel e shows the directions of the forces maintaining αIIb(G966/C^α) and β3(I721/C^α) x-positions near the end of the simulation (0.993446 μs). For reference, phosphate atoms of lipid headgroups are shown as grey spheres.

Limiting fluctuation amplitudes appears easier than improving packing that is invariably restricted by the Trp968 side chain geometry. For example, limiting d_min(H^α1+H^α2) to ≤5.495 Å matches mean d_min between the wild-type and W968V-substituted TM complex (Figure 4a) without significantly reducing the W968/C^α-I693/C^α distance (Figure S4d). Individually, the H^α1 and H^α2 contacts contributed 0.09 and 0.07 Å, respectively, to the mean d_min increase relative to W968V. H^α1 was mainly responsible for increased baseline packing whereas H^α2 primarily gave rise to enhanced distance fluctuations (Figure 4b). H^α2 is in danger of rotating away from the β3 helix face whenever the first helix turn has to accommodate Trp in an “unfavorable” configuration (Figure S4c). We therefore hypothesized that a force acting on the first TM helix turn may stabilize Trp dynamics and prevent H^α2 rotations. To test this possibility, we extended the MD simulation by 1 μs and incorporated forces that pulled the first αIIb TM helix residue, Ile966/C^α, parallel to the membrane surface while fixing the position of the last β3 TM helix residue, Ile721/C^α, in this direction (Figure 4d). After moving a distance of 20 Å in 20 ns, pulling was stopped and forces only maintained the obtained Ile966/C^α and Ile721/C^α coordinates in the original force coordinate. Physiologically, such a situation might arise from transient forces acting on the ectodomains and/or membrane with the cytosolic β3 tail permanently linked to the cytoskeleton.^29,30

The applied force initially pulled H^α2 away from its partnering helix face (Figure 4d). In response, the complex reoriented relative to the applied force and rotated H^α2 closer to the helix-helix interface and H^α1 slightly away from it (Figure 4c,e). Relative to the absence of force, this increased d_min(H^α1) by 0.02 Å whereas d_min(H^α2) decreased by 0.08 Å, resulting in a net packing improvement of 0.06 Å. The force also led to an increased mean W968/C^α-I693/C^α distance of 0.80 Å. However, the overall complex r.m.s.d. was unchanged compared to the absence of force (Figure S4e), indicating that the TM helix-helix interface remained unperturbed. The H^α2 site now achieved a mean d_min(H^α2) of 2.47 ± 0.30 Å, which surpassed the value for W968V of 2.48 ± 0.28 Å, and benefitted from reductions in fluctuation amplitudes (Figure 4b,c). Consequently, the safeguarding motif was activated by and buffered against a pulling force acting on the first αIIb TM helix turn by stabilizing Gly972/H^α2 positioning in the helix-helix interface.

Shear stress activates the safeguarding motif

To identify a physiological trigger of the safeguarding motif, we tested its buffering capacity against shear stress (frictional force). We again used integrin αIIbβ3-expressing CHO cells, and exposed the immobilized cells to shear (laminar) flow that created a frictional force of 40 dyne/cm² (4 N/m²; Figure 5a). Activated receptors were detected by their ability to bind fibrinogen. The percentage of receptors activated by flow, termed AI_flow, was calculated in relation to fibrinogen binding obtained in the presence of the inhibitor eptifibatide and the activator LIBS6 (Figure S5). Up to three receptor variants could be quantified and, aside from wild type, we evaluated W968V- and G972A-substituted integrin αIIbβ3. In CHO cells, AI_THD is a direct measure of TM complex stability (Figure 3a). AI_flow increased from W968V- to wild-type to G972A-substituted receptors following the order of their AI_THD values (Figure 5b). In this assay, contributions from intracellular signaling to receptor activation were excluded,³¹ which reveals that integrin αIIbβ3 activation by shear stress is proportional to TM complex stability. As such, shear stress dissociated the TM complex, which invariably transmits a signal into the cell (outside-in activation) and identifies a direct vulnerability of platelets to hydrodynamic force.

Figure 5. Shear stress-based activation of integrin αIIbβ3 and variants.

(a) Schematic diagram of the fibrinogen-binding flow assay. Cells attached to a type I collagen-coated flow chamber were incubated with media containing biotinylated fibrinogen under static or flow conditions. (b) The percentage of receptors activated by shear flow stress of 40 dyne/cm², termed AI_flow, is correlated to the extent of talin-mediated receptor activation, AI_THD (Figure 3a). Error bars represent the standard error.

If the safeguarding motif responded to shear stress, integrin αIIbβ3 is expected to tolerate more stress than predicted based on its TM complex stability (Figure 1b). We compared wild type to W968V and G972A variants that possess little safeguarding capacity (Table 1) but exhibit markedly different AI_THD values (Figure 3a). The AI_THD-AI_flow correlation identified wild type below the W968V-G972A trend line (Figure 5b), i.e., absorbing more frictional force than expected based on its resting TM complex stability. Hence, shear stress activated the safeguarding motif. AI_flow of wild type is 44% lower than expected (Figure 5b). This corresponds to a buffering of approximately 0.4 kcal/mol (0.44×ΔG°_B) and establishes the WxxxG safeguarding motif as a buffer against frictional force.

Discussion

The inability to fully suppress pathological thrombosis by drugs interfering with physiological integrin αIIbβ3 activation suggests that the receptor is at risk of activation by yet an unidentified pathway. Here, thermodynamic measurements and talin-based activation assays revealed an intrinsic destabilization of the TM complex that maintains the inactive receptor state by approximately 1.0 kcal/mol (Figure 1b and Table 1). Although destabilization originates from overpacking of αIIb(W968)-β3(I693) at the extracellular membrane border, it mainly arises from the consequential underpacking of the αIIb(G972) contact (WxxxG sequence motif). Structural analysis and MD simulations revealed this destabilization to arise not only from steric barriers to optimal Gly972 packing but also from fluctuations of Gly972/H^α2 contacts arising from αIIb(W968)-β3(I693) strain (Figure 4b). The safeguard is not triggered by talin-based inside-out signaling (Figure 3a) and, conceptually, only an event that mitigates αIIb(W968)-β3(I693) destabilization while leaving the TM complex mostly intact is expected to result in safeguarding (Figure 1b). Based on the location of αIIb(W968)-β3(I693) at the extracellular membrane border (Figure 1a,c), we then hypothesized that an external stimulus could trigger safeguarding. MD simulations and receptor activation pattern under shear flow indicate that (external) pulling and frictional forces can activate the safeguarding motif. Accordingly, perturbations in the structure of the membrane and/or the inactive ectodomain that impact the TM complex appear buffered. The discovery of the safeguarding motif indicates that most vertebrates had experienced evolutionary pressure to buffer integrin αIIbβ3 (Figure 1d). Over evolutionary relevant periods, life spans are not limited by cardiovascular disease,^32,33 implying that integrin αIIbβ3 was at risk of activation from blood flow hydrodynamics already under physiological conditions.

The asymmetric triggering of the WxxxG safeguarding motif by pathological but not physiological stimuli effectively increases the free energy barrier between inactive and active integrin αIIbβ3 states by up to 1.0 kcal/mol without incurring an increased risk of bleeding (Figure 1b). This mode of action identifies a universal and efficient way to protect a receptor functional state against unwarranted activation. Furthermore, its existence establishes integrin αIIbβ3 as mechano-sensitive under physiological conditions. When compared to mechano-sensitive ion channels,³⁴ the WxxxG safeguarding motif is structurally relatively simple and easy to integrate into the TM helix-helix interface (Figure 2b), raising the possibility that safeguarding motifs are frequent occurrences. Nonetheless, in the 23 other human integrins, an aliphatic residue replaces Trp968 (Figure S1) and essentially eliminates buffering. Some of these integrins are present on platelets or leukocytes^4,5,35 and may also activate in response to hydrodynamic forces. Unwarranted adhesion through these integrins must be less cataclysmic and a degree of erroneous outside-in signals may be neutralized within intracellular signaling networks.

While most vertebrates have adopted the WxxxG safeguarding motif, it is altered in a few species (Figure 1d). In the desert rodent Jaculus jaculus, a W968L substitution will achieve a larger ΔG°_A and largely eliminate safeguarding in analogy to W968V (Table 1). A permanently higher barrier to thrombosis is implemented (Figure 1b) at the cost of enhancing the propensity to bleed. On the other hand, the Burmese python (Python bivittatus) and spotted gar (Lepisosteus oculatus) incorporate G972A, which reduces buffering (smaller ΔG°_B) and enhances receptor activation (smaller ΔG°_A) to diminish bleeding tendencies at heightened danger of unwarranted thrombosis. The safeguarding and allosteric properties of integrin αIIbβ3 thus change, establishing that organisms can successfully vary the properties of their cardiovascular system.

Liquid shear flow creates wall shear stress that can activate integrin αIIbβ3 on the surface of immobilized cells (Figure 5a).³¹ At a level of 40 dyne/cm², the degree of receptor activation correlated with the stability of the TM complex (Figure 5b) indicating its accompanying dissociation, which is a hallmark of outside-in signaling.³⁶ For floating platelets shear stress will be greatly reduced,³⁷ making it an unlikely candidate for causing direct integrin αIIbβ3 activation. The cardiovascular system branches from large arteries to small arterioles and capillaries. As blood passes through such branch points, it has to accelerate to flow in the narrower vessels (Figure 6).^32,38 At any given point in this transition, the velocity of the fluid preceding this point is slower than the fluid ahead of it, creating an elongational pulling force.^38,39 The elongational force surpasses the shear force, which is also increasing in this context.³⁸ In analogy to the distortion of polymers,^38,39 these forces may distort the shape of platelets, i.e., their plasma membrane including immersed proteins, and activate integrin αIIbβ3 in analogy to shear stress when immobilized (Figure 6). Hence, we hypothesize that peak elongational forces gave rise to the evolution of the safeguarding motif in integrin αIIbβ3.

Figure 6. Hypothesis of integrin αIIbβ3 activation by hydrodynamic force.

Blood passing from large to small vessels accelerates its flow. This acceleration not only increases flow velocity and accompanying wall shear stress³² but also creates a localized elongational pulling force.^38,39 On free floating platelets, the elongational pulling force is expected to activate integrin αIIbβ3 at the site of vessel narrowing. Once this site is cleared, the force dissipates allowing the receptor to return to its inactive state.

The narrowing of blood vessels (stenosis) is also a hallmark of cardiovascular disease^32,33 with pathological thrombosis frequently observed at such sites.^32,40 For example, localized vascular stenosis can causes platelet aggregation in vivo at the downstream stenosis margin in an agonist-independent but integrin αIIbβ3-dependent manner.⁴¹ Stenosis can easily cause hydrodynamic forces to exceed physiological margins,^32,33 which will invariably exceed the buffering capacity of the safeguarding motif. It appears that outside-in activation of integrin αIIbβ3 by hydrodynamic elongational forces underlies its pathological activation. Accordingly, integrin αIIbβ3 plays an active rather than indirect role in the processes leading to fatal thrombosis.^32,33 Ideally, a next generation of anti-thrombotic drugs would enlarge ΔG°_B (Figure 1b). However, given the complexity of this task, drugs that moderately expand ΔG°_A appear promising for at risk patients despite an increased risk of bleeding. Drugs and compounds that target the TM complex^42,43 may be most effective.

Materials and Methods

Peptide preparation

Peptides encompassing human integrin αIIb(A958-P998) and β3(P685-F727) with β3(C687S) were prepared applying published protocols.^44,45 The mutations depicted in Table 1 were introduced by QuikChange mutagenesis. To prepare disulfide-linked αIIb(W968V)β3 dimer, αIIb(A963CG/W968V) and β3(G690CG) peptides were prepared. To diminish the conformational constraint of the disulfide bond, a glycine residue was incorporated in αIIb and preserved in β3. Covalent dimers were assembled as described.⁴⁶ ~70%-²H/99%-¹³C/99%-¹⁵N-labeled peptides were prepared using 99%-¹³C-glucose, 99%-¹⁵NH₄Cl and 99% D₂O precursors. 99%-²H/99%-¹⁵N-labeled peptides were produced utilizing 99% d₇-glucose, 99% ¹⁵ND₄Cl and 99% D₂O. Freeze-dried peptides were taken up in 320 μL of 350 mM 1,2-dihexanoyl-sn-glycero-3-phosphocholine (DHPC), 105 mM 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC), 6% D₂O, 0.02% w/v NaN₃ buffered by either 25 mM NaH₂PO₄/Na₂HPO₄, pH 7.4 or 25 mM HEPES·NaOH, pH 7.4. The following NMR samples were prepared: 1.2 mM ²H/¹³C/¹⁵N-αIIb(W968V), 0.8 mM ²H/¹³C/¹⁵N-αIIb(W968V) + 1.2 mM β3, 1.2 mM αIIb(W968V) + 0.8 mM ²H/¹⁵N-β3, 0.8 mM ²H/¹³C/¹⁵N-αIIb(A963CG/W968V)–²H/¹³C/¹⁵N-β3(G690CG), 0.8 mM ²H/¹⁵N-αIIb(A963CG/W968V)–β3(G690CG), and 0.8 mM αIIb(A963CG/W968V)– ²H/¹⁵N-β3(G690CG).

Isothermal titration calorimetry

ITC measurements of the peptide combinations listed in Table 1 were carried out using a Microcal VP-ITC calorimeter. Specifically, 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 1,2-dihexanoly-sn-glycero-3-phosphocholine (DHPC), 17 mM 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC), 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 complex stoichiometry was fixed at the experimentally verified ratio of 1:1²⁰ and the reaction enthalpy (ΔH°), entropy (ΔS°), and K_XY were calculated from the measured heat changes, δH_i, as described previously.²⁰

NMR spectroscopy

Starting from the ¹H^N, ¹⁵N, ¹³C^α, ¹³C^β, and ¹³C′ assignments of the αIIbβ3 TM complex and αIIb TM monomer,^10,44 backbone assignments of αIIb(W968V), αIIb(W968V)β3, and αIIb(A963CG/W968V)–β3(G690CG) were achieved employing HNCA, HNCO, HNCACB and NOESY-TROSY experiments. To detect intra- and intersubunit NOEs, ¹⁵N-edited NOESY-TROSY experiments using ²H/¹⁵N-αIIb(A963CG/W968V)–β3(G690CG) or αIIb(A963CG/W968V)–²H/¹⁵N-β3(G690CG) dimers were acquired with mixing times of 150 and 175 ms. Side chain assignments were taken from the αIIbβ3 and αIIb(A963C)–β3(G690C/A711P) TM complexes and found little changed in the aforementioned NOESY spectra. For 0.3 mM ²H/¹³C/¹⁵N-αIIb(A963CG/W968V)–²H/¹³C/¹⁵N-β3(G690CG) in 143 mM DHPC, 43 mM DMPC, 25 mM HEPES·NaOH, pH 7.4, H-N residual dipolar couplings (¹D_NH), were measured in a compressed polyacrylamide gel of 320 μl volume and initial diameter of 6 mm using the ARTSY pulse sequence.^47,48 The gel was polymerized from a 4.0% w/v solution of acrylamide (AA), 2-acrylamido-2-methyl-1-propanesulfonate (AMPS) and bisacrylamide (BIS) with a monomer-to-crosslinker ratio of 63:1 (w/w) and a molar ratio of 95:5 of AA to AMPS. For 1.2 mM ²H/¹³C/¹⁵N-αIIb(W968V), ¹D_NH, ¹D_Cα_C′, and ¹D_C′N were measured in a gel of identical dimensions but polymerized from a 4.2% w/v solution with a AA+AMPS/BIS ratio of 49:1 (w/w) and a molar ratio of 94:6 of AA to AMPS. ¹D_Cα_C′, and ¹D_C′N were quantified using HNCO-based experiments.^49,50 For ²H/¹³C/¹⁵N-αIIb(W968V), ³J_C′Cγ and ³J_NCγ couplings were also determined.⁵¹ All NMR experiments were carried out on a cryoprobe-equipped Bruker Avance 700 spectrometer at 35 and 40 °C for monomer and dimer, respectively.

Structure calculation of the integrin αIIb(W968V)β3 TM complex

Structure calculations were carried out by simulated annealing, starting at 3000 K using the program XPLOR-NIH.⁵² Backbone torsion angle restraints were extracted from ¹⁵N, ¹³C^α, ¹³C^β, and ¹³C′ chemical shift patterns using TALOS-N.⁵³ Within experimental uncertainties, H-N RDCs measured for the αIIb(A963CG/W968V)–β3(G690CG) TM dimer fitted the αIIb(W968V) and β3 TM monomer structures (data not shown). This congruence permitted the use of H-N, C^α-C′, N-C′ RDCs measured for these monomers to further restrict the individual αIIb(W968V) and β3 backbone conformations in the dimer. NOE peak intensities were referenced to αIIb(T981/H^N)-αIIb(T981/H^O), αIIb(I982/H^N)-αIIb(T981/H^O), β3(S699/H^N)-β3(S699/H^O) and β3(V700/H^N)-β3(S699/H^O) distances and classified into four groups (1.8–2.9, 1.8–3.5, 1.8–5.0, and 1.8–6.0 Å). An employed torsion angle potential of mean force⁵⁴ was biased to use the experimental χ₁ angles detected in the monomeric αIIb(W968V) and β3 TM segments, which mostly corresponded to canonical values. The side chains of αIIb(Trp967), αIIb(Trp988), αIIb(Lys989), αIIb(Phe992), β3(Trp715) and β3(Lys716) were restricted to snorkeling positions.⁵⁵ Moreover, the αIIb(Arg995)-β3(Asp723) salt bridge, whose absence destabilizes the αIIbβ3 TM complex by 1.5 ± 0.2 kcal/mol,¹⁵ was implemented. Aside from standard force field terms for covalent geometry (bonds, angles, and improper dihedrals) and nonbonded contacts (Van der Waals repulsion), dihedral angle restraints were implemented using quadratic square-well potentials. In addition, a backbone-backbone hydrogen-bonding potential was employed.⁵⁶ A quadratic harmonic potential was used to minimize the difference between predicted and experimental residual dipolar couplings (RDC; Δ¹D). The final values for the force constants of the different terms in the simulated annealing target function were as previously described.¹⁰ Table S1 summarizes the structural statistics for all 20 calculated structures.

Talin-dependent integrin αIIbβ3 activation

Chinese hamster ovary (CHO) cells were purchased from the Korean Collection for Type Cultures. CHO cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM, HyClone) supplemented with 10% (v/v) fetal bovine serum (HyClone), 2 mM L-glutamine (WELGENE Inc.), non-essential amino acids (Thermo Fisher Scientific), and penicillin-streptomycin (HyClone). CHO/αIIbβ3 cells were generated by infecting CHO cells with lentiviruses bearing human integrin αIIb and β3 subunits as previously described.²² Alternatively, W968A, W968V, G972A, and W968V/G972A-substituted αIIb subunits were used to generate CHO/αIIb(W968A)β3, CHO/αIIb(W968V)β3, CHO/αIIb(G972A)β3, and CHO/αIIb(W968V/G972A)β3 cells. Infected cells were sorted and pooled based on their ability to bind the integrin αIIbβ3 complex-specific antibody, D53, and maintained in the same condition as CHO cells described above

The integrin-expressing cells were transiently transfected with GFP-tagged talin head domain (THD) using Lipofectamine 2000 Transfection Reagent (ThermoFisher Scientific), and incubated for 24 h at 37 °C. Cells were detached by trypsinization, washed twice with serum-free DMEM, and incubated with the integrin αIIbβ3 ligand-mimetic antibody PAC-1 (5 μg/ml) for 30 min at room temperature.^57,58 After washing, cells were further stained with allophycocyanin-conjugated anti-IgM antibody (Thermo Fisher Scientific) for 30 min at 4 °C. Stained cells were collected and analyzed by BD FACSCalibur flow cytometer. MATLAB R2014a (MathWorks) was used to calculate the mean fluorescence intensities (MFIs) of PAC-1 binding to those cells at different level of GFP-THD expression as previously described.⁵⁹ Moreover, as reference for zero and full activation, MFIs were determined in the presence of 10 μM Eptifibatide and 5 mM MnCl₂, respectively. The activation index, termed AI, at a cellular GFP-THD concentration was calculated as AI(GFP-THD) = 100%·(MFI_THD − MFI_Eptifibatide) / (MFI_Mn2+ − MFI_Eptifibatide). For each integrin variant, three independent pairs of duplicate experiments were conducted and averaged. Using gnuplot, mean AI values and standard deviations were fit to a one-site competitive binding curve, AI(GFP-THD) = A2+(A1-A2) / (1+ 10^(log₁₀(AI(GFP-THD)/A0)), to extract EC₅₀=A0 and AI_THD=A2 with asymptotic standard errors (Figure S3). AI_THD denotes the maximal achievable activation level at saturating THD concentration.

Molecular dynamic simulations

Using the coordinates of the αIIbβ3 and αIIbβ3(W968V) TM complex structures, αIIb(Leu956-Pro998) and β3(Glu686-Phe727) were immersed in POPC bilayers and solvated with TIP3P model water molecules⁶⁰ employing the programs VMD 1.9.3⁶¹ and SOLVATE 1.0. Na⁺ and Cl⁻ ions were added to a concentration of 100 mM each and to neutralize the system. Initially, the complexes were oriented to align the αIIb TM helix (Ile966-Lys989) along the bilayer normal and center it to the bilayer middle. Simulations consisted of 197/196 lipids, 16353/16341 water molecules, 31/31 Na⁺ ions, and 35/35 Cl⁻ ions for αIIbβ3/αIIbβ3(W968V), respectively. All-atom MD simulations were carried out using CHARMM22 and CHARMM27 force fields for proteins and lipids, respectively, in the context of the program NAMD 2.13.^62–64 A uniform integration time step of 1 fs and periodic boundary conditions were employed. Electrostatic interactions were calculated using the Particle Mesh Ewald algorithm with a grid size of <1 Å.⁶⁵ Only water molecules were treated rigidly using the SETTLE algorithm.⁶⁶ The melting of lipid tails with all other atoms fixed was followed by minimization and equilibration with protein constrained and equilibration with protein released, each for a period of 0.5 ns. During the simulations, the area per lipid was kept constant and the cell dimension was variable. Simulations were carried out at 310 K and at constant pressure of 1 atm. Constant temperature was maintained using Langevin dynamics,⁶⁷ with a damping coefficient of 1.0 ps⁻¹. Constant pressure was enforced using a Nosé-Hoover Langevin piston with a period of 200 fs and time constant 50 fs. If applicable, constant velocity pulling of αIIb(I966/C^α) was performed for 20 ns at 0.00001 Å/fs with a force constant of 2.5 kcal·A²/mol while harmonically constraining β3(I721/C^α) as illustrated in Figure 4d. To test whether parameter means between two simulations were meaningfully different, two-tailed Welch’s tests at a significance level of 0.001 were evaluated.

Shear stress-induced fibrinogen binding assay

CHO/αIIbβ3, CHO/αIIb(W968V)β3 and CHO/αIIb(G972A)β3 cells were seeded and allowed to spread on collagen-coated flow chamber (μ-slide, ibidi, 80166). The chamber was washed twice with Hank’s Balanced Salt Solution (HBSS), and HBSS containing 20 μg/ml of biotinylated fibrinogen was applied to the chamber for 5 min at a flow rate of 10 mL/min to induce a shear stress of 40 dyne/cm². To calculate nonspecific binding, cells in the flow chamber were exposed to the same fibrinogen solution at the same flow rate but in the presence of Eptifibatide (10 μM). To calculate the maximum fibrinogen binding for each cell line, cells in the flow chamber were incubated with the biotinylated fibrinogen-containing HBSS for 5 min under static condition in the presence of 10 μg/ml integrin αIIbβ3-activating antibody LIBS6.⁶⁸ The flow chamber was gently washed three times with HBSS and the cells were directly lysed with 1X SDS-PAGE loading buffer containing 3% β-mercaptoethanol. Samples were boiled at 95 °C for 5 min and subjected to Western blotting using either anti-integrin β3 antibody (Santa Cruz Biotechnology, sc-365679) or Horseradish peroxidase (HRP)-conjugated streptavidin (Sigma-Aldrich, S2438). The detected band intensities (Figure S5), termed I, were quantified and the activation index for receptors under flow, termed AI_flow, was calculated as 100%·(I_flow − I_Eptifibatide) / (I_LIBS6 − I_Eptifibatide). Four independent experiments were performed.

Abbreviations:

CHO

Chinese hamster ovary

THD

talin head domain

TM

transmembrane