Rifampicin-dependent antibodies bind a similar or identical epitope to glycoprotein IX–specific quinine-dependent antibodies

Rifampicin is a widely used agent for the treatment of tuberculosis. Adverse effects from the administration of rifampicin can include interstitial nephritis, thrombocytopenia, and hemolytic anaemia. The first incidence of rifampicin-induced thrombocytopenia was reported in 1970 by Blajchman et al.1 Since then, many other cases have been reported.2,3 Several investigators have demonstrated rifampicin-dependent antiplatelet antibodies2,3but have not characterized the platelet antigen of these antibodies. Kakaiya et al2 demonstrated that rifampicin did not bind platelets directly. They suggested that the antibody first reacts with the drug to form a drug-antibody complex that then binds to platelets, resulting in their clearance through an “innocent bystander” mechanism. Martinez et al4 suggested that the interaction of the offending drug with blood cell membrane proteins leads to the formation of an antigenic complex to which the drug-dependent antibodies bind. Our observations favor the second mechanism.

In this study, we have characterized the antibody from a patient who developed rifampicin-induced thrombocytopenia. Using flow cytometry, monoclonal antibody–specific immobilization of platelet antigens (MAIPA) assay, and immunoprecipitation, we demonstrated that the antibody reacted with the glycoprotein (GP) Ib-IX complex on the platelet surface. The antibody epitope was further mapped to the GPIX subunit of the complex. More specifically, the antibody binding site was located at a site where the epitopes of 3 other drug-induced antibodies (quinine-, quinidine-, and ranitidine-induced) have been previously located. These data imply that this site on GPIX is strongly immunogenic when it combines with drugs. A careful analysis of this GPIX region may provide useful information about the pathophysiology of drug-induced thrombocytopenia in general.

Bovine serum albumin (BSA), phenylmethylsulfonyl fluoride (PMSF), ethylenediaminetetraacetic acid, disodium salt (EDTA), bacitracin, benzamidine, dithiothreitol (DTT), dimethylsulfoxide, iodoacetamide, propidium iodide, and 2,2′-azinobis 3-ethylbenzthiazolinesulfonic acid (ABTS) were purchased from Sigma (St Louis, MO); 3,3′,5,5′-tetramethylbenzidine dihydrochloride from Kirkegaard & Perry (Gaithersburg, MD); sulfosuccinimidobiotin from Pierce (Rockford, IL); Western blot chemiluminescence reagents from DuPont (Boston, MA); and the sheep anti-mouse–coated Dynabeads and the sheep anti-rabbit immunoglobulin G (IgG)-coated Dynabeads from Dynal (Oslo, Norway). Rifampicin was purchased from Marion Merrell Dow (Sydney, Australia). All chemicals were of analytical reagent grade.

The goat anti-mouse immunoglobulin and the horseradish peroxidase (HRP)-conjugated goat anti-human immunoglobulin (Jackson, West Grove, PA), rabbit anti-mouse HRP-conjugated antibody (Dako, Carpentaria, CA), and the streptavidin-HRP conjugate (Amersham, Bucks, United Kingdom) were purchased as indicated.

All monoclonal antibodies (mAb) were of the IgG class. MOPC21 (Becton & Dickinson, San Jose, CA), a murine IgG₁ myeloma protein, was used as a control immunoglobulin, and SZ1, SZ21, and SZ22 were purchased from Immunotech (Marseille, France). The sheep anti-mouse fluorescein isothiocyanate (FITC)-conjugated secondary antibody was purchased from Silenius (Hawthorn, Australia), and the rabbit anti-human FITC-conjugated secondary antibody from Dako.

AK2 and FMC25 mAbs, which are directed against epitopes on various parts of the human GPIb-IX complex, were obtained from Dr M. Berndt (Melbourne, Australia).5 AP2, a mAb that is directed against an epitope on GPIIb-IIIa, was a kind gift from Dr T. J. Kunicki (La Jolla, CA).

Mrs N.P., a 66-year-old woman, was diagnosed with pulmonary tuberculosis and was begun on rifampicin and isoniazid. Four months later, she noted spontaneous bruising. Physical examination revealed petechiae on her legs but no other bleeding. Blood counts showed a severe thrombocytopenia: platelets 9 × 10⁹/L, hemoglobin 165 g/L, and white blood cells 5.0 × 10⁹/L. A diagnosis of drug-induced thrombocytopenia was made, and all drugs she was taking were stopped. She was begun on 50 mg of prednisone daily. A test for drug-dependent antibodies using an antigen-capture assay—MAIPA—demonstrated a rifampicin-dependent antiplatelet antibody with specificity against GPIb-IX complex, but no isoniazid-dependent antibody was detected. A bone marrow aspirate revealed normal numbers of megakaryocytes and erythroid and myeloid precursors, consistent with a thrombocytopenia due to increased peripheral platelet destruction. Her platelet count rose to 124 × 10⁹/L, 189 × 10⁹/L, and 214 × 10⁹/L on days 4, 6, and 7, respectively, after withdrawal of antituberculosis drugs. Prednisone was then stopped.

Chinese hamster ovary (CHO) DUK^- (dihydrofolate reductase–negative [DHFR^-]) cells and mouse L (tk^-) cells stably transfected with complementary DNA (cDNA) encoding the GPIb-IX subunits in various combinations were produced in one of our laboratories (J.A.L.) as previously described.6 The cloning of the cDNAs for GPIbα, GPIbβ, and GPIX has been reported previously.7-9 The 3 cDNAs (each containing the entire coding sequence and the 3′-untranslated region) were cloned separately into the eukaryotic expression vector pDX (a kind gift from Dr K. Berkner, Seattle, WA), in which transcription is driven by the adenovirus major late promoter and the SV40 enhancer.

The CHO cells were transfected with the following combinations of GPIb-IX subunit cDNAs: (a) GPIbα, GPIbβ, and GPIX, (b) GPIbα and GPIbβ, (c) GPIbα and GPIX, and (d) GPIbβ and GPIX. The L cells were transfected with (a) GPIbα and GPIbβ and (b) GPIbβ and GPIX. Expression of the GPIb-IX subunits in the cell lines was substantiated by Northern blot analysis to detect messenger RNA (mRNA) and by flow cytometry and enzyme-linked immunosorbent assay (ELISA) to ensure protein expression of the subunits on the cell surface. In all cell lines, the appropriate GPIb-IX subunits were expressed on the cell surface except in the CHO cells transfected with GPIbα and GPIX cDNA. In this cell line, GPIbα was expressed on the cell surface, but GPIX was located in the cytoplasm when detected by flow cytometry.

The ELISA was performed as previously described.10 

The MAIPA assay was performed as previously described with minor modifications.11 Group O platelets (2 × 10⁷ per tube) were washed once in phosphate-buffered saline (PBS)/1% EDTA buffer, resuspended in 100 μL PBS/2% (w/v) BSA, and added to 50 μL of patient serum with or without 5 μL of rifampicin (final concentration, 70 μg/mL). The positive control for the assay was normal pooled platelets at a concentration of 2 × 10⁷/100 μL plus 50 μL of patient serum known to contain an anti-PL^A1 antibody.

Flow cytometry was performed on a FACStar Plus flow cytometer (Becton & Dickinson) fitted with a 100-mW air-cooled argon ion laser using the 488-nm green line for fluorescence excitation. The cell emission spectra were collected on FL1 (green) using a band pass filter 530 DF 30. Dead cells were excluded using propidium iodide on FL3 using a 700 LP filter. Cells or group O platelets for flow cytometry were prepared as described previously.10 Primary antibody incubation, mAb (10 μg/mL) or patient serum (1:20 dilution), was performed for 10 minutes at room temperature in the presence or absence of rifampicin (700 μg/mL). In experiments carried out in the presence of rifampicin, the working or wash buffer contained the drug beyond this stage at a concentration of 700 or 600 μg/mL, respectively.

The platelets (2 × 1010) were incubated with 168 μmol of sulfosuccinimidobiotin, and the free biotin was quenched with excess glycine. The platelets were then incubated with one of the following: a mouse IgG₁ control (MOPC21; 10 μg/mL), the GPIX-specific mAb (GRP; 10 μg/mL), or AB or patient serum (40 μL of each with or without the drug at 700 μg/mL) for 30 minutes. In experiments performed in the presence of rifampicin, the working or wash buffer contained the drug beyond this stage at a concentration of 700 or 600 μg/mL, respectively. After 3 washes, the platelets were lysed by resuspension in 500 μL of 0.01 mol/L triethanolamine-buffered saline containing 0.5% Triton X-100, 0.05% Tween-20, and protease inhibitors (PMSF 2 mmol/L, leupeptin 10 μM/L, aprotinin 10 μM/L, EDTA 5 mM/L) and incubated for 1 hour. Following centrifugation at 12 000g for 30 minutes, the supernatant (450 μL) was collected; 1 × 10⁷ sheep anti-mouse–coated Dynabeads, or 2.25 × 107 Dynabeads coated with goat anti-rabbit IgG linked to rabbit anti-human IgG, were added to the mAb or the human serum immunoprecipitates, respectively. After incubation for 2 hours, the Dynabeads were washed 5 times using the MPC-E-1 magnet before resuspension in 10 μL 2 × Laemmli buffer12 and storage at −80°C until analysis by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE).

The 20-μL samples for SDS-PAGE analysis were treated with 2 μL of 0.5-mol/L DTT and boiled for 5 minutes. Free DTT was quenched by 4 μL of 0.5-mol/L iodoacetamide. SDS-PAGE was performed according to the method of Laemmli12 using 12% Tris-glycine gels (Bio-Rad, Hercules, CA). After electrophoresis, the proteins were transferred to polyvinylidine difluoride (PVDF) membrane. Membranes were blocked overnight in 5% (w/v) skim milk (Diploma) and the following morning washed 5 times in PBS/0.05% Tween-20. Streptavidin HRP diluted 1:2000 in 2% BSA/PBS/0.05% Tween-20 was incubated with the membrane for 60 minutes. After 5 washes, the presence of a signal was detected using the Western blot chemiluminescence reagents according to the manufacturer's instructions.

The patient serum was analyzed using the MAIPA assay. Antibody binding was only seen in the presence of rifampicin. The serum showed rifampicin-dependent antibody binding to platelets when the GPIbα-specific mAb, AK2, was used to capture the antigen (GPIb-IX complex). Similarly, strong rifampicin-dependent antibody binding occurred when the non cross-blocking GPIX-specific mAb, FMC25, captured the antigen (Figure 1). When a GPIIb-IIIa–specific mAb, AP2, SZ21, or SZ22, was used as the capture antibody, a negative result was obtained. These data indicate that the rifampicin-dependent serum contains an antibody that reacts with the GPIb-IX but not GPIIb-IIIa complex.

When immunoprecipitation experiments were performed using human AB or patient serum with and without rifampicin, the GPIb-IX complex components were isolated only in the presence of the patient serum and rifampicin (Figure 2). When the experiment was repeated in the presence of a nonimmune mouse IgG₁ (MOPC21) and the GPIX-specific mAb, GRP, immunoprecipitation of the GPIb-IX complex components was observed with GRP in both the presence and absence of the drug, but no specific band was observed with MOPC21. The immunoprecipitation of the GPIb-IX subunits by the patient serum in the presence of rifampicin indicates that the drug-dependent antibody has specificity for the GPIb-IX complex.

Flow cytometry was used to test the binding of the patient antibody to platelets and CHO cells stably transfected with GPIbα, GPIbβ, and the GPIX cDNAs. A negative control of AB serum detected no specific antibody binding in the presence of the drug on platelets (Figure 3) or CHO cells (not shown). Rifampicin-dependent antibody binding was observed on both the platelets and the CHO αβIX cells. These results confirm that the rifampicin-dependent antibody binds to the GPIb-IX complex.

Flow cytometry was also used to test the binding of the patient antibody to mouse L cells that had been stably transfected with either the human GPIbβ and GPIX cDNA (L βIX cells) or the human GPIbα and GPIbβ cDNA (L αβ cells). Binding was not observed in the absence of rifampicin. In the presence of rifampicin, the drug-dependent antibody bound only to L βIX cells and not to the L αβ cells (Figure 4). These data indicate that the rifampicin-dependent antibody binds specifically to the GPIX subunit of the GPIb-IX complex—GPIX is the only component not present on the L αβ cells.

The MAIPA assay is an antigen-capture ELISA that uses an mAb to capture GPIb-IX after the human drug-dependent antibody has attached to the GP complex at a site distant from the mAb-binding site (Figure 5A). If the human antibody and the murine mAb-binding sites coincide or are in close proximity, the human antibody will cross-block the binding of the mAb, the antigen is not captured, and a negative result is obtained (Figure 5B).

When the experiment was repeated using the anti-GPIX mAb, SZ1, a negative result was obtained because the rifampicin-dependent antibody bound to platelet GPIb-IX blocked the antigen capture by SZ21 (Figure 6A). As we have shown previously (Figure 1), antigen capture was not inhibited when other non–cross-blocking mAbs (GPIbα-specific, AK2, and GPIX-specific, FMC25) were used, and a positive result was obtained with each of these mAbs. These data confirm that the specificity of the antibody in the patient's serum is GPIX rather than GPIbα or GPIbβ. The same antibody panel was used to assay the binding to platelets of the GPIX-specific quinine-dependent antibody, and a comparable result was obtained (Figure 6B). These results indicated that the epitopes of the rifampicin- and quinine-dependent antibodies were identical or located in close proximity to each other (and also to that of mAb SZ1).

In this study, we have shown that the antibody of a patient with rifampicin-induced thrombocytopenia interacts drug-dependently with the platelet GPIb-IX complex, and not the GPIIb-IIIa complex, using flow cytometry, the MAIPA assay, and immunoprecipitation.

GPIb is a major sialoglycoprotein on the platelet cell surface, with approximately 25 000 copies per platelet.13 GPIb is composed of 2 subunits, the α subunit of approximately 143 kd that is disulphide-bonded to a smaller β subunit of about 24 kd. GPIb is noncovalently linked to GPIX that has a molecular mass of 20 kd. The 3 glycoproteins have leucine-rich motifs, and they exist as a heterodimeric complex in the platelet membrane.5,6,14 The subunits of GPIb-IX are encoded by separate genes.7-9 15 

To determine which component of the GPIb-IX complex, the rifampicin-dependent antibody, was binding to, we used CHO and mouse L cells stably transfected with various components of this complex as described in “Methods.” Drug-dependent binding of the patient antibody was observed on the CHO αβIX cells that contained all 3 components of the complex. When the mouse L cells were transfected with 2 of the 3 components of the complex, binding was only observed on the L βIX cells in the presence of the drug. Binding was not observed on the L αβ cells, indicating that neither of these components contained the epitope recognized by the drug-induced antibodies. These results indicate that the rifampicin-dependent antibodies bind an epitope on GPIX on the platelet surface. The CHO cell line transfected with GP Ibα and GPIX cDNA was not tested because it expressed only GPIbα on the cell surface, but GPIX was present in the cytoplasm (see “Methods”).

The rifampicin-dependent antibody was able to inhibit the binding of the mAb SZ1, specific for GPIX,16 in the competitive MAIPA assay. The binding of this mAb was also inhibited by binding of the quinine-dependent antibody in the same assay, shown in this study and previously reported by our group.17 It is interesting that 2 other drug-induced antibodies could also block the binding of the mAb SZ1 to platelets. Gentilini et al18 recently reported that the ranitidine-induced antibody blocked the binding of SZ1 to platelets, and we have observed the same inhibitory effect previously on the binding by quinidine-induced antibodies.17 These observations suggest that these 4 drug-induced antibodies and the mAb SZ1 bind to either the same site or sites very close to each other. Because the binding sites of 4 drug-induced antibodies have been mapped to this region of GPIX, it is possible that the epitopes of many other drug-induced antibodies may also be localized to this site. The reason for the colocalization of the binding sites of these 4 antibodies is at present unclear because the drugs involved are chemically or structurally dissimilar, except for quinine and its optic isomer, quinidine.

In summary, this is the first study to map the epitope of the rifampicin-dependent antibody to the GPIX subunit of the GPIb-IX complex—more specifically, to a domain that is also the common binding site of 3 other drug-induced antibodies. This region of GPIX obviously plays an important role in epitope formation for drug-induced antibodies. Further elucidation of the characteristics of this region on GPIX is clearly required, and it will provide useful insights into the mechanisms of drug-induced thrombocytopenias.

We wish to thank Ms Sue Evans for assistance with the MAIPA assay.