Platelet Fragmentation Requires a Specific Structural Conformation of Human Monoclonal Antibody against β3 Integrin*

We have described an autoantibody against β3 (GPIIIa49–66), a region of platelet integrin αIIbβ3 that is unique. It induces platelet fragmentation in the absence of complement via antibody activation of platelet NADPH oxidase and 12-lipoxygenase to release reactive oxygen species, which destroy platelets. To study the mechanism of anti-GPIIIa antibody-induced platelet fragmentation, we screened a human single chain Fv antibody library with the GPIIIa49–66 peptide. Nine monoclonal antibodies were identified that were capable of binding to GPIIIa49–66. Surprisingly, binding avidity for GPIIIa49–66 did not correlate with activity of induction of platelet fragmentation. We therefore investigated the requirements for platelet fragmentation. Mutations were introduced into the heavy chain complementary-determining region-3 of clones 11, 43, and 54 by site-directed mutagenesis. The capability of these clones to induce platelet fragmentation or bind to GPIIIa49–66 subsequently changed. Molecular modeling of these clones with their mutants revealed that the ability to induce platelet fragmentation is affected by the side chain orientation of positively charged amino acids in the heavy chain of residues 99–102. Thus, a structural change in the conformation of anti-GPIIIa49–66 antibody contributes to its binding to the β3 integrin and subsequent antibody-induced platelet fragmentation and aggregate dissolution.

Patients with early HIV-1 3 infection commonly develop an immunologic thrombocytopenia (HIV-1-ITP) with shortened platelet survival because of an autoantibody specific against an epitope of integrin subunit ␤3 (GPIIIa) on the surface of platelets, GPIIIa49 -66 (CAPESIEFPVSEAR-VLED) (1,2). Anti-GPIIIa49 -66 autoantibody induces thrombocytopenia in mice and correlates inversely with platelet count in HIV-1-ITP patients. Rabbit polyclonal antibody raised against the GPIIIa49 -66 peptide induces platelet fragmentation in the same manner as the antibody from HIV-1-ITP patients (3). The development of this antibody in HIV-1-ITP patients is due to molecular mimicry of epitopes on polymorphic regions of HIV protein: Nef, Env, and Gag (4). The antibody is unique in that it induces complementindependent platelet fragmentation in vitro by the generation of reactive oxygen species released through activation of 12-lipoxygenase and NADPH oxidase (3,5). Because anti-GPIIIa antibodies against other regions of ␤3 do not induce platelet fragmentation (3), it is likely that a unique mechanism is employed in anti-GPIIIa49 -66 antibody-induced platelet fragmentation.
The extracellular domain of the ␤3 integrin subunit consists of a head and stalk region (6). The head includes a ␤A domain with a metal ion-dependent adhesion site motif and one hybrid domain. The stalk consists of a PSI (plexin/semaphorin/integrin) domain, four tandem epidermal growth factor repeats, and a ␤ tail domain (7). Although the function of the PSI domain is incompletely understood, a Pl A2 polymorphism of the ␤3 integrin (Leu-33 to Pro-33) has been linked to this region, which leads to an increased risk of thrombotic events (8,9). It has been suggested that the Pl A2 allele is associated with greater affinity of the ␣IIb␤3 integrin for fibrinogen (10,11). In addition, the Pl A2 polymorphism can result in a severe form of alloimmune thrombocytopenia (12,13). Thus, the PSI domain can be considered as a domain with strong immunogenicity as well as the potential to regulate the binding of ␣IIb␤3 integrin to fibrinogen. We have reported previously that the anti-platelet autoantibody developed in HIV-1-ITP patients is immunodominant against GPIIIa49 -66 (a region that crosses the PSI and hybrid domains) (3)(4)(5). However, it is unclear how antibody interacts with this region between the PSI and hybrid domains to induce the downstream signaling and oxidative platelet fragmentation.
To investigate the detailed mechanism of antibody-induced binding and fragmentation, monoclonal antibody is required because some polyclonal antibodies often have cross-reactivity with other antigens. However, attempts at raising monoclonal antibody against human GPIIIa49 -66 in mice were unsuccessful. This is likely due to homology sequences shared with human GPIIIa49 -66 leading to mouse immune tolerance. We therefore switched to the phage surface display antibody technology system.
Phage surface display antibody technology has been successfully employed to generate antibody that specifically blocks the activated form of ␣IIb␤3 integrin (14) and to investigate the anti-␣IIb␤3 autoantibody profile from patients with autoimmune thrombocytopenic purpura or Glanzmann thrombasthenia after receiving blood transfusions (15). In the present study, we have employed human scFv (single chain fragment of the variable region) Tomlinson I ϩ J libraries to generate human anti-GPIIIa49 -66 monoclonal antibody. Each library contains over 100 million different scFv fragments cloned in the phage-mid vector pIT2. The library sizes of I and J are 1.47 ϫ 10 8 and 1.37 ϫ 10 8 constructs, respectively (16 -18). Both libraries are based on a single human framework for VH (V3-23/ DP-47 and JH4b) and V (O12/O 2 / DPK9 and J1) with side chain diversity incorporated at positions in the antigen-binding site that make contacts to antigen in known structures and are highly diverse in the mature repertoire. Thus, these libraries can be used to screen antibodies against virtually any given antigen.
Therefore, we used these libraries to better understand the mechanism of antibody-induced oxidative platelet fragmentation with respect to interaction with the PSI and hybrid domains of the ␤3 integrin subunit. In the present study, we show that 1) eight scFv monoclonal antibodies induce platelet fragmentation; 2) there is no correlation between binding activity and ability to induce platelet fragmentation; 3) the presence and position of positively charged amino acids in the complementary-determining region-3 (CDR3) of the heavy chain can affect platelet fragmentation as well as binding; 4) these scFv monoclonal antibodies are capable of destroying platelet aggregates; and 5) one of the scFv antibodies tested decreases the platelet count and prolongs the bleeding time in mice. These results strongly suggest that a specific structural conformation is required for antibody-induced platelet fragmentation and that appropriate scFv antibodies might be useful clinically in the dissolution of platelet thrombi.

EXPERIMENTAL PROCEDURES
Materials-All reagents were obtained from Sigma unless otherwise designated. Human scFv Tomlinson I ϩ J libraries were kindly provided by MRC Geneservice (Cambridge, UK).
Screening of the Phage Display Peptide Library-GPIIIa49 -66-biotin peptide (130 nM) was incubated with 10 12 -10 13 phage from the Tomlinson I or J library in 2% nonfat milk in PBS at room temperature for 3 h. Avidin-agarose beads were then added overnight at 4°C, centrifuged, and washed with 0.1% Tween 20 and PBS 20 times, and positive phages were eluted with trypsin/PBS by shaking for 10 min at room temperature.  3). B, ELISA after three rounds of panning. Individual clones from library J were measured for their ability to bind to the GPIIIa49 -66 peptide coated on the microplate. The concentration of phage antibodies was normalized, and the dilutions (black and gray columns) were 1:3 and 1:9 in PBS, respectively (n ϭ 3). C, binding ability of clones 43 and 54 with respect to clone 11, measured by standard curve. x axis numbers 1-8 refer to dilutions of 1:2-1:9, respectively (n ϭ 3). 13CG2 is a control scFv antibody.
Eluted phages were then titered and amplified following the protocol provided by MRC Geneservice for the next round of panning. After three rounds of panning, positive phage clones were randomly picked up from the plate.
Induction of Platelet Particle Formation-Gel-filtered human platelets were prepared from platelet-rich plasma obtained from blood collected in 0.38% sodium citrate utilizing a Sepharose 2B column preincubated with Tyrode's buffer (pH 7.4). 1 ϫ 10 7 gel-filtered platelets/ml were labeled with either anti-GPIIb-fluorescein isothiocyanate-labeled monoclonal antibody (3B2) or anti-GPIIIa-fluorescein isothiocyanate-labeled monoclonal antibody (Ancell, Bayport, MN) to 10 g/ml for 30 min at 4°C, centrifuged at 1000 ϫ g for 6 min at room temperature, and resuspended in Tyrode's buffer. 10 l of fluorescein isothiocyanate-labeled platelets (10 7 /ml) were then incubated with human scFv anti-GPIIIa49 -66 antibody (5-40 g/ml) and 75 l of Tyrode's buffer for 0 -4 h at 37°C and stored in an ice bucket prior to measurement of the percentage of platelet particles by flow cytometry. Further particle formation was arrested at 0°C.
Fluorescently labeled platelet particles were measured by flow cytometry, employing a FACScan (BD Biosciences) as reported previously (3). Debris and dead cells were excluded using scatter gates. Only cells with low orthogonal light scattering were included in the sorting gates. Gates were adjusted for control platelets by exclusion of other blood cells. Fluorescently labeled intact platelets were monitored in the right upper quadrant with the y axis measuring forward scatter and the x axis measuring fluorescence. A shift in fluorescent particles from the right upper quadrant to the left upper and lower quadrants reflected the percentage of platelet particle induction of 10,000 counted platelets/particles.
Destruction of Platelet-Fibrinogen Aggregates in Vitro-1 ϫ 10 7 platelets were incubated with 100 mg/ml fibrinogen and 10 M ADP for 30 min with shaking at 37°to create platelet aggregates. Excess reagents were removed by centrifugation at  (n ϭ 3). CtlIgG, control IgG. D, inhibition of functional clones 54 and 11 from inducing platelet fragmentation by nonfunctional clone 43 and irrelevant clone CG2. Serial dilutions of clones 11, 43, 54, and CG2 were made with PBS. For inhibition assays, serial dilutions of clones 11 and 54 were mixed with fixed amounts of clone 43 or CG2 to make the ratio between 11 or 54 with 43 or CG2 as 1:1, 1:2, 1:4, and 1:8 (n ϭ 2). E, prolongation of bleeding time with 2-4-h preinfusion of functional clone 11 compared with nonfunctional clone 43 (n ϭ 11). pended in PBS on a counting chamber, and the number of platelets/aggregate was counted. Anti-GPIIIa49 -66 antibody was then added for various time intervals, and the remaining number of platelets/aggregate was enumerated.
Bleeding Time-The mouse tail vein was severed 2 mm from its tip and blotted every 30 s on a circular sheet of filter paper to obtain an objective measurement. The variation of the bleeding time was recorded after absence of blood on the filter paper. Bleeding time differences were recorded by an unbiased observer and confirmed by two other observers.
Site-directed Mutagenesis-Two mutation primers (forward and backward) targeted for amino acids at positions 99 and 101 were synthesized by Sigma Genosys. The 5Ј and 3Ј fragments of mutant heavy chain cDNA were amplified using the mutant primers. The full-length mutant heavy chain cDNA fragment was assembled by splicing overlap extension PCR. The mutant heavy chain cDNA was then cloned into the pIT2 expression phage vector between the restriction enzyme cutting sites of SfiI and XhoI.
Molecular Modeling-The heavy and light chain amino acid sequences of clones 11, 43, and 54 and their mutants were used to generate the molecular model through WAM-Web antibody modeling. The side chains of amino acids at positions 99 and 101 in the heavy chain have been labeled with different colors using Rasmol software.

Generation of Human scFv Anti-GPIIIa49 -66 Monoclonal
Antibodies-Tomlinson I ϩ J libraries were used to screen against a biotin-conjugated GPIIIa49 -66 peptide. After each round of screening, binding activities of phage clones were examined by ELISA (Fig. 1A). After a second round of panning, positive phage clones from library J were enriched. Because the binding activities of phage clones from library I were not increased after three rounds of panning, we focused on phage clones from library J. After three rounds of panning, scFv clones were randomly chosen from the bacterial plate, and individual scFv antibodies were made from each clone. The binding activity for GPIIIa49 -66 was examined by ELISA (Fig. 1B). All nine clones were found to bind to GPIIIa49 -66 with varying activity, with the absorbance reading varying from 3 to 40 times background (Fig. 1B). Two clones tested did not bind to ␤3 Ϫ/Ϫ intact mouse platelets, indicating specificity of binding to intact ␤3 platelets (data not shown).
Effect of scFv Antibodies on Platelet Fragmentation and Destruction of Platelet Aggregates-We then tested the ability of each clone to induce platelet fragmentation. We found that not all clones induced platelet fragmentation ( Fig. 2A) and that the binding activity (Fig. 1B) did not correlate with platelet fragmentation. This suggested that a specific conformation induced by the antibody is more important than the avidity of binding. For example, clone 43 bound to GPIIIa49 -66 (Fig. 1B) but did not induce platelet fragmentation. In contrast, clone 54 had weak binding to GPIIIa49 -66 but induced significant platelet fragmentation ( Fig. 2A). We confirmed the findings with clones 43 and 54 by measuring the binding and ability to induce platelet fragmentation against a standard curve created with clone 11, a potent binding and functional antibody. Fig. 1C demonstrates similar binding for clones 43 and 54, which is about 6-fold less potent compared with clone 11. Fig. 2B demonstrates similar platelet fragmentation function for clones 11 and 54, which is about 24-fold compared with clone 43. We also tested the functional ability of clones 37, 41, 49, and 54 to disaggregate/destroy ADP-induced in vitro platelet aggregates. All clones tested destroyed platelet aggregates in the same manner as antibody from an HIV-1 patient (Fig. 2C). We also tested the ability of the absent functional fragmentation clone 43 to inhibit strong functional fragmentation clones 54 and 11. As noted in Fig. 2D, clone 43 inhibited fragmentation induced by clones 54 and 11, whereas no such inhibition was noted with irrelevant clone CG2. One functional clone tested, clone 11, increased the in vivo bleeding time 3-fold (n ϭ 11; p ϭ 0.001 compared with 43) (Fig. 2E) and decreased the platelet count by ϳ30% from 1.096 Ϯ 0.039 to 0.740 Ϯ 0.020 million/l; p Ͻ 0.003.
Sequence Analysis of the Heavy and Light Chains of Antibody Clones-Nine clones had sequences for both heavy and light chains. The alignment of sequences is shown in Fig. 3 Function of Mutants of Clones 11,43, and 54-A series of mutant clones was generated by site-directed mutagenesis of positively charged amino acids. The amino acids in HCDR3 of clones 11 (Lys-100, Ser-101), 43 (Tyr-99, Arg-101), and 54 (Arg-99, Ser-101) were exchanged to generate mutants 11M1, 43M, and 54M (Fig. 4A). In clone 11, Lys-100 and Ser-101 were mutated to Ser-100 and Lys-101 to generate mutant 11M1. Lys-100 and His-102 were mutated to Gly to generate mutant 11M2 (Fig. 4A). In clone 43, Tyr-99 and Arg-101 were mutated to Arg-99 and Tyr-101 to generate mutant 43M. In clone 54, Arg-99 and Ser-101 were mutated to Ser-99 and Arg-101 to generate mutant 54M. The binding activities of mutants 11M1, 11M2, 43M, and 54M were then examined by ELISA (Fig. 4B). The binding activity of clones 11M2 and 43M decreased compared with their wild-type sequences, whereas 54M increased and 11M1 remained the same compared with their wild-type sequences. In addition, 43M gained the ability to induce platelet fragmentation, whereas 11M1, 11M2, and 54M decreased their platelet fragmentation function (Fig. 4C). Of particular note was 11M2, which still had some binding activity as well as function without Lys-100 and His-102 of HCDR3. These data suggest that a specific conformation is required and that the positively charged amino acids in HCDR3 may affect antibody function.
Molecular Modeling-To examine the structural differences between the wild-type and mutant clones, molecular models of clones 11, 43, and 54 and their mutant clones were generated by computer programs WAM and Rasmol (Fig. 5). According to the molecular models, although clones 43M and 54M have different side chain orientations of Arg compared with their wild-type sequences, we found that 43M has similar Arg side chain orientation to 54. 54M has similar Arg side chain orientation to 43. When correlated with the functional assay of clones 43, 54, 43M, and 54M, these observations suggest that a specific conformation is required for inducing platelet fragmentation. In contrast, comparing clone 11 with clones 43 and 54, we have not found significant differences in the side chain orientation of Lys-100 and His-102 in clone 11 with Lys-101 and His-102 in clone 11M1. This is consistent with the fact that there is no significant difference in function between 11 and 11M1.

DISCUSSION
␣IIb␤3 integrin function has been intensively studied (6). Although there is still disagreement as to how the ␣IIb␤3 integrin transforms from a low affinity to a high affinity state, it is generally accepted that the ␣IIb␤3 integrin transmits conformational changes bidirectionally: inside-out and outside-in signaling (19). Clustering of the ␣IIb␤3 integrin subunit is required for outside-in signal transmission. It has been proposed that during the conformational change of ␣IIb␤3 integrin, the reorganization of hydrogen bonds in the interface between the ␣7 helix of the ␤A domain allows the hybrid domain and the rigidly connected PSI domain to swing out. This results in a 70-Å separation of the ␣IIb and ␤3 stalks at their "knees," a feature noted in electron microscopic images of active forms of ␣IIb␤3 integrin in the presence or absence of ligand (7). We hypothesize that our antibody may induce the PSI domain to swing out and cluster the ␤3 subunit. This is supported by our recent discovery that a biotin-labeled peptide capable of binding to GPIIIa49 -66 requires anti-biotin antibody-induced clustering to induce oxidative platelet fragmentation. 4 How such a movement leads to activation of 12-lipoxygenase and NADPH oxidase in platelets is still not clear.
In the present study, we have clearly shown that nine monoclonal antibodies against GPIIIa49 -66 can be generated by panning with Tomlinson I ϩ J scFv libraries. It is unlikely that these antibodies target other molecules on the platelet surface,  as could be the case with polyclonal antibody. Further specificity of our scFv monoclonal antibody was demonstrated by its inability to react with ␤3 Ϫ/Ϫ mouse platelets. Of great interest was the observation that binding avidity for the GPIIIa49 -66 peptide did not correlate with the ability to induce platelet fragmentation. Clone 43 does not induce platelet fragmentation despite its avid binding to the GPIIIa49 -66 peptide, whereas the other eight clones induce fragmentation. It is likely that the 43 binding conformation is different compared with the antibodies that induce fragmentation. Those antibodies that induce platelet fragmentation are indeed doing this through the ␤3 integrin because they fail to induce platelet fragmentation with platelets from ␤3 Ϫ/Ϫ mice. Clones 11 and 43 bind to wild-type mouse platelets by flow cytometry in a similar manner as patient antibody (data not shown). In addition, mutation analysis of clones 11, 43, and 54 shows that changing the positions of positively charged amino acids in the HCDR3 region alters the function of antibody-induced platelet fragmentation, suggesting that the positively charged amino acid side chain orientation may affect antibody-induced platelet fragmentation.
GPIIIa49 -66 (CAPESIEFPVSE-ARVLED) contains five negatively charged amino acids, four glutamic and one aspartic acid. It is therefore not surprising that there are many positively charged amino acids (Arg, Lys, and His) in the CDRs of both heavy and light chains of our antibodies. Compared with other anti-␣IIb␤3 antibodies, our antibodies do not contain RGD sequences; therefore, it is unlikely that our antibodies can cross-react with the RGD binding site of ␣IIb␤3 integrin (20). Although we have shown that the positively charged amino acids in the HCDR3 play a role in inducing platelet fragmentation, this result does not exclude the possible role(s) of other CDRs in heavy and light chains. In fact, after we mutated both positively charged Human Anti-GPIIIa Antibody FEBRUARY 8, 2008 • VOLUME 283 • NUMBER 6 amino acids in 11 HCDR3, 11M2 still retained some platelet fragmentation capability, suggesting that other regions also contribute to induce platelet fragmentation. Nevertheless, our data establish the concept that a specific conformation with ␤3 is required for antibody-induced platelet fragmentation.
The ␣IIb␤3 integrin has been utilized as a therapeutic target for thrombosis and occlusion due to coronary artery stenting (21,22). Side effects of these drugs are thrombocytopenia and bleeding. Development of a different approach to inhibit arterial platelet thrombi by disaggregation/lysis of platelet thrombi with the specific scFv anti-GPIIIa49 -66 antibody may be of clinical value.