Electron microscopy shows that binding of monoclonal antibody PT25-2 primes integrin αIIbβ3 for ligand binding

The murine monoclonal antibody (mAb) PT25-2 and its antigen-binding fragment (Fab) were shown by Tokuhira et al¹  to induce αIIbβ3 to bind fibrinogen and initiate platelet aggregation. We and others have used it in studies of αIIbβ3 structure and function.^2-9  Mutation analysis involving swaps of murine and human sequences suggested that it binds to the αIIb Pro334-Ala339 loop in the fifth β-propeller repeat,¹⁰  which is far from the Arg-Gly-Asp (RGD)-binding pocket. Thus, the mechanism by which PT25-2 induces the receptor to bind ligand is not clear. To obtain a better understanding of the underlying mechanism, we assessed the effect of PT25-2 binding on exposing ligand-induced binding site (LIBS) antibody epitopes on αIIbβ3 and on the conformation of αIIbβ3 by negative-stain electron microscopy (EM). We also determined a cryo-EM structure of the αIIbβ3 headpiece in complex with the Fab of PT25-2. Collectively, the data support a model in which PT25-2 binding prevents αIIbβ3 from adopting the fully inactive, bent-closed conformation and thus primes the integrin for ligand binding by biasing it toward the more active, extended-closed conformation.

Human participant studies reported in this manuscript were conducted under a protocol approved by the Rockefeller University Institutional Review Board. The study was conducted in accordance with the Declaration of Helsinki.

Recombinant clasped αIIb and β3 full-length ectodomain constructs were cloned into the pMSCV-IRES-DsRed and the pMSCV-IRES-GFP II vectors, respectively, expressed in HEK293S–GnTI^−/− cells and purified by sequential metal-affinity, anion-exchange, and gel-filtration chromatography steps. We selected the clasped αIIbβ3 construct instead of the construct containing an additional engineered disulfide bond between αIIb Leu959Cys and β3 Pro688Cys that we previously used¹¹  because the latter is stabilized in the bent conformation, whereas the former has greater freedom to sample a variety of conformations between the fully bent and extended ones (supplemental Information).

The Fab fragment of PT25-2 was generated and purified using the Pierce Mouse IgG₁ Fab and F(ab')₂ Preparation Kit (supplemental Information).

The amino acid sequence of the variable regions of PT25-2 was determined from the mRNA of the hybridoma cell line expressing PT25-2¹ by reverse transcriptase-polymerase chain reaction and DNA sequencing of the cloned polymerase chain reaction products. The sequences were deposited in the GenBank with accession numbers MW424434 and MW424435 (supplemental Information).

The amino acid sequence of PT25-2 was also assessed by C₁₈ reverse-phase liquid chromatography-mass spectrometry (LC-MS/MS) (supplemental Information).

Purified αIIbβ3 was mixed with PT25-2 Fab at a 1:3 molar ratio (1.1 µM/3.3 µM) in a buffer containing 150 mM NaCl, 20 mM HEPES, pH 7.4, 2 mM CaCl₂, and 1 mM MgCl₂, incubated for 2 hours at 4°C, and subjected to gel-filtration chromatography (Superdex Increase 200). The fractions containing αIIbβ3–PT25-2 Fab complexes were identified by sodium dodecyl sulfate-polyacrylamide gel electrophoresis, followed by Coomassie blue staining and then imaged by negative-stain EM.

Binding of the LIBS mAbs AP5, directed at an N-terminal epitope on the β3 plexin-semaphorin-integrin domain,¹²  and LIBS1 and LIBS6, both directed at the β3 leg domain,^13,14  was performed as previously described¹⁵  (supplemental Information).

αIIbβ3 alone and in complex with PT25-2 Fab was negatively stained as described¹⁶  and imaged with a CM10 electron microscope. Particles were picked with Electron Micrograph Analysis (EMAN)2,¹⁷  and averages were calculated in System for Processing Image Data from Electron Microscopy and Related Fields (SPIDER)¹⁸  (supplemental Information).

Specimens were vitrified with a Vitrobot Mark IV and imaged with a Titan Krios electron microscope using SerialEM¹⁹  (supplemental Information).

Image stacks were motion-corrected in MotionCor2,²⁰  the contrast transfer function (CTF) parameters determined with CTFFIND4,²¹  and the particles picked with Gautomatch. The particle images were subjected to 2-dimensional (2D) classification and several cycles of 3-dimensional classification in REgularised LIkelihood OptimisatioN (RELION)-3.²²  After 3-dimensional refinement, postprocessing, CTF refinement, and Bayesian polishing, the final map reached a resolution of 3.3 Å (supplemental Information).

The crystal structure of αIIbβ3 (Protein Data Bank [PDB]: 3FCS)²³  and a homology model for the PT25-2 Fab generated with Phyre2²⁴  were rigid-body fitted into the cryo-EM density map in University of California–San Francisco Chimera.²⁵  Residues were manually fitted and adjusted in Coot.²⁶  The model was improved by iterative cycles of refinement, followed by minor manual adjustments. To validate the model, a cross-validation procedure was performed. The model was refined against half map 1. Then, Fourier shell correlation curves were calculated between the refined model and half map 1 (work), half map 2 (free), and the combined map. The cryo-EM map has been deposited at the Electron Microscopy Data Bank under accession code EMD-23245, and the coordinates have been deposited at the Protein Data Bank under accession code 7LA4.

The rate of binding of mAb PT25-2 was compared with the rates of binding of mAbs 10E5 and 7E3 by a modification of methods that were previously described.^27,28  Washed platelets (2 × 10⁸/mL) were incubated with the Alexa⁴⁸⁸-labeled mAbs at 10 μg/mL for the indicated times, and then the samples were diluted with HEPES-buffered modified Tyrode's solution (HBMT; 10× volume) to slow the reaction and immediately analyzed by flow cytometry (FACSCalibur). The data were normalized by expressing the results as the percentage of antibody bound at 1 hour, which is the time required to achieve near-saturating levels of mAb 7E3, the mAb that binds slowest.

Four independent runs of molecular dynamics simulations were carried out starting from the cryo-EM structure of the αIIbβ3–PT25-2 complex and using the Desmond MD program²⁹  as implemented in the Schrödinger 2020-3 software package (supplemental Information).

To explain the negative impact of αIIb human-to-murine Arg335Lys and His338Gln mutations on PT25-2 binding, the change in αIIbβ3–PT25-2 binding free energy between wild-type and mutant receptor, $ΔΔG(bind)$⁠, was calculated using the molecular mechanics-generalized born surface area method (supplemental Information).

We first expressed and purified the ectodomain of human αIIbβ3 (αIIb residues 1-963, β3 residues 1-690) that was stabilized by C-terminal segments containing coiled coils and a disulfide bond between the coiled coils (supplemental Figure 1A).²³  We assessed the quality of the αIIbβ3 ectodomain by negative-stain EM. Raw images showed a mono-dispersed particle population, but the particles varied substantially in size and shape (Figure 1A). After picking and classifying ∼10 000 particles into 100 classes, the averages showed the integrin in conformations varying from the bent-closed conformation that resembles a V shape to the extended-closed conformation whose face-on view resembles a figure 8 (Figure 1B). We then prepared the PT25-2 Fab fragment and used it to form a complex with the αIIbβ3 ectodomain (supplemental Figure 1B). Negative-stain EM raw images of the αIIbβ3–PT25-2 Fab complex showed a similarly diverse particle population as for the αIIbβ3 ectodomain alone (Figure 1C), but classification of ∼10 000 particles into 100 classes yielded different class averages (Figure 1D). Of the class averages showing the αIIbβ3 ectodomain with clear density for bound PT25-2 Fab, none showed the integrin in the canonical bent-closed conformation, and many showed the integrin in various extended states (Figure 1D). The density for the Fab and the αIIbβ3 headpiece is clearly defined in most of these averages, but the density for the integrin legs is often blurred. This result suggests that binding of the PT25-2 Fab to the integrin headpiece either induces the αIIbβ3 legs to become very mobile and/or prevents αIIbβ3 from adopting the bent-closed conformation. Of note, none of the averages could be interpreted as unequivocally showing the PT25-2 Fab-bound integrin in a fully activated, extended-open conformation.

PT25-2 increased the binding of the LIBS mAbs AP5, LIBS1, and LIBS6 to washed platelets to the same extent or more than T6, a thrombin receptor-activating peptide (SFLLRN), indicating that even in the absence of added ligand, PT25-2 is capable of making regions of the β3 subunit accessible to LIBS mAbs that are ordinarily occluded in the inactivated αIIbβ3 receptor (Figure 2). However, both T6 and PT25-2 were variably less effective than eptifibatide or EDTA in exposing the epitopes for LIBS mAbs.

Despite the conformational variability of the integrin legs, we sought to use cryo-EM to determine how the PT25-2 antibody binds to the αIIbβ3 headpiece. We vitrified the αIIbβ3–PT25-2 Fab complex and imaged the sample using a K2 Summit direct electron detector on a Titan Krios electron microscope (supplemental Figure 2). Image processing resulted in a density map of the PT25-2-bound αIIbβ3 headpiece at 3.3-Å resolution (Figure 3A; supplemental Figures 2 and 3; supplemental Table 1), which allowed us to build an atomic model for αIIb residues 1 to 452, comprising the β-propeller domain, β3 residues 56 to 432, comprising the βI and hybrid domains, and PT25-2 residues 1 to 116 from the heavy chain, and residues 1 to107 from the light chain (Figure 3B; supplemental Table 1).

The PT25-2 Fab uses the variable domains of its light chain (LC) and heavy chain (HC) to bind to the αIIb β propeller with a buried surface area of 914 Ų while forming no interactions with the β3 subunit (Figure 3). Three regions of αIIb form the epitope for PT25-2. Two loops of αIIb form the key interactions with PT25-2 (Figure 3C). One αIIb loop (residues 332-344) is part of repeat 5 of the β propeller and interacts with both the PT25-2 HC and LC. αIIb Arg335 forms five hydrogen bonds, with HC Thr30, Asp31, Asn33, Asn52 and Ser99, and αIIb Ala339 and Leu340 form hydrogen bonds with LC Asn92 and Tyr32, respectively. In addition, although αIIb His338 is not involved in hydrogen bonding, it forms hydrophobic interactions with HC Leu59 and LC Leu94. Consistent with this interaction, a previous study found that the αIIb Arg335Lys/His338Gln double mutation completely abrogates PT25-2 binding.¹⁰  The second αIIb region (residues 273-281) is a β strand in repeat 4 and interacts exclusively with the PT25-2 LC, with αIIb Gln275 and Arg276 forming hydrogen bonds with LC Arg53, and αIIb His278 with LC Tyr50. Asp271 in the neighboring β strand forms a salt bridge with LC Arg53. The third region, the calcium-binding loop from repeat 4 (residues 240-255), contains Asp245, which forms hydrogen bonds with HC Arg103 and LC Tyr49, as well as Glu243 and Gly246 that form hydrogen bonds with HC Tyr100 and Tyr101. The binding of the PT25-2 Fab has relatively little impact on the structure of the αIIb β propeller other than producing an ∼5-Å shift in the apex of repeat 5 (supplemental Figure 4).

The PT25-2 antibody is known to induce αIIbβ3 to bind ligand.¹  Because ligand binding involves both extension of the headpiece from the legs and swing-out of the β3 hybrid domain away from the β3 βI domain, we analyzed the structure for insights as to how PT25-2 induces ligand binding.

A composite structure of a PT25-2 Fab-bound αIIbβ3 ectodomain in the bent-closed conformation (based on the overlay of the head domains in the cryo-EM and crystal structures) immediately reveals substantial steric clashes between the Fab and the αIIb leg domains (Figure 4A; supplemental Figure 5). In particular, the PT25-2 HC clashes with the calf-1 domain and the LC with the calf-2 domain. The clash score based on the program MolProbity³⁰  for the composite structure of the PT25-2-bound αIIbβ3 ectodomain is 26.86 with 750 overlapping atom pairs as compared with 1.60 and just 1 overlapping atom pair for the ectodomain by itself. This result shows that the PT25-2 antibody cannot bind to αIIbβ3 in the bent-closed conformation and, conversely, that αIIbβ3 with bound PT25-2 antibody cannot adopt the bent-closed conformation. The clashes are produced directly by the epitope-binding variable domains of the antibody, which explains why the PT25-2 Fab fragment is as effective as the intact antibody in inducing ligand binding.¹ 

The head and leg domains of the αIIb subunit do not form any direct interactions with each other that would stabilize the fully bent-closed conformation of the αIIbβ3 ectodomain. Instead, the bent-closed conformation is stabilized by the lower leg of the β3 subunit, in particular the EGF-3 and EGF-4/βTD domains, which are sandwiched between the lower leg (the calf-1 and calf-2 domains) of the αIIb subunit on one side and the hybrid domain of the β3 headpiece on the other side (Figure 4B). Through this domain configuration, the calf-1 and calf-2 domains of the αIIb subunit hold the β3 EGF-3 and EGF-4/βTD domains in place, which interact with the hybrid domain, thus forming the only direct interactions of the integrin legs with the headpiece. When the PT25-2 Fab binds to the αIIb β propeller and prevents the αIIb lower leg from assuming the position it usually has in the bent-closed conformation, the αIIb lower leg interaction with the β3 EGF-3 and EGF-4/βTD domains is likely disrupted, thus weakening the interaction of the β3 leg with the β3 headpiece and preventing the αIIbβ3 ectodomain from adopting the fully bent-closed conformation.

Our negative-stain EM results are consistent with this proposed mechanism. With a bound PT25-2 Fab, the αIIbβ3 ectodomain also exists in a continuum of closed conformations, but none of the averages shows it in the canonical bent-closed conformation (Figure 1D). Although the angle between the legs and the headpiece for the αIIbβ3 ectodomain by itself in the bent-closed conformation is ∼45° (Figure 1B,E, average 1), the smallest the corresponding angle can become in the αIIbβ3 ectodomain when PT25-2 Fab is bound is ∼80° (Figure 1D,F, average 1), illustrating that the bound Fab prevents the legs from closely associating with the headpiece. In addition, although some averages show indications of two legs, in most averages of PT25-2 Fab-bound αIIbβ3 ectodomain the legs are smeared out, presumably as a result of them no longer being able to interact with the headpiece and thus becoming more mobile (supplemental Information provides more details).

The PT25-2 Fab binds αIIb more than 30 Å away from the ligand-binding pocket, and thus any change in the pocket would be the result of allostery. Comparison of our cryo-EM structure of the αIIbβ3 headpiece in complex with the PT25-2 Fab with the crystal structure of the αIIbβ3 ectodomain by itself (PDB: 3FCS)²³  (Figure 5A,C) or with the αIIbβ3 ectodomain in complex with the fibrinogen peptide (PDB: 2VDO)³¹  (Figure 5B-C) shows that Fab binding does not cause swing-out of the hybrid domain. We found minor differences in the locations of the metal ion-dependent adhesion site (MIDAS) Mg²⁺ (0.7 Å) and the adjacent to MIDAS Ca²⁺ (1.1 Å) relative to those in the unliganded αIIbβ3, but consider these to be within the range of thermal fluctuations (Figure 5).

The structure of the αIIbβ3–PT25-2 Fab complex indicates clashes between the antibody and the leg regions when the receptor is in its bent-closed conformation, and we therefore considered it likely that the antibody can only bind to receptors in at least partially extended conformations. If this is the case, we would expect the antibody’s on-rate to be slower than that of an antibody that can bind equally well to all conformations, as epitomized by the binding of our mAb 10E5 to the exposed cap domain of αIIb.^27,32  Thus, we would expect the antibody’s on-rate to be more like that of mAb 7E3, which binds to an epitope on the β3 I domain that is partially hidden in the bent-closed conformation.^11,27,28,33 

Figure 6 shows the results of 3 different experiments with platelets from 3 different donors in which the binding of mAb PT25-2 was directly compared with the binding of mAbs 10E5 and 7E3. To normalize the data, they are expressed as the percentage of the binding observed at 1 hour, the time we previously identified as resulting in maximal binding of mAb 7E3,^27,28  the mAb that binds slowest. At every time point through the first 3 minutes, the PT25-2 values were intermediate between those of mAbs 10E5 and 7E3. Thus, the binding rate of PT25-2 appears to be intermediate between those of 10E5 and 7E3.

To provide a structural and energetic basis for the known negative impact of αIIb Arg335Lys and His338Gln mutations on PT25-2 binding, we calculated differences in αIIbβ3–PT25-2 binding free energy between wild-type and either single or double mutants. Although the His338Gln mutation has a small impact on the calculated αIIbβ3-PT25-2 binding free energy (⁠$ΔΔG(bind)$ = −1.11 ± 0.19 kcal/mol), both the Arg335Lys and double Arg335Lys+His338Gln mutants show a significant decrease of their calculated $ΔΔG(bind)$ values (6.71 ± 0.30 and 5.71 ± 0.36 kcal/mol, respectively), implying a larger impact of the Arg335Lys mutation on the association between αIIbβ3 and PT25-2.

Our structural and functional data provide insights into the mechanism by which mAb PT25-2 induces the αIIbβ3 receptor to bind ligand. They also provide information about the conformations of αIIbβ3 on platelets and the equilibrium among conformations. The structure of the complex showed that PT25-2 forms interactions with both αIIb Arg335 and His338, providing a structural basis for the loss of PT25-2 binding when these were mutated to the respective murine residues Lys and Gln.¹⁰  Analysis of binding free energy values identified the Arg335Lys mutation as more deleterious for PT25-2 binding than the His338Gln mutation. Our cryo-EM structure supports an especially important role for Arg335 because it interacts with 5 residues in PT25-2. We extend on previous studies by Puzon-McLaughlin et al¹⁰  by identifying additional binding interactions for the nearby residues Ala339, Leu340, as well as interacting sites on other regions: Asp271 on one β strand, residues Gln275, Arg276, and His278 on a neighboring β strand, and calcium-binding loop residues Glu243, Asp245, and Gly246. We also mapped all the PT25-2 HC and LC residues from all the CDRs that make interactions with these residues.

Overlay of the αIIbβ3–PT25-2 Fab complex with the crystal structure of the αIIbβ3 ectodomain demonstrates substantial clashes of the Fab with the αIIb calf domains of the integrin in the bent-closed conformation, which is thought to be the dominant conformation on unactivated platelets. This finding suggested that PT25-2 can only bind to αIIbβ3 when the receptor is in an extended conformation. We tested this notion directly by analyzing the complex by negative-stain EM and found that, indeed, when complexed with PT25-2, αIIbβ3 cannot adopt the bent-closed conformation but exists in a range of extended conformations. Notably, there was no evidence of swing-out of the β3 hybrid domain, which is characteristic for the extended-open conformation that is adopted by integrins that are fully activated, as for example by RGD peptide binding. This finding is not surprising considering that PT25-2 does not engage the RGD-binding pocket and that the swing-out motion is thought to be triggered by the ligand carboxyl group engaging the MIDAS metal ion.^32,34,35  This is also consistent with the finding that PT25-2 does not itself initiate outside-in signaling.¹ 

We compared the kinetics of PT25-2 binding to unactivated platelets versus the binding of 2 mAbs that we previously characterized, mAb 10E5, which binds very rapidly, presumably because its epitope is always available on the surface of αIIbβ3, and mAb 7E3, which binds slowly, presumably because it selectively binds to conformations that are adopted by the receptor relatively infrequently.^27,28  We found that PT25-2 binds more slowly than 10E5, but more rapidly than 7E3, suggesting that, although the PT25-2 epitope is on the surface of the αIIb β propeller, access to the epitope is restricted in the bent-closed conformation. We conclude that a wider range of partially extended αIIbβ3 conformations can accommodate PT25-2 binding than 7E3 binding and that binding of PT25-2 prevents αIIbβ3 from adopting its fully bent conformation.

Our findings help to explain the functional data on PT25-2 previously reported. In their original description of PT25-2,¹  Tokuhira et al found that (1) both intact PT25-2 and PT25-2 Fab can induce platelet aggregation of platelet-rich plasma or washed platelets in the presence of fibrinogen; (2) PT25-2 induces fibrinogen binding to platelets; (3) ∼100 000 PT25-2 molecules bind per platelet; (4) dissociating αIIbβ3 with EDTA at 37°C results in loss of PT25-2 binding; (5) PT25-2–induced platelet aggregation does not result in ATP release or calcium mobilization and is not inhibited by prostaglandin E₁ (PGE₁) or aspirin; and (6) PT25-2 binding is not affected by thrombin stimulation or the binding of an RGD peptide.¹  In additional studies, Litinov et al⁶  used single-molecule analysis to demonstrate that PT25-2 produces a dose-response increase in the rupture force necessary to break the interaction between single αIIbβ3 and fibrinogen molecules using concentrations of PT25-2 up to 100 μg/mL.

Studies of mutations that prevent PT25-2 from inducing ligand binding have been particularly instructive. Takagi et al³⁶  reported that cross-clamping αIIbβ3 by engineering a disulfide bond between the αIIb β-propeller Arg320 and the β3EGF-4 leg domain Arg563 with cysteine mutations resulted in loss of the ability of the combination of Mn²⁺ and PT25-2 to induce fibrinogen binding. Although reduction with dithiothreitol (DTT) alone did not rescue the ability to bind fibrinogen, DTT in combination with Mn²⁺ and PT25-2 produced fibrinogen binding. We previously demonstrated that limiting extension of αIIbβ3 around the αIIb genu by engineering a disulfide bond between αIIb Arg597 and Tyr645 with cysteine mutations prevents PT25-2 from activating recombinant αIIbβ3 to bind fibrinogen or the activation-dependent monoclonal antibody PAC1.³  Reducing the disulfide bond with DTT produced activation by itself, and combining DTT and PT25-2 fully rescued the ability to bind both ligands. We also previously performed similar experiments to limit extension of β3 by engineering Ser367Cys/Ser551Cys, Gly382Cys/Thr564Cys, and Val332Cys/Ser674Cys double mutations designed to stabilize the hybrid/EGF-3, the hybrid/EGF-4, and the βI/βTD interfaces, respectively.⁹  Each of these disulfide bonds resulted in the loss of fibrinogen binding in response to PT25-2, and DTT resulted in each binding fibrinogen.⁹  These data suggest that PT25-2 activation requires that the receptor be able to undergo extension of both subunits. We also found that limiting the swing-out motion of the β3 hybrid domain by engineering a disulfide bond between the αIIb β-propeller loop Lys321 and either β3 Glu358 or Arg360 in the β3 hybrid domain dramatically reduces the ability of PT25-2 to initiate binding of both fibrinogen and PAC1 to αIIbβ3. As with the mutant-limiting extension, reducing the disulfide bond variably rescued the ability to bind both ligands, and adding PT25-2 to DTT further enhanced the binding of both ligands.²  These data suggest that PT25-2–induced ligand binding requires that the receptor be able to undergo swing-out of the hybrid domain. We also studied the impact of modifying the interaction at the interface between the αIIb calf-2 domain and β3 EGF-4/βTD domains, an interaction that we implicated in propagating inside-out signaling through the ectodomain, and found crucial in achieving ligand binding.⁸  Disrupting this interface resulted in constitutive fibrinogen binding, whereas stabilizing it with an engineered disulfide bond eliminated the constitutive activation produced by truncating mutants of both αIIb and β3, and reduced by half ligand binding induced by PT25-2. Thus, because we found that PT25-2 binding likely destabilizes the αIIb calf-2 domain interactions with the EGF-4/βTD domains, it may also weaken the interactions of the β3 leg with the β3 headpiece, facilitating extension.

In conclusion, our single-particle EM analysis provides structural data that confirm the results from prior mutational analysis and identify additional sites on the αIIb β propeller to which PT25-2 binds. They also support a model in which PT25-2 binds a specific epitope that is only accessible when αIIbβ3 is in one of its extended-closed conformations. Thus, as receptors adopt an extended conformation and bind PT25-2, their RGD-binding pockets become accessible and the receptor is primed to bind ligands. The PT25-2-bound αIIbβ3 structure does not, however, show a swung-out β3 hybrid domain, suggesting that this conformational change either occurs downstream of receptor extension or stochastically in response to ligand binding. This model is also consistent with data on receptor extension induced by agonist activation as judged by the binding of mAb 7E3,^27,28  and all of the studies in which preventing receptor extension by mutagenesis limited hybrid domain swing-out.^2,3,8,9 

From a translational perspective, theoretically, agents patterned after PT25-2 may be of value as topical hemostatics in patients who have αIIbβ3, but who are unable to activate it via inside-out signaling, such as patients with mutations in the genes coding for kindlin-3 (type-III leukocyte adhesion deficiency)³⁷  or the calcium-sensing guanine nucleotide exchange factor,³⁸  or patients with variant Glanzmann thrombasthenia with mutations that affect signal transduction.³⁹  Similarly, improved understanding of the transition from the unactivated to the extended-closed and extended-open conformations may identify new targets for antithrombotic agents.

The authors thank Stanka Semova and Svetlana Mazel of the Rockefeller University Flow Cytometry Resource Center for assistance with cell sorting and Alexandra E. Pagano and Henrik Molina from the Rockefeller Proteomics Resource Center for assistance with LC-MS/MS determination of the amino acid sequence of PT25-2. The authors also thank Mark Ebrahim and Johana Sotiris at the Evelyn Gruss Lipper Cryo-EM Resource Center of The Rockefeller University for assistance with cryo-EM data collection.

Simulations and analyses were run on resources available at the Icahn School of Medicine at Mount Sinai through the Office of Research Infrastructure of the National Institutes of Health under awards S10OD018522 and S10OD026880, as well as the Extreme Science and Engineering Discovery Environment under MCB080077 (M.F.), which is supported by National Science Foundation grant ACI-1548562. This work was also supported in part by grant 19278 from the National Heart, Lung, and Blood Institute, and grant UL1TR001866 from the National Center for Advancing Translational Science, National Institutes of Health.

Contribution: D.N. designed, performed, and analyzed experiments to produce PT25-2 Fab and recombinant αIIbβ3 and determined the sequence PT25-2; M.B. designed, performed, and analyzed EM experiments; A.S. designed, performed, and analyzed computational studies; J.L. designed, performed, and analyzed LIBS and antibody binding studies; T.K. and M.H. provided the PT25-2 antibody and hybridoma and critiqued the manuscript; M.F. designed, oversaw, and analyzed computational studies and wrote the article; T.W. designed, oversaw, and analyzed EM studies and wrote the article; and B.S.C. designed, oversaw, and analyzed the protein experiments and had primary responsibility for writing the article.

Conflict-of-interest disclosure: B.S.C. has royalty interests in abciximab (Centocor) through the Research Foundation of the State University of New York and the VerifyNow Assays (Accumetrics/Instrumentation Laboratories), and is an equity holder and consultant to Scholar Rock, CeleCor Therapeutics, and Pulmoquine Therapeutics. The remaining authors declare no competing financial interests.

Correspondence: Barry S. Coller, Allen and Frances Adler Laboratory of Blood and Vascular Biology, The Rockefeller University, 1230 York Ave, New York, NY 10065; e-mail: collerb@rockefeller.edu.