Anti-CD36 autoantibodies in thrombotic thrombocytopenic purpura and other thrombotic disorders: identification of an 85 kD form of CD36 as a target antigen

Authors


Dr Duane R. Schultz University of Miami School of Medicine, Department of Medicine (R-102), Division of Immunology, P.O. Box 016960, Miami, FL 33101, U.S.A.

Abstract

The presence of anti-CD36 antibodies in plasma of patients with thrombotic thrombocytopenic purpura (TTP), idiopathic thrombocytopenic purpura (ITP), and heparin-induced thrombocytopenia without/with thrombosis (HIT/HITT) has been examined by immunoblots, and a monoclonal antibody capture assay, the platelet-associated IgG characterization assay (PAICA). Results with PAICA showed that 73% (8/11) of patients with TTP were positive, and 71% (10/14) by immunoblots. With ITP, 20% (6/30) were positive by PAICA and 19% (3/16) by immunoblots; HIT, 30% (3/10) were positive by PAICA and 60% (6/10) by immunoblot; HITT, 50% (2/4) by PAICA and 100% (4/4) by immunoblot. Purification of CD36 by fast protein liquid chromatography (FPLC) from Triton X-100 extracts of normal platelet membranes resulted in the isolation of two different forms: the classic 88 kD form, and a second, lighter 85 kD form. Our data indicated that the patients' plasma autoantibodies reacted strongly with the 85 kD form. Conventional monoclonal and polyclonal antisera produced to the 88 kD form reacted strongly with the 88 kD form but weakly with the 85 kD form. These results confirm the possible importance of anti-CD36 antibodies in the pathophysiology of TTP and other thrombocytopenias and demonstrate the presence of a previously unrecognized target antigen for these antibodies.

The human CD36 is a single-chain integral membrane polypeptide (also known as GPIV; IIIb) expressed by platelets as well as many other cells, cell lines and tissues ( Okumura & Jamieson, 1976; Asch et al, 1987 ; Tandon et al, 1989b ; Greenwalt et al, 1990 , 1992; Daviet & McGregor, 1997). Numerous functions have been described for CD36, including a role as a cell surface receptor interactive with a large number of ligands, in cytoadhesion reactions, and implication in intracellular signal transduction ( Jaffe et al, 1982 ; Leung, 1984; Silverstein et al, 1984 , 1989; Majack et al, 1988 ; Kieffer et al, 1989 ; Tandon et al, 1989a , b; Goodet al, 1990 ; Murphy-Ullrich et al, 1992 ; Savill et al, 1992 ; Greenwalt et al, 1992 ; Endemann et al, 1993 ; Abumrad et al, 1993 ; Daviet et al, 1997 ; Kehrel et al, 1998 ). The molecular mass of CD36 on different cells is variable; for example, Mr values of 88 000, 85 000 and 78 000 have been reported for human platelets, human mammary epithelial cells, and human erythroblasts, respectively ( Tandon et al, 1989b ; Greenwalt et al, 1990 ; Kieffer et al, 1989 ).

A cDNA clone encoding CD36 was isolated from a human placenta cDNA library and expressed in COS cells by Oquendo et al (1989 ), resulting in a polypeptide with 471 residues and a predicted Mr of approximately 53 kD. The identification of 10 potential N-linked glycosylation sites accounted for the difference in Mr between the polypeptide and the 83 kD species found by immunoprecipitation with a pool of mouse monoclonal anti-CD36 antibodies. Tandon et al (1989b ) purified CD36 from solubilized Triton X-114 extracts of human platelet membranes, and reported the 88 kD protein as having 26% carbohydrate, of which approximately two-thirds were in alkali-labile O-glycosidic linkages. Treatment of human platelet CD36 with endoglycosidase F resulted in a protein with an apparent Mr of 57 000 after removal of N-linked oligosaccharides ( Greenwalt et al, 1990 ).

Platelet membrane CD36 may be important immunologically in the aetiology of thrombosis ( Vermylen et al, 1997 ). Burns & Zucker-Franklin (1982) showed that an IgG in TTP plasma mediated time-dependent immune destruction of human cultured endothelial cells. Pfueller et al (1990 ) described a patient with thrombosis and thrombocytopenia whose plasma had potent platelet aggregating activity caused by an IgG directed against platelet CD36. Further support for these observations came from a report by Tandon et al (1994 ) which showed a high frequency of anti-CD36 antibodies in patients with TTP. Specifically, they showed, using plasma from TTP patients as the source of autoantibodies and radio- or enzyme-labelled protein A as the reporter molecule for the patient IgG, that 85% were CD36 positive by immunoprecipitation, 75% were positive by protein blotting, and 60% were positive by dot blots, using purified CD36 as the antigen. Furthermore, platelets from a Naka-negative donor (constitutively lacking CD36) ( Yamamoto et al, 1990 ) were tested by Tandon et al (1994 ). Approximately 50% of their TTP plasma samples that caused 70% release of 14C-serotonin from control platelets failed to induce release from Naka-negative platelets, indicating that CD36 was the major target for antibodies in those TTP plasmas. In direct contrast to the data of Tandon et al (1994 ), in a recent report by Raife et al (1997 ) antibodies to CD36 were not found in 49 acute-phase plasma samples from TTP patients by an ELISA and flow cytometric assays.

In order to clarify this important aspect of the possible aetiology of TTP we have examined plasmas from TTP patients for antibodies reactive with platelet membrane CD36. Furthermore, these studies were expanded to test for anti-CD36 antibodies in patients with idiopathic thrombocytopenic purpura (ITP) and heparin-induced thrombocytopenia without and with thrombosis (HIT, HITT). Additionally, a lighter glycoform of CD36 (85 kD) was separated from the classic 88 kD form chromatographically and shown to react predominantly with autoantibodies from most TTP patients as well as patients with ITP, HIT and HITT.

MATERIALS AND METHODS

Reagents

Chemicals, molecular weight standards, and horseradish peroxidase (HRP) labelled goat, mouse and rabbit secondary antisera were purchased from Sigma Chemicals (St Louis, Mo.) unless otherwise specified. Mouse mAb OKM5 (IgG1, 10 μg/ml), reactive with both an 88 kD monocyte membrane protein and platelet CD36, was purchased from Ortho Diagnostics Systems Inc. (Raritan, N.J.). Fluorescein isothiocyanate (FITC)-labelled mouse mAb reactive with platelet marker CD41 (clone SZ22, IgG1, 50 μg/ml), which reacts with activated/non-activated platelet membrane GP IIb–IIIa, and FA6-152 (IgG1, 200 μg/ml) reactive with platelet CD36, were from Immunotech (Luminy, France). Rabbit polyclonal antiserum (IgG, 1 mg/ml) produced to SDS-denatured human platelet integral membrane glycoprotein IV (CD36) and mouse mAbs Mo91 (IgG1, 550 μg/ml) and 131.5 (IgG1, 2.9 mg/ml) reactive with human platelet CD36 were generous gifts from Dr G. A. Jamieson (American Red Cross, Rockville, Md.). A biotinylated goat anti-rabbit IgG (1 mg/ml) was purchased from Chemicon International Inc. (Temecula, Calif.). Horseradish peroxidase-labelled ExtrAvidinTM (2 mg/ml) was purchased from Sigma Chemical.

All antisera were used at concentrations that were determined to be optimum for the immunoassays to which they were applied.

Patient selection

Four groups of patients were investigated. All were volunteers, and all gave informed consent according to institutional guidelines.

(i) Thrombotic thrombocytopenic purpura. This group consisted of 14 patients. All had the classic triad of TTP: fluctuating neurologic dysfunction, severe thrombocytopenia, and microangiopathic haemolytic anaemia. The TTP was in the active state when blood was drawn for the study.

The plasma from one of the TTP patients (patient 1) was a potent source of autoantibodies reactive with CD36, and was used as a reagent to test for CD36. These autoantibodies were reactive mainly with the 85 kD form of CD36, and served as an indicator to distinguish the 85 kD form from the classic 88 kD form of the molecule. A brief case history of patient 1 follows. He was a white man, aged 25 years, who presented with headache, nausea and altered mental status. Blood tests showed severe thrombocytopenia (platelet count 13 × 109/l; normal range 200–350 × 109/l); anaemia (haemoglobin 6.9 g/dl; normal range 12–16 g/dl); haematocrit 20% (normal range 40–50%), white blood cells 22.5 × 109/l (normal range 3.6–9.6 × 109/l); reticulocyte count 9.5% (normal range 0.5–1.5%); total bilirubin 83.8 μmol/l (normal range 3.4–22.2 μmol/l); LDH, 2.8 u/l (normal range 0.3–0.6 u/l). His blood smear showed features of microangiopathic haemolytic anaemia. This patient had a history of recurrent TTP for 1 year prior to transfer to Jackson Memorial Hospital, Miami, Fla., and a diagnosis of recurrent TTP was made. He responded well to exchange plasmapheresis with fresh frozen plasma infusion, and in 10 d the platelet count had increased to 107 × 109/l, the LDH fell to 0.9 u/l, and his altered mental status cleared. He was discharged with a normal platelet count. He relapsed after 14 months, with a similar initial presentation, and was treated again with exchange plasmapheresis and plasma infusion, whereby he fully recovered and was discharged. His plasma was collected during the exacerbations of TTP, and stored frozen at −70°C for these studies.

(ii) Chronic idiopathic thrombocytopenic purpura. This group consisted of 30 patients. All patients had active ITP with thrombocytopenia at the time of the study. All met the diagnostic criteria of chronic ITP ( Ahn et al, 1974 ): increased megakaryocytes in the bone marrow and the absence of splenomegaly.

(iii) Heparin-induced thrombocytopenia without thrombosis ( Kelton et al, 1988 ; Chong et al, 1989 ; Amiral et al, 1992 ). 10 patients were investigated. All developed thrombocytopenia (i.e. >50% reduction of platelet count following heparin therapy) without clinical or laboratory evidence of thrombosis. All received unfractionated heparin intravenously except one individual who received low molecular weight heparin subcutaneously. All blood samples were drawn when the patients were thrombocytopenic.

(iv) Heparin-induced thrombocytopenia with thrombosis. Four patients were investigated. All were thrombocytopenic following heparin therapy and suffered from arterial thrombosis with gangrene in the extremities. One developed strokes, deep vein thrombosis, and pulmonary emboli.

Because of the paucity of some plasma samples, it was not always possible to test each sample by both the platelet-associated IgG characterization assay (PAICA) (see below) and immunoblots ( 1 Table I).

Table 1. Table I. Autoantibodies reactive with platelet membrane CD36 by immunoblots (I.B.) and PAICA in plasma samples from patients with TTP, ITP, HIT and HITT.Thumbnail image of

Purification of CD36

Originally it was decided to purify human platelet CD36 in order to produce an inhouse polyclonal antiserum in goats for use in the PAICA. The procedure used here is similar to that described elsewhere ( Tandon et al, 1989b ; McGregor et al, 1989 ). The Triton X-100 platelet extract was applied to a 4.5 × 9 cm column of Q-Sepharose FF equilibrated with 50 m M Tris/5 m M EGTA, pH 7.4 at 4°C. When the column effluent OD280 was <0.2, column buffer containing 0.5 M NaCl was applied and a 200 ml eluate fraction (0.15 M NaCl) was collected and concentrated 10-fold by ultrafiltration using a 10 000 MW cut-off membrane (Amicon Inc., Beverly, Mass.). Following dialysis against column buffer, the concentrated eluate was centrifuged. The supernatant fluid was passed through a 0.2 μm filter and applied to a Mono Q 10/10 anion exchange column equilibrated with the same column buffer. A computer-controlled linear 0–0.5 M NaCl gradient was applied, protein was monitored at OD280, and fractions were sampled for reactivity v TTP patient 1. plasma (autoantibody source) in immunoblots. Reactive fractions were pooled, concentrated, dialysed against 50 m M Na acetate, pH 5.0, and applied to Mono S 10/10 cation exchanger equilibrated in the same buffer. Following a 0–1 M NaCl linear gradient, reactive fractions were pooled, dialysed against 20 m M ethanolamine, pH 9.1, and applied to Mono Q in this alkaline buffer. Proteins were eluted with a 0–0.5 M NaCl step-gradient (0–35%, then 5% increments to 55%). For the final step, reactive fractions were pooled, concentrated to 1.5 ml, and applied to Sephacryl S-200 Superfine (1.5 × 89 cm) equilibrated in 50 m M Tris/0.3 M NaCl/5 m M EGTA, pH 7.4.

Preparation of plasmas

Group O blood from normal volunteers free of medications such as aspirin for at least 10 d, was drawn into citrate (VacutainersTM) using a 21-gauge butterfly needle. The first tube drawn was discarded or used for other purposes. Platelet-rich plasma (PRP) was obtained by centrifugation at 160 g for 10 min at 25°C. Further centrifugation at 1500 g for 6 min in a microfuge yielded platelet-poor plasma (PPP). Flow cytometry showed that this removed >99% of platelets. Plasma was frozen in 1.4 ml aliquots and stored at −70°C until further use.

Platelet-associated IgG characterization assay (PAICA)

Plasma autoantibodies produced to CD36 were detected using a modification of the PAICA described by Macchi et al (1996 ). For this procedure a lysate of normal platelets preincubated with patient PPP was prepared as follows: citrated normal human PRP (see above) containing 5 m M EDTA was centrifuged at 2000 g for 3 min at room temperature (RT). The pelleted platelets were washed twice with PBS/1% EDTA and suspended to a concentration of 1 × 109/ml. 100 μl were incubated with 300 μl of patient PPP and 50 μl PBS/2% BSA for 30 min at RT. The platelets were diluted, pelleted, and washed three times with 1 ml PBS/1% EDTA. Finally, they were resuspended, solubilized in 1% Triton X-100 in PBS at 4°C for 30 min and the lysate was obtained by centrifugation (12 000 g, 30 min, 4°C).

The PAICA was carried out in microtitre plates (Costar Corp., Cambridge, Mass.) coated with 100 μl/well of goat anti-mouse IgG (3 μg/ml) in 15 m M Na bicarbonate buffer, pH 9.5. After overnight incubation at 4°C, the plates were washed three times with PBS/0.05% Tween-20, pH 7.2 (PBS-T), and blocked with 250 μl/well PBS-T/BSA for 1 h at RT. The plates were washed three times with PBS-T and 100 μl/well of a 10 μg/ml solution of mouse mAb FA6-152 (anti-CD36) in PBS-T/BSA was added. After 1 h at RT, the plates were washed three times as above. Then, 100 μl of platelet lysate/well were added and incubated for 1 h at RT. After three washes with PBS-T, 100 μl of HRP-labelled F(ab′) fragments of Fc-specific goat anti-human IgG (1 μg/ml) were added to each well. Following a 1 h incubation at RT, the plates were washed three times in PBS-T and 100 μl of chromogenic 3,3′,5,5′-tetramethylbenzidine (TMB) substrate (Sigma Chemical, St Louis, Mo.) was added and incubated for 30 min. The reaction was stopped by the addition of 100 μl of 1 N H2SO4. The OD at 450 nm was read using the SLT Spectra Automatic Plate Reader (SLT Lab. Instruments, Grödig/Salzburg, Austria) and the ratio of patient plasma OD to normal plasma pool was calculated. A ratio >2.5 was defined as a positive test for patient IgG autoantibodies reactive with CD36.

Enzyme-linked immunosorbent assay (ELISA)

Microtitre plates were coated with 100 μl of the highly purified 85 kD CD36 protein (5 μg/ml) and incubated overnight at 4°C. The supernatant fluid was removed and the plates were washed three times with PBS/0.05% Tween-20, pH 7.2 (PBS-T) and blocked with 250 μl/well PBS-T/BSA for 1 h at RT. The plates were washed three times with PBS-T, and 100 μl/well of a battery of 4 μg/ml of anti-CD36 solutions were added, including mAbs FA6-152, OKM5, 131.5, Mo91, and control SZ22 in PBS-T/BSA. After 1 h at RT the plates were washed three times as above. Next, 100 μl (2 mg/ml) of HRP-labelled goat anti-mouse IgG diluted 1:2000 in PBS was added and the plates were incubated at 25°C for 1 h. The second incubation was followed by three washes with PBS-T after which 100 μl of TMB substrate was added to the wells. The chromophore development was terminated after 30 min at 25°C with 2 N H2SO4. The plates were read at a wavelength of 450 nm as described above.

Platelet isolation and extraction

Human platelet concentrates outdated for clinical use were obtained from the American Red Cross, South Florida Region. The pooled and washed platelets were suspended in 100 ml of 50 m M Tris/5 m M EGTA/1 m M EDTA/2 m M PMSF/1 m M leupeptin/1% Triton X-100, pH 7.4, and incubated overnight at 4°C with gentle agitation. The soluble lysate was obtained by centrifugation at 10 000 g for 30 min at 4°C.

Electrophoresis and immunoblots

Reagents, supplies and equipment used for polyacrylamide gel electroporesis (PAGE) with or without SDS, and for the electrotransfer of proteins to 0.2 μm nitrocellulose paper were purchased from Bio-Rad Laboratories, Hercules, Calif. Procedures have been described for PAGE ( Laemmli, 1970) and electrotransfer ( Towbin et al, 1979 ).

For immunoblot (Western) analysis, after proteins were separated by PAGE in 8.5% gels and transferred to nitrocellulose paper, the blots were rinsed twice for 5 min each in 20 m M Tris/0.5 M NaCl, pH 7.4 (TBS), and blocked for 1 h in 1% BSA in TBS containing 0.05% Tween 20 (TTBS). Primary antibodies were diluted in BSA-TTBS as follows: human plasmas, 1:3 or 1:7; rabbit polyclonal anti-CD36, 1:1500; mouse mAb, 1:500. Blots were overlaid with primary antibodies and incubated at RT for 2 h with gentle agitation. The blots were removed and washed three times for 5 min each in TTBS, overlaid with the appropriate HRP-labelled secondary antibody, and incubated for an additional 1 h at RT. Following two 5 min washes with TTBS and one 5 min wash with TBS, the blots were incubated for 20 min in 4-chloro-1-naphthol colour development reagent (0.06% in TBS containing 0.024% H2O2). The blots were rinsed in water to stop development.

RESULTS

Autoantibodies reactive with CD36 in plasma of TTP, ITP, HIT and HITT patients

Using both the PAICA and immunoblots, 8/11 (73%) and 10/14 (71%) of the TTP plasmas were positive, respectively, for autoantibodies reactive with CD36 in a Triton X-100-soluble extract of normal platelets ( Table I). Insufficient quantities of plasma were available from three TTP patients, for whom only immunoblots were carried out. The plasma of one TTP patient was negative in the PAICA and positive by immunoblot, probably justifying the use of both assays for each patient's plasma in clinically difficult cases when sufficient quantities of plasma are available. These data confirmed the report of Tandon et al (1994 ).

Furthermore, 6/30 (20%) and 3/16 (19%) of the ITP plasmas were positive by PAICA and immunoblot, respectively, for autoantibodies reactive with CD36 in the Triton X-100-soluble extract of normal platelets. In the similarly tested group of HIT without thrombosis, 3/10 (30%) and 6/10 (60%) were positive, and with thrombosis 2/4 (50%) and 4/4 (100%) were positive by PAICA and immunoblot, respectively. As with the TTP patients, the PAICA and immunoblots did not always give the same results with individual plasmas, necessitating the use of both methods if ample quantities of plasma were available.

The results were negative with 12 normal plasmas tested by PAICA and 10 normal plasmas tested by immunoblot.

Studies on antibodies reactive with CD36

The rabbit polyclonal antibodies used in these studies were produced to highly purified SDS-denatured 88 kD CD36 ( Tandon et al, 1989b ). This monospecific antiserum was compared to plasma containing autoantibodies from TTP patient 1 in immunoblots, using electrophoresis in gels in the presence or absence of SDS ( Figs 1A and 1B). The results show that the rabbit antiserum reacted with CD36 on the blot using SDS, but no reaction was observed without SDS (A). The same results were obtained with mouse mAb Mo91 (results not shown). In contrast, the blots developed with the autoantibodies were positive with or without SDS (B), with bands visible at the top of the gel in the absence of SDS. In this figure it should be noted: (1) the bands developed with autoantibodies in immunoblots were below the 88 kD protein (c 85 kD) in (A), and (2) the bands were sharper and relatively narrow with the autoantibodies (B) compared to the more diffuse bands developed with the rabbit antiserum (A).

Figure 1.

for the rabbit polyclonal antiserum resulted in the same pattern as shown in (A) (not shown).

Characterization of 85 kD CD36

The purification scheme for CD36 gained its final form after preliminary efforts to follow published methods ( Tandon et al, 1989b ; McGregor et al, 1989 ). However, with patient 1 TTP plasma as the indicator antibody, some discrepancies were disclosed between our target molecule and published characteristics of CD36. For example, as a purification step, Tandon et al (1989b ) used wheat germ agglutinin affinity chromatography to purify CD36. In our hands, this step was not effective in adsorbing the protein reactive with the TTP plasma and most of the CD36 was found in pass-through fractions (data not shown). This information coupled with the results of non-SDS-PAGE (Fig 1B) led to the use of anionic Mono Q at high pH to eliminate low molecular weight contaminants of CD36 (Fig 2). Another difference was that the highly purified CD36 shown in Fig 2, lane 5, had a calculated Mr of 85 kD. Finally, to show that the rabbit polyclonal anti-human CD36 was reactive with the purified 85 kD protein, as it was expected to be, required an additional step in the immunoblot analysis. Instead of HRP-labelled goat anti-rabbit IgG secondary antibody, a biotinylated goat anti-rabbit IgG was used. After 1 h at RT, the blot was washed and overlaid with HRP-ExtrAvidinTM for 1 h at RT. Washing and substrate development proceeded as described above.

Figure 2.

00 Superfine gel filtration.

Therefore the 85 kD protein in the eluate fractions was purified to homogeneity using the plasma from patient 1 as a source of autoantibodies, and the final step of purification yielded a single band by both SDS-PAGE analysis and the corresponding immunoblot (Fig 2, lane 5). The procedure was repeated two additional times with the same results. The yield, starting with 20 units of outdated platelets (twice) and 18 units (one), was approximately 0.5 mg of highly purified CD36 for each separate purification.

The reaction of various antisera with the purified 85 kD CD36 was as follows: positive reaction with the autoantibodies by both immunoblots and ELISA; the OKM5 mAb did not react by either immunoblots or ELISA; the mAbs FA6-152, 131.5 and Mo91 were reactive in ELISA. According to the manufacturer (Immunotech), FA6-152 is not reactive in immunoblots.

Temporal studies of anti-CD36 autoantibodies from TTP patients

Two separate samples of plasma from TTP patient 1 were compared in an immunoblot, the first was obtained when the patient was in crisis, the second was collected approximately 1 month later in remission. A prominent band was observed at 85 kD, using a Triton X-100 extract of normal platelet membranes and the patient's plasma as the source of autoantibodies. The band was scarcely visible when the patient was in remission (data not shown). Therefore it appeared that clinically useful information might be obtained regarding the status of the patient's anti-CD36 autoantibodies before, during, and after treatment.

A more detailed study of a second patient (patient 2) was possible because of multiple plasma samples collected over the course of 143 d. The results in Fig 3 show comparative immunoblots together with platelet counts, haemoglobin and haematocrit determinations. Starting with day 1, the 85 kD CD36 was scarcely visible in spite of a low platelet count (13 × 106/ml), low haemoglobin (9 g/dl) and low haematocrit (28%). Since CD36 is found on a number of other cell membranes and tissues in addition to platelets ( Greenwalt et al, 1992 ), it is possible that the anti-CD36 autoantibodies were reactive with many different cells and tissues, and not available in high concentration in the plasma. After 24 d the patient had a severe thrombotic crisis, and the platelet count dropped to 6 × 106/ml, the haemoglobin to 6.7 g/dl, and the haematocrit to 19%. A prominent band was visible in the immunoblot. After 143 d, when the patient was in remission, the band was scarcely visible, the platelet count increased to 317 × 106/ml, the haemoglobin to 13.6 g/dl, and the haematocrit to 43%.

Figure 3.

6.

Immunoblots were used for other patients with TTP to follow the course of their disease when multiple plasma samples were available (data not shown). The data indicated, as depicted in Fig 3, that testing for autoantibodies produced to 85 kD CD36 in immunoblots could produce useful information regarding the immune status of the TTP patient, before, during, and after treatment.

DISCUSSION

Our findings indicated that >70% of the TTP patients investigated in this study had plasma antibodies reactive with the platelet membrane glycoprotein CD36. There was a close correlation between the two methods used to test the plasmas, the PAICA and immunoblots (Western). We routinely use the PAICA as a rapid screening assay because it is sensitive enough to detect changes shown by immunoblots. Previously, a Triton X-100 extract of pooled normal platelets was used as a source of CD36. This source, as was shown in this study, is a mixture of the 85 and 88 kD CD36. It is now recommended that the antigen should be enriched with the 85 kD CD36 for immunoassays. These results with patients' plasmas confirmed the study of Tandon et al (1994 ), and refuted the report of Raife et al (1997 ).

In addition to TTP plasmas, a close correlation was also found between the PAICA and immunoblots for detecting anti-CD36 antibodies in plasmas of patients with ITP: 20% were detected with PAICA and 19% with immunoblots. He et al (1994 ), using mAb FA6-152 and antigen capture methods, found antibodies to GPIV (CD36) in 38% of 47 patients with chronic ITP. The form of CD36 was not investigated.

A disparity was found in PAICA and immunoblots when testing 10 plasmas from HIT patients, with 30% anti-CD36 positive by PAICA and 60% by immunoblots. In tests of patients with HITT, the plasmas of only four cases were available, and 50% were positive by PAICA and 100% by immunoblots. Currently we are collecting plasmas from patients with HIT and HITT for additional tests.

Also in agreement with the report of Tandon et al (1994 ), our data indicated that antibodies directed against CD36 may be involved in initiating thrombotic complications because the target CD36 antigen is present on both platelets and endothelial cells. As we ( Valant et al, 1998 ) and others ( Tandon et al, 1994 ) have shown, TTP plasma causes platelet activation. In addition, TTP plasmas have platelet-neutrophil aggregate promoting activity ( Valant et al, 1998 ). It was possible to test for these activities in TTP plasma because of the development of sensitive flow cytometric methods ( Valant et al, 1998 ).

A valuable source of autoantibodies for this study was the plasma of TTP patient 1. A detailed description of the clinical and laboratory findings of this patient are found in Materials and Methods. During the course of purifying platelet membrane glycoprotein CD36 from Triton X-100 extracts of pooled normal outdated platelets, a second molecular weight form of CD36 was discovered. In addition to the classic 88 kD form purified and characterized by Tandon et al (1989b ) and others ( Oquendo et al, 1989 ; McGregor et al, 1989 ), a lighter 85 kD form of CD36 was highly purified by FPLC technology. Without the use of the autoantibodies, the 85 kD form would have been much more difficult to distinguish from the classic 88 kD form of CD36. The one polyclonal and two monoclonal antisera produced privately (American Red Cross, Rockville, Md.) and the two monoclonal antisera produced commercially reacted strongly with the 88 kD form. All of these antisera reacted weakly or required modifications to produce more sensitive reactions (e.g. biotin-ExtrAvidinTM technology) with the 85 kD form. As we have shown, all patients' autoantibodies collected for this study reacted principally with the 85 kD form.

Clinically, it may be useful to test for autoantibodies to the 85 kD CD36, and this was shown in Fig 3. There appeared to be a correlation between platelet count, haemoglobin, haematocrit, and the detection of antibodies to the 85 kD form of CD36 by immunoblots. Further characterization of this 85 kD form of CD36 is currently under investigation.

The concept that thrombosis is an immune event mediated by specific antibodies has been reviewed by Vermylen et al (1997 ). From our investigation of platelet CD36, two questions relate to antibody-mediated thrombosis: (1) what is the origin of the platelet 85 kD CD36, and (2) are elevated levels of autoantibodies reactive with this form important in the pathophysiology of TTP, ITP and HIT/HITT? It is well documented that native platelet CD36 is 88 kD ( Tandon et al, 1989b ; Oquendo et al, 1989 ; McGregor et al, 1989 ), and is heavily glycosylated, containing 26% carbohydrate ( Tandon et al, 1989b ). Approximately two-thirds were in alkali-labile O-glycosidic linkages ( Tandon et al, 1989b ), and 10 potential N-linked glycosylation sites were located in the extracelluar portion ( Oquendo et al, 1989 ). Sialic acid was identified as a substantial part of the carbohydrate composition ( Tandon et al, 1989b ). The 85 kD form of CD36 may arise as an epiphenomenon which accompanies the pathologic events causing the thrombocytopenia characteristic of the three different conditions. Partially lysed and/or aggregated platelets may not be completely cleared, allowing indigenous sialidases to desialylate oligosaccharides of CD36. Terminal sialic acid (S.A.) removal may enhance accessibility of the immune system to CD36. There are many reports showing that S.A.s mask sequences recognized by autoimmune antibodies (reviewed by Pilatte et al, 1993 ). S.A.s are capable of reducing or preventing accessibility of penultimate recognition sites to the immunosurveillance system ( Schauer, 1985). For example, the T antigen on human erythrocytes treated with neuraminidase renders these cells polyagglutinable because of the anti-T antibody in sera of most adults ( Bird, 1977). Compared to the 88 kD CD36, we have shown that the 85 kD CD36 is deficient in S.A. (unpublished observations), which could possibly explain why this form is the optimal target of the autoantibodies.

For the second question, the concept that antibodies mediate thrombogenic activity is now well appreciated ( Vermylen et al, 1997 ). Binding of antibodies to specific epitopes of cell membranes may cause activation of platelets, leading to activation of prothrombotic events via FcγRII receptors and complement pathway activation. Firm binding of immune complexes to the FcγRII of platelets leads to signal transduction, thromboxane A2 production and granule release ( Vermylen et al, 1997 ). The findings of a higher incidence of anti-CD36 antibodies in TTP and HITT and lower incidence in HIT and ITP ( Table I) fit well with this concept of antibody-mediated thrombosis. Anti-CD36 antibodies may be the crucial cofactor that works in conjunction with other thrombogenic factors to potentiate platelet activation, endothelial cell damage, and enhance prothrombogenic activity. In HITT and HIT, heparin-dependent antibodies interact with platelet factor 4 to form immune complexes that may bind FcγRII, leading to activation of platelets. The presence of additional antibodies such as anti-CD36 may potentiate prothrombotic activity prior to thrombosis. Anti-CD36 antibodies alone or in conjunction with other platelet-activating factors may play a role in the pathogenesis of TTP and HITT. In contrast, in ITP, most anti-platelet antibodies are directed to GPIIb, IIIa ( Karpatkin, 1997), and they seldom activate platelets. The presence of anti-CD36 antibodies in the small percentages of ITP patients may fail to induce thrombogenic activity. In only one of our cases of ITP were recurrent thrombotic complications reported, and anti-CD36 antibodies were detected in this patient.

Acknowledgements

The authors gratefully acknowledge the generous gifts of antisera, the helpful discussions, and the critical review of Dr G. A. Jamieson of the American Red Cross, Rockville, Md.

Ancillary