Hemophilia A mutations associated with 1-stage/2-stage activity discrepancy disrupt protein-protein interactions within the triplicated A domains of thrombin-activated factor VIIIa

Effective hemostasis is mediated by the regulated and sequential activation of serine proteases in the coagulation cascade. Factor VIII (FVIII) functions in the intrinsic pathway of blood coagulation as a cofactor to accelerate by 100 000-fold the activation of factor X by factor IXa that occurs on a phospholipid surface in the presence of calcium ions.1 A quantitative or qualitative deficiency of FVIII leads to the phenotype of the bleeding disorder hemophilia A. The FVIII amino acid sequence, deduced from the cloned complementary DNA (cDNA), identified that the molecule is synthesized as a single-chain polypeptide having the domain structure A1-A2-B-A3-C1-C2.2,3 On secretion from the cell, FVIII is processed to a heterodimer consisting of a carboxy terminal-derived light chain (LC) of 80 kd in a metal ion-dependent association with a variably proteolyzed (90-200 kd) amino terminal-derived heavy chain (HC) fragment. The FVIII A domains share almost 40% amino acid identity with each other and to the factor V (FV) A domains, as well as with the A domains in the copper-binding protein ceruloplasmin,4 suggesting the A domains may be involved in copper ion binding. Indeed, studies have detected 1 mole copper ion per mole of both FV and FVIII protein.5-7 

In vivo, FVIII activity is regulated by proteolytic activation. On thrombin cleavage of FVIII there is a rapid 50-fold increase and subsequent first-order decay of procoagulant activity. The activation coincides with proteolysis of both the HC and LC of FVIII.8 Thrombin activates FVIII through proteolytic cleavage after ^ARG740, ^ARG1689, and^ARG372 generating an FVIIIa heterotrimer consisting of the A1 subunit (50 kd) in a copper ion–dependent association with the thrombin-cleaved LC (73 kd) and a free A2 subunit associated with the A1 subunit through a weak ionic interaction. The FVIIIa heterotrimer exhibits a pH-dependent dissociation of the A2 subunit from the A1/A3-C1-C2 heterodimer that correlates with loss of procoagulant activity.9-11 Porcine FVIIIa exhibits an increased affinity for its A2 subunit compared to human FVIIIa and accordingly demonstrates an increased specific activity.10,12 In addition, a novel genetically engineered FVIII molecule, in which the A2 subunit remains covalently linked to the LC even after thrombin activation, has increased specific activity in vitro and abrogates the first-order decay of procoagulant activity as observed for native FVIII.13 These observations highlight the role of A2 dissociation in limiting FVIIIa activity in vitro. However, it is not yet known whether spontaneous A2 subunit dissociation or further proteolysis limits FVIIIa procoagulant activity in vivo. Insight into this question has been gained by the recent characterization of an unusual clinical phenotype of hemophilia A.

Patients with hemophilia A have been described whose plasmas display a discrepancy between their FVIII activities, where the 1-stage (1-st) activity assay displays greater activity than the 2-stage (2-st) activity assay.14-18 Initially, the molecular mechanism for one of these mutations, a missense mutation,^ARG531^HIS, was characterized to result from increased rate of inactivation of FVIIIa by increased rate of A2 subunit dissociation.19 On thrombin activation, the^ARG531^HIS A2 subunit exhibited a 4-fold increased rate of dissociation from the A1/A3-C1-C2 heterodimer. The increased instability of the ^ARG531^HISheterotrimer would reduce its specific activity in both 1-st and 2-st assays. However, the specific activity of^ARG531^HIS is disproportionately reduced in the 2-step procedure, possibly due to the incubation phase in the first step. This hemophilia A phenotype therefore supports previous in vitro studies that have suggested that nonproteolytic regulation of FVIIIa activity, via spontaneous A2 subunit dissociation, is important in vivo.

Examination of a homology model of the A domains of FVIII20 identified ^ARG531 to lie at the interface of the A1 and A2 subunits. We hypothesize that the ARG to HIS amino acid substitution destabilizes protein-protein interactions between the A1 and A2 subunits. We also hypothesize that if this region is important for FVIIIa stability, other mutations within this region would have similar phenotypes. Interestingly, there are 3 described hemophilia A mutations either directly contacting ^ARG531 (^ALA284^GLU, ^ALA284^PRO) or closely adjacent to ^ARG531^HIS at the A domain interface in the tightly packed hydrophobic core (^SER289^LEU). All patients with these mutations have the phenotype of 1-st/2-st discrepancy. FVIII cDNA constructs were prepared containing either the ^ALA284^GLU or^SER289^LEU mutations and the proteins were expressed and purified from transfected COS-1 monkey cells. FVIII harboring these missense mutations were secreted at similar levels to FVIII wild-type (WT), had reduced specific activity by 1-st clotting activity, and at least a 2-fold lower activity by 2-st analysis (COAMATIC), similar to the reported clinical data. Analysis in an optical biosensor demonstrated that on thrombin cleavage, A2 dissociation was 3-fold faster for both^SER289^LEU and ^ALA284^GLUcompared to FVIII WT. Thus, these mutations within the A1 subunit of FVIIIa demonstrate a similar destabilization of the A domain interface as the originally characterized ^ARG531^HISmutation within the A2 subunit. These results support a role for this region in stabilizing the A domain interface of FVIIIa and support the validity of the molecular model of the A domains. The amino acid residues 281 to 290 within human FVIII are conserved in porcine, canine, and murine FVIII. Further characterization of the key protein-protein interactions within this region will advance genetic engineering strategies aimed at improving the stability of the FVIIIa heterotrimer.

Anti-HC FVIII monoclonal antibody (F-8) conjugated to CL-4B Sepharose was a gift from Debra Pittman (Genetics Institute, Cambridge, MA). FVIII-deficient and normal pooled human plasma samples were obtained from George King Biomedical (Overland Park, KS). Activated partial thromboplastin (Automated aPTT reagent) and CaCl₂were purchased from General Diagnostics Organon Teknika (Durham, NC). Anti-LC FVIII monoclonal antibody ESH-8 was purchased from American Diagnostica (Greenwich, CT). Dulbecco modified Eagle medium (DMEM) and fetal bovine serum were purchased from Gibco BRL (Gaithersburg, MD). COAMATIC was purchased from DiaPharma (West Chester, OH).

Mutagenesis was performed within the mammalian expression vector pMT₂21 containing the FVIII cDNA (pMT₂VIII). Oligonucleotide-directed mutagenesis was used to create a SpeI-KpnI polymerase chain reaction fragment in which codon 284 was mutated from CGC to CAC, predicting an amino acid substitution of glutamic acid for alanine, and codon 289 was mutated from TCG to TTG, predicting an amino acid substitution of leucine for serine. The mutagenized and SpeI-KpnI digested fragments were then ligated into SpeI-KpnI digested pMT₂VIII. The resulting mutant plasmids were designated ^ALA284^GLU and^SER289^LEU. The plasmid containing the wild-type FVIII cDNA sequence was designated FVIII WT. The plasmid containing a B-domain–deleted (BDD) FVIII was designated BDD-FVIII. The SpeI-KpnI fragment from ^SER289^LEU was ligated into SpeI-KpnI– digested BDD-FVIII and the resultant plasmid designated BDD-^SER289^LEU. The mutant plasmid^ARG531^HIS was prepared previously.19 All plasmids were purified by centrifugation through cesium chloride and characterized by restriction endonuclease digestion and DNA sequence analysis.

Plasmid DNA was transfected into COS-1 cells by the diethylaminoethyl-dextran method as previously described.22 Conditioned medium was harvested at 64 hours after transfection in the presence of 10% fetal bovine serum.

Purified ^ALA284^GLU,^SER289^LEU, BDD-FVIII, and BDD-^SER289^LEU proteins were obtained from 200 mL conditioned medium from transiently transfected COS-1 cells by immunoaffinity chromatography23 yielding 750 to1500 ng per purification. FVIII WT protein was purified in parallel from stably transfected Chinese hamster ovary cells. The proteins eluted into the ethylene glycol–containing buffer were dialyzed and concentrated against a polyethylene glycol (molecular weight, ∼15 000-20 000)–containing buffer24 and stored at −70°C.

FVIII activity was measured by (1) a 1-st aPTT clotting assay on an MLA Electra 750 fibrinometer (Medical Laboratory Automation, Pleasantville, NY) by reconstitution of human FVIII-deficient plasma or (2) a modified 2-st assay using the COAMATIC chromogenic assay according to the manufacturer's instructions. FVIII antigen was quantified by an anti-FVIII LC sandwich enzyme-linked immunosorbent assay (ELISA) method using the IMUBIND commercial kit according to the manufacturer's instructions (American Diagnostica).

The kinetics of the A2 subunit dissociation from thrombin-activated FVIII WT and mutant proteins were studied by surface plasmon resonance using the Biacore biosensor (Piscataway, NJ), which measures protein binding and subsequent dissociation in real time.25 Binding of 1 ng protein/mm² of the biosensor chip produces a resonance signal of 200 Arc seconds. Monoclonal antibody ESH8 (50 μg/mL) in 10 mM sodium acetate, pH 5.0, was covalently coupled at 16.5 ng/mm² to the activated carboxymethyldextran-coated biosensor cuvette via amino groups using succinimide ester chemistry.25 The carboxymethyldextran chip and reagents for its activation, N-ethyl-N′-(dimethylaminopropyl)carbodiimide hydrochloride and N-hydroxysuccinimide, and deactivation, ethanolamine, were purchased from Affinity Sensors (Cambridge, United Kingdom). FVIII WT and mutant protein binding to ESH8 and their dissociation from the antibody and activation by thrombin were measured in 200 μL 20 mM Hepes, 0.15 NaCl, 5 mM CaCl₂, 0.01% Tween 20, pH 7.4. The chip was regenerated by addition of 0.1 M glycine, pH 3.0, for 3 minutes, resulting in complete dissociation of FVIII proteins from the capture ESH8. Identical signals for reference binding of the FVIII proteins to immobilized ESH8 were obtained before and after regeneration.

The values of the rate constants for dissociation (k_off) of FVIII WT or mutant proteins from immobilized ESH8 and those for dissociation of the A2 subunits on thrombin activation of immobilized FVIII proteins were calculated using the dissociation model with offset described by the following equation: R = R_i × exp(−k_offA2t) + R_F.26In the equation R is the observed surface plasmon resonance signal during A2 dissociation; k_offA2, the kinetic rate constant for the A2 dissociation on thrombin activation of FVIII WT or mutant proteins; R_i, the initial resonance signal when thrombin was injected; R_f, the final signal observed after all A2 dissociated (the final plateau achieved). R_i, R_f , and k_offA2 were used as the independent parameters to fit observed resonance signal R versus time (t) by nonlinear regression analysis using Sigmaplot (Jandel Scientific, San Raphael, CA).

The synthesis and secretion of FVIII WT and^ARG531^HIS, ^ALA284^GLU, and ^SER289^LEU mutants were compared after transient DNA transfection into COS-1 monkey cells. FVIII antigen determinations by ELISA were performed on conditioned media collected from 24 to 64 hours after transfection and were similar for all mutants compared to FVIII WT suggesting no significant defect in synthesis or secretion.

FVIII activity (Figure 1) was measured by a 1-st aPTT clotting assay or by a modified 2-st method using the COAMATIC chromogenic assay. Activities were normalized to their expressed FVIII antigen levels as determined by ELISA and are presented as a percent of FVIII WT activity. Similar to the patient phenotypes and data reported previously, ^ARG531^HIS had discrepant activity as measured by the 1-st aPTT clotting assay (44% ± 15% of FVIII WT) compared to the 2-st chromogenic assay (COAMATIC; 19% ± 4% of FVIII WT). For the^ALA284^GLU mutation, 1-st aPTT (26% ± 6%) and 2-st chromogenic assay (6% ± 4%) activities showed discrepancy. For the ^SER289^LEUmutation, 1-st aPTT (32% ± 11%) and 2-st chromogenic assay (13% ± 5%) activities also showed discrepancy. The activity measurements for the patient plasmas represent all of the reported data obtained from the HAMSTeRS hemophilia A mutation database.27 The slightly higher 2-st activity results obtained with the recombinant-derived mutant proteins are consistent with those obtained from the few reported patient plasmas where a chromogenic assay rather than the classical 2-st assay was used.18 Following immunoaffinity purification of FVIII WT and mutant proteins from the conditioned medium, similar discrepancy in 1-st/2-st activities were obtained (data not shown).

The relative rates of A2 dissociation for plasma-derived (pd) FVIII, FVIII WT, and mutant proteins were determined using an optical biosensor (Figure 2). An anti-LC antibody, ESH8, was covalently immobilized on a carboxymethyldextran-coated biosensor chip. Similar amounts (1.12 ng/mm²) of immunoaffinity purified FVIII WT or mutant proteins were bound to ESH8 antibody. Unbound material was removed by washing with buffer and free (nonproteolytic) dissociation from antibody was measured. The values of dissociation rate constants, k_off  = 8.9 ± 0.23 × 10⁻⁵/s, were similar for FVIII WT and mutant proteins. Subsequently, thrombin was added to a final concentration of 1 U/mL. Because the FVIII preparations are bound to an ESH8-coated chip via the C2 domain of their LCs, this interaction is not disturbed by thrombin cleavage and the A1 subunit remains associated with the LC through the copper ion–dependent linkage between the A1 and A3 domains. Thus, the thrombin-induced release of the A2 subunit from the heterotrimer can be measured as a dissociation curve registered by the optical biosensor. Control experiments with thrombin concentrations ranging from 0.05 to 2.25 U/mL demonstrated that maximal thrombin cleavage under these experimental conditions occurs at a thrombin concentration of 0.75 U/mL (data not shown). The k_off values and half-lives of A2 dissociation for the A2 subunits are presented in Table1. Results for^ARG531^HIS have been reported previously19 and are presented for comparison purposes. Both the ^ALA284^GLU and^SER289^LEU mutants displayed a 3-fold greater rate of A2 dissociation compared to pdFVIII WT. Because A2 dissociation correlates with inactivation, the mutant thrombin-activated heterotrimers would be expected to inactivate approximately 3-fold faster compared to plasma-derived FVIII or FVIII WT. The single exponential model used by us to describe dissociation of the A2 subunit observed for FVIII WT in this study is similar to that previously reported for FVIIIa decay.12 24 The minor difference in thrombin-mediated A2 dissociation rates for pdFVIII compared to recombinant FVIII WT did not reach statistical significance (P = .24, n = 5).

Because these mutations were expressed within the FVIII WT construct, they contain a B domain. The B domain is also released on thrombin activation and contributes to the loss of mass observed by the optical biosensor. Although the release of B-domain fragment(s) is immediate on thrombin activation we prepared BDD forms of FVIII and the^SER289^LEU mutant so that only A2 subunit dissociation would be measured by the optical biosensor. Both BDD-FVIII and BDD-^SER289^LEU were expressed from COS cells by transient transfection and the purified proteins subjected to the same analysis as the full-length proteins. BDD-^SER289^LEU exhibited similar 1-st/2-st discrepancy as ^SER289^LEU in activity assays (data not shown). Purified BDD-^SER289^LEUwas then subjected to optical biosensor analysis and displayed 5-fold greater rate of A2 dissociation compared to the BDD-FVIII control. Because the amount of each immobilized FVIII was approximately 310 RU when thrombin was added, the reduction of the resonance signal on dissociation of A2 is 28% and 27.6% for BDD-FVIII and BDD-^SER289^LEU, respectively. This is identical to the expected reduction of the resonance signal (27.6%) on complete dissociation of the A2 subunit calculated using molecular weights of 170 kd and 123 kd for BDD-FVIII and its A1/A3-C1-C2 dimer. In contrast, B domain–containing samples in panels A and B of Figure 2demonstrated an initial rapid loss of resonance signal immediately following addition of thrombin (consistent with loss of B-domain fragments) that was not observed for the BDD samples in panel C. However, all samples subsequently exhibited a relatively slower loss of resonance signal (consistent with A2 subunit dissociation) that was used to calculate the k_off for dissociation.

To address the effect of the increased rate of A2 dissociation on the discrepant activity of the mutants in the 2-st assay, FVIII WT,^ARG531^HIS, ^ALA284^GLU, and ^SER289^LEU proteins were assayed in the chromogenic assay under conditions of increased duration of the first stage (Figure 3). Conditioned media samples collected at 64 hours following transfection were used for the assay. Antigen concentrations were similar for FVIII WT and the mutants and ranged between 10 and 20 ng/mL. The manufacturer's protocol was modified whereby duplicated samples were incubated in the first stage for 2, 4, 8, and 16 minutes in parallel with FVIII WT. The first stage was initiated for the 16-minute samples first, followed by the 8-, 4-, and 2-minute incubation samples. The second stage was then initiated for all samples simultaneously and then measured together. During the first-stage incubation, the FVIII samples are activated due to the presence of thrombin in the reagent mix. The resultant FVIIIa is then available to exert cofactor activity with factor IXa to generate factor Xa from unactivated factor X. The total amount of factor Xa generated in the first stage is then quantified at the second stage by the amount of chromophore released from the chromogenic reagent by factor Xa. A standard curve was performed in parallel with dilutions of pooled normal plasma and incubated for 2 minutes during the first stage. The y-axis of Figure 3 is thus the relative FVIII activity of the samples following the various one-stage incubations as compared to this standard curve. We hypothesized that both FVIII WT and the mutants with increased rate of A2 dissociation would become fully activated in the first-stage incubation and exhibit FVIII activity as determined by the total amount of factor Xa generated. However, with increased rate of inactivation due to A2 dissociation, the mutants would be unable to sustain FVIII activity and the total amount of factor Xa generated would not increase at the same rate as FVIII WT. Indeed, FVIII WT was observed to have continued cofactor activity (as demonstrated by increasing factor Xa generation) throughout even the longest first-stage incubation of 16 minutes. In contrast, each of the mutants, though showing continued cofactor activity for up to 4 minutes of first-stage incubation fail to contribute to any further factor Xa generation beyond 4 minutes. This is consistent with complete spontaneous inactivation of the mutants after 4 minutes. Thus the relative amount of factor Xa generation decreases with increasing duration of first-stage incubation compared to FVIII WT.

Despite the 4 decades of availability of both the 1-st and 2-st assays for measuring FVIII activity, there is still no consensus on which method most accurately represents FVIII cofactor function in vivo. Although there is little assay discrepancy between the methods when comparing human plasmas or various FVIII-containing concentrates, considerable variability has been described when not comparing FVIII molecules with similar molecular composition and structure. For instance, recombinant FVIII shows a reduced potency in the 1-st assay compared to the 2-st assay (chromogenic) when plasma FVIII is used as the standard.28 Similar discrepancy is observed following infusion of recombinant FVIII into patients with hemophilia, prompting recommendations that recombinant FVIII be used to prepare the standards for this type of analysis.

In addition, a naturally occurring patient mutation was described whereby the 1-st assay reports reduced activity compared to the 2-st assay.29 The mutation ^GLU720^LYSwithin FVIII results in a mild hemophilia A phenotype with detectable circulating FVIII antigen. Patients with this mutation exhibit normal levels of FVIII cofactor activity as measured by the chromogenic assay method but have reduced levels of procoagulant activity when measured by 1-st assay. Recombinant ^GLU720^LYS was purified and found to require higher concentrations of FIXa for full cofactor activity as measured by FXa generation, whereas lower concentrations of FIXa (as would be expected in a 1-st assay) resulted in 2- to 3-fold reduced activity. It was proposed that this mutation at residue 720, adjacent to one of 3 putative FIXa binding sites within the A2 domain (residues 698-712), could result in reduced affinity for FIXa and therefore contribute to the higher FIXa concentration required for optimal cofactor activity.

The 1-st/2-st discrepancy described above relates to reduced FVIII activity in the 1-st compared to the 2-st assay. However, the hemophilia A phenotype analyzed in this report relates to FVIII activity that is reduced in the 2-st assay compared to the 1-st assay. Early reports characterizing this clinical phenotype were attributed to type 2N von Willebrand disease (vWD) where a mutation in the FVIII binding site within von Willebrand factor leads to reduced affinity for FVIII in plasma.30,31 The reduced FVIII activity in the 2-st assay was due to increased adsorption of FVIII to Al(OH)₃ during the pretreatment phase of the 2-st assay. This has been subsequently confirmed in an additional recent report of type 2N vWD.32 This type of 1-st/2-st activity discrepancy was first characterized in hemophilia A pedigrees in 198333 suggesting it was related to specific FVIII mutations. Subsequently, several specific mutations were described.18 Biochemical analysis of a recombinant-derived form of one of these mutations, ^ARG531^HIS, led to the observation that the thrombin-activated mutant protein had an increased rate of spontaneous inactivation secondary to an increased rate of A2 subunit dissociation.19 This suggested a role for this particular residue in contributing to the stabilization of the FVIIIa heterotrimer. An additional report describes analysis of patient plasma from a hemophilia A patient with the^ARG531^HIS mutation.34 Using a FVIII ELISA specific for heterodimeric FVIII, this patient had reduced amounts of heterodimeric FVIII, suggesting the mutation may also reduce the stability of the triplicated A domains even before thrombin activation, further highlighting the role of protein-protein interactions within the A-domain interface.

Molecular modeling of the environment of the ^ARG531 residue, using the homology-derived FVIII A domain model20predicted that the ^ARG531 arginine side chain in A2 participates in an H-bond with the main chain carbonyl oxygen of^ARG282 in A1, providing both a plausible explanation for the molecular pathology of the hemophilic FVIII variant and support for the accuracy of the model. However, the model also permitted predictions concerning the hemophilic variants^ALA284^PRO and^ALA284^GLU, which also show the 1-st/2-st discrepancy18 35: here the small hydrophobic side chain of^ALA284 packs closely against the ^ARG531 side chain at the A1-A2 boundary, very close to the^ARG531-^ARG282 H-bond (Figure4A). Replacement of the small alanine side chain with that of either proline or glutamic acid (as in the hemophilic FVIII variants) may be predicted to disrupt the stabilizing^ARG531-^ARG282 H-bond. We have tested this prediction by expression and characterization of the^ALA284^GLU variant.

In addition we have expressed and studied^SER289^LEU, another hemophilia mutation in A1 with a similar 1-st/2-st discrepancy, placed close in the model to the A1-A2 domain interface. In this case, inspection of the FVIII A domain model predicts an interdomain H-bond between the A1 and A3 domains, between the side chains of ^SER289 and ^TYR1979 (Figure 4B). Replacement by the bulky leucine side chain is predicted to abolish this interdomain H-bond but also to disrupt packing very close to the interface between all 3 A domains at the pseudo-threefold axis of the model.

Our in vitro characterization of both of these FVIII missense mutations located within the A1 subunit (at residues 284 and 289) confirms the hemophilia A clinical phenotype of 1-st/2-st discrepancy and demonstrates the same molecular phenotype of increased rate of A2 subunit dissociation on thrombin cleavage. These observations provide evidence that the packing of the triplicated A domains is critical to FVIII function and that interdomain protein-protein interactions are required for stabilizing both the unactivated and activated forms of the molecule. These observations also provide first experimental evidence that validates the molecular model of the triplicated A domains based on their homology to ceruloplasmin.

The insights gained from studying this patient phenotype and the molecular characterization of these particular mutations has extended the observations made from analyzing the^ARG531^HIS mutation. Two additional reports have added to our observations. A family pedigree has been described in which all members have 2-st FVIII activities that are one half of that measured by 1-st assay.36 This pedigree included affected males as well as homozygous females. Molecular analysis identified a missense mutation at residue 694 within the A2 subunit where ILE is substituted for ASN. This mutation predicts loss of intradomain hydrogen bonds as well as introduces severe side-chain packing constraints within an area adjacent to the A2-A3 interface. The second report describes a hemophilia A patient with the same 1-st/2-st discrepancy phenotype attributed to a missense mutation at residue 1954 within the A3 subunit where LEU is substituted for a HIS.37 This patient's 1-st activity assay was within normal limits though the 2-st activity assay and chromogenic assay were only 18% and 35% of normal, respectively. This residue is in close proximity to ^TYR1979, and to residues ^ARG698 and ^MET1947, both previously indicated to lie at the A2-A3 interface and associated with the same 1-st/2-st discrepant mild hemophilia phenotype under discussion.18 35 This report is particularly interesting in that the hemophilia phenotype was more accurately reflected by the 2-st assay rather than the more widely used 1-st assay. Thus, mild hemophilia A variants can be missed when screened in laboratories that exclusively use the 1-st aPTT-based activity assay. These observations, as well as those reported here, showing that the 2-st assay more closely reflects in vivo FVIII clotting activity, support that FVIII activity is limited by A2 dissociation in vivo.

There are reports of this particular discrepant hemophilia phenotype associated with predictions of disrupted interactions at or close to the A1-A2, A1-A3, and A2-A3 domain interfaces. Figure5 shows a global view of the FVIII A domain model viewed down the pseudo-threefold axis, showing the predicted positions of the amino acids mentioned above, which are associated with the 1-st/2-st discrepancy in hemophilic subjects. There is a clear clustering of all these residues at or very close to the interface of the A2 domain with the A1 and A3 domains.

The proposed pseudo-threefold axis at the hydrophobic core of the A1-A2-A3 interface is tightly packed and predicts several interdomain interactions. These lie at the interface of A1-A2, A2-A3, and A1-A3. Sufficient disruption of these protein-protein interactions along the pseudo-threefold axis would lead to destabilization and increased propensity for dissociation particularly after thrombin cleavage whereby covalent bonds flanking the A2 subunit are lost.

The validation of the molecular modeling of the A domains and the characterization of these putative interdomain protein-protein interactions now allows the hypothesis that increased stabilization of this interface would lead to greater stability of FVIII and increased cofactor activity. There is now opportunity to explore complementary mutations that might allow better packing and stabilization of the A domains to test this hypothesis. Novel bioengineering strategies may ultimately enable strategic interdomain cross-linking to provide further stabilization of FVIIIa. Further characterization of FVIIIa-FIXa sites of interaction combined with both molecular modeling and crystallography observations may ultimately allow the design of a peptidomimetic for FVIIIa.