Molecular insight into human platelet antigens: structural and evolutionary conservation analyses offer new perspective to immunogenic disorders

BACKGROUND: Human platelet antigens (HPAs) are polymorphisms in platelet membrane glycoproteins (GPs) that can stimulate production of alloantibodies once exposed to foreign platelets (PLTs) with different HPAs. These antibodies can cause neonatal alloimmune thrombocytopenia, posttransfusion purpura, and PLT transfusion refractoriness. Most HPAs are localized on the main PLT receptors: 1) integrin αIIbβ3, known as the fibrinogen receptor; 2) the GPIb-IX-V complex that functions as the receptor for von Willebrand factor; and 3) integrin α2β1, which functions as the collagen receptor. STUDY DESIGN AND METHODS: We analyzed the structural location and the evolutionary conservation of the residues associated with the HPAs to characterize the features that induce immunologic responses but do not cause inherited diseases. RESULTS: We found that all HPAs reside in positions located on the protein surface, apart from the ligand-binding site, and are evolutionary variable. CONCLUSION: Disease-causing mutations often reside in highly conserved and buried positions. In contrast, the HPAs affect residues on the protein surface that were not conserved throughout evolution; this explains their naive effect on the protein function. Nonetheless, the HPAs involve substitutions of solvent-exposed positions that lead to altered interfaces on the surface of the protein and might present epitopes foreign to the immune system.

P olymorphisms in platelet (PLT) membrane glycoproteins (GPs) are responsible for alloantibody production upon exposure to PLTs with different human PLT antigens (HPA). These antibodies can cause neonatal alloimmune thrombocytopenia (NAIT), posttransfusion purpura, or PLT transfusion refractoriness after exposure to unmatched PLTs. The molecular basis of the 24 serologically defined antigens had been resolved 1 and found to be a single-nucleotide polymorphism (SNP) resulting in an amino acid substitution in 23 of 24 antigens: 12 HPAs are grouped in six biallelic systems (HPA-1 to -5 and -15) in which alloantibodies against both the common (designated a) and the rare (designated b) alleles were observed. For the rest, only alloantibodies against the rare allele have been detected. The only non-SNP is HPA-14bw, an in-frame triplet deletion coding for one amino acid (b3-Lys611). 2 Overall, there are 17 different HPAs. A nomenclature was devised by the International Society of Blood Transfusion and the International Society of Thrombosis and Haemostasis (see Table 1).
Most HPAs are localized to the main PLT receptors, namely, integrin aIIbb3 that is also known as the fibrinogen receptor, the complex GPIb-IX-V that functions as the receptor for von Willebrand factor (VWF) and integrin a2b1-the collagen receptor. Exceptional is HPA-15, which is carried by the glycosylphosphatidylinositollinked protein CD109 3 that was found to be a part of the transforming growth factor (TGF)-b receptor system and functions as a negative regulator of its signaling. 4 Integrins are adhesion receptors that mediate vital bidirectional signals within the cell. 5 They form heterodimers of an a and a b subunit, both Type I membrane proteins with large extracellular segments. 6,7 There are 24 known heterodimers in mammals, composed of 18 a and eight b subunits. 5 The NH2-terminal ectodomain of the a and b subunits assembles into an ovoid "head" and two "legs" composed of several domains. The remaining segments form two tails that span the plasma membrane. 6,7 Integrin b3 subunit is known as PLT membrane GPIIIa that forms heterodimers with aIIb and av integrins subunits; aIIbb3 is the most abundant PLT receptor with approximately 80,000 copies per PLT. The structure of the inactive extracellular region of avb3 was the first one reported (PDB ID 1JV2 7 ). The structure of the entire ectodomain of aIIbb3 was determined in the inactive conformation (PDB ID 3FCS 8 ). The active conformation of aIIbb3 was determined only for part of the ectodomain (e.g., PDB ID 1TXV 6,9 and 3FCU 8 ). For the structural analyses, we used the crystal structures relevant to each of the HPAs.
Eleven HPAs are located on the most abundant receptor on PLTs surface: integrin aIIbb3. Nine of the HPAs are mapped to the b3 subunit that was found to be the most polymorphic GP on the PLT membrane. 10 Some of the HPAs are frequent in the Caucasian population (HPA-3 and HPA-15) but most of them are rare and some are private polymorphisms, restricted to one family  11 Only 10% of HPA-1b-homozygous women with an HPA-1a-positive fetus develop anti-HPA-1a alloantibodies during pregnancy. 12 Maternal responsiveness to HPA-1a shows strong association with DRB*0101 allele encoding human histocompatibility leukocyte antigen DR52a. It was shown that short peptide containing HPA-1a (Leu33) bind to recombinant DR52a molecules, whereas the HPA-1b (Pro33) version does not. Interestingly, although most HPAs are nonsynonymous SNPs and HPA-14w is even a single-amino-acid deletion, they probably inflict a minor effect on PLT receptors function since they do not cause inherited diseases, and even their influence on thrombus formation is controversial. 13 The major complication is their involvement in immunogenic response.
In this article we analyzed the structural features and evolutionary conservation of the HPAs to define the characteristics that lead to immunologic problems but not to inherited disease. We demonstrate that all HPAs studied here involve residues located on the surface of the protein far from the ligand-binding site and are not evolutionary conserved. These criteria imply that the substitutions in these positions present a minor effect on the structure and function of the protein and therefore, do not represent disease-causing mutations. Moreover, their location on the surface suggests that they could play a role in presenting epitopes.

Evolutionary conservation analyses
The evolutionary conservation analyses were calculated using the Bayesian method 14 implemented in the ConSurf Web server (http://consurf.tau.ac.il) 15 or the ConSeq Web server (http://conseq.tau.ac.il/). 16 The conservation scores range from 1 to 9, indicating variable-to-conserved positions.

a-Integrins
A multiple sequence alignment (MSA) of the b-propeller domain in various a-integrins was constructed as described previously. 17 This alignment was used to calculate the conservation scores for the HPAs in the b-propeller domain of a2. An MSA of the thigh and calf domains was taken from the a-integrin family in the Pfam database (http://pfam.sanger.ac.uk/). This alignment corresponds to Residues 481 to 921 in aIIb and was used to calculate the conservation scores for the HPAs in the calf domain of aIIb. The last two strands of the calf2 domain constitute a very variable region among a-integrins and therefore were not included in the Pfam alignment. Since one of the HPAs in a2 is located on the C-terminus of the calf2 domain, we generated a smaller MSA via ConSeq, using a2 human as the query sequence to collect 50 homologs from the Swiss-Prot database 18 that were aligned using MUSCLE. 19 We note that the conservation scores for HPA-13 located in calf1 were identical when using in the Pfam and ConSurf generated alignment.

b-Integrin
Human b3 (Swiss-Prot entry ITB3_HUMAN) was used as query to collect homologous sequences from the Swiss-Prot database 18 using PSI-BLAST. 20 The resulted 27 sequences were aligned using CLUSTALW 21 with default parameters. This intermediate alignment was used to generate a hidden Markov model, 22 which was subsequently utilized to collect remote homologous sequences from the UniProt database 23 that were aligned using MUSCLE. 19 From the 105 hits found, redundant (above 95% sequence identity) and fragmented sequences, as well as sequence variants and mutants, were discarded along with sequences that included irregular characters or ones that were sequenced by the whole genome shotgun project therefore considered as preliminary data. The resulted alignment contained 66 b-integrins and was used to calculate the evolutionary conservation scores in b-integrins. To support the analysis, we also used an alignment generated by ConSurf using 300 homologous sequences collected from the UniProt database 23 and aligned using MUSCLE. 19 The conservation scores were very similar.

Leucine-rich repeat family of proteins
Human GPIba (PDB ID 1SQ0, Chain B) 24 was used as query to calculate the evolutionary conservation scores via the ConSurf Web server. 15 The calculations were based on an MSA constructed from 200 homologous sequences from the Swiss-Prot database 18 collected using PSI-BLAST and aligned with MUSCLE. 19 Since the structure of human GPIbb (Swiss-Prot entry: GP1BB_HUMAN) is not available, we used the ConSeq Web server 16 to calculate the evolutionary conservation scores using 90 sequences collected from the Swiss-Prot database with PSI-BLAST 20 and aligned by MUSCLE. 19

CD109
Since the structure of human CD109 is not available (Swiss-Prot entry: CD109_HUMAN), the ConSeq Web server 16 was used to calculate the conservation scores of human CD109 based on 66 sequences collected from the Swiss-Prot database 18 using PSI-BLAST 20 and aligned by MUSCLE. 19

Structural analyses
The analyses were based on the crystal structures of aVb3 (PDB ID 1JV2), 7 aIIbb3 (PDB ID 1TXV, 6,9 3FCS, 8 and 3FCU), 8 and GPIba (PDB ID 1SQ0). 24 The structure of a2 was modeled based on the crystal structure of aIIbb3 (PDB ID 3FCS 8 ) using NESTs 25 with default parameters. The alignment between aIIb and a2 was derived from an MSA of 50 homologs sequences of a-integrins generated by ConSurf as described above. The b-propeller domain, lacking the I domain, was modeled separately using an alignment with aIIb derived from an MSA of the b-propeller domain of a-integrins constructed as described previously. 17 The solvent-accessible surface area (SASA) was calculated using the SURFV program with a probe sphere of radius 1.4 Å and default parameters. 26 The percentage of the surface exposure of each residue in the monomer was calculated from the total solvent-accessible area on a Gly-X-Gly tripeptide (where X represents each of the 20 amino acids). We consider a residue to be buried if less than 5% of its surface is accessible to the solvent. 27 More importantly, we examined the location of the residue in relation to the surface of the protein.

Statistical analysis
Data are expressed as mean Ϯ standard deviation. Kruskal-Wallis test was used to analyze the difference between groups using computer software (GraphPad PRISM5, GraphPad Software, Inc., San Diego, CA). A p value of less than 0.05 was considered as a significant difference. In case that the SASA was calculated in various structures, the mean value was used for the statistical analysis.

HPA polymorphisms in integrin b3
Integrin b3 ectodomain is composed of eight domainsthe plexins, semaphorins, and integrins (PSI); bA; hybrid; four epidermal growth factor (EGF) domains; and a b-tail domain (bTD). There are nine HPAs located in integrin b3; seven of them (HPA-4, -7, -8, -10, -11, -14, and -16) are visible in the inactive structures of both avb3 (PDB ID 1JV2) and aIIbb3 (PDB ID 3FCS). Structural analyses of the crystal structures revealed that all HPAs are exposed to the solvent (Table 1) and, moreover, are located on the surface of the protein and thus do not play a role in the stabilization of the protein fold.
Integrin activation is a multistep process that involves transitions between multiple conformations. 6 Correspondingly, the heterodimer is highly flexible. The avb3 and aIIbb3 crystal structures might have captures slightly different conformations of the b3 that led to differences in the solvent accessibility of the HPA-related residues (Table 1). Some HPAs are located close to the interface with adjacent domains on the b subunit. For example, Arg633, the polymorphic residue of HPA-11, and Arg636, the polymorphic residue of HPA-8, both located in the bTD, are close to the hybrid domain. In the inactive complex aVb3, Arg633 forms a salt bridge with Asp393 from the hybrid domain and is surrounded by Leu389, Gly388, and Cys374. In the structure of aIIbb3, also in the inactive conformation, there are some conformational variations with respect to the inactive b3 in the aVb3 complex. In aIIbb3, Arg633 is facing away from the hybrid domain, yet might be still bound with Glu378 on the hybrid domain. Arg636 is in close proximity to Met387 and Gly388 from the hybrid domain in the aIIbb3 complex, while in the aVb3 structure, Arg636 is close to Asn376. Arg489 of HPA-6, located on the EGF-1 domain, is close to the Calf1 domain of the aIIb subunit (in the inactive conformation) but is not in direct contact.
Five polymorphic residues involved in the formation of HPA-1, -4, -7, -10, and -16 are visible in the crystal structures of both the active (PDB IDs 3FCU 8 or 1TXV 6,9 ) and the inactive (PDB ID 3FCS) 8 conformations of integrin aIIbb3, enabling analysis in both states. Different solvent exposures of HPA-4 and -7 were detected due to structural changes in the backbone between the two conformations, as well as different rotamers (side-chain conformation; Table 1). Nevertheless, in both states, all HPA-related residues were found to be located on the surface of the protein, exposed to the solvent. Moreover, all of them are located far away from the ligand binding site and are not in direct contact with the a subunit in the structurally determined conformations (Fig. 1A). Interestingly, Leu33 of the most immunogenic HPA-1a is totally exposed ( Table 1) and found on the tip of a loop connecting two antiparallel b-sheets in the PSI domain. This loop is flexible and particularly long in integrins. 6,7 The loop, and particularly Leu33, is close to the EGF domains in the inactive conformation (Fig. 1A).
Evolutionary conservation analysis of b-integrins from different species revealed that most HPAs occupy evolutionary variable positions (Table 1). Moreover, some of the polymorphic residues (HPA-1b, -4b, -6b, -10b, and -11b) can be found in the corresponding positions in other b-integrins ( Table 2 in bold), supporting the benign effect of the amino acid substitutions on PLTs normal function. The third allele in HPA-1 (Leu33Val) is a particularly naive substitution that was also found in other b-integrins, showing a conservative change, in an exposed location. Interestingly, differences in reactivity of anti HPA-1a antibodies were observed, some bind also to b3-Val33 while others showed significantly reduced reactivity. 28 The residue related to HPA-7 (Pro407) is unique among other HPAs because it shows an intermediate evo- presented as space-filled atoms. It is clear that HPA-2 polymorphic residue is evolutionary variable, surface exposed, and located far from the interface with the ligand VWF. lutionary conservation. Nonetheless, this position can be occupied by a variety of amino acids and is not restricted to proline in b-integrins from other species ( Table 2). Pro407 is located on a loop at the hybrid domain, facing the solvent, and is not in close proximity to other domains in the structurally determined conformation of integrin b3. The substitution (Pro407Ala) probably has a minor effect on the structure.
HPA-14, the single-deletion polymorphism, Lys611del, is located within the bTD, at the C-terminus of the ectodomain. The region encompassing Lys611 forms a helix in the aVb3 structure; thus such a deletion might disrupt the structure. However, this helix is partially unwound in the aIIbb3 inactive conformation (PDB ID 3FCS) raising uncertainties about its stability and contribution to the overall fold.
Recently, two additional low-frequency SNPs, not yet approved as HPAs, were reported within integrin b3: Lys137Gln and Glu628Lys. 29 Both positions are highly evolutionary variable and located on the surface of the protein (Fig. 1A), consistent with the other HPAs.

HPA polymorphisms in integrin aIIb
Integrin aIIb (PLT membrane GPIIb) comprises the b-propeller, thigh, and two calf domains. Two HPAs (HPA-3 and -9) were found in the aIIb integrin. Both related residues (Ile843 and Val837, respectively) are located in the calf2 domain (Fig. 1B), within a highly evolutionary variable region that shows many insertions and deletions among a-integrins. This region is disordered in the crystal structures, suggesting that it is flexible. Val837 is visible in the inactive structure of aIIbb3. 8 It is located on a loop facing the solvent, far from the b subunit or other domains in aIIb (Fig. 1B). Residues 840 to 873 are missing in the crystal structure of aIIbb3 (due to the disorder of that region); thus we could not examine Ile843.
Recently, a low-frequency HPA was found in aIIb corresponding to Thr619Met substitution 29 at the calf1 domain. As observed for the other HPAs, this position is highly evolutionary variable (ConSurf score = 1; see Materials and Methods) and located on the surface of the protein (Fig. 1B).

HPA polymorphisms in integrin a2
Integrin a2 (PLT membrane GPIa) forms hetrodimers with integrin b1 and, similarly to the integrins discussed above, is important for PLT function. Three polymorphic residues in the a2 integrin form antigens, HPA-5, -13, and -17, localized to the b-propeller, calf1, and calf2 domains, respectively. Since the only crystal structure of a2 is for the I-domain in the b-propeller (PDB ID 1AOX 30 ), we modeled the rest of the structure using aIIb (PDB ID 3FCS) 8 as a template.
The second common allele causing NAIT (HPA-5b), affecting Lys505, is located on one of the loops of the b-propeller domain, on the surface of the protein, far from the interface with the b subunit (Fig. 1C). Lys505 corresponds to Arg355 in aIIb; this position, as well as the surrounding residues on the loop, is evolutionary variable ( Table 1). The rare HPA-13b is related to the substitution of Thr799 located within a b-strand in the calf1 domain, facing the solvent (Fig. 1C). This position, which corresponds to Asp636 in aIIb, shows intermediate conservation, yet is occupied by the polymorphic methionine in other a-integrins, suggesting tolerance to such substitution. HPA-17, affecting Thr1087, is located at the C-terminus of the calf2 domain. Thr1087 shows intermediate conservation among the closest 50 sequences homologous to the human a2 (Table 1), yet is located in a region that is highly variable between a-integrins. Integrin a2 forms a complex with b1. Taking into account that the interactions between the a and b subunits resemble those of aIIbb3 and aVb3, all three HPAs on a2 are located on the opposite side from the interface with the b subunit.

HPA polymorphisms in the GPIb-V-IX complex
The GPIb-V-IX complex, another PLT receptor, is composed of four Type I membrane spanning proteins that belong to the leucine-rich repeat (LRR) family of proteins: GPIba, GPIbb, GPIX, and GPV. HPA-2 and HPA-12 are localized to the two first GPs, respectively. Crystal structures are available only for the LRR domain of GPIba (e.g., PDB ID 1SQ0). HPA-2, affecting Thr145, is located in this region, on the surface of the protein, facing the solvent. It is located far from the interface with its ligand, the VWF (Fig. 2D).
The GPIb-V-IX complex is known to regulate thrombin function by distinct interactions. 31 There are two crystal structures of thrombin bound to GPIba (PDB IDs 1P8V 32 and 1OOK 31 ). The two structures present different interpretations to how the two proteins interact. 33 In one of these structures (PDB ID 1P8V), 32 Thr145 is located at the interface with one of the thrombin molecules. Recently a model of GPIbb ectodomain was constructed, based on decorin crystal structure (PDB ID 1XKU 34 ), but is not yet available for analysis. 35

HPA polymorphisms in CD109
CD109 antigen, also known as PLT-specific Gov antigen, is a TGF-b1 binding protein for which no structure is available. The closest homolog with known structure has 26% sequence identity, which is not enough for accurate structural modeling. Tyr703 (Residue 682 in the mature protein), associated with HPA15, is evolutionary variable.

Polymorphisms versus disease-causing mutations
Most HPAs are localized to integrin aIIbb3, specifically to the b3 subunit. 10 Several missense mutations in this integrin were reported to cause severe bleeding disorder named Glanzmann thrombasthenia (GT) (http:// sinaicentral . mssm . edu / intranet / research / glanzmann / menu). Single-nucleotide sequence alterations responsible for both HPAs and GT result in nonsynonymous amino acid substitutions. The interesting question is why some amino acid substitutions cause disease while others are naive polymorphisms. 36,37 Moreover, some SNPs are immunogenic, while others were not reported to cause an immune response. Our working hypothesis is that mutations in residues that are evolutionary conserved and are deeply buried within the fold of the protein or located near the active site, near the ligand binding site or at the interfaces with other domains, are most likely to damage the structure and function of the protein. On the other hand, substitutions in residues that are located on the surface of the protein and are evolutionary variable will be more tolerable.
Having this in mind, we examined the HPAs in terms of evolutionary conservation and structural characteristics. In comparison we analyzed mutations causing GT or Bernard-Soulier syndrome (BSS) previously reported by us (Table 3) and validated (by frequency in any population) SNPs found in NCBI SNP database that were not reported to cause an immune response (Table 4).
Our structural analysis revealed that the SNPs responsible for the HPAs affected nonconserved positions located on the surface of the protein, far away from the ligand binding site and not in direct contact with other subunits. In contrast, disease-causing mutations within integrins often disturb highly conserved and buried positions in the proteins' cores 38 (Table 4, Fig. 2), which imply that they are essential for maintaining the basic structure common to all integrins. Similarly, a missense mutation leading to a rare bleeding disorder, BSS, was mapped to an evolutionary conserved and hydrophobic residue at the core of the LRR region in the GPIb-IX-V complex. 39 In contrast, Thr145 responsible for HPA-2, located at the same repeat region (Fig. 1D), is exposed to the solvent and is evolutionary variable. Consequently, we raised the question whether immunogenicity of HPAs, as opposed to SNPs not reported to cause an immune response, is the consequence of the presence of evolutionary variable and exposed residues in HPAs. Analyses of integrin aIIbb3 polymorphisms that were not published as immunogenic showed that some polymorphic residues are evolutionary variable and exposed similarly to HPAs, while others are buried. Only one polymorphism was shown to be evolutionary conserved, but not buried as found for missense mutations (Table 4). Statistical analysis showed that HPAs differed significantly from causing disease missense mutations in both variables: evolutionary conservation and SASA (Fig. 2). In contrast, HPAs differed from nonimmunogenic polymorphisms in SASA but not in their conservation score, suggesting that the immunogenicity is dependent mostly on the structural location of the residue. Interestingly, it was found that nonimmunogenic polymorphisms differed significantly from missense mutations in the evolutionary score but not in the SASA (Fig. 2).
Taken together, we can conclude that although disease-causing mutations and SNPs responsible for HPAs can be localized to the same domain, they differ in their evolutionary conservation and solvent exposure. Interestingly, although most HPAs result in substituting chemically dissimilar residues, they have no dramatic effects on the apparent function of the proteins. The observations that these positions are evolutionary variable and surface exposed explain their tolerance to substitutions for residues of various properties. Actually, in most cases, the amino acid of the polymorphic residue was found in equivalent positions within homologous proteins ( Table 2). It is interesting to note that many of the HPArelated positions are occupied by positively charged  residues. Substitutions would therefore lead to alteration of the surface electrostatic potential. This might be related to the formation of unique interfaces that lead to antigenicity and to the production of antibodies. The antigenicity of the polymorphisms seems to be dependent on the amino acid residue, as shown for HPA-1. The immunization against HPA-1a (Leu33) is known to be the most common cause of NAIT (80%) while the HPA-1b allele (Pro33) is responsible for only approximately 4% of NAIT cases. This discrepancy in the frequencies is probably due to a different antigenicity of the alleles, as shown by the observation that peptides containing the Leu33 polymorphism bind to recombinant DR52a molecules, whereas the Pro33 version does not. 12 Similarly, HPA-5b allele is immunogenic and the HPA-5a is not (10% in NAIT compared to 0%). Although the likelihood of antigen incompatibility between mother and fetus is very high in HPA-3 and HPA-15 systems, due to their high heterozygosity rate, they are less likely to cause NAIT compared to HPA-1 and HPA-5 systems. Interestingly, HPA-1 and -5 are associated with very low evolutionary conservation (1 = variable) and very high residue SASA (Table 1).

Some HPAs are nonlinear epitopes
Leu33 involved in HPA-1 is located in a loop of the PSI domain of integrin b3, totally exposed to the solvent in both the active and the inactive conformations of aIIbb3 (Table 1). HPA-1 was shown to be a part of a nonlinear epitope 40 sensitive to disulfide bond reduction or cysteine substitution in the PSI domain. 41 This suggests that the epitope is dependent on the intact tertiary structure of its domain. Moreover, substitutions such as Arg93Gln or Pro407Ala (of HPA-7), both localized to the hybrid domain, were shown to be critical for binding of anti-HPA-1a antibodies. 42 This indicates that the HPA-1a epitope is dependent on residues in the neighboring hybrid domain. The sensitivity of the HPA-1a epitope to polymorphisms located in neighboring region can explain its heterogeneous nature as demonstrated by the third allele Val33, which reacts with some anti-HPA-1a but not with others. 12 In addition to HPA-1, epitopes of other HPAs were also shown to be nonlinear 43 and affect posttranslation process: For instance, Arg636Cys (constructed HPA-8) alters the N-linked glycosylation pattern 44 and deletion of Lys611 (HPA-14b) modifies the disulfide bond pattern, both mapped to the bTD of integrin b3. Similarly, HPA-3 determinant, caused by Ile843Ser substitution in integrin aIIb, is dependent upon O-linked carbohydrate. 45 The dependency of the immune response on the glycosylation state can explain the heterogeneous nature of this epitope. It has been shown that some anti-HPA-3a require sialic acid or O-linked oligosaccharide residues for reactivity, whereas others do not. 45

Association between HPA epitopes and activation states of integrins
Integrins such as aIIbb3 were shown to be in a bent inactive conformation on the resting PLT. After physiologic activation, the integrins shift to the extended active conformation leading to exposure of multiple new epitopes named LIBS (ligand-induced binding site). 6 The activation mechanism of the integrins involves multiconformations of both subunits; 6 few were recapitulated in the available crystal structures. [6][7][8] We found some differences in the SASA of HPA-related residues in the various conformations of the b3 subunit due to structural changes in the backbone as well as different rotamers (side-chain conformation; Table 1). It is possible that structural changes and rearrangements of interdomain orientation accompanying the activation process lead to alterations in the presentation of the HPAs' epitopes. Correspondingly, some HPA-related residues are mapped to the same regions harboring the LIBS epitopes that become exposed on activation. For example, few LIBS epitopes, as well as HPAs-1, -6, -8, -11, and -14 of b3 integrin, are located in the PSI, bTD, and EGF domains. 8 Similarly, the epitopes for HPA-3 and -9 at the aIIb integrin, as well as the epitope to the LIBS antibody PMI-1, 8,45 are mapped to the same region-an unstructured loop in the calf2 domain of aIIb. This loop is cleaved during biosynthesis, thus enabling breathing movements that can shift the equilibrium toward integrin extension. 8 The antigenic determinant of HPA-3 is similar to that of the LIBS antibody PMI-1, suggesting that the activation state might also affect exposure of the HPA-3 epitope. 45 The rearrangements of interdomain orientation during activation impinge on the presentation of the HPA epitopes. Then again, some HPA polymorphisms can have an effect on the dynamic of the conformational changes involved in the activation process. For example, Arg633 in b3, the residue substituted in HPA-11, was shown to be important for constraining aIIbb3 in a low-affinity state. Correspondingly, the polymorphism Arg633His can enhance the separation of domains. Interfering with interdomain interactions can increase protein flexibility and initiate the extension of the integrin. 46 Another example is HPA-1b where a Pro33 in b3 subunit has been reported to cause hypercoagulability effects 47 due to increased aggregability, 48 different sensitivity to agonists, 49 or enhancement of outside-in signaling. 50 It is possible that the polymorphism might partially mimic the activated conformation. Our analyses show that both Arg633 and Leu33 in b3 are located close to the interface with other domains, implying that the substitutions might affect interdomain rearrangements. Taken together, these data suggest that although the residues that form the HPAs are probably not involved in stabilizing the three-dimensional fold, their surface location might shape the rigidity of the protein, and influence the interdomain orientation, thus making the protein more prone to activation.
In conclusion, our analyses provide a unique perspective to the molecular basis of amino acid substitutions leading to the formation of epitopes for auto-/ alloantibodies. The HPAs involve SNPs at evolutionary variable positions located on the protein surface. These polymorphisms do not cause diseases but lead to immunologic disorders. Interestingly, some HPAs might also affect the activation process of the protein. The integrins form hetrodimers, whereas each subunit comprises multiple domains. This allows a regulation mechanism that involves multiple conformations of the protein on the cell surface. Our analyses show that some HPAs involve substitutions at positions located at the interface between domains and thereby might affect interdomain rearrangement and the activation dynamics.