The Inhibitor Antibody Response Is More Complex in Hemophilia A Patients Than in Most Nonhemophiliacs With Factor VIII Autoantibodies

THE X-LINKED GENETIC disorder hemophilia A results from mutations that reduce the activity or prevent the normal synthesis and secretion of blood coagulation factor VIII (fVIII). Hemophilic patients are treated with plasma-derived or recombinant fVIIIs (pdfVIII, rfVIII) to correct their clotting deficiency. However, approximately 25% of patients infused with fVIII mount an IgG immune response, which leads to its inactivation.1 The presence of such inhibitory antibodies is most frequent in individuals whose fVIII genes contain large deletions, nonsense mutations, or inversions and who are thereby probably not tolerant to infused fVIII.2 Hemophiliacs with missense mutations or small deletions of fVIII have a much lower inhibitor risk.2 Rarely, anti-fVIII inhibitor autoantibodies are found in individuals with normal fVIII.3 

The fVIII protein consists of two series of repeated homologous domains and a unique B domain arranged in the following order: A1-A2-B-A3-C1-C2. Small regions rich in acidic amino acid residues (AR) are located between domains A1 and A2 (AR1), A2 and B (AR2), and B and A3 (AR3).4 The fVIII heavy chain consists of A1-AR1-A2-AR2-B, and the light chain (LCh) consists of AR3-A3-C1-C2. The presence of all A and C domains is essential for the activity of fVIII as a cofactor for factor IX in the intrinsic pathway of blood coagulation,5,6 whereas the B domain is not required.7 8 

In earlier studies, we demonstrated that anti-A2 and anti-C2 antibodies were present in 68% (19 of 28) of inhibitor plasmas tested by immunoprecipitation assays9,10 and that there were also rarer anti-A3 and anti-C1 antibodies detected by immunoblotting.11 In other studies, 2.9% (two of 68) and 46% (38 of 82) of inhibitor plasmas contained antibodies that bind to AR1^12,13 or AR3,14 respectively. These results establish that the immune response to fVIII is oligoclonal and that there are at least six different epitopes for anti-fVIII antibodies.

We have demonstrated that anti-A2 and anti-C2 domain antibodies are fVIII inhibitors because their ability to inactivate fVIII can be prevented by preincubation with saturating concentrations of the recombinant A2 or C2 domains.9,15 We also concluded that some patients have antibodies to AR3-A3-C1 because their inhibitor titer was neutralized incompletely by a combination of A2 and C2, but completely by A2 and LCh.15 To rule out the possibility that the greater neutralization by LCh than by C2 was due to an abnormal structure of the recombinant C2 that was used, we also demonstrated that similar molar concentrations of each were required for maximal neutralization.15 Synthetic peptide fVIII 351-365, corresponding to part of AR1, was able to neutralize rare human antibodies binding to this region; therefore, anti-AR1 antibodies are also inhibitors.12 

The contribution of the different antibodies to the plasma inhibitor titer had been examined in a limited number of patients (14)^9,15; however, among these there was considerable heterogeneity in their number and epitope specificity. We report in this study a more extensive analysis of the inhibitors in plasmas from 55 patients. Our purpose was to determine if the multiplicity or the domain specificity of inhibitors varies among individuals treated with different fVIII products or in which the inhibitors arose as autoantibodies or as a complication of hemophilia A. Comparisons were, therefore, made between hemophilic and autoantibody patients treated with low or intermediate purity pdfVIII products and between hemophiliacs treated exclusively with pdfVIII or the rfVIII products Recombinate16 or Kogenate.17 

Antibodies.Monoclonal antibody (MoAb) ESH8 IgG was purchased from American Diagnostica (Greenwich, CT). Its fVIII epitope was localized to C2 domain amino acid residues 2248-2285 by immunoblotting assay.15 The epitope of MoAb 413 was localized by immunoblotting to A2 domain residues 373-606.18 A smaller area critical for MoAb 413 binding to A2 was further localized to residues 484-508 by replacing these residues with the corresponding porcine residues and demonstrating that MoAb 413 was not an inhibitor of this hybrid fVIII.19 The MoAb 413 was purified from ascites fluid on a protein G Sepharose column (Gammabind G, Pharmacia Biotech, Uppsala, Sweden), as described.18 Citrated human plasmas from inhibitor patients were used in most experiments, and IgG prepared as for MoAb 413 was used as indicated. Plasma inhibitor titers were measured in the Bethesda assay20 and expressed as Bethesda units (BU)/mL. Inhibitor IgG of patient NS was further affinity purified on an fVIII LCh column as described.15 The total yield of IgG and BU in the flow through were 78% and 80%, respectively, of the starting material. The yield of eluted LCh-specific IgG and Bethesda titer were 2% and 12% of the input, respectively. The specific activity was increased 6.3-fold to 3,900 BU/mg IgG by affinity purification.

Factor VIII.The highly purified rfVIII Recombinate was generously provided by Baxter Healthcare, Glendale, CA. Its specific activity was approximately 3,000 U/mg when fVIII activity was measured in a one-stage clotting assay. FVIII LCh was purified from Recombinate as described.9 For some experiments, LCh was similarly purified from pdfVIII (3,750 U/mg), and it gave identical results.

Neutralization assay.FVIII cDNAs encoding the A2 or C2 domains preceded by a heterologous signal peptide were cloned into baculovirus transfer vector pVL941 (kindly provided by Dr Max Summers, University of Texas, Austin, TX), and they were expressed as soluble, secreted polypeptides in Sf-9 insect cells.9,18 The A2 domain polypeptide concentration in the growth medium was 1 to 2 μg/mL, as measured by enzyme-linked immunosorbent assay (ELISA).9 The growth medium was concentrated 20-fold by ammonium sulfate precipitation and dialyzed in Tris buffered saline, pH 7.5, before use. The C2 concentration in the growth medium was 35 μg/mL, and it was not further concentrated. The fVIII domain preparations were stored frozen at −80°C without protease inhibitors, which were not required because we observed no degradation of the secreted, radiolabeled proteins in previously described immunoprecipitation assays.9,18 Neither A2, C2, or LCh at the highest concentrations used in the neutralization assays had any effect on fVIII activity. The growth media containing the secreted polypeptides or the purified fVIII LCh were used in assays that tested the ability of increasing concentrations of each to neutralize the inhibitor titer (3 to 4 BU/mL) of patient plasmas, as previously described.9 15 

Semiquantitative immunoprecipitation assays of antibody binding to recombinant A2 and C2 domains.The A2 and C2 cDNA constructs described above were also cloned into plasmid pMT2,21 generously provided by Genetics Institute, Cambridge, MA. The DNA was used to transfect COS-1 cells by the diethylaminoethyl (DEAE)-dextran method, the transfected cells were labeled with ³⁵S-methionine (Tran³⁵S-label, ICN Biomedicals, Costa Mesa, CA),9 and the growth medium was used for immunoprecipitation analysis of the binding of antibodies from inhibitor plasmas to the radiolabeled polypeptides. The immune complexes were captured with protein G Sepharose and washed extensively, followed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and autoradiography.9 

Patient plasmas to be tested by the inhibitor neutralization assay.In this study, we tested plasmas from 23 hemophilic (HA) and 21 autoantibody (AA) inhibitor patients who had been treated with low and intermediate purity pdfVIII products, and most of whom were older children or adults. Eleven hemophilic inhibitors from patients treated only with rfVIII products were also evaluated. These patients were infants and small children enrolled in clinical trials for testing of the safety and efficacy of Recombinate16 or Kogenate.17 The hemophiliacs were all classified as severe based on <2% fVIII activity. From 43 of the 55 patients, we were able to obtain additional information about their clinical history. The average age of the patients at the time the plasma sample was obtained and the average length of fVIII treatment is shown in Table 1. Four of the autoimmune patients had associated cancers, three had a variety of other diseases, one was postpartum, and the others had no additional diseases. Ten of the autoantibody patients had never been treated with fVIII, eight had received immunosuppressants for inhibitor therapy, and one (EM) received porcine fVIII. Plasmas from single time points were tested for each patient because serial samples were available only from those treated with rfVIII.

Most hemophilic inhibitor patients have multiple antibodies that contribute to the plasma inhibitor titer.Each plasma was diluted to 3 to 4 BU/mL and mixed 1:2 with increasing concentrations of A2, C2, or LCh. After 2 hours incubation, each mixture was diluted 1:2 into normal pooled plasma for an additional 1 hour, and the inhibitor titer was measured by the Bethesda method.20 When the second incubation step was extended to 2 hours, the results were similar. All plasmas tested had inhibitor titers ≥10 BU/mL in order to achieve the required titer in the final incubation of antibody, neutralizing fragment, and normal pooled plasma. A positive control of inhibitor with no competing fVIII fragment was included. Duplicate determinations were done for each point, and the variation between them was ≤5%. The percent inhibitor neutralization was calculated as (1-[BU/mL inhibitor+fragment/ BU/mL inhibitor control]) × 100. The plasmas were tested for inhibitor neutralization by all fVIII fragments in the same experiment, and the results were verified in ≥1 duplicate experiments. When fVIII fragments were partially neutralizing, all values within the plateau region were averaged. The plateau was defined as the point beyond which increasing fVIII fragment concentrations did not lead to greater neutralization. The lower limit of detection was 5% to 10% neutralization and the upper limit ≥95%.

The neutralization results for four hemophilic inhibitor patients are shown in Fig 1 and those for the others are summarized in Table 2. The R7611 inhibitor (Fig 1A) was partially neutralized by A2 (74%) and C2 (27%). Neutralization by the LCh (23%) was similar to that of C2, indicating that the only LCh inhibitors were anti-C2 antibodies. When A2 and C2 at saturating concentrations were combined, complete neutralization was observed (not shown), demonstrating that antibodies with at least two different epitopes made up 101% of the inhibitor titer. Inhibitor KB (Fig 1B) was completely neutralized by the LCh, partially neutralized by C2, and not neutralized by A2; therefore, the major inhibitory antibodies are directed only against the LCh. Plasmas such as KB were considered to have two distinct anti-LCh inhibitors if the percentage neutralization by LCh was >10% more than that for C2, and it could be repeated in duplicate experiments. Smaller differences could not be reproduced. For the data we obtained, this cutoff value appeared to be reasonable because only four of the 46 plasmas with anti-LCh antibodies (UJ, AA, FM, and CHI) were within 2 points (8% to 12% neutralization, LCh- C2) of the cutoff. GK1831 (Fig 1C) inhibitor binding to at least three different regions was observed, as shown by partial neutralization with A2, C2, and LCh. The greater neutralization by LCh than by C2, although not as dramatic as in Fig 1B, was also observed. For plasmas such as GK1831 with both LCh and heavy chain inhibitor epitopes, the sum of the maximal percentage neutralization by LCh plus heavy chain polypeptides cannot theoretically exceed 100. For GK1831, this total was 96% (heavy chain 40% plus LCh 56%). The average value of maximal neutralization for all plasmas tested (Table 2) except WD was 96% ± 9%.

The WD inhibitor (Fig 1D) contained antibodies to a rare epitope (amino acid residues 355-362) within AR1.22 Fifty-nine percent of its inhibitor titer was neutralized by synthetic peptide 355-362, 56% by A2, 34% by LCh, and <10% by C2. The percentage neutralization by A2 plus peptide 355-362 of 115% and additional neutralization of 34% by LCh for a total of 149% lead us to assume that one of the determinations was incorrect, particularly because this characteristic was not seen for any other plasma. When we calculated the lowest polypeptide concentrations, which gave maximal inhibitor neutralization, we obtained the following values for WD: A2, 2.2 nmol/L; peptide 355-362, 92 nmol/L; and LCh, 50 nmol/L. Although the maximal peptide concentration is expected to be high because its conformation is probably not identical to the native sequence in fVIII, the high concentration of LCh required suggests that antibodies are binding to it with considerably lower affinity than to A2. This possibility is consistent with the lower slope of the LCh neutralization curve. Due to the inefficient neutralization by LCh compared with A2, we have currently listed the epitope specificity of WD as A2 and peptide 355-362. For the other patients in Fig 1, the concentrations of all fVIII polypeptides required for maximal neutralization were more similar. R7611 was optimally neutralized by 9.1 nmol/L C2 and 6.7 nmol/L A2; KB by 6.5 nmol/L C2 and 12.5 nmol/L LCh; and GK1831 by 2.1 nmol/L A2, 4.2 nmol/L C2, and 3.1 nmol/L LCh. No other inhibitors tested required the high concentration of LCh needed for WD neutralization. The similar concentrations for maximal neutralization by LCh and C2 of all plasmas, except WD, reinforces our earlier conclusion15 that the recombinant C2 polypeptide does not have an abnormal conformation or require other regions of the LCh for formation of the inhibitor epitope(s).

Although the percentage inhibitor neutralization varied ≤5% in duplicate measurements, there was as much as 30% unexpected variation between some points, as seen in the plateau regions for A2 neutralization of R7611 (Fig 1A) as well as A2, C2, and LCh neutralization of GK1831 (Fig 1C). At higher A2 and C2 concentrations, the neutralization appeared to decrease, but this was not always seen in different experiments (KB, Fig 1B; WD, Fig 1D) or in repetitions of the same experiment. We have not been able to define the source of this variation in our complex assay. For this reason, all results have been confirmed in two to four duplicate experiments.

The results of inhibitor neutralization assays for hemophiliacs treated with rfVIII are summarized in Table 3. A comparison of the inhibitor specificity for hemophiliacs treated with pdfVIII or rfVIII is shown in Table 4. Table 4 (section A) lists the percentages of plasmas with antibodies to single or multiple regions of fVIII. The average percentage of neutralization by anti-heavy chain plus anti-LCh polypeptides was 103% ± 10%. Both groups of hemophiliacs had a heterogeneous immune response to fVIII infusion and similar percentages of inhibitors that bound to 1, 2, or 3 different fVIII regions, ie, A2, C2, peptide 355-362, AR3-A3-C1, or their combinations. However, there were some differences in the frequencies and epitope specificities of the antibodies in the two groups. Indistinguisable percentages of plasmas had antibodies to A2 or C2 (Table 4 [section B]), but antibodies to AR3-A3-C1 were common in patients treated with pdfVIII (61%) but rare in patients treated with rfVIII (18%). The combination of anti-C2 plus anti-AR3–A3-C1 antibodies was not found in the rfVIII plasmas, but it was present in 35% of pdfVIII plasmas. Conversely, the combination of anti-C2 plus anti-A2 antibodies was more common in the rfVIII (45%, Table 4 [section C]) than in the pdfVIII (17%) group.

Antibodies to single fVIII domains account for the inhibitor titer in most autoantibody patients.Similar neutralization experiments with inhibitor plasmas from anti-fVIII autoantibody patients generated considerably different results (Table 5). Sixty-two percent (13 of 21) of these patients had antibodies to one domain, C2 or A2, which accounted for all of the inhibitor titer (Table 4 [section A]). The most unusual result was the preponderance of anti-C2 antibodies as the sole inhibitor in 48% (10 of 21) of autoantibody plasmas compared with 2.9% (1 of 34) in hemophiliacs treated with rfVIII or pdfVIII (Tables 2, 3, and 5). The average percentage of neutralization by anti-heavy chain plus anti-LCh polypeptides was similar (93% ± 8%) to the other patient groups. The number of autoantibody plasmas with inhibitors binding to one polypeptide was, therefore, overrepresented and those binding to two or three polypeptides was underrepresented compared with the hemophiliacs. The percentage of plasmas with anti-A2 antibodies (Table 4 [section B]) was lower (33%) than for both hemophilic groups (61%, 91%), while that for anti-C2 antibodies was equal for all three groups. Antibodies against AR3-A3-C1 were less frequent (29%) in the autoantibody than in the pdfVIII hemophilic plasmas, but not significantly different from the rfVIII plasmas. Nine autoantibody patients who were never treated with fVIII (indicated by an asterisk in Table 5) had inhibitor epitope specificity patterns distributed approximately equally among those of single or mulitple polypeptides. Therefore, fVIII treatment is not necessary for generation of a complex immune response in these patients. Nine of the 15 patients for whom we had treatment information were given immunosuppressants. Five of these had only anti-C2 antibodies as the major inhibitor, one with A2+C2, two with C2+AR3-A3–C1, and one with A2+C2+AR3-A3–C1. Although these data suggest that immunosuppressants do not reduce the complexity of the immune response to fVIII, the numbers are too small to be statistically distinguishable.

The presence of antibodies that do not contribute significantly to the inhibitor titer.Immunoprecipitation (IP) assays for anti-A2 and anti-C2 antibodies were performed in all inhibitor plasmas tested above to determine if additional antibodies that do not contribute to the inhibitor titer are present. A representative example of an immunoprecipitation of four plasmas with ³⁵S-A2 or C2 is shown in the autoradiograms of Fig 2. The plasma samples were used at a 1:3 dilution so that low levels of antibodies could be detected. The negative control of no antibody demonstrated that the A2 and C2 bands were not due to comigrating cellular proteins and that these domains did not bind nonspecifically to the protein G Sepharose used to precipitate the immune complexes. Because this assay is semiquantitative, the samples were scored only as positive or negative for each patient in Tables 2, 3, and 5. If one examines those patients whose inhibitors are neutralized solely by either A2 or C2, it is evident that most of them (11 of 18) have additional antibodies against the other domain.

In autoantibody patient NS, the inhibitor titer was neutralized only by A2, but anti-C2 antibodies were also present (Table 5). We affinity purified the IgG on an immobilized fVIII LCh column, as described in Materials and Methods, to determine if these antibodies are inhibitory. The LCh-specific IgG accounted for 15% of the combined inhibitor titer of the NS IgG in the flow-through and in the eluate. These results suggest that the anti-LCh inhibitors are present at lower concentrations in the total IgG than the anti-A2 inhibitors. The unfractionated IgG inhibitor was neutralized 89% by A2 and ≤5% by LCh. However, the LCh-specific IgG was neutralized ≤5% by A2, 45% by C2, and 75% by LCh. Based on these results, we propose that there are at least three different types of inhibitor antibodies in the NS plasma, but because of its higher concentration, the anti-A2 IgG is the sole inhibitor measured in the neutralization assay. The anti-LCh antibodies appear to have a specificity not unlike that in other patients where these antibodies predominate, ie, anti-C2 plus anti-AR3-A3-C1.

We have showed that the combinations of antibodies, which contribute significantly to the plasma inhibitor titer of hemophilia A patients, is more complex than that of normal individuals with fVIII autoantibodies. We define complexity as the minimum number of different epitopes recognized by the inhibitors from one plasma. For both groups of patients, the antibody or combinations of antibodies that make up the inhibitor titer usually consist of those with A2, AR3-A3–C1, or C2 domain epitopes. This conclusion is based on the ability of polypeptides containing the inhibitor epitopes, alone or in combinantion, to completely neutralize the inhibitor activity. Antibodies with an epitope in AR1 rarely (one of 55) contributed to the inhibitor titer, although they were also found in six other plasmas (H. Nakai, unpublished results, August 1995). The neutralization assay we have used to make these determinations is limited by its relatively low sensitivity; therefore, we have been able to test only high responder inhibitors (≥10 BU/mL), and we do not know if low responders have the same properties.

That antibodies, which bind to the LCh outside C2, are inhibitory was inferred from the greater neutralization of many plasmas by the LCh than by C2. Because C2 is part of the LCh, this suggested the existence of at least two distinct LCh inhibitor epitopes in C2 and AR3-A3–C1. We do not believe that our interpretation is an artifact due to an abnormal conformation of the C2 domain because in both the present and a previous study15 equimolar concentrations of the C2 domain and the LCh were required for the neutralization of monoclonal and human antibodies with a predominant anti-C2 inhibitor. We have recently expressed the recombinant A3-C1 region as a soluble, secreted protein (D. Zhong, D. Scandella, unpublished results, April 1996); therefore, direct neutralization of inhibitors by this fragment in the future should provide final confirmation for the existence of an inhibitor epitope within this region of fVIII.

A comparison of the individual inhibitor patterns of autoantibody patients and hemophiliacs showed that the major difference between them is the large percentage (48%, 10 of 21) in the former group of anti-C2 antibodies as the sole detectable inhibitor. Only one of 34 hemophiliacs treated with pdfVIII or rfVIII had this characteristic. In addition, significantly fewer autoantibody plasmas contained anti-A2 and anti-AR3–A3-C1 antibodies, and the occurrence of A2 plus C2 antibodies as the major inhibitors was also less frequent. It is interesting that eight autoantibody patients had complex inhibitor patterns indistinguishable from those of the hemophiliacs, yet four of them had never been treated with fVIII. Stimulation of the immune response with exogenous fVIII is, therefore, not required to generate a complex response. Among nine autoantibody patients given immunosuppressants for inhibitor therapy, four had complex inhibitor patterns of two or three different antibodies, while five had only anti-C2 antibodies. Although the data are limited, there is so far no indication that immunosuppressants decreased the complexity of the inhibitor response.

When hemophiliacs treated with pdfVIII or rfVIII therapeutic products were compared, their inhibitor patterns were equally complex (Table 4, section A). In contrast, the presence of anti-AR3–A3-C1 antibodies was threefold more common in pdfVIII-treated patients. Further differences were (1) that 35% of the pdfVIII group, but none (zero of 11) of the rfVIII group, had the combination of anti-C2 plus anti-AR3–A3-C1 antibodies as >95% of the inhibitors, and (2) 45% of the rfVIII group had the combination of anti-A2 and anti-C2 inhibitors, but only 17% of the pdfVIII. Because these patient groups varied in both the fVIII product used for therapy, in age, and in years of fVIII exposure (Table 1), it is not clear from our data which factor(s) may be responsible for the observed differences in epitope specificity of the major inhibitors.

The information we have obtained is from single time points in the history of each inhibitor patient. As we tested 55 patient plasmas, the results can probably be considered as representing most of the types of inhibitor patterns that normally exist. It is important to note, however, that the epitope specificity of the anti-fVIII antibodies for an individual is not static over time, as described in the immunoblotting study of Fulcher et al,23 and that switching from one inhibitor pattern to another may occur. We have examined the patterns in serial samples of two patients treated with rfVIII over periods of 4 or 10 months. The five samples all had anti-A2 and anti-C2 antibodies as the predominant inhibitors. Although the epitope specificity did not change, the relative contribution of these antibodies to the inhibitor titer did change (R. Prescott, unpublished results, November 1994).

For patient NS, we showed that only anti-A2 inhibitors were detectable in the plasma, but after affinity purification on a LCh column, anti-C2 and anti-AR3–A3-C1 inhibitors were also found. The detection of a particular inhibitory antibody, therefore, depends not only on its epitope specificity, but also on its concentration in the plasma. In the 18 patients whose inhibitor was specific for one fVIII domain, antibodies to at least one other domain were present in 11, as measured by immunoprecipitation assays. These may be nonihibitory antibodies, which were previously postulated to exist in most inhibitor plasmas on the basis of competition assays,24 but it seems likely that some of them are inhibitors. Because the antibody profiles in individuals vary over time,23 inhibitors present at lower concentrations at one time may become clinically significant as the dominant inhibitors at a later time. Potential therapies targeted at eliminating inhibitors must, therefore, be capable of eradicating all of them, regardless of their titer.

Although we have showed that the inhibitor patterns of hemophiliacs and autoantibody patients are not identical, the epitopes of at least some of the inhibitory antibodies are probably similar. Our earlier experiments in which one inhibitor ¹²⁵I-Fab′ was competed for A2 domain binding by Fab′ from other hemophilic and autoantibody plasmas demonstrated that they shared an A2 epitope.18 Hemophilic and autoantibodies with C2 domain specificity act by preventing the binding of fVIII to phospholipid surfaces and to von Willebrand factor,15,25,26 and they compete for binding to C2 of the MoAb ESH8, which has an overlapping epitope.10 These results imply that the C2 epitope is also shared among different inhibitors. The epitope similarities and the probable low number of major inhibitor epitopes shown by our studies suggest that strong inhibition of fVIII activity can only be achieved by antibody binding to a few specific sites, which are major functional epitopes. Strategies for making mutant fVIIIs, which do not bind inhibitors, such as those proposed by Lollar et al,19 27 may thereby be feasible, particularly for those patients who have a simple inhibitor profile.

We thank the following physicians for sending us their patient plasmas and for the data, which are summarized in Table 1: Craig Kessler, Leon Hoyer, Gilbert White II, Edward Tuddenham, Liberto Pechet, Emily Czapek, Margaret Ragni, Jessica Lewis, David Green, and George King, Inc.