Reduction of the inhibitory antibody response to human factor VIII in hemophilia A mice by mutagenesis of the A2 domain B-cell epitope

Inhibitory antibodies to factor VIII (fVIII) arise in response to infusions of fVIII in approximately 25% of previously untreated patients with hemophilia A.^1-4  These antibodies increase the morbidity and mortality and the cost of care of these patients. Treatment options include bypassing the requirement for fVIII using activated coagulation factor concentrates or recombinant fVIIa and using porcine fVIII.⁵  Additionally, inhibitory antibodies can be eliminated in some patients using immune tolerance induction regimens that include daily infusions of large doses of fVIII.⁶  All these treatment modalities are expensive and are not always successful.

Developing a method to reduce the incidence of inhibitory antibody development would improve the management of patients with hemophilia A. One possible strategy is to alter the fVIII molecule and render it less immunogenic. The incidence of inhibitory antibody development is higher in patients with severe hemophilia A, presumably because they recognize fVIII as immunologically foreign in the absence of endogenous protein. Immune responses to soluble foreign protein antigens usually involve T-helper (T_H) cell–dependent activation, proliferation, and differentiation of B cells. B cells recognize epitopes on the native, intact antigen through membrane immunoglobulin (mIg) molecules. The structure of these B-cell epitopes could play a role in the intrinsic immunogenicity of a protein. Thus, mutagenesis of B-cell epitopes without disruption of biologic function is a strategy to reduce the incidence of inhibitor development in patients with hemophilia A.

FVIII contains a sequence of domains designated A1-A2-B-ap-A3-C1-C2.⁷  The B domain can be deleted without loss of biologic activity,^8,9  and B domain–deleted fVIII is effective in the treatment of hemophilia A.³  Most inhibitory antibody activity is directed against epitopes in the A2 or C2 domain.^10-17  The A2 epitope has been localized to a single, continuous sequence bounded by residues R484-I508.¹⁵  Antigenic residues within this segment that are recognized by human inhibitory antibodies include R484, Y487, S488, R489, P492, V495, F501, and I508.¹⁸  Single alanine mutations at any of these sites results in the expression of functional fVIII molecules.¹⁸  In the current study, we prepared B domain–deleted (BDD) human fVIII molecules containing mutations at R484A/R489A or R484A/R489A/P492A, and we compared the immunogenicity of these molecules with nonmutated human fVIII in hemophilia A mice.

Citrated human hemophilia A plasma and normal pooled human plasma (FACT) were purchased from George King Biomedical (Overland Park, KS). Activated partial thromboplastin time reagent (Automated APTT) was purchased from Biomerieux (Durham, NC). Murine antihuman fVIII monoclonal antibodies ESH4 and ESH8 for ELISA were purchased from American Diagnostica (Stamford, CT). Murine anti–human fVIII monoclonal antibodies ESH4 and ESH8 for ELISA were purchased from American Diagnostica (Stanford, CT). Murine anti–human fVIII monoclonal antibody ESH5 for immunoaffinity chromatography was generously provided by Dr. Duncan Pepper (Scottish National Blood Transfusion Program, Edinburgh, Scotland, United Kingdom). ESH5 was coupled to Sepharose CL 4B (Sigma Chemical, St. Louis, MO) according to instructions provided by the supplier. SP-Sepharose Fast Flow and Source Q or Mono Q ion-exchange resins were purchased from Amersham Biosciences (Uppsala, Sweden). Synthetic oligonucleotides were purchased from Invitrogen Life Technologies (Carlsbad, CA). Restriction enzymes were purchased from New England Biolabs (Beverly, MA) or Promega (Madison, WI). A cell line derived from baby hamster kidney cells was a generous gift from Dr R. T. A. MacGillivray.¹⁹  Exon 16–disrupted (E16) hemophilia A mice²⁰  in a C57BL/6 background were obtained from Dr Leon Hoyer (Holland Laboratories, American Red Cross), and a breeding colony was established. Nine- to 12-week-old male and female mice were used in the experiments. Novel fVIII DNA sequences generated by polymerase chain reaction (PCR) were confirmed by dideoxy sequencing using an Applied Biosystems 373a automated DNA sequencer (Applied Biosystems, Foster City, CA) and the PRISM dye terminator kit (Applied Biosystems).

The cDNA encoding a human BDD form of fVIII, designated HSQ,²¹  was prepared as described previously.²²  The HSQ cDNA encodes an amino acid sequence identical to the commercial fVIII product Refacto. It contains a linker sequence, S F S Q N P P V L K R H Q R, between the A2 and the ap-A3 domains. The linker corresponds to the first 5 and the last 9 amino acids of the B domain and contains a R H Q R recognition sequence for intracellular PACE/furin processing.²³  This produces an A1-A2/ap-A3-C1-C2 heterodimer as the dominant secreted species, which is considered the physiologic form of fVIII.²¹  The cDNAs for porcine and murine BDD fVIII, designated POL and MSQ, respectively, which also contain PACE/furin recognition linker sequences, were prepared as described previously.^22,24  The cDNA for a BDD hybrid human/porcine fVIII molecule, designated HP9, which contains insertion of the porcine segment corresponding to residues 484-508 in the A2 domain of human fVIII (Figure 1), has been described previously.¹⁵  A BDD hybrid human/murine fVIII cDNA, HM1, encoding an fVIII molecule that is murine except for a human segment corresponding to residues 484-508 of the A2 domain (Figure 1), was constructed by splicing-by-overlap extension mutagenesis using MSQ as the template.

Construction of a cDNA encoding BDD R489A human fVIII has been described¹⁸  and was modified further by the insertion of DNA encoding the SFSQNPPVLKRHQR linker sequence. Resultant cDNA was used as a template for the production of cDNAs encoding R484A/R489A and R484A/R489A/P492A BDD human fVIII by splicing-by-overlap extension mutagenesis (Figure 1). For the R484A/R489A mutant, 5′-CAC GGA ATC ACT GAT GTC GCC CCT TTG TAT TCA GCC AGA-3′ and 5′-TCT GGC TGA ATA CAA AGG GGC GAC ATC AGT GAT TCC GTG-3′ were used as the mutagenic sense and antisense primers, respectively. For the R484A/R489A/P492A mutant, 5′-ACT GAT GTC GCC CCT TTG TAT TCA GCC AGA TTA GCC AAA-3′ and 5′-TTT GGC TAA TCT GGC TGA ATA CAA AGG GGC GAC ATC AGT-3′ were used as the mutagenic sense and antisense primers, respectively.

Recombinant fVIII molecules were expressed in baby hamster kidney–derived cells in serum-free medium using the ReNeo expression vector as described previously.²⁵  HSQ, HSQ R484A/R489A, HSQ R484A/R489A/P492A, MSQ, POL, and HM1 were purified by SP-Sepharose Fast Flow and Source Q or Mono Q ion-exchange chromatography, essentially as described previously for HSQ, MSQ, and POL.^22,24  HP9 was purified using the following procedure. Ammonium sulfate was added to 5.9 L cell culture medium at 4° C to 65% saturation and allowed to stir overnight. The precipitate was collected by centrifugation, dialyzed against 0.15 M NaCl, 0.02 M HEPES (N-2-hydroxyethylpiperazine-N′-2-ethanesulfonic acid), 5 mM CaCl₂, 0.01% Tween-80, pH 7.4, and applied to a 1.5 × 10 cm ESH5-Sepharose column equilibrated in the same buffer. HP9 was eluted with 1 M NaCl, 5 mM Mes (4-morpholineethanesulfonic acid), 2.5 mM CaCl₂, 50% ethylene glycol (vol/vol), pH 6.0. Fractions from the fVIII activity peak were diluted one fifth into 0.04 M HEPES, 5 mM CaCl₂, 0.01% Tween-80, pH 7.4, and were further purified using Mono Q ion-exchange chromatography.

Purified proteins were at least 90% pure as judged by sodium dodecyl sulfate–polyacrylamide gel electrophoresis and contained heterodimeric fVIII as the dominant species. Concentrations were calculated using the absorbance at 280 nm purified proteins and molar extinction coefficients that were estimated from the respective deduced amino acid compositions.²⁶  Specific activities of the purified proteins were calculated using protein concentration and fVIII coagulant activity, measured as described in “fVII coagulation assays and ELISAs” against a human fVIII plasma standard. The following specific activities were obtained: HSQ, 1300 U/nmol; HSQ R484A/R489A/P492A, 1270 U/nmol; HSQ R484A/R489A, 1460 U/nmol; HP9, 1100 U/nmol; HM1, 540 U/nmol. Specific activities of the POL and MSQ preparations have been previously reported as 2050 U/nmol and 660 U/nmol, respectively.^22,24  Specific activities were converted to units per microgram using a molecular weight of 165 000 for all fVIII species in the study.

Preparations of HSQ, HSQ R484A/R489A/P492A, and HSQ R484A/R489A were diluted to 0.17 mg/mL in 0.4 M NaCl, 20 mM HEPES, 5 mM CaCl₂, 0.01% Tween-80, pH 7.4, and were stored in small aliquots at –80° C. Samples were diluted to 4 μg/mL in sterile normal saline immediately before injection. Eighty-three hemophilia A mice were divided into HSQ (n = 24), HSQ R484A/R489A/P492A (n = 23), HSQ R484A/R489A (n = 24), buffer-injected control (n = 5), and noninjected control (n = 7) groups. Mice were warmed under a 75-W lamp for 3 to 4 minutes to dilate tail veins before injection. Silver nitrate was used to cauterize bleeding after injections and tail snips. Recipients of different test materials were mixed within each cage and remained in their original cages to avoid fighting. Approximately equal numbers of males and females were in each group. Mice received 6 injections of 10 μg/kg body weight (approximately 80 U/kg) at 14-day intervals, followed by a final injection of 25 μg/kg body weight (approximately 200 U/kg) 2 weeks after the sixth dose. Blood was collected by tail snip into 0.1 vol 3.8% trisodium citrate under methoxyflurane anesthesia 13 days after the fourth injection and by terminal cardiac puncture 13 to 14 days after the final injection. Samples were held on ice before centrifugation at 3000g for 15 minutes at 4° C to collect plasma. During the course of the experiment, 0, 4, 5, 1, and 2 mice died in the HSQ, HSQ R484A/R489A/P492A, HSQ R484A/R489A, buffer-injected control, and noninjected control groups, respectively. The overall mortality rate during the study was 14%, which is consistent with the increased mortality rate of hemophilia A mice associated with handling.²⁰ 

HSQ or HSQ R484A/R489A/P492A was diluted to 50 U/mL (approximately 6.25 μg/mL) in sterile saline for injection immediately before use. Hemophilia A mice were anesthetized with metofane, weighed, warmed under a 75-W lamp for 3 minutes to dilate veins, and injected with 100 U/kg (approximately 12.5 μg/kg) by tail vein. Silver nitrate was used to cauterize the injection site. Because serial sampling from hemophilia A mice is difficult, 3 mice were used for each time point for each fVIII construct. At 0.25, 0.5, 1, 2, 4, 8, or 24 hours, mice were anesthetized by intraperitoneal injection of a mixture of 300 mg/kg ketamine and 75 mg/kg xylazine, and blood was collected by cardiac puncture into 0.1 vol 3.8% trisodium citrate. Plasma samples were prepared by centrifugation and stored at –70° C. FVIII activity was determined by chromogenic assay as described in “fVIII coagulation assays and ELISAs.”

Hemostatic efficacy of HSQ R484A/R489A/P492A in hemophilia A mice was measured using a tail vein transection model, as described previously.²⁷  Briefly, mice were anesthetized with 1.5 mg/kg droperidol and 75 mg/kg ketamine intraperitoneally, warmed under a heat lamp to dilate the tail veins, and injected with HSQ R484A/R489A/P492A. After anesthesia was deepened using methoxyflurane, the distal 1 cm of tail was transected, and the stump was placed in a test tube containing 150 mM NaCl at 37° C. At 2 hours, surviving mice were caged and death at 24 hours was determined. The up-and-down method for small samples^28,29  was used to estimate the dose that produces 50% survival (ED₅₀). An initial dose of fVIII was given to a single mouse as an a priori estimate of the ED₅₀. If the mouse survived the 24-hour test period, another mouse was tested and the dose was decreased. If the subject died, the dose was increased in the next subject. This continued until a nominal sample size of 6 was reached. A constant log-dose increment or decrement of 0.1, corresponding to a dilution factor of 1.26, was used. The ED₅₀ was calculated using the equation [Log ED₅₀ = x_f + kd], where x_f is the logarithm of final test dose, d is the log-dose increment or decrement, and k is obtained from a table based on maximum likelihood estimates.²⁸ 

fVIII activity was measured by 1-stage clotting assay,³⁰  as described previously,²²  using normal human plasma (FACT) as the standard. fVIII inhibitor titers were measured by a modified Bethesda assay in which fVIII constructs (HSQ, HSQ R484A/R489A, HSQ R484A/R489A/P492A, HP9, or POL) were added to human hemophilia A plasma to a final concentration of 0.8 to 1.2 U/mL and were incubated with varying concentrations of inhibitor for 2 hours at 37° C. One Bethesda unit (BU) is defined as the amount of inhibitory activity that produces 50% inhibition of fVIII activity in the 1-stage clotting assay. The 50% inhibition point was identified by interpolation using only data points falling within a range of 40% to 60% inhibition. An average of at least 3 data points in this range was used for each determination. Because of the smaller volumes required, fVIII activity was measured in the pharmacokinetic study by a chromogenic assay (Coamatic; Chromogenix/Diapharma Group, West Chester, OH) according to instructions supplied by the manufacturer.

Antibodies to HSQ, HSQ R484A/R489A/P492A, MSQ, and HM1 were measured by enzyme-linked immunosorbent assay (ELISA) using immobilized purified antigen bound to microtiter plates, followed by the addition of diluted test plasma and then detection using alkaline phosphatase–conjugated goat antimurine immunoglobulin G (IgG), as described previously.²⁴  Absorbance values obtained from 7 dilutions of test plasma were plotted against the logarithm of the plasma dilution, and the resultant sigmoidal curves were fit to the 4-parameter logistic equation by nonlinear regression using the Levenberg-Marquardt algorithm. The ELISA titer was defined empirically as the dilution of plasma that returned an absorbance value of 0.3 derived from the fitted curves.

To reduce the immunogenicity of human fVIII, 2 BDD fVIII mutants were developed that contained alanine substitutions of antigenic amino acids in the A2 domain. These mutants, designated HSQ R484A/R489A and HSQ R484A/R489A/P492A (Figure 1), were compared with a control BDD fVIII molecule (HSQ) in hemophilia A mice using an intravenous immunization protocol described in “Materials and methods.” A critical question in this experimental model is whether hemophilia A mice recognized the R484-I508 A2 sequence targeted by human anti-fVIII inhibitory antibodies. We used 2 approaches to address this problem. First, plasma from the HSQ treatment group was tested for reactivity against a hybrid human/porcine fVIII molecule, designated HP9, which is human except for the insertion of a porcine sequence within the R484-I508 segment (Figure 1). The rationale behind this experiment was that if immune plasma contains antibodies that recognize the R484-I508 epitope, it should react less well with the HP9 molecule than with HSQ because of the incomplete cross-reactivity of fVIII inhibitors with human and porcine fVIII. Average Bethesda titers in the HSQ treatment group against HSQ and a BDD porcine fVIII construct designated POL were 680 and 31 BU, respectively, corresponding to a cross-reactivity of 4%, demonstrating that porcine fVIII is poorly cross-reactive in this model (data not shown). Figure 2A shows paired data comparing the inhibition of HSQ and HP9 by individual mouse plasma in the Bethesda assay. Bethesda titers of most of the pairs were decreased using HP9 as the target antigen compared with HSQ. The difference between the HP9 and the HSQ groups was statistically significant (P < .0001; paired t test). Figure 2A also shows that all the plasmas recognized HP9, indicating the presence of inhibitory epitopes directed toward human epitopes outside the R484-I508 A2 epitope.

Immunodominance of the human R484-I508 segment in hemophilia A mice also was tested by a converse experiment in which the same plasma was tested by ELISA for reactivity against a hybrid human/murine fVIII molecule, designated HM1, which is murine except for the human R484-I508 segment (Figure 1). The average ELISA titer against BDD murine fVIII, MSQ, was 19% of that against HSQ (data not shown), demonstrating that the plasma cross-reacted poorly with murine fVIII and, thus, that human/murine hybrid fVIII molecules can be used to study differential antigenicity. The HM1 hybrid would be expected to be more antigenic than MSQ because of the presence of the human R484-I508 segment. Figure 2B shows that ELISA titers of most of the plasma were increased with HM1 rather than MSQ as the target antigen. The difference between the HM1 and the MSQ groups was statistically significant (P < .02; paired t test).

Additionally, the plasma from mice immunized with HSQ showed a significant correlation between the reduction in antigenicity of HP9 compared with HSQ and the increase in antigenicity of HM1 compared with MSQ (Figure 3). Thus, the degree to which individual mice developed an immune response to the R484-I508 epitope could be detected similarly in both assay systems. This indicates that the reduced antigenicity of HP9 and the increased antigenicity of HM1 are not artifactual, for example, because of differences in antigen binding to microtiter plates.

Hemophilia A mice in the HSQ, HSQ R484A/R489A, and HSQ R484A/R489A/P492A treatment groups received six 10-μg/kg intravenous injections at 14-day intervals and a final injection of 25 μg/kg, as described in “Materials and methods.” Plasma from each was obtained 2 weeks after the last injection and was tested for inhibitory anti-fVIII antibodies using a modified Bethesda assay in which isologous antigen was added to human hemophilia A plasma (eg, HSQ was added to human hemophilia A plasma to test plasma from mice immunized with HSQ). Figure 4 shows the Bethesda titers for individual mice. Titers of 2 mice each in the HSQ R484A/R489A/P492A and HSQ R484A/R489A groups were less than 0.8 BU and were not plotted. Bethesda titers were lower in the HSQ R484A/R489A/P492A group than in the control HSQ group (P = .01; Mann-Whitney U test). In contrast, inhibitor levels in the HSQ R484A/R489A group were not significantly different from those in the control group. Plasma from the HSQ and HSQ R484A/R489A/P492A groups also was tested for anti-fVIII antibodies by ELISA using immobilized isologous antigen (Figure 5). ELISA titers of the 2 mice in the HSQ R484A/R489A/P492A group that tested negative in the Bethesda assay were less than 50 and were not plotted. There was no significant difference between the groups.

Hemostatic efficacy of HSQ R484A/R489A/P492A was measured in hemophilia A mice using a tail vein transection model.²⁷  Under the conditions used, tail vein transection is uniformly fatal within 24 hours in untreated mice. The estimated dose of HSQ R484A/R489A/P492A that produces 50% survival (ED₅₀) was measured using the up-and-down method,^28,29  as described in “Materials and methods.” In this method, a single subject is given a dose that is an a priori estimate of the ED₅₀. If the subject survives, another subject is tested, and the dose is decreased. If the subject dies, the dose is increased in the next subject, and so on, until a nominal sample size of 6 is reached. Results are shown in Figure 6. The method yielded an ED₅₀ of 47.5 U/kg (95% confidence interval [CI], 34.9-64.6 U/kg) for HSQ R484A/R489A/P492A. For HSQ, an ED₅₀ of 57.7 U/kg (95% CI, 42.4-78.5 U/kg) has been published.²⁷  Published data for HSQ and the HSQ R484A/R489A/P492A data in the current study were collected contemporaneously on matched littermates. Comparison indicates that the hemostatic efficacies of HSQ and of HSQ R484A/R489A/P492A are indistinguishable in this model. The clearance of HSQ and HSQ R484A/R489A/P492A also was compared in hemophilia A mice. Mice were injected with 100 U/kg fVIII, and samples were taken at 0.25, 0.5, 1, 2, 4, 8, and 24 hours, as described in “Materials and methods.” There was no significant difference between the HSQ and the HSQ R484A/R489A/P492A groups at any of the time points (data not shown).

We tested the hypothesis that mutagenesis of an immunodominant fVIII B-cell epitope can reduce the inhibitory antibody response to fVIII. The A2 epitope was selected for study because it contains known antigenic amino acids that can be mutated without loss of procoagulant function.^15,18  Two BDD fVIII mutants containing mutations at R484A/R489A and R484A/R489A/P492A, respectively, were compared with BDD human fVIII in hemophilia A mice.

We used an immunogenicity model developed by Qian et al³¹  in which fVIII was given intravenously to hemophilia A mice using a dosing schedule that mimicked clinical use. Mice received 10 μg/kg (approximately 80 U/kg) fVIII every 2 weeks for 6 doses and a final dose of 25 μg/kg (approximately 200 U/kg) 2 weeks later. These doses corresponded to the high end of the clinical spectrum on a body weight basis. The high-titer inhibitory antibody response that occurred was consistent with observations by other groups using similar dosing regimens.^31-34 

The E16 knock-out hemophilia A mice used in this study, along with the genetically similar, phenotypically indistinguishable E17 knock-out mice, develop immune responses to human fVIII that require CD4⁺ T-cell help^31,32,34  and B7/CD28 and CD40/CD40L costimulation pathways.^35-38  IgG1, IgG2a, and IgG2b anti-fVIII antibodies are produced, indicating a combined T_H1 and T_H2 response.^34,37-39  fVIII-specific interferon-γ⁺ (IFN-γ⁺),³⁸  interleukin-4⁺ (IL-4),³⁴  and IL-10⁺ T cells^34,38  have been identified in immunized mice, which also is consistent with a combined T_H1 and T_H2 response. In human hemophilia A, anti-fVIII antibodies are a mixture of IgG1, IgG2, and IgG4 antibodies,^40-43  again indicating a combined T_H1 and T_H2 response.

T-cell proliferation assays using spleen cells from immunized hemophilia A mice in response to synthetic peptides that overlap the entire human fVIII sequence indicate that the A1, A2, A3, B, C1, and C2 domains all contain T-cell epitopes.³⁴  Similar results have been obtained using peripheral blood mononuclear cells from hemophilia A patients with inhibitory antibodies to fVIII.⁴⁴  Thus, immune responses to fVIII in human and murine hemophilia A appear to be similar. However, for hemophilia A mice to be a valid model for studying the immunogenicity of fVIII, it is necessary that they recognize the A2 R484-I508 epitope. This was confirmed in the current study by evaluating the antigenic properties of hybrid human/porcine and human/murine fVIII molecules in hemophilia A mice immunized with human BDD fVIII. HP9, which is human except for the insertion of a porcine sequence within the R484-I508 segment, was less antigenic than human BDD fVIII (Figure 2A). In contrast, HM1, which is murine except for the insertion of a human sequence within the 484-508 segment, was more antigenic than murine BDD fVIII (Figure 2B). Furthermore, there was a correlation between the reduced antigenicity of HP9 and the increased antigenicity of HM1 (Figure 3), indicating that the antigenicity of the A2 epitope in individual mice is detected similarly in both assay systems.

The immune response to blood-borne soluble proteins such as fVIII develops in the spleen. Antigen-presenting cells, primarily dendritic cells, take up antigen nonspecifically and present proteolytically degraded antigenic peptides on the surfaces of major histocompatibility complex (MHC) class 2 molecules. Antigen presentation to T-cell receptors on CD4⁺ cells and engagement of costimulatory receptors result in T-cell activation. CD4⁺ T cells provide help to B cells, which initially recognize antigen through the mIg component of the B-cell receptor (BCR) complex. The BCR binds the native protein, recognizing the antigenic surface that ultimately will be targeted by mature antibody. The antibody evolves by affinity maturation, and the contacts underneath the antibody footprint change. B cells also are antigen-presenting cells. However, in contrast to dendritic cells, B cells take up antigen specifically and with high affinity using the BCR. Given their ability to bind antigen at low concentrations, B cells are considered particularly efficient antigen-presenting cells as the immune response matures and the concentration of free antigen decreases because of the formation of antigen-antibody complexes.⁴⁵ 

The underlying hypothesis in this study is that initial BCR-dependent antigen recognition and antigen presentation are functions of the structure of the antigen itself. This contrasts with the view that the entire surface of a protein is potentially immunogenic and that host regulatory mechanisms, especially T-helper functions, direct the outcome of the antibody response.⁴⁶  In part, the latter reasoning is based on the fact that it is possible to raise antibodies to nearly the entire surface of a protein, at least when small globular proteins are used as immunogens in the presence of strong adjuvants. Additionally, the vertebrate immune system has evolved mechanisms to produce an almost unlimited antibody repertoire in any individual. However, if a distinction is made between possible epitopes and dominant epitopes, the hypothesis that a hierarchy of immunodominant epitopes exists as a function of the intrinsic immunogen structure seems tenable. Conceivably, the coevolution of immunoglobulin genes and pathogens leads to a bias toward structures that antibodies recognize more easily.

In support of this hypothesis, the R484A/R489A/P492A mutant, but not the R484A/R489A mutant, produced lower inhibitory anti-fVIII antibodies in hemophilia A mice compared with the control group receiving human BDD fVIII. This suggests that P492 is particularly important in the immunogenicity of fVIII. An x-ray structure of the A2 domain is unavailable. However, a model of the fVIII A1-A2-A3 domains has been developed⁴⁷  by homology to the triplicate A domain structure of ceruloplasmin.⁴⁸  In the model, the R484-I508 segment projects as a large loop between 2 β strands. The loop is exposed at the surface and does not interact with the A1 or A3 domains, suggesting that it has significant conformational flexibility. Proline has less conformational freedom than any other amino acid residue because its side chain is fixed by a covalent bond to the main chain. Conceivably, the loss of structural constraints associated with the P492A mutation is associated with a decrease in the intrinsic immunogenicity of the R484-I508 loop. Although it has been proposed that mobile antigenic sites are inherently more immunogenic than less mobile segments, this hypothesis has not been firmly established.⁴⁹ 

Although our hypothesis is that the immunogenicity of fVIII can be reduced at the level of the BCR, we cannot exclude the possibility that the R484A/R489A/P492A mutant altered a T-cell epitope. The relationship between B- and T-cell epitopes has been investigated in several antigenic systems. The B-cell epitope of some antigens is protected from proteolytic degradation and MHC class 2 presentation by binding to the BCR, leading to a bias against T- and B-cell–epitope overlap.⁵⁰  However, occasionally B- and T-cell epitopes do overlap.⁵¹  Additionally, in some cases, antibody binding to antigen can enhance the presentation of a T-cell determinant while simultaneously suppressing the presentation of another T-cell determinant within the same antibody footprint.⁵²  Overlap of B- and T-cell epitopes was identified in a patient with mild hemophilia A that was produced by an R2150H missense mutation in the fVIII C1 domain.⁵³  The patient developed antibodies to wild-type fVIII, but not self-fVIII, indicating that the B-cell epitope included R2150. T-cell clones from the patient recognized synthetic peptides encompassing the R2150 site. However, this patient is atypical because most hemophilia A patients who develop inhibitory antibodies produce no fVIII to which a state of immune tolerance can be developed.

The R484-I508 region in fVIII can be subjected to extensive mutagenesis yet retain full procoagulant activity.¹⁸  In the current study, the specific activity of the R484A/R489A/P492A mutant was not significantly different from that of BDD human fVIII (see “Materials and methods”). The R484-I508 region also interacts with the low-density lipoprotein receptor-related protein in mice and participates in the clearance of fVIII.⁵⁴  However, the in vivo clearance and the hemostatic efficacy of the R484A/R489A/P492A mutant are indistinguishable from BDD human fVIII (Figure 6; see “Results”). This indicates that residues in the R484-I508 epitope that are involved in fVIII procoagulant activity and clearance differ from antigenically important residues. Alternatively, the R484-I508 segment may not be involved in functionally important macromolecular interactions, and antibodies to this region may be inhibitory because of steric hindrance.

Inhibitory antibodies to fVIII were measured using a modified Bethesda assay, which also detects anti-C2 and other inhibitory antibodies in addition to anti-A2 antibodies.^13,16  Antibody neutralization assays indicate that human anti-A2 and anti-C2 antibodies contribute approximately equally to the overall inhibitor titer.¹³  Our results in the murine system are consistent with this observation because mutagenesis of the A2 epitope resulted in only a partial reduction of the inhibitor titer. Combined mutagenesis of the A2 and C2 domains may further reduce the immunogenicity of human fVIII.

In contrast to the Bethesda assay results, there was no significant difference between the R484A/R489A/P492A mutant and BDD human fVIII groups in anti-fVIII antibodies measured by ELISA, which detects inhibitory and noninhibitory antibodies (Figure 5). Noninhibitory anti-fVIII antibodies have been identified in patients with hemophilia A,⁴³  though their relative contribution to the total antibody response is unknown. Our findings suggest that ELISA is insensitive to reductions in immunogenicity of the A2 inhibitor epitope because of the background resulting from noninhibitory antibodies and inhibitory non-A2 antibodies.

The incidence of inhibitory antibody development to human BDD fVIII was greater than 90% in the hemophilia A mice (Figure 4). In contrast, most patients with hemophilia A do not develop inhibitory antibodies, despite repeated exposure to fVIII, suggesting that a state of immune tolerance develops. It is possible that lower doses of fVIII in the hemophilia A mouse model similarly would produce an immune group and a tolerant group. If so, it would be important to determine whether the R484A/R489A/P492A mutant reduces the incidence of inhibitory antibody development as a predictor of its potential clinical use as a low-immunogenicity fVIII product.

Normal C57BL/6 mice mount lower immune responses to human fVIII than E16 hemophilia A mice, indicating that they are partially tolerant because of sequence similarity between human and murine fVIII.³¹  E16 and E17 mice contain targeted disruptions in the region of the fVIII gene that encodes the A3 domain.^20,55  Both strains produce a circulating, nonfunctional polypeptide that includes the A2 domain.⁵⁶  However, E16 hemophilia A mice are intolerant of the A2 domain (Figures 2, 3), possibly because of nonnative folding of the truncated fVIII polypeptide chain, which prevents the induction of tolerance. Alternatively, differences in amino acid sequences between human and murine A2 domains may be responsible for the observed antihuman A2 antibodies.

The study of protein immunogenicity has been heavily influenced by experimental systems in which strong immune responses are elicited. In contrast, little effort has been made to reduce the immunogenicity of proteins. Analyzing immunologic mechanisms that underlie modification of the fVIII molecule to reduce its immunogenicity may lead to new insight into immune responses to proteins.