A thrombin–cross-reactive anticardiolipin antibody binds to and inhibits the anticoagulant function of activated protein C




To test the hypotheses that some thrombin-reactive anticardiolipin antibodies (aCL) may bind to protein C (PC) and/or activated PC (APC), and that some of the PC- and APC-reactive aCL may inhibit PC activation and/or the function of APC.


We studied the reactivity of patient-derived monoclonal aCL with PC and APC. We examined the effects of the reactive antibodies on PC activation and on the activity of APC in plasma coagulation.


Five of 5 patient-derived, thrombin-reactive monoclonal aCL bound to PC and APC. In addition, 1 patient-derived monoclonal antiprothrombin antibody (APT) that displayed aCL activity and reacted with thrombin also bound to PC and APC. Of these 6 PC- and APC-reactive aCL/APT, all failed to inhibit PC activation, but 1 (CL15) shortened the plasma coagulation time in the presence of exogenous APC and thus inhibited the anticoagulant function of APC.


Most of the thrombin-reactive aCL in patients with antiphospholipid syndrome may bind to PC and APC. Of the APC-reactive aCL, some (like CL15) may inhibit the anticoagulant function of APC and are thus likely to be prothrombotic in the host.

Immunologic studies of antiphospholipid antibodies (aPL) in the antiphospholipid syndrome (APS) have shown that aPL represent a heterogeneous group of immunologically distinct antibodies that recognize various phospholipids (PLs), PL-binding plasma proteins, and/or PL–protein complexes (1–6). The involved plasma proteins include β2-glycoprotein I (β2GPI), prothrombin (PT), protein C (PC), and protein S. Mechanistically, aPL of different binding specificities are thought to promote thrombosis via different mechanisms. For example, it was suggested that antibodies against β2GPI and/or its complexes with cardiolipin (CL) interact with endothelial cells (ECs) and monocytes and induce a tissue factor (TF)–dependent procoagulant state (7–10). On the other hand, antibodies against PC, protein S, or PLs in complexes with either PC or activated PC (APC) and protein S may inhibit activation of PC and/or function of APC (6, 11–15). Since APC proteolytically inactivates the activated factors V and VIII (denoted as Va and VIIIa, respectively), the reduced activation of PC and/or reduced APC function may lead to a procoagulant effect and thrombotic events. Of note, congenital, heterozygous PC deficiency increases the risk of venous thrombosis ∼5–10-fold (16).

Recently, during our studies of anti-PT antibodies (APT), we found that the IS6 monoclonal IgG APT (derived from a patient with primary APS) bound to CL and thrombin, and that 5 of 7 patient-derived monoclonal IgG anticardiolipin antibodies (aCL) cross-reacted with PT and thrombin (17). Of these 6 thrombin-reactive IgG monoclonal antibodies (mAb), CL24 could interfere with inactivation of thrombin by antithrombin (AT) (17). Therefore, CL24 and two other thrombin-reactive mAb were analyzed for their binding affinities to thrombin and PT by competitive inhibition. The results showed that soluble thrombin was more effective than PT in inhibiting the binding of all 3 mAb to either PT or thrombin in solid phase. Importantly, thrombin could inhibit all tested mAb from binding to thrombin and PT, while PT could only inhibit mAb from binding to PT, but not from binding to thrombin. These results demonstrated that these 3 mAb were more specific for thrombin than for PT. Based on these inhibition data, the relative Kd values of these “anti-thrombin antibodies” are ∼1.7–7.5 × 10−6M (17).

The discovery of thrombin-reactive aCL raised a possibility that such aCL may also react with PC, which contains a trypsin/thrombin-like serine protease domain. A comparison of PC and thrombin at the protein level shows that these proteins share a similarity of 50.5% with an identity of 40.8% (Figure 1). The homologous amino acid sequences reside mainly in the trypsin/thrombin-like serine protease domains of thrombin and PC, as well as in that of APC, which is generated from PC after cleavage by the thrombin–thrombomodulin complex between residues R211 and L212 in the heavy chain of PC (based on the amino acid positions of the PC precursor) (Figure 1).

Figure 1.

Amino acid sequence comparison of human thrombin and human protein C (PC). The amino acid sequence of thrombin (residues 328–622 of prothrombin, with the accession name of “tbhu” in the Protein Information Resource [pir] databases) was compared with that of PC (residues 43–461 of the PC precursor, pir:kxhu) using the Gap program. PC is a two-chain molecule, consisting of a light chain (residues 43–197) and a heavy chain (residues 200–461). Dots are introduced to maximize the similarity between the two sequences. The heavy chains (beginning with I364 in thrombin and D200 in PC) and trypsin/thrombin-like serine protease domains (encompassing residues 364–613 in thrombin and 212–445 in PC) in both proteins are marked. Vertical lines are placed between amino acids that are the same, colons are placed between amino acids whose comparison values are ≥0.5, and dots are placed between amino acids whose comparison values are ≥0.1 and <0.5. Four regions of the most homologous amino acid sequences are underlined (the sequences L422–H429 and R433–R436 for thrombin are considered one region).

Similarly, thrombin and APC are homologous at the structural level. As can be seen in their 3-dimensional (3-D) structures in a ribbon model (Figures 2A and B), thrombin and APC share three α-helices on the left side, two β-sheets at the bottom side, and a homologous active site at the center that consists of the identical catalytic triad residues (H406, D462, S568 for thrombin and H253, D300, S405 for APC) (18, 19). Moreover, the space-filling models of both molecules show that most of the amino acid residues in the highly homologous regions are on the surface (Figures 2C and D). The two most homologous regions (colored in red and magenta, respectively) are in the active site cleft and contain His and Ser of the catalytic triad residues in both proteins (Figure 2). The third region (colored in orange) is in close proximity to the catalytic center. The fourth homologous region (colored in green) is a part of the domain that corresponds to the exosite I in thrombin and the loops 60–70 in APC, which are implicated in interaction with their macromolecular substrates (20, 21).

Figure 2.

Three-dimensional structures of thrombin and activated protein C (APC) in A and B, ribbon models and C and D, space-filling models. Human thrombin and APC share homologous structures, and their surfaces contain regions of the highly homologous amino acid sequence. The homologous regions (in descending order of regional similarity) are shown in red (D562–P571 for thrombin and D399–P408 for APC), magenta (W400–L408 for thrombin and W247–M255 for APC), orange (G586–C594 for thrombin and G418–C426 for APC), and green (L422–H429, R433–R436 for thrombin and L260–Y267, R271–K274 for APC). A and C, Thrombin consists of a light chain (residues 328–363, light gray) and a heavy chain (residues 364–622, gray). A, The ribbon model of thrombin showing a direct view into the active site cleft, with the catalytic residues of thrombin (H406, D462, S568) at the center. In addition, residues implicated in fibrinogen recognition (R425) and antithrombin interaction (E592) (see refs. 20 and36) are marked. B, The ribbon model of APC is shown with an orientation similar to that of thrombin in A. The catalytic residues of APC (H253, D300, S405) are at the center, and the residues R263 and E424 in APC that correspond, respectively, to R425 and E592 in thrombin are marked. C, The space-filling model of thrombin showing a direct view into the active site cleft in the standard orientation (having the cleft running horizontally from left to right), with the exosite I on the right side. D, The space-filling model of APC is shown with an orientation similar to that of thrombin in C.

This similarity of thrombin to PC and APC led us to hypothesize that some of the thrombin-reactive monoclonal aCL may bind to PC and/or APC and thus interfere with PC activation and/or APC function. We now report that 6 of 6 thrombin-reactive monoclonal IgG aCL/APT bind to both PC and APC, and, of these, CL15 inhibits the anticoagulant activity of APC.


Computer sequence analysis and 3-D structure modeling.

The amino acid sequence of human thrombin (residues 328–622 of PT, with the accession name of “tbhu” in the Protein Information Resource [pir] databases [accessed August 29, 2002 online at http://pir.georgetown.edu/]) was compared with that of human PC (residues 43–461 of the PC precursor, pir:kxhu) using the Gap program in the Genetics Computer Group software package (22). The Gap program uses a scoring matrix with matches scored as 1.5 and mismatches scored according to the evolutionary distance between the amino acids. The 3-D structures of thrombin and APC in the ribbon and space-filling models were generated based on the coordinates of human thrombin (1A5G) (18) and human APC (1AUT) (19), using RasMol software (version 2.7; Bernstein & Sons, Bellport, NY).

Patient-derived monoclonal aCL and APT.

Seven IgG monoclonal aCL and 1 IgG monoclonal APT were analyzed in the present study. The aCL included CL1, CL15, CL24, IS1, IS2, IS3, and IS4 (23), and the single APT was IS6 (24). Their generation and characterization had been reported previously (23, 24).

Enzyme-linked immunosorbent assays (ELISAs) for antibodies against PC and APC.

High-binding ELISA plates (Costar, Cambridge, MA) were coated with 5 μg/ml of human PC or human APC (both from Haematologic Technologies, Essex Junction, VT) in Tris buffered saline (TBS; 20 mM Tris HCl, 150 mM NaCl, pH 7.4) containing 2.5 mM CaCl2. After incubating overnight at 4°C, plates were blocked with TBS/2.5 mM CaCl2 containing 0.3% gelatin. Then test mAb or control IgG (1 μg/ml) in TBS/2.5 mM CaCl2 containing 0.1% gelatin were distributed to wells in duplicate and incubated for 1 hour at room temperature. After washing with TBS/2.5 mM CaCl2, bound human IgG was detected with horseradish peroxidase–conjugated goat anti-human IgG (γ chain–specific; BioSource International, Camarillo, CA) and peroxidase substrate tetramethylbenzidine (Kirkegaard & Perry, Gaithersburg, MD).

Preparation of PL liposomes.

The PLs 1,2-dilinoleoyl-sn-glycero-3-phosphatidylethanolamine (PE), 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphatidylserine (PS), and 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphatidylcholine (phosphatidylcholine) were purchased from Avanti Polar Lipids (Alabasta, AL). The PL liposomes were prepared by ultrasonication of a 1 mg/ml mixture of 40% PE, 20% PS, and 40% phosphatidylcholine in TBS (15).

Assay of PC activation.

The assay was performed as described by Xu et al (25), with minor modifications. Briefly, EAhy926 (endothelial hybridoma) cells (2 × 104/well) were cultured in Dulbecco's modified Eagle's medium (Invitrogen, Carlsbad, CA) supplemented with 20% fetal calf serum in a 24-well culture plate. After incubation for 3 days, cells in wells were washed twice with Hanks' balanced salt solution (HBSS) containing 3 mM Ca2+, 0.6 mM Mg2+, and 0.1% bovine serum albumin. Thereafter, 25 μl of PC (400 nM) was mixed with 150 μl of test mAb or control IgG (133.3 μg/ml) in the HBSS buffer, and the mixtures were added to cells and incubated for 1 hour. The activation reactions were initiated by adding 25 μl of human thrombin (40 nM). The final concentrations of PC, thrombin, and IgG were 50 nM, 5 nM, and 100 μg/ml, respectively.

After 15 minutes at 37°C, the reactions were stopped by adding 50 μl of hirudin at 75 units/ml in HEPES buffer (20 mM HEPES–HCl, 150 mM NaCl, pH 7.5). The amidolytic activities of APC were then measured by transferring 200 μl of the supernatants from each well to a 96-well plate and then adding 100 μl of APC chromogenic substrate S-2366 (0.6 mM; Chromogenix, Orangeburg, NY) in the HEPES buffer to each well. The rate of substrate hydrolysis (APC activity) was determined based on optical density (OD) measured at 405 nm using a microtiter plate reader. All rates were in a linear range as a function of time. The concentration of APC was determined by comparison with a standard curve of amidolytic activity versus APC concentration constructed with freshly prepared, fully activated PC. Under these conditions, <10% of PC was activated during the assay.

Functional assays of APC activity.

The effects of APC-reactive mAb on APC function were studied initially with regard to the amidolytic activity of APC. Briefly, 50 μl of APC (20 nM) was mixed separately with 50 μl of test mAb, normal human IgG, or monoclonal isotype control IgG3 (all at 200 μg/ml) in the HBSS buffer for 1 hour at room temperature. One hundred microliters of S-2366 (0.4 mM) was then added to each reaction mixture. After 5 minutes, OD was measured and the APC activity was determined as described above.

Subsequently, the effects of APC-reactive mAb on APC activity in terms of plasma coagulation were examined according to the method described by Smirnov et al (15), with minor modifications. Experiments were conducted using TBS containing 0.1% gelatin in microtiter plates at 25°C. The assay was initiated by incubating 10 μl of APC (170 nM) with 10 μl of test mAb, normal human IgG, or monoclonal isotype control IgG3 (all at 850 μg/ml) for 5 minutes at room temperature. To each reaction mixture was then added 20 μl of human normal pooled plasma (Chromogenix), 10 μl of the PL liposomes (425 μg/ml), and 10 μl of a factor X activator (85 ng/ml; purified from Russell's viper venom [Chromogenix]). After a 1-minute incubation, clotting was initiated by adding 25 μl of 20 mM CaCl2, and OD at 405 nm was measured every 3 or 5 seconds for real-time kinetics in an iEMS microplate photometer (ThermoLab Systems, Helsinki, Finland). The clotting time was determined as the time that the OD (above the plasma background) increased by 10% of the OD increase in fully clotted plasma, as described by Smirnov et al (26). The final concentrations of APC, PL liposomes, factor X activator, and IgG were 20 nM, 50 μg/ml, 10 ng/ml, and 100 μg/ml, respectively, unless stated otherwise.

Characterization of the binding affinity of mAb to PC and APC.

A competitive inhibition assay was used to study the binding affinity of selected mAb to PC and APC. Briefly, each mAb (1 μg/ml or 2 μg/ml) was preincubated for 1.5 hours with various concentrations of either PC or APC. The mixture was then distributed to the PC- or APC-coated wells in duplicate. After incubation, bound IgG was measured. The inhibition data of each mAb were used to calculate its relative Kd toward PC and APC (27).

Statistical analysis.

The effects of all mAb on APC activity in plasma were analyzed using analysis of variance followed by Dunnett's multiple comparison test. When CL15 was analyzed alone with the control IgG, Student's 2-tailed t-test was used. P values less than 0.05 were considered significant.


Reactivity of monoclonal aCL to PC and APC.

To test our hypothesis that thrombin-reactive aCL may bind to PC and/or APC, we first analyzed 7 IgG monoclonal aCL for their reactivity with PC. The results revealed that 5 of 5 thrombin-reactive monoclonal aCL bound to PC, while neither of 2 non–thrombin-reactive aCL (i.e., IS1 and IS2) bound to PC (Figure 3A). In addition, we analyzed the IS6 monoclonal APT that also reacted with CL (24) and found that it bound to PC. Of note, all 5 thrombin-reactive aCL also bind to PT, while neither of the 2 non–thrombin-reactive aCL bind to PT (17). Of these 6 PC-reactive monoclonal aCL/APT, IS3 and CL1 at 1 μg/ml displayed the strongest binding activity.

Figure 3.

Binding of thrombin-reactive monoclonal anticardiolipin antibodies/antiprothrombin antibodies to protein C (PC) (A) and activated PC (APC) (B). Microtiter wells were coated with PC or APC, and the test monoclonal antibodies (mAb) and control IgG were analyzed at a concentration of 1 μg/ml. IS1 and IS2 are IgG1, and the other mAb are IgG3. Bound IgG was measured and expressed in optical density (OD). Values are the mean and SEM (2 experiments).

Subsequently, we analyzed all 8 mAb against APC. Figure 3B shows that 6 of 6 thrombin-reactive aCL/APT bound to APC, but none of the non–thrombin-reactive aCL bound to APC. The binding patterns of tested mAb to APC were similar to those to PC.

Effects of PC-reactive mAb on PC activation.

Upon discovery of these 6 patient-derived, PC/APC-reactive IgG mAb, we studied the effects of these mAb on PC activation in an EC-based functional assay. None of these mAb affected the activation of PC on the EC surface (data not shown).

Effects of APC-reactive mAb on APC activity.

We then examined the effects of these 6 mAb on the amidolytic activity of APC, using a small chromogenic substrate of APC, S-2366. None of the 6 mAb affected the amidolytic activity of APC (data not shown). Since S-2366 is a small molecule, the assay might not reflect the inhibitory effects of some APC-reactive mAb on proteolytic inactivation of factors Va and VIIIa by APC. Accordingly, we examined effects of these APC-reactive mAb on APC in plasma coagulation.

To this end, the test mAb were first examined for their background activity (in the absence of APC) in plasma coagulation initiated by the addition of a factor X activator. As can be seen in Figure 4A, except for CL15, none of the test mAb, or the control human IgG (including polyclonal IgG and monoclonal IgG3), affected the baseline clotting time (i.e., in the presence of buffer only). CL15 at 100 μg/ml had a slightly extended clotting time of 123 seconds, compared with 109 seconds (for buffer only) or 103 seconds (for the IgG control), in the presence of 50 μg/ml of PL liposomes. Of note, CL15 displayed strong lupus anticoagulant activity in a kaolin clotting time test (28).

Figure 4.

Reduction of plasma coagulation times by CL15 in the presence of exogenous APC. Test mAb and control IgG were first examined to determine the effect of their background activity on clotting times A, in the absence of APC and B, in the presence of 20 nM exogenous APC. APC was preincubated with test mAb; normal pooled plasma, phospholipid liposomes, and factor X activator were then added to the mixture. Clotting was initiated by CaCl2, and OD at 405 nm was measured every 3 or 5 seconds. Clotting times (in seconds) were determined as described in Materials and Methods. Values are the mean and SEM for each test sample (2–3 experiments in A, 5–6 experiments in B). C, The effects of CL15 were examined in the presence of APC at the indicated concentrations (0–20 nM). Values are the mean ± SEM (2 experiments). D, CL15-mediated inhibition of APC function was analyzed with the indicated concentrations of CL15 (0–50 μg/ml). Values are the mean and SEM (3 experiments). ∗ = P < 0.05 versus normal human IgG and monoclonal isotype control IgG3. See Figure 3 for definitions.

The above assays were then repeated in the presence of 20 nM APC (final concentration), which increased the baseline clotting time (for buffer only) to 452 seconds. Under this condition, CL15 reduced the clotting time to 287 seconds (a 37% reduction from 452 seconds), while none of the other test mAb or control IgG affected the clotting time significantly (Figure 4B). These data show that CL15 inhibits the anticoagulant function of APC.

Subsequently, we examined CL15 in the presence of 0–20 nM APC. As can be seen in Figure 4C, CL15 caused a significant reduction in clotting times in the presence of APC at 15 nM and 20 nM, and the degree of reduction in clotting times was increased as the concentration of APC was increased. In addition, CL15-mediated inhibition of APC function at various concentrations of CL15 (0–50 μg/ml) was analyzed. The results showed that CL15 at only 25 μg/ml significantly reduced the plasma coagulation time in the presence of exogenous APC (Figure 4D). Considering that the plasma IgG concentration is ∼10 mg/ml, 25 μg/ml is equivalent to 0.25% of total plasma IgG, suggesting that the observed inhibition of APC function by CL15 may be physiologically relevant.

The binding properties of two selected mAb to PC and APC.

Why did only CL15 inhibit APC function? Was this because the binding affinity of CL15 to APC is higher than that of all other mAb? In addition, why did all mAb (including CL15) fail to interfere with PC activation? Was this because all mAb bind to PC with low affinities that are physiologically irrelevant? To test these hypotheses, we determined the binding affinities of CL15 and IS3 to PC and APC by competitive inhibition. We elected to study only two mAb because of the prohibitory cost of APC; IS3 was chosen because it displayed the highest OD in binding to PC and APC on the plate (Figure 3). The results showed that PC and APC were comparable in inhibiting both mAb from binding to the corresponding antigens on the plate (Figure 5). Based on the inhibition data, the relative Kd values of these mAb to PC were calculated to be 1.6 × 10−6M and 3.6 × 10−6M for CL15 and IS3, respectively; the relative Kd values of CL15 and IS3 to APC were 2.2 × 10−6M and 3.8 × 10−6M, respectively.

Figure 5.

Competitive inhibition of CL15 and IS3 binding to PC (A) and APC (B). Results shown are representative of those from 2 experiments (mean ± SD). See Figure 3 for definitions.

These data show that the binding affinity of CL15 to APC is only slightly higher than that of IS3 to APC, and they therefore suggest that the affinity difference between CL15 and IS3 probably does not account for their difference in affecting plasma coagulation time in the presence of exogenous APC. On the other hand, the binding affinity of CL15 to PC is even slightly higher than that of CL15 to APC.


In studies to test the hypotheses that thrombin-reactive aCL may bind to PC and/or APC and consequently interfere with PC activation and/or anticoagulant function of APC, we found that 6 of 6 patient-derived, thrombin-reactive IgG monoclonal aCL (including IS6) bound to both PC and APC, while 2 of 2 non–thrombin-reactive monoclonal aCL did not. More important, of these 6 PC- and APC-reactive mAb, CL15 at 25 μg/ml (equivalent to 0.25% of total plasma IgG) shortened plasma coagulation time in the presence of exogenous APC, indicating an inhibition of the APC activity by CL15. Of note, it is generally accepted that the diagnostic cutoff for IgG aCL in APS patients is 20 IgG phospholipid units, equivalent to 20 μg/ml of IgG aCL. CL15 binds to APC with a relative Kd value of 2.2 × 10−6M. Combined, these data suggest that many thrombin-reactive aCL in APS bind to PC and APC, and that a few such antibodies (like CL15) inhibit APC function.

The key coagulation cascade has the following 4 steps: 1) there is the expression of TF, which binds and activates factor VII (generating factor VIIa); 2) the TF–VIIa complexes activate factors IX and X (generating factors IXa and Xa); 3) factor Xa works with factor Va to convert PT to thrombin, while factor IXa works with factor VIIIa to generate more factor Xa; and 4) thrombin converts fibrinogen to a fibrin network, and feedback amplifies the cascade by activating more of factors V and VIII (16). The coagulation cascade is subjected to 3 major feedback regulation mechanisms: 1) the TF pathway inhibitor inhibits TF–VIIa complexes from activating factors IX and X; 2) AT binds to thrombin and to factors IXa and Xa and inactivates their enzyme activity; and 3) PC is activated by the thrombin–thrombomodulin complex on the EC surface, and the APC then forms a complex with protein S on PL surfaces and proteolytically inactivates factors Va and VIIIa (16, 29). In this context, CL15 may inhibit APC function by interfering with its binding to protein S and/or PL, or by interfering with its subsequent cleavage of factors Va and/or VIIIa.

Studies have shown that familial thrombophilia is associated with a single point mutation (present in 20–40% of unrelated patients with thrombophilia) in the factor V gene (G1691A), resulting in an abnormal factor V (R506Q) that lacks one of the normal APC cleavage sites and is thus resistant to APC inactivation; these phenomena are termed APC resistance (APCR) (30–32). Subsequently, the aforementioned aPL-mediated inhibition of APC function has been frequently referred to as “acquired APCR.” In most studies of acquired APCR in APS, the target antigens of the involved aPL and the associated mechanisms are not well defined. Recently, it was reported that the EY2C9 and GR1D5 monoclonal IgM aCL bound to PC in the presence of both CL and β2GPI, but not in the absence of either CL or β2GPI (33). Moreover, the study also showed that soluble PC could inhibit EY2C9 from binding to coated PC in the presence of both CL and β2GPI, and that EY2C9 shortened the clotting time of normal plasma in the presence of APC in an activated partial thromboplastin time test (33). These findings led the study's investigators to suggest that some aCL might bind to PC via β2GPI.

Taken together with the CL15 data, these results suggest that there are at least two different species of aPL in APS which could inhibit APC function: one binding directly to APC and the other binding indirectly to PC via β2GPI. In the future, it will be important to study these two species of aPL and determine the prevalence and the pathologic significance of each aPL in APS.

In addition to aPL-mediated inhibition of APC function, acquired APCR was observed in 2 patients with thrombosis (34, 35). In the first case, the patient had two IgG paraproteins, but the antigens recognized by the involved antibodies were not defined (34). In the second case, the patient's IgG was shown to react with APC but not PC (35), and thus it differed from CL15, which binds to both PC and APC. These data suggest that more than one epitope is recognized by anti-APC antibodies that inhibit anticoagulant activity of APC and are associated with thrombosis.

It is noteworthy that only 1 of 6 APC-reactive mAb from APS patients inhibits the anticoagulant activity of APC. Therefore, it is likely to be fruitless to assess the clinical significance of all anti-APC antibodies in APS by association studies of the presence of anti-APC antibodies to APS, or to study the functional activities of affinity-purified polyclonal anti-APC antibodies from patients. Instead, it will first be necessary to delineate the APC epitopes recognized by various anti-APC antibodies with or without acquired APCR activities, such as CL15 (which inhibits APC activity) and IS3 (which does not inhibit APC activity but which binds to APC with an affinity similar to that of CL15) (Figures 4 and 5). If the APC epitopes recognized by only prothrombotic anti-APC antibodies are defined, then specific assays for the disease-relevant anti-APC antibodies may be developed and used to study the roles of prothrombotic anti-APC antibodies in thrombosis in patients with APS.


We thank Mrs. Paifei Chen for technical assistance.