Identification of Divalent Metal Ion-dependent Inhibition of Activated Protein C by cr2-Macroglobulin and cr2-Antiplasmin in Blood and Comparisons to Inhibition of Factor Xa, Thrombin, and Plasmin*

The half-life of activated protein C (APC) was 31 min in citrated blood and 18 min in whole blood. Immunoblotting analysis of citrated blood identified APC-protein C inhibitor (APC-PCI) and APC-al-anti-trypsin complexes. Whole blood contained two addi- tional APC-inhibitor complexes, one stimulated by Ca"+ and another by Mg2+. The former was identified as APC-a2-macroglobulin (APC-a2M) while the latter was not identified. APC-a2-antiplasmin complexes (APC-a2AP) were identified, comigrating with APC-PC1 complexes. Purified a2M and a2AP inhibited APC in the presence of Ca2+ (k2 = 99 and 100 M" s-', respectively. Inhibition of APC and Factor Xa by azM and inhibition of APC by azAP was stimulated by Ca2+, Mn"+, and Mg". Inhibition of thrombin by a2M and of plasmin by a2AP was not altered by EDTA or Ca2+, suggesting divalent metal ions affect APC and Factor Xa rather than the inhibitors. k2 values for the APC inhibitors and their plasma concentrations suggest that PC1 and al-antitrypsin are the more important APC inhibitors and that cr2M and a2AP are metal ion-de-pendent auxiliary inhibitors. Inhibitors can account for the in vivo half-life of

Protein C is a vitamin K-dependent zymogen (1, 2) which, when proteolytically cleaved to form activated protein C (APC),' serves as an anticoagulant regulator of coagulation pathways by inactivation of the cofactors Va and VIIIa (3-5). Evidence for its physiologic importance stems from reports of potentially fatal purpura fulminans in homozygous protein Cdeficient infants (6) and association in some families of venous thrombotic disease with heterozygous deficiency of protein C (7). Evidence of protein C activation during intravas-* This study was supported in part by Grant HL-31950 from the National Institutes of Health and a fellowship (to A. G.) from the California Affiliate of The American Heart Association. A preliminary report has appeared in abstract form ( The abbreviations used are: APC, activated protein C; a2AP, 012antiplasmin; nlAT, a]-antitrypsin; n2M, a2-macroglobulin; 1-2581, Ndansyl-(p-guanidine)-Phe-piperidine-HC1; kn, second order association rate constant; PCI, protein C inhibitor; S-2366, pyro-Glu-Pro-Arg-p-nitrophenylanilide; TBS, Tris-buffered saline; SDS, sodium dodecyl sulfate; BSA, bovine serum albumin. cular coagulation has been presented (8). APC had an antithrombotic effect and prevented death due to septic shock when administered in several animal models (9)(10)(11)(12). The possibility of future therapeutic use of APC in humans has led to increased interest in its physiologic inhibition by protease inhibitors in blood. Recent work revealed that major inhibitors of APC in plasma include PC1 and alAT (13-16), both of which inhibit APC relatively slowly compared with inhibition of other coagulation enzymes by plasma inhibitors. The existence of other APC inhibitors has been suggested (16). In addition to the two previously identified plasma inhibitors, here we report that a 2 M , a,AP, and possibly another protein inhibit APC in whole blood in a divalent metal ion-dependent manner, making the half-life of APC in whole blood without divalent metal ion chelators significantly shorter than in citrated blood or plasma. To determine whether the unusual metal ion dependence was due to an effect on APC or to an effect on azM and a2AP, we compared inhibition of APC to that of Factor Xa, thrombin, and plasmin.

EXPERIMENTAL PROCEDURES
Reagents-Human protein C was purified and activated as described (9). A previously described monoclonal antibody to the light chain of protein C (C3) was used for immunoblotting and for solid phase assay of APC activity (9,15,17,18). This antibody C3 recognizes with equal efficiency protein C, APC, or complexes of APC with PC1 or n,AT. For some immunoblots, goat polyclonal antibody to protein C was used (17). Rabbit antibody to PC1 was prepared as described (19,20). Human anM was a kind gift from Dr. Steven Gonias (University of Virginia Medical School). Human a2AP was obtained from Athens Research and Technology. Recombinant Arg'"nlAT was a generous gift from Drs. Michael Courtney and Ranier Bischoff (Transgene, Strasbourg, France). Monoclonal antibody to a2M was obtained from Chemicon. Other antibodies were prepared in rabbits and obtained from Behring. Polyclonal antibody to a2AP was affinity-purified by absorption to a2AP coupled to Affi-Gel 10 (Bio-Rad) according to manufacturer's instructions. The bound antibody was eluted with 3 M NaSCN in Tris-buffered saline (TBS) and immediately dialyzed against TBS. Cephalin andp-nitrophenyl guanidinobenzoate were obtained from Sigma, protein A-Sepharose from Pharmacia LKB, Biotechnology Inc., heparin from Elkins-Sinn, and human Factor Xa, thrombin, and plasmin from Enzyme Research Laboratories.
General Methods-Electrophoresis and immunoblotting were performed as previously described (17,19). The concentrations of a2M and n,AP were determined from absorbance a t 280 nm, using E"" of 8.9 (21) and 6.7 (22), respectively. The n2M was determined to be 91% active by titration against active site-titrated thrombin (92% active), and the a,AP was determined to be 67% active by titration against active site-titrated plasmin (14% active), according to procedures previously described (23).
The activity of thrombin titrated with n2M was determined by clotting of fibrinogen (Kabi) (24), and the amidolytic activity of plasmin titrated with a2AP was determined using the peptide substrate S-2366 (Kabi).
Kinetics for Inhibition of APC and Other Proteases-For determination of the second order association rate constant (k,) for inhibition of APC by a2M, 0.5 p~ APC was incubated at 37 "C with 2.6 p~ n,M and 5 mM CaCI? or other additions as specified in TBS containing 0.05% bovine serum albumin. Aliquots of 4 pl were removed over time, and the residual APC anticoagulant activity was tested by its ability to prolong the activated partial thromboplastin time. The control curve was obtained using dilutions of APC incubated without tr,M. Clotting mixtures contained 100 pl of pooled normal plasma (George King Bio-Medical, Inc.) preincubated 200 s with the test sample and a cephalin-silica activator (100 p1 of Thrombosil, Ortho). Clotting was initiated by addition of 100 pl of 25 mM CaCI2. For inhibition of APC by a2AP, loss of amidolytic activity toward S-2366 was measured over time by assaying aliquots of mixtures containing APC (63 nM) and a,AP (from 4.6 to 14 p~) .
Pseudo-first order rate constants (hi) were determined from the slopes of semilog plots of activity versus time, and k2 was calculated from hz = kl/[inhibitor].
The theoretical half-life of APC in blood was calculated from t,,, = 0.69/kz [inhibitor], using the plasma concentration of nnM, a2AP, or other inhibitor.
For determination of the k, for inhibition of Factor Xa by a2M, 0.18 unit/ml Factor Xa was incubated a t 37 "C with 1.12 p~ a2M in TBS, 1% BSA containing 5 mM CaCI2 or other addition to be specified. At selected times from 5 to 15 min, reactions were quenched by dilution by addition of 20 volumes of TBS-BSA, followed by immediate determination of residual Factor Xa activity in a Xa onestage clotting assay as follows. Normal plasma (100 pl) was preincubated with 100 pl of test sample and 100 p1 of 0.2 mg/ml cephalin for 200 s. Clotting was initiated by addition of 100 pl of 30 mM CaCI,. Standard dilutions of Factor Xa were incubated without a2M for the same times as the test samples, and these clot times were used to construct a standard curve.
For determination of the k? for inhibition of thrombin by a,M, 0.5 p M a,M was preincubated for 20 min a t 37 "C with 5 mM CaCI2 or other additions as specified in TBS, 1% BSA. Thrombin (22.5 units/ ml) was added and incubation continued for selected times from 1 to 3 min. The reaction was quenched by addition of a 25-fold volume of 11% polyethylene glycol-8000, 20 mM CaCI2. An immediate test for clotting activity (24) was made using an equal volume of 4 mg/ml fibrinogen, and thrombin activity was quantitated based on a standard curve constructed using control thrombin dilutions.
For determination of the k, for inhibition of plasmin by a2AP, 2.8 pg/ml plasmin a t 5 "C was incubated with 7.8 pg/ml a,AP a t 5 "C in TBS, 1% BSA with 5 mM CaCI, or other addition as specified. Aliquots were removed a t 30-s intervals and immediately assayed for amidolytic activity by diluting into 60 volumes of 0.8 mM S-2366,0.05 M Tris-HCI, 0.1 M NaC1, 0.1% BSA, p H 8.2.
Inhibition of APC in Blood and Plasma-For half-life studies of APC activity in blood, blood was collected from seven different normal donors into separate siliconized polypropylene tubes containing volume of 0.11 M trisodium citrate or TBS. APC was added to a final concentration of approximately 2 pg/ml in the fluid phase, and the mixture was incubated a t 37 "C. Aliquots were taken over time into tubes containing buffer with benzamidine (Kodak) to give a final concentration of 30 mM. Blood was centrifuged for 1 min in a Beckman Microfuge B (8,900 X g), and the supernatant plasma was assayed for APC activity in the wells of microtiter plates coated with monoclonal antibody C3 against protein C and blocked with TBS, 1% casein, 0.02% NaN, (9,18). After incubation of aliquots containing APC in the wells and subsequent washing, the wells were incubated with 100 pl of 0.8 mM S-2366, the change in absorbance a t 405 nm was recorded over time using a Biotek 312 enzyme-linked immunosorbent assay plate reader, and the observed amidolytic activity was compared with APC standards. In parallel studies, aliquots of the incubation mixtures containing blood and APC were taken over time without addition of benzamidine and centrifuged for immediate direct determination of APC amidolytic activity in 0.8 mM S-2366 containing 20 p~ of the thrombin inhibitor 1-2581 (Kabi) (17,19). Under these conditions 1-2581 does not inhibit APC but neutralizes any background activity due to thrombin.
In some experiments to compare inhibition of APC in blood uersus plasma, blood was collected into tubes with APC in TBS to give a final concentration of 16 pg/ml, prostaglandin E-1 at a final concentration of 1 pglml, and ' /9 volume of either 0.11 M sodium citrate or TBS. Half of the sample containing citrate and half of the sample containing TBS were centrifuged 4 min a t 5,000 X g in a Dade microcentrifuge a t room temperature. The whole blood mixtures and the cell-free mixtures were incubated at 37 "C and aliquots taken over time for determination of APC anticoagulant activity, APC amidolytic activity, and APC antigen immunoblotting pattern as described above.
Identification of APC-Inhibitor Complexes-In order to remove specific APC-inhibitor complexes from incubation mixtures of blood and APC by immunoabsorption, the mixtures were incubated for a specified time and then made 30 mM in benzamidine using a 500 mM benzamidine solution and centrifuged 1 min in a Beckman Microfuge B. The supernatant plasma in 7-pl aliquots was mixed with 15 pl each of IgG fractions of various antisera against known protease inhibitors. The samples were incubated 1 h at 3'7 "C and 2 h a t 4 "C. Each sample was then added to 20 pl of washed, packed protein A-Sepharose beads in a 1.5-ml conical microcentrifuge tube and mixed. The mixture was incubated 16 h a t 4 "C, with occasional mixing during the last hour of incubation. The samples were centrifuged 1 min, and 20 pl of each supernatant solution was subjected to immunoblotting analysis.

Effect of Metal Ions in Blood on APC Half-life and Inhibitor
Complexes-To study the influence of divalent metal ions on the inhibition of APC by blood, APC was mixed with freshly drawn blood in the absence or presence of citrate, and the loss of APC activity was measured as a function of time. As calculated from the APC amidolytic activity data in Fig The microplate assay for APC activity was described under "Experimental Procedures." The lower panel shows an immunoblot for protein C and APC in blood incubated with APC in the presence and absence of citrate. A representative individual sample from the experiment in the upper panel is shown. The immunoblot was from a 5% nondenaturing polyacrylamide gel and employed monoclonal antibody to protein C (C3) followed by "'I-protein C (15,17). of Activated I'rotpin C i.e. in the absence of citrate, was not due to any metal iondependent reactions involving blood cells under these conditions. Moreover, half-life determinations based on APC anticoagulant, activity assays gave 20 min for APC in whole blood without citrate, 33 min in blood with citrate, and in blood from which cells were removed, 17 min without citrat,e and 28 min in the presence of citrate. Thus, the inhibition of APC anticoagulant activity in blood coincided with loss of amidolytic activit,y, was significantly influenced by divalent metal ions, and was not. significantly influenced by blood cells. Other experiments in which recombinant hirudin (Ciba-Geigy, 2 pg/ml) was included in the blood mixtures suggested that inhibition of APC activity in blood was not influenced by any thrombin that might be present.
T h e lowerpand of Fig. 1 shows an immunoblotting analysis using a nondenaturing gel for protein C antigen in blood samples incubated with APC. When APC was incubated with blood in the presence of citrate (lanes 6-10), there were two apparent major bands of AI'C-inhibitor complexes formed, similar to previous reports (14,15) for citrated plasma incubated with APC. Immunoblotting for PC1 or for trlAT (data not shown) confirmed that these bands contained APC-PC1 and APC-trIAT, as previously reported (14,15). However, in the reaction mixhres using whole hlood, ie., without citrate ( Fig. 1, louwpanel, lanes 1-.5), two additional bands containing APC antigen, designated x and y , were visible following incubation of APC with blood. Other experiments (data not shown) demonst.rated that addition of recombinant hirudin (2 pg/ml) to the mixt.ures of blood and APC had no effect on bands x and y or any of the other APC bands.
To investigate whether formation of these bands was divalent. metal ion-dependent, metal ions and other addkions were made to reaction mixtures of APC with citrated normal plasma or citrated protein S-depleted plasma, and the mixtures were electrophoresed on nondenaturing gels and subjected to immunohlotting for protein C antigen. As seen in Fig. 2 (lanes 1 and 6), the only two apparent bands detected in the absence of added metal ions migrated at positions previously assigned to APC-PC1 and APC-n,AT complexes. Addition of 12 mM calcium ions led to the appearance of a new band (band x ) of very low electrophoretic mobility (Fig.  2, lanm 2 and 7 ) . Other experiment,s using protein S-depleted plasma revealed that. t.his low mobility band x seen in Fig. 2 (lane 7 ) was dependent on calcium ions but not on the presence of protein S. In separate experiments (data not shown), band x formation was also stimulated by magnesium or strontium ions but to a lesser extent than bv calcium ions. Addition of a metal ion mixture to give final concentrations of 1.8 mM MgCI,, and 0.1 mM each ZnCI,, MnCI,, CoCI?, and CuCI,, led to the appearance of another new band, designated band .v (Fig. 2, lanr 3 ) . In other experiments appearance of b a n d y was dependent on magnesium ions and not on anv of the other divalent ions tested (data not shown). Addition of phospholipids or protein S to citrated plasma did not change the overall pattern of complex formation (Fig. 2, lnncs 4 and -5). A diminution of the band ascribed to AI'C-PC1 complexes was observed when calcium ions were added (Fig. 2, compare lanes 2, 3 , and 7 with lanes 1 and 6 ) , but not when calcium ions were added in combination with phospholipids ( Fig.  2,  lanes 4, 5, and 8). When APC was incubated with human serum (Fig. 2, lane 9 ) . bands comigrating with hands x and .v were observed. Thus, the divalent metal ions, calcium and magnesium, stimulated the formation of APC complexes in blood, plasma, and serum.
Idmtification of AI'C Cornplexed with tr,M and (r,AP in Blood-Band x was identified as the complex AI'C-tr,M by its removal from an incubation mixture of blood and AI'C using monoclonal antibody to n,M, followed by protein A-Sepharosc (Fig. 3, compare lanes 3 and 9 with lane 2 ) . Complexes of APC-cyIAT were removed in a similar manner by adsorption using anti-trlAT antibodies (Fig.  3 , l a m 4 ) . Adsorption of reaction mixtures using antibodies to nl-antichvmotrypsin, C1 inhibitor, inter-tr,-protease inhihitor, and /jJ-glvcoprotein I, followed by protein A-Sepharose, did not remove band . v or any of the other bands (Fig. 3 , lancs 5-8). In other experiments, adsorpt.ion using antibodies to plasminogen activator inhibitor-1 did not remove any of the prominent bands containing APC antigen (data not shown).
When the reaction mixtures of blood and APC without citrate were adsorbed using affinity-purified antibodv to tr,AI' followed by protein A-Sepharose, the band previously identified as APC-PC1 was diminished (Fig. 3  antigen had indeed been removed (data not shown). Control experiments showed that the affinity-purified antibody to a2AP did not recognize PC1 (data not shown). Adsorption of the reaction mixture of blood and APC simultaneously using antibodies to both a2AP and PC1 completely removed the band(s) in this region (Fig. 3, lune 13), showing that both APC-PC1 and APC-a2AP complexes were originally present in the immunoblot band labeled APC-PCI (Fig. 2). This was further confirmed using a reaction mixture of PCI-depleted plasma with APC, calcium, and magnesium ions (Fig. 3, lunes  14 and 15). The PCI-depleted plasma formed complexes in the region of APC-PC1 (Fig. 3, lune 151, and it was devoid of PC1 antigen when analyzed by immunoblotting with antibodies to PC1 (data not shown). The identity of these complexes as APC-a,AP was established by their complete removal upon adsorption with affinity-purified antibody to a2AP (Fig. 3,  lune 14). Thus, APC complexes with both a2M and a2AP in whole blood.
An additional band containing APC antigen was seen on immunoblots, migrating just below bund y in some experiments, e.g., in Fig. 3 (lanes 10 and 12). This band was faint when samples were electrophoresed immediately after incubation of the reaction mixture containing APC, but it increased in intensity when samples were frozen and thawed or handled for extended periods of time prior to immunoblotting analysis, suggesting that this band may arise after proteolysis or degradation of one of the other species containing APC. This band may be related to APC-a2AP, since antibody to a2AP removed this band in Fig. 3, lanes I1 and 13. Moreover, as seen below in Fig. 6, a band of this mobility was detected on immunoblots of plasma/APC mixtures with antibodies against a2AP.
Inhibition of APC by Purified a2M-Inhibition of APC by human a2M was studied using purified proteins in the presence of 5 mM calcium ions or 5 mM EDTA. From the slopes of pseudo-first order plots of APC activity against time (Fig.  4, upper panel), the second order association rate constant k2 for inhibition of APC anticoagulant activity by a2M in the presence of calcium ions was 99 20 M-' s" ( n = 3). Inhibition of APC by a2M in the presence of EDTA was negligible. Similar values for k2 were obtained by monitoring the inhibition of APC amidolytic activity by a2M; however, there was often a lag phase of up to 30 min before inhibition of APC amidolytic activity by a2M was observed, and maximum inhibition did not exceed 80%. To see whether APC bound to a2M retained some activity against small molecules, but not macromolecules, incubation mixtures of APC and a2M were tested to measure how much of the remaining APC amidolytic activity was inhibited by Arg'"alAT. After 50 min of incubation of APC with a2M, 26% of the remaining APC amidolytic activity was protected from Arg'"alAT, after 70 min of incubation, 50% was protected, and after 16 h of incubation, 100% was protected. Thus, APC bound to a2M retained approximately 20% of its amidolytic activity.
Complex formation of APC with purified a2M was directly demonstrated using immunoblotting. Fig. 4 (lower panel) shows an immunoblot for APC antigen from a nondenaturing gel (left) and also an immunoblot from a denaturing gel (right) of incubation mixtures of APC with purified human a2M. Bands (Fig. 4, lower left) with the same mobility as bund x in Figs. 1 and 2 and APC-aZM in Fig. 3 were apparent after 5 min of incubation in the presence of calcium ions but were barely detectable after 16 h of incubation in the presence of EDTA. The denaturing SDS gel immunoblot revealed a doublet of heat-stable, detergent-stable bands of molecular weight in excess of 200,000 formed after APC was incubated with a2M for 2 5 min in the presence of calcium ions. Samples from incubation mixtures of blood and APC immunoblotted from denaturing SDS gels had an identical pattern of immunoblot bands representing complexes of molecular weight over 200,000 (data not shown). The molecular weight of these complexes on reduced denaturing SDS gels was 130,000-150,000 (data not shown). In the nondenaturing gel blot (Fig.  4, lower panel), the total measured radioactivity per lane fell by 33% for the 120-min incubation of APC and a2M in the presence of calcium ions, but the radioactivity per lane remained constant during 120 min of incubation of APC in the presence of calcium ions with a2M and remained constant during 120 min of incubation of APC and a2M in the presence of EDTA. This indicates that APC in the APC-azM complexes was underrepresented in the immunoblots, either because APC was not fully accessible to the antibody or because the large APC-02M complexes did not transfer efficiently to the nitrocellulose paper, or both. In the denaturing SDS gel blot (Fig. 4, lower panel), the total radioactivity per lane also decreased for the 30-min incubation with APC and a2M in the presence of calcium ions, but not in the presence of EDTA. Thus, although immunoblotting analysis qualitatively demonstrates complexation of APC with purified a2M, it does not allow a quantitation of these complexes.
T o learn whether the unusual metal ion dependence of inhibition of APC by a2M might be due to APC or to a2M, k2 values for inhibition by a2M of APC, Factor Xa, and thrombin were obtained using kinetic studies and then compared under various conditions. The data in Table I show that 5 mM CaC12 stimulated APC inhibition at least 7-fold and Factor Xa inhibition 4-fold when compared with 0.2 mM EDTA. Fig. 5 shows that the k2 for inhibition of APC had a similar dependence on calcium ion concentration as the k2 for inhibition of the homologous vitamin K-dependent Factor Xa. The k2 values obtained for inhibition of Factor Xa are in reasonable agreement with those previously reported (25), which were  determined in the presence of 4 mM CaClZ. The patterns for calcium ion stimulation of a2M inhibition of APC and Factor Xa were similar, although no final plateau for stimulation was observed for APC inhibition by a2M. Inhibition of both APC and Factor Xa by azM was stimulated by magnesium ions to a lesser extent than by calcium ions (Table I). Manganese ions were more stimulatory for Factor Xa inhibition than for APC inhibition. In clear contrast to APC and Factor Xa, inhibition of thrombin by a,M was unaffected by the divalent metal ions or chelating agents that were tested (Table I). This was true whether the ions tested were preincubated for 20 min with a2M or whether the addition was made immediately prior to the thrombin (values in parentheses in Table I). The k, values obtained for thrombin inhibition are in reasonable agreement with those previously reported (26).
Inhibition of APC By Purified aAP-Inhibition of APC by purified a2AP was studied. Fig. 6 (left panel) is a pseudo-first order plot for inhibition of APC amidolytic activity upon incubation with 14 p M a2AP in the presence of calcium and magnesium ions or in the presence of EDTA. The rate of inhibition of APC by a2AP was four times higher in the presence of divalent metal ions than in the presence of EDTA. Fig. 6 (right panel) demonstrates that calcium ions were more effective than magnesium ions in stimulating this inhibition of APC. Heparin at 1 unit/ml diminished inhibition of APC by a2AP. Kinetic analyses showed that the association rate constant k, for inhibition of APC by a2AP was 100 k 9 M" s" in the presence of calcium and magnesium ions and 28 M" s" in the presence of EDTA.
T o determine whether this unusual metal ion dependence of inhibition of APC by azAP was due to APC or to azAP, inhibition of APC was compared with that of plasmin by a2AP. The data in Table I1 show that inhibition of APC by a2AP was stimulated 3-4-fold by 5 mM CaC12, that it was less strongly metal ion-dependent than inhibition of APC by a2M, and that inhibition was stimulated by other divalent metal ions, including manganese and magnesium to a lesser extent. In clear contrast to APC, inhibition of plasmin by azAP was unaffected by any of the divalent metal ions or chelating agents that were tested (Table 11).
Complex formation of APC with a2AP was shown using immunoblotting analysis. Purified azAP formed complexes with APC (Fig. 7, last lane) that comigrated with APC-PC1 complexes and with the APC-a2AP complexes identified in incubation mixtures of plasma or blood with divalent metal ions and APC (Fig. 7, left panel). In immunoblots from denaturing 4-15% gradient gels, APC-a2AP complexes also comigrated with APC-PC1 complexes (data not shown). Purified APC-a2AP and APC-a2AP complexes formed in plasma containing divalent metal ions reacted with antibodies to azAP, and these complexes were not formed in plasma from a patient congenitally deficient in a2AP (27) (Fig. 7, right  panel, a A P d ) . The quantity of APC-a2AP complexes in the 1-h incubation mixtures shown in Fig. 7 (right panel) did not appear to be significantly different in the presence or absence of divalent metal ions. However, in other experiments (not shown) divalent metal ions stimulated the formation of these complexes at incubation times of less than 30 min. Thus, purified aZAP inhibits APC and forms complexes identifiable on immunoblots of reaction mixtures of APC with the purified inhibitor or plasma. Immunoblots for PC1 antigen (not shown) revealed that fewer APC-PC1 complexes were formed in blood or plasma in the presence than in the absence of divalent metal ions a t all time points, even though these metal ions do not affect the rate of inhibition of APC by purified PC1 (20). The unidentified magnesium-ion dependent bund y of APC complexes was not related to either a2AP or to PCI, since it was formed in both PCI-depleted plasma and plasmadeficient in a2AP (Fig. l , left panel).

Inhibition of APC by the plasma inhibitors aIAT and PC1
is not divalent metal ion-dependent in purified systems or in plasma (14,20,28), and the inhibition of plasma blood coagulation enzymes by plasma protease inhibitors has not been previously shown to be stimulated by divalent metal ions. However, as described here, we found a difference in the halflife of APC in whole blood in the presence and absence of citrate, a divalent metal ion chelator, and this led us to discover the existence of two additional types of apparent APC-inhibitor complexes that are observed in blood or serum when citrate is absent. One of these complexes is identified here as APC-anM, and the other one has not yet been identified or indeed proven to be associated with inhibition of APC. In addition, divalent metal ion-stimulated complexes of APC-anAP are shown here to form when APC is added to plasma, blood, or serum, and these complexes comigrate on immunoblots with APC-PC1 complexes.
The immunoblotting observations prompted kinetic studies using purified proteins. Purified human a2M and a2AP inhibit APC in calcium ion-stimulated reactions with second order association rate constants of 99 and 100 M" s-I, respectively, compared with 11 M-' s-l for inhibition of APC by alAT and 6.0 X 10:' M" s" for inhibition of APC by PC1 (14,20,28). alAT (40 p~) , and PC1 (88 nM), the calculated half-life of APC in blood considering each inhibitor separately is 38, 110, 26, and 22 min, respectively. This kinetic estimate of relative reactivity of inhibitors in plasma is consistent with the suggestion from the immunoblotting data that alAT and PC1 are primary inhibitors, whereas a2M and a2AP are auxillary inhibitors of APC in blood. The observed half-life of APC in whole human blood without citrate of 18-21 min is approximately what the combination of calculated half-lives would imply and is similar to the reported half-life of 23 min for APC infused into humans (29) and is slightly longer than the in vivo half-life of 10-14 min for APC infused into baboons (9). The discovery of divalent metal-ion dependent inhibition of APC by a2M, anAP, and possibly by one other inhibitor in blood helps to explain the differences between the in vivo half-life of APC and the in vitro half-life of 31 min in citrated plasma. Thus, inhibition of APC by protease inhibitors in blood could entirely account for the regulation and removal of APC activity in vivo.
APC has approximately nine calcium ion binding sites (30), and calcium ions and phospholipids are required for optimal rates of inactivation of Factors Va and VIIIa (3,5). Most of the calcium ion binding sites reside in the NHn-terminal domain containing nine y-carboxyglutamyl (Gla) residues, but APC with the Gla domain removed contains a t least one high affinity binding site for calcium or manganese ions that may alter APC conformation (30,31). Calcium ions are not required for inhibition of APC by PC1 (20), alAT (14, 28), or plasminogen activator inhibitor-1 (32) or for inhibition of other coagulation serine proteases by plasma protease inhibitors. Consequently, finding a strong influence of calcium ions on inhibition of APC by a2M and arAP was unexpected.
T o assess whether the divalent metal ion influences either the enzyme or the protease inhibitor in these reactions, the inhibition of Factor Xa and thrombin by a2M and of plasmin by arAP was studied. Inhibition of the vitamin K-dependent homologue, Factor Xa, is stimulated by divalent metal ions, whereas inhibition of thrombin by a2M or of plasmin by arAP is not divalent metal ion-dependent. Thus, the unusual divalent metal ion dependence of inhibition of APC and Factor Xa is probably due to a property of the enzymes rather than due to a property of the inhibitors, anM and a2AP.
a2M, a tetramer of subunits of M , 180,000, inhibits a wide variety of proteases by a mechanism in which the protease is entrapped in a cagelike structure (33,34). The protease active site is not altered, and it may still exhibit partial enzymatic activity toward small substrates. The protease usually cleaves a "bait" region of the a2M, causing a change in conformation which stimulates covalent attachment between reactive thioester groups of anM and NH, groups of the protease and which closes the cage. Such a mechanism for a2M and APC interactions is consistent with the observation here that APC amidolytic activity is diminished but cannot be entirely inhibited and that, over time, the remaining observed amidolytic activity becomes less susceptible to inhibition by Arg:'"alAT, a very efficient macromolecular inhibitor that neutralizes >98% of APC (28). APC in APC-a2M complexes is probably less accessible to anti-protein C antibodies as well, since the total detectable anti-protein C antibody per lane on immunoblots decreased as APC-a2M complexes increased in incubation mixtures of APC and anM. Since APC-a2M complexes on immunoblots were stable to heat and detergent treatment of samples taken after 5 min of incubation of APC with blood or with purified anM, it appears that some of the APC-a2M complexes are covalent. In fact, the APC-a2M complexes were equally intense on denaturing SDS gel immunoblots as on nondenaturing gel immunoblots. The apparent molecular weight of APC-a2M complexes on reduced SDS gels of 130,000-150,000 suggests that the light or the heavy chain of APC is linked to fragments of a2M in the range of M , 90,000-120,000. Such fragment sizes have been reported for cleaved anM (35), suggesting that APC cleaves a2M. a2M inhibits thrombin (26) and Factor Xa (25,36); however, we are not aware of any previous reports of divalent metal ion-dependent inhibition by a2M of these or any other proteases. Investigators previously found that only about 13% of "'I-Factor Xa incubated with mouse or human citrated plasma bound to anM, whereas 90% of "'I-Factor Xa infused into mice was bound to aZM within 2 min (36). The difference between these in vitro and in vivo data could be explained as due to divalent metal ion stimulation of the inhibition of Factor Xa by a z M , as seen here. It was reported that purified a2M and azM in plasma do not inhibit APC, but the experiments described by these authors (37) apparently did not include the addition of divalent metal ions to reaction mixtures. Perhaps the divalent metal ion-dependent conformation of APC can fit into the cagelike structure of a Z M in a manner to cleave the bait region, or perhaps APC in this conformation is more reactive with some macromolecular substrates.
a2AP, a glycoprotein of M , 70,000, is a member of the serpin superfamily (38, 39) and is homologous to PC1 and alAT. It interacts very efficiently with plasmin with a kz of 4 x lo7 "' s-', forming 1:l complexes that are predominantly covalent. Inhibition of APC by a2AP is stimulated by divalent metal ions in a purified system and in blood. Interestingly, after 30 min of incubation of plasma with APC, complexes observed on immunoblots are more intense in the absence of divalent metal ions than in their presence (Fig. 6, and data not shown). Proteolysis of a2AP or its complexes in plasma containing divalent metal ions may be responsible for this observation. We are unaware of previous reports of divalent metal ion stimulation of inhibition of proteases by a2AP.
Complexes of APC with alAT and PC1 were identified in plasma of patients with disseminated intravascular coagulation (8), in baboons infused with APC (40), and in chimpanzees infused with phospholipid vessels containing Factor Xa (41). In the latter report, additional APC complexes were detected on immunoblots but not identified. We have recently identified APC-a2M and APC-a2AP complexes in plasma of baboons infused with APC (42). Thus, alAT, PCI, a2M, and a2AP each complex in vivo with infused APC, and each inhibitor can function physiologically to neutralize APC.
In summary, divalent metal ions in blood significantly enhance the rate of inhibition of APC by the protease inhibitors aZM and a2AP, and the inhibition of APC by protease inhibitors in blood may adequately account for the clearance and i n vivo half-life of APC.