Some Human Inhibitor Antibodies Interfere With Factor VIII Binding to Factor IX

MoAbs CLB-CAg A and NMC-VIII/5 were kindly provided by Drs Jan van Mourik (Central Laboratory, Netherlands Red Cross) and Midori Shima (Nara Medical College, Japan), respectively. The preparation of IgG from plasmas of severe hemophilia A patients with inhibitors was previously described.6 Inhibitor titers were determined in the Bethesda assay.11 

A cDNA construct encoding the signal peptide of interferon-β in frame with the A3 and C1 domains (amino acids 1690-2172) and two stop codons was assembled by overlap extension12 and cloned into plasmid pMT2,13 which also encodes dihydrofolate reductase (dhfr). The pMT2 A3-C1 construct was used to transfect dhfr⁻ Chinese hamster ovary (CHO) cells, a stable dhfr⁺ cell line expressing A3-C1 was isolated, and CHO cells containing amplified copies of A3-C1, verified by Southern blotting, were selected in increasing concentrations of methotrexate13 to 0.32 μmol/L. The C2 cDNA including the stop codon was attached to a 5′ sequence encoding the prepro polypeptide of tissue plasminogen activator,14 five histidines, and a thrombin cleavage site by overlap extension,12 and the resulting DNA was cloned into baculovirus vector pVL1393 (Invitrogen, Carlsbad, CA). Using the Bac to Bac expression system (Life Technologies, Gaithersburg, MD), the C2 cDNA was transferred to a baculovirus plasmid by in vivo site specific transposition in Escherichia coli. The C2-baculovirus plasmid was purified and used to transfect Sf9 insect cells for generation of baculoviruses expressing C2. Secreted C2 was quantitated by enzyme-linked immunosorbent assay (ELISA)14 as 27 μg/mL. All cDNA constructs were sequenced to verify the absence of DNA errors.

The amplified A3-C1 cell line, 2 × 10⁶ in 100-mm tissue culture dishes, was grown in Dulbecco's modified Eagle medium (DMEM) α-medium (Life Technologies) with 10% fetal calf serum for 48 to 72 hours to 90% confluency. The cells were washed two times with methionine-free DMEM and labeled in 4 mL of this medium with 0.5 mCi Tran³⁵S-Label (ICN Biomedicals, Costa Mesa, CA) for 4 hours at 37°C. The culture medium was collected, and 1% NP40 (Sigma, St Louis, MO) was added. Cell lysates were suspended in 50 mmol/L Tris, pH 7.4, 150 mmol/L NaCl (TBS), 1% NP40. The growth medium and cellular fractions were centrifuged at 10,000 rpm for 10 minutes at 4°C. Aliquots of the supernatants were stored at −80°C.

Plasma derived fVIII was purified from fVIII concentrate Method M (American Red Cross),15 and heavy and light chains were further purified as described.15,16 The A1 and A2 domains were purified from plasma fVIII after thrombin cleavage¹⁷, and recombinant C2 polypeptide was purified from Sf9 insect cell growth medium.14 Ammonium sulfate (38%) was added to the Sf9 medium containing 0.7 mg (His)₅-C2, and the C2 precipitate was dialyzed into 50 mmol/L phosphate, 0.3 mol/L NaCl, 0.05% Tween 20, pH 8.0, and passed over a 2-mL Ni²⁺-NTA agarose column (Qiagen, Valencia, CA). The column was washed with the above buffer adjusted to pH 5.8 followed by elution of (His)₅-C2 with a pH gradient of 5.6 to 4.0 in the same buffer. Recovery of (His)₅-C2 was 65%. Recombinant fVIII and B domain were generously provided by Baxter/Hyland (Glendale, CA).

Polypeptides were radiolabeled with 5 μL Na ¹²⁵I (100 mCi/mL; Amersham, Arlington Heights, IL) by immobilized lactoperoxidase (Worthington, Freehold, NJ) as described.17 Specific radioactivities ranged from 5 to 13 μCi/μg for all polypeptides. They were aliquoted and stored at −80°C for up to 1 month.

Details were previously described.17 Briefly, each¹²⁵I-polypeptide (0.75 nmol/L) was incubated with inhibitor IgG or plasma dilutions at 4°C overnight. Immune complexes were precipitated with Protein G Sepharose (Pharmacia Biotech, Piscataway, NJ) and washed to remove unbound ¹²⁵I-polypeptide. The negative control contained all components except antibody. Dose-response binding curves were done for each IgG, and data points from the linear portion of each curve were used to calculate immunoprecipitation (IP) units per milliliter: [(Bound/Total¹²⁵I-fVIII Polypeptide − Background) × Plasma Dilution × 20 (to Convert to IP Units/mL)]. IP of³⁵S-labeled A3-C1 polypeptide in cell culture medium with inhibitor was performed as described above, except that bound³⁵S-labeled A3-C1 was detected by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) (10% acrylamide) and autoradiography.

To determine the concentration of A3-C1, aliquots of the CHO cell culture medium containing A3-C1 were used to compete the binding of¹²⁵I-recombinant fVIII to MoAb CLB-CAg A IgG in competitive IP assays, a modification of the above IP method. Recombinant fVIII was used to generate a standard curve.

Purified recombinant (His)₅-C2 was immobilized on Ni⁺²-NTA agarose at 250 μg per mL of resin via the (His)₅ tag. Inhibitor IgGs were diluted to 2 mg/mL in HBS (20 mmol/L HEPES, 0.15 mol/L NaCl, pH 7.4, 0.01% Tween 80), mixed with resin at a 2:1 vol/vol ratio, and incubated for 18 hours at 4°C. Upon removal of the resin by centrifugation, the supernatants were tested for the presence of anti-C2 antibodies by IP assay using¹²⁵I-labeled C2 polypeptide.

Details were previously described.16 Briefly, plasma fVIII (2 nmol/L) was incubated with increasing concentrations of MoAb CLB-CAg A or inhibitor IgG for 30 minutes at 37°C. Aliquots were diluted 10-fold into HBS (above) with 5 mmol/L CaCl₂, 2 nmol/L fIXa (Enzyme Research Laboratories, South Bend, IN), and 20 μmol/L phosphatidylserine/phosphatidylcholine vesicles (75/25, wt/wt). fVIII was actived with 40 nmol/L thrombin for 30 seconds, and 300 nmol/L factor X was added. Aliquots were withdrawn after 15, 30, 45, and 60 seconds, and factor X activation was stopped with 0.05 mol/L EDTA. Factor Xa generation was measured by cleavage of synthetic substrate S-2765, 0.3 mmol/L (Pharmacia Hepar, Franklin, OH) using a V_max microplate reader (Molecular Devices, Menlo Park, CA). A purified factor Xa standard (Enzyme Research Laboratories) was used to convert absorbance (410 nm) into factor Xa concentration. The molar concentrations of fIXa, factor X, and factor Xa were calculated using molecular masses of 45 kD, 57 kD, and 45.3 kD, respectively.

The kinetics of light-chain binding to ligands was determined by surface plasmon resonance using the IAsys biosensor (Fisons, Cambridge, UK). Anti-C2 MoAb NMC-VIII/5 (50 μg/mL) in 10 mmol/L sodium acetate, pH 5.0, was covalently coupled to an activated carboxymethyldextran coated biosensor cuvette (Affinity Sensors, Paramus, NJ) via amino groups using succinimide ester chemistry.18 fIXa was inactivated for use in the binding studies by reaction with the active site specific reagent DEGR-CK (Calbiochem, La Jolla, CA) as described.19 All measurements of fIXa-DEGR binding to fVIII light chain were performed in 200 μL HBS, 5 mmol/L CaCl₂at 37°C. Dissociation of bound fIXa-DEGR was initiated by substitution with 200 μL buffer lacking fIXa-DEGR. Binding of MoAb or inhibitor IgG to immobilized light chain in the presence or absence of the synthetic peptide was performed under the above conditions. The cuvette was regenerated by addition of 0.1 mol/L glycine, pH 3.0, for 3 minutes, resulting in complete dissociation of light chain from MoAb.

The previously observed greater neutralization of inhibitors by light chain than by C2⁶ suggested the presence of one or more additional inhibitor epitopes in the A3-C1 domains of the light chain. To determine if antibodies from inhibitor plasmas bind to this region, we constructed an expression plasmid, pMT2,13 encoding the A3 and C1 domains of the light chain but not its amino terminal acidic region, residues 1649-1689. The interferon-β (IFN-β) signal sequence cDNA was linked in frame to the 5′ end of the A3-C1 cDNA for potential secretion of the recombinant polypeptide. A stable CHO cell line containing multiple copies of the A3-C1 cDNA was established (Materials and Methods). Expression of recombinant A3-C1 was analyzed by IP of the growth medium and the cellular fraction of the³⁵S-Met labeled CHO cell line (Fig 1) with MoAb CLB-CAg A IgG (epitope A3 residues 1804-1820¹⁰) followed by SDS-PAGE. Similar concentrations of ³⁵S-A3-C1 with the expected molecular weight (∼60 kD) were seen in the extracellular (lane 1) and the intracellular (lane 2) fractions, demonstrating approximately 50% secretion. Additional bands seen in the cell fraction were not fVIII specific because they were also present when MoAb CLB-CAg A was omitted (Fig 1, lane 3). The A3-C1 concentration in the growth medium was determined to be 0.76 μg/mL in a competitive IP assay (Materials and Methods).

The binding of ³⁵S-A3-C1 to decreasing concentrations of IgG purified from each of 10 inhibitor plasmas was measured by IP followed by SDS-PAGE (Materials and Methods), and representative results are shown in Fig 2. The binding of the positive control, anti-A3 MoAb CLB-CAg A IgG10 (Fig2B), and human inhibitors RI (Fig 2A) and MU (Fig 2C) was dose dependent, and controls without antibody were negative. As both complex antibodies and impure A3-C1 were used, the results in Fig 2 cannot be used for quantitative comparisons among the IgGs. The binding of the eight other human IgGs was also dose dependent in this assay (not shown). For a more extensive analysis of the frequency of A3-C1 binding by IgG in inhibitor plasmas, eight were diluted 1:5 and tested in single-point IP assays. Binding was positive in seven and negative in one (not shown). A3-C1 binding by 17 of 18 human inhibitors from both experiments demonstrated that specificity for A3-C1 is common.

For further characterization of antibodies directed against A3-C1, we selected three IgGs from hemophilia A patients with high Bethesda titers. In previous inhibitor neutralization assays6 the MU, RI, and MS IgGs were neutralized to a greater extent by light chain than by C2 (Table 1), suggesting that they contained 44%, 48%, and 79% anti–A3-C1 and 15%, 26%, and 30% anti-C2 inhibitors, respectively. MU and RI were not detectably neutralized by the A2 domain (≤10%) but MS was 27% neutralized. In addition, binding of each IgG to the individual fVIII heavy-chain domains, the intact light chain, and the isolated C2 domain by semi-quantitative IP assays (Materials and Methods) was also tested. Figure 3 contains representative binding curves for RI binding to the isolated fVIII fragments, and the total results are summarized in Table 1. The combined anti–heavy-chain titers were ≤0.1 of the anti–light-chain titers. The percent anti–A3-C1 antibodies (LCh-C2) was 61%, 37%, and 91% and anti-C2 antibodies was 39%, 64%, and 9% for MU, RI, and MS, respectively. These percentages are not the same as those determined by the inhibitor neutralization assay due to the presence in plasmas of both inhibitory and noninhibitory antibodies.20 

Because the neutralization assays indicated that 15% to 30% of the inhibitor titers were caused by anti-C2 antibodies, each IgG was depleted of these antibodies (Materials and Methods) to measure only the effects of anti–A3-C1 antibodies in the following experiments. IP assays of ¹²⁵I-C2 binding by each IgG before and after the depletion showed the >300-fold removal of anti-C2 antibodies in all three IgGs (Table 1).

Although the fIXa binding site consists of amino acid residues within the A2⁹ and A3 domains,10 the light chain binds to fIXa with moderate affinity (kd =14.8 nmol/L).21 Due to the fact that inhibitors RI, MU, and MS had measurable levels of anti-A2 antibodies (Table 1) that might complicate the interpretation of the results, we chose to examine the effect of inhibitor IgG on fIXa binding to the light chain only. For detection of protein-protein interactions we used a biosensor technique based on the surface plasmon resonance phenomenon which measures protein binding in real time. The change in the refractive index due to association of a fluid-phase ligand with an immobilized ligand is measured by generation of a signal of 200 Arc seconds per 1 ng of protein bound per mm² of the biosensor cuvette. MoAb NMC-VIII/5 (C2 epitope 2170-2327⁸) was immobilized to the cuvette, and the light chain (20 μg/mL) was maximally bound to 35 fmol/mm².

To determine if all immobilized light-chain molecules were accessible for fIXa binding, we measured their maximal binding capacity. To prevent the possible fIXa-mediated cleavage of immobilized light chain at Arg1721,22 we used an active-site modified fIXa (fIXa-DEGR), which lacks proteolytic activity19 but retains fVIII binding properties.23 

Curve 1 in Fig 4 shows the resonance response of the time course of fIXa-DEGR association with light chain and dissociation of the complex upon replacement of fIXa-DEGR with buffer (shown only for curve 1). No binding of fIXa-DEGR to the cuvette was observed in the absence of light chain (data not shown). The association rate constant (k_on = 1.2 × 10⁴ M⁻¹s⁻¹) and the dissociation rate constant (k_off = 3.4 × 10⁻⁴ s⁻¹) for fIXa-DEGR/light-chain interaction were derived from the best fit of these kinetic data to a model describing single-phase association and dissociation processes using the Fast Fit 1.0b computer program (FISONS). Because the concentration of fIXa-DEGR (1,000 nmol/L) is >25 times the corresponding equilibrium dissociation constant (kd = 28 nmol/L, calculated as k_off/k_on) for its interaction with light chain, the amount of fIXa-DEGR bound at equilibrium (36.6 fmol/mm²) represents the maximum binding capacity of immobilized light chain. This result shows that the immobilized light chain (35 fmol/mm²) was able to bind fIXa-DEGR with 1:1 stoichiometry. In addition, the similar values of the kd for fIXa-DEGR/light chain determined in our assay (28 nmol/L) and that determined for fIXa/fVIIIa interaction in solution (42 nmol/L)24 suggest that the light-chain binding properties were not significantly altered by immobilization. The absence of phospholipid is probably a major contributor to the relatively high kd values for fVIII -fIX interaction obtained in both our (28 nmol/L) and previous experiments (42 nmol/L).24 This interpretation is consistent with reported kds of 2.3 and 46 nmol/L for fVIIIa/fIXa association in the presence and absence of phospholipid, respectively.24 

Inhibitor or MoAb IgGs were bound to the light chain at varying IgG/light-chain molar ratios calculated from the resonance signals produced by bound IgG (not shown), which was followed by replacement of the IgG solution with buffer containing 1,000 nmol/L fIXa-DEGR. The binding of fIXa-DEGR to light chain in the presence or absence of bound RI IgG approached equilibrium after 50 minutes, and the fIXa-DEGR binding progressively decreased upon saturation of light chain with RI IgG (Fig 4A). Similar results were obtained for IgG from MU, MS, and MoAb CLB-CAg A (not shown). Because of the multiple binding components, these experiments more closely represent a stoichiometric titration rather than equilibrium binding. In a control experiment, dissociation of bound IgG was less than 3% during 50 minutes, which showed that the IgG/LCh ratio did not change significantly during fIXa-DEGR binding to LCh.

FIXa-DEGR binding to light chain was inhibited more than 95% by MoAb CLB-CAg A at a stoichiometric ratio (Fig 4B). As the MoAb binds to A3 residues 1804-1820, an fIX binding site, the binding of MoAb and fIXa to light chain is mutually exclusive. MU, RI, and MS IgGs maximally inhibited fIXa-DEGR binding by 95%, 91%, and 88%, respectively, at IgG/light chain ratios of 1.5 (Fig 4B). The nonstoichiometric ratios for the human inhibitors may be due to the additional presence of lower concentrations of antibodies directed against regions of A3-C1 other than the fIXa binding site.

As the inhibitory effect of MoAb CLB-CAg A is caused by direct competition with fIXa for fVIII binding,10 we tested the hypothesis that the human anti–A3-C1 antibodies have the same mechanism of inhibition. We used synthetic A3 peptide 1804-1819 to compete for human IgG binding to light chain. MoAb CLB-CAg A (10 nmol/L) or inhibitor IgGs MU (50 nmol/L), RI (1,000 nmol/L), and MS (1,000 nmol/L) were incubated with increasing concentrations of peptide for 1 hour at 37°C before addition to the biosensor cuvette containing light chain immobilized on MoAb NMC-VIII/5. The above antibody concentrations were chosen as each inhibited the factor Xase assay (see below) by approximately 80%. The concentration of anti-fVIII antibodies was different in each IgG; therefore, the total IgG required in our experiments also varied widely. Stoichiometric titration, measured as above, was achieved within 30 minutes at all peptide concentrations. The binding of human inhibitor MU, RI, and MS, or MoAb CLB-CAg A IgG to light chain was inhibited ≥73% in a dose-dependent mannner by the peptide (Fig5). The maximal inhibition of binding was 94%, 92%, 76%, and 73% for CLB-CAg A, MU, RI, and MS IgG, respectively. The less complete inhibition of MS and RI binding by the peptide may be due to insufficient peptide concentrations or to the presence of a minor population of antibodies with an epitope(s) outside the 1804-1819 sequence. The millimolar concentrations of peptide required for inhibition in each case suggests that the peptide-antibody affinities are low.

A randomized, soluble peptide (PVKETYKFNKKTVFNV) of the fVIII sequence 1804-1819 was used as a control for MU and CLB-CAgA (not shown) binding to LCh. There was no inhibition up to 4 nmol/L peptide. Because peptide 1804-1819 contains four positively charged Lys residues and one negatively charged Glu residue, its effect on antibody binding may be caused by the overall positive charge. This possibility was excluded by lack of inhibition of MU (Fig 5; RI and MS, negative but not shown) in the above assay by an unrelated control peptide of A2 domain residues 417-428, containing six positively charged and two negatively charged amino acids.

To determine if anti–A3-C1 antibodies are able to inhibit fVIII function in a purified system, we tested their effect in a chromogenic factor Xase assay using purified components. This assay (Materials and Methods) measures the ability of activated fVIII (fVIIIa) in the presence of phosphatidylserine/phosphatidylcholine vesicles (PSPC) to act as a cofactor for fIXa in the activation of factor X.16Because fVIIIa activity cannot be directly measured in this assay, we determined the initial rate of factor X activation using a chromogenic peptide substrate, S-2765. Under our conditions, this rate was linearly proportional to the fVIIIa activity.

Only inhibitor MU IgG was tested because it had the lowest ratio of anti-A2 to anti–light-chain antibodies (2.5 × 10⁻⁵) and no other anti–heavy-chain antibodies by IP assays. Due to further depletion of anti-C2 antibodies (Materials and Methods), the ratio of anti-C2 to anti–light chain antibodies was decreased to 6.7 × 10⁻⁵ (Table 1). The MU IgG used in this experiment therefore contained more than 99.9% anti–A3-C1 antibodies. MU and CLB-CAg A IgGs both inhibited the factor Xase assay by more than 90% (Fig 6).

The greater neutralization of many inhibitor plasmas by the fVIII light chain than by C2⁶ suggested that these plasmas may contain inhibitor antibodies with epitopes in A3-C1. To examine this possibility, we first demonstrated by an IP assay that 17 of 18 human inhibitor IgGs bound to a recombinant ³⁵S-A3-C1 polypeptide. The IgG from three such inhibitors with >10-fold higher anti–light chain than anti–heavy chain antibody titers was further depleted of anti-C2 domain antibodies to prevent their interference in subsequent assays. One IgG (MU) with barely detectable anti–heavy chain antibodies completely inhibited the intrinsic factor Xase assay; therefore, some anti–A3-C1 antibodies are likely to have inhibitor activity.

MoAb CLB-CAg, A which binds to the A3 domain peptide 1804-1818, a fIX binding site, also inhibits fVIII activity.10 No other binding site for a physiological ligand of fVIII has been localized to A3-C1. Therefore, we tested the possibility that some human anti–A3-C1 antibodies inhibit fVIII-fIX binding, which is required for assembly of the factor Xase complex. In an assay of fIXa binding to light chain, the three inhibitor IgGs bound to light chain inhibited subsequent fIXa binding by ≥88%, compared with 97% by CLB-CAg A. The inhibition of the human and MoAb CLB-CAg A IgG binding to light chain by synthetic peptide 1804-1819 suggested that the epitopes of these antibodies overlap the fIX binding site. These data are consistent with a model of direct inhibition of fIX binding to fVIII by these antibodies.

We found that peptide 1804-1819 inhibited binding of inhibitor MU IgG and MoAb CLB-CAg A to light chain by 88% to 90%. This result implies that MU anti–A3-C1 IgG is mainly directed against the fIX binding site, which is consistent with the observed 1.1:1.0 molar ratio of bound MU IgG to light chain. In contrast, binding of the inhibitor IgGs RI and MS to light chain was inhibited only 76% and 73%, respectively, by the peptide. This may be due to insufficient concentrations of peptide for complete inhibition of binding or to the presence of a minor population of anti–A3-C1 antibodies that do not bind to the fIXa site.

Although we tested only three anti–A3-C1 inhibitors, all of which competed for fIX binding to fVIII, this may be a common inhibitory mechanism as suggested by our earlier finding that 20 of 34 inhibitor plasmas were more completely neutralized by light chain than by its C2 domain.6 This remains to be confirmed by characterization of additional inhibitors.

We thank Dr Jan van Mourik (Netherlands Red Cross, Amsterdam) for providing the MoAb CLB-CAg A, which was crucial for our work.