Membrane Binding by Prothrombin Mediates Its Constrained Presentation to Prothrombinase for Cleavage*

Background: Prothrombin variants lacking membrane binding have probed the contribution of the substrate-membrane interaction in thrombin formation by prothrombinase. Results: Loss of membrane binding yields modest changes in rate but affects the pathway for substrate cleavage. Conclusion: Membrane binding by the substrate constrains the presentation of prothrombin for cleavage by prothrombinase. Significance: New insights into how the action of prothrombinase on prothrombin is regulated. Long-standing dogma proposes a profound contribution of membrane binding by prothrombin in determining the rate at which it is converted to thrombin by prothrombinase. We have examined the action of prothrombinase on full-length prothrombin variants lacking γ-carboxyglutamate modifications (desGla) with impaired membrane binding. We show an unexpectedly modest decrease in the rate of thrombin formation for desGla prothrombin but with a major effect on the pathway for substrate cleavage. Using desGla prothrombin variants in which the individual cleavage sites have been singly rendered uncleavable, we find that loss of membrane binding and other Gla-dependent functions in the substrate leads to a decrease in the rate of cleavage at Arg320 and a surprising increase in the rate of cleavage at Arg271. These compensating effects arise from a loss in the membrane component of exosite-dependent tethering of substrate to prothrombinase and a relaxation in the constrained presentation of the individual cleavage sites for active site docking and catalysis. Loss of constraint is evident as a switch in the pathway for prothrombin cleavage and the intermediate produced but without the expected profound decrease in rate. Extension of these findings to the action of prothrombinase assembled on platelets and endothelial cells on fully carboxylated prothrombin reveals new mechanistic insights into function on physiological membranes. Cell-dependent enzyme function is probably governed by a differential ability to support prothrombin binding and the variable accumulation of intermediates from the two possible pathways of prothrombin activation.

Long-standing dogma proposes a profound contribution of membrane binding by prothrombin in determining the rate at which it is converted to thrombin by prothrombinase. We have examined the action of prothrombinase on full-length prothrombin variants lacking ␥-carboxyglutamate modifications (desGla) with impaired membrane binding. We show an unexpectedly modest decrease in the rate of thrombin formation for desGla prothrombin but with a major effect on the pathway for substrate cleavage. Using desGla prothrombin variants in which the individual cleavage sites have been singly rendered uncleavable, we find that loss of membrane binding and other Gla-dependent functions in the substrate leads to a decrease in the rate of cleavage at Arg 320 and a surprising increase in the rate of cleavage at Arg 271 . These compensating effects arise from a loss in the membrane component of exosite-dependent tethering of substrate to prothrombinase and a relaxation in the constrained presentation of the individual cleavage sites for active site docking and catalysis. Loss of constraint is evident as a switch in the pathway for prothrombin cleavage and the intermediate produced but without the expected profound decrease in rate. Extension of these findings to the action of prothrombinase assembled on platelets and endothelial cells on fully carboxylated prothrombin reveals new mechanistic insights into function on physiological membranes. Cell-dependent enzyme function is probably governed by a differential ability to support prothrombin binding and the variable accumulation of intermediates from the two possible pathways of prothrombin activation.
Thrombin, the pivotal proteinase of blood coagulation, is produced by the proteolytic activation of prothrombin (1). Prothrombinase, the physiological catalyst for this reaction, is an enzyme complex assembled by reversible protein-protein and protein-membrane interactions between the proteinase (factor Xa), its cofactor (factor Va), and membranes containing phosphatidylserine (2). Membrane binding by the cofactor and proteinase, mediated by specific domains, greatly enhances both the kinetics and thermodynamics of prothrombinase assembly (1,2). Reactions between membrane-bound Xa and Va proceed with very high rate constants because of approximation arising from dimensional and orientational restrictions (3). Tight binding interactions, further enhanced by linkage effects, allow the membrane-bound proteins to bind each other at subnanomolar concentrations (4). Prothrombin also binds reversibly to these membranes in a Ca 2ϩ -dependent fashion by virtue of ␥-carboxyglutamate residues at its N terminus (5). Consequently, the reversible interaction between substrate and membranes is also expected to result in its accelerated and oriented delivery to membrane-bound prothrombinase (1,6). These features are considered necessary for normal hemostasis, allowing accelerated thrombin formation on activated platelets or other cells expressing phosphatidylserine on their outer leaflet and localized at the site of vascular damage (7).
The essential role for membrane binding in blood coagulation is obvious from bleeding associated with deficiencies in the vitamin K-dependent reactions required for the post-translational carboxylation of specific glutamates to form ␥-carboxyglutamate (8). Warfarin and its derivatives, which interfere with ␥-carboxylation, are widely used for the therapeutic control of thrombosis and need careful dosing to avoid bleeding (9). However, interference with ␥-carboxylation impacts Ca 2ϩ binding and membrane binding of all of the vitamin K-dependent coagulation proteins (9). In the case of prothrombinase, this would impair membrane binding by Xa as well as prothrombin, thereby impacting the assembly of prothrombinase as well as membrane-dependent substrate delivery. Thus, warfarin effects do not permit incisive inferences regarding the importance of the substrate-membrane interaction in function.
The contribution of membrane binding by prothrombin toward thrombin formation has been extensively investigated (6). Although some controversies linger, it is widely accepted that membrane binding by prothrombin plays an essential role in affecting the rate of thrombin formation. Ideas in the field are dominated by an influential model proposed almost 30 years ago based on co-concentration effects of prothrombinase and prothrombin in a microscopic subspace bordering the membrane surface (10,11). The model predicts a catastrophic decrease in rate, at least by a factor of 3,500, associated with a loss in membrane binding by the substrate (10). The proposed magnitude of functional loss is not fully consistent with kinetic studies done in solution or with various prothrombin fragments lacking the membrane binding domain (12,13). However, some of those interpretations could be compromised by unanticipated effects associated with proteolytic elimination of ϳ30% of the polypeptide structure of prothrombin.
A complexity, insufficiently considered in the foregoing, arises from the fact that the conversion of prothrombin to thrombin requires the action of prothrombinase on two sites in two sequential enzyme-catalyzed reactions (14). For prothrombinase assembled on phospholipid vesicles containing an optimal fraction of phosphatidylserine, the reaction exclusively proceeds by initial cleavage at Arg 320 followed by cleavage at Arg 271 , yielding meizothrombin (mIIa) 3 as an intermediate (for clarification, see Scheme 2) (15,16). The basis for such ordered cleavage, despite the fact that both sites are accessible for proteolysis, lies in the constrained way the substrate is tethered to the enzyme through exosite interactions (17,18). These interactions, between enzymic sites removed from the active site and sites on the substrate distant from the cleavage site, facilitate the constrained presentation of the cleavage sites for docking at the active site of Xa within prothrombinase (14,18,19). Although such ideas have centered on protein-protein contacts, the substrate-membrane interaction in the vicinity of prothrombinase would also constitute a component of exosite binding that might further contribute to the constrained presentation of the substrate to prothrombinase. This idea is supported by previous studies hinting at an altered order of bond cleavage of prothrombin species isolated from the blood of warfarin-treated cows and by studies done in the absence of membranes (13,20).
We have now used newer developments in the understanding of substrate recognition by prothrombinase as a framework to investigate this problem. We employed a series of full-length recombinant prothrombin variants lacking ␥-carboxylation (desGla, dG) to permit an assessment of the contribution of membrane binding by the substrate to all four possible cleavage reactions. Our findings yield surprising insights into how membranes regulate function with bearing on prothrombinase func-tion on physiologic membranes that probably express insufficient amounts of phosphatidylserine for robust prothrombin binding.
Proteins-Prothrombin, factor X, and factor V were isolated from human plasma by established procedures (23,24). Factor Xa and factor Va were purified and quality-controlled following preparative activation of factor X by the purified activator from Russell's viper venom or of factor V by thrombin as before (15,25). Thrombin and prethrombin 2 (P2) were purified following preparative proteolysis of prothrombin as described (26). Recombinant tick anticoagulant peptide (rTAP) was prepared as before (27).
tamate following base hydrolysis (28). Uncarboxylated F12 (dG-F12) was obtained by preparative activation of dG-II S195A by prothrombinase and purified using procedures similar to those described for carboxylated F12. Uncarboxylated prothrombin (dG-II WT , 32 M) in 20 mM Hepes, 0.15 M NaCl, 5 mM CaCl 2 , and 5 M CoCl 2 was preparatively activated with 0.42 M ecarin in the presence of ATA-FPRck (0.16 mM), and the resulting inactivated meizothrombin (dG-mIIa i ) was isolated as before (15). A fraction of dG-mIIa i was reacted with an excess of Alexa 488 maleimide in the presence of 0.1 M NH 2 OH, and the resulting singly labeled fluorescent adduct (dG-mIIa Alexa488 ) was purified as described (15,29). The cDNA encoding Ecarin with Ser 170 replaced with Pro, a His 6 extension at the COOH terminus, and flanked by HindIII and EcoRI sites was synthesized based on the published sequence (30). Digestion with these enzymes allowed for cloning into pcDNA 3.1ϩ (Invitrogen) and subsequent transfection of AV-12 cells to obtain stable clones. Highest producing clones were identified by measurements of prothrombin activation and verified by Western blotting using a mouse anti-His-4 antibody (Qiagen). Ecarin was expressed on a large scale in serumfree medium essentially following procedures described for the prothrombin variants (15). Purification was done by an initial capture step with Q-Sepharose Fast Flow (GE Healthcare), followed by chromatography using Poros HQ-150 (Applied Biosystems) using buffers and gradient conditions described for prothrombin purification (15). The pool containing Ecarin activity was dialyzed into 20 mM Hepes, 10 mM imidazole, 0.15 M NaCl, pH 7.4, and applied to a 1-ml HisTrap FF column (GE Healthcare) charged with Ni 2ϩ . Bound protein was eluted with the same buffer containing 500 mM imidazole, dialyzed into 20 mM Hepes, 0.15 M NaCl, 5 M CoCl 2 , pH 7.4, and stored frozen in aliquots.
Cleavage of Prothrombin Variants-Reaction mixtures (800 l) containing either 5 or 1.  6.4% (w/v) SDS, 32% (v/v) glycerol, 0.04% (w/v) bromphenol blue, 50 mM EDTA, 50 mM dithiothreitol, pH 6.8, supplemented with 36 M FPRck. Following heating at 85°C for 5 min, samples were subject to electrophoresis using 10% Tris-glycine gels (Invitrogen), and protein bands were visualized by staining with Coomassie Brilliant Blue R250 for experiments done at the high substrate concentration or by the Colloidal Blue stain and infrared imaging for the low substrate concentration (34). Quenched samples from the low concentration data set were also visualized following quantitative Western blotting as described previously (34). For measurements of proteolytic activity, aliquots (10 l) were quenched by mixing with 40 l of Assay Buffer containing 1 M rTAP and 50 mM EDTA in place of CaCl 2 . The concentration of proteinase formed was determined from initial velocities of S2238 hydrolysis as described previously (35).
Kinetic Studies-Initial velocity measurements of proteinase formation from dG-II Q271 were determined discontinuously as described previously (15). Reaction mixtures (200 l) containing the indicated concentrations of dG-II Q271 , 30 M PCPS, and 30 nM Va at 25°C were initiated with 0.5 nM Xa. Aliquots (10 l) were removed at 0, 0.5, 1.0, 1.5, 2, and 3 min following initiation and quenched by mixing with 40 l of Assay Buffer containing 1 M rTAP and 50 mM EDTA in place of CaCl 2 . The concentration of proteinase formed in the quenched samples was inferred from the rate of S2238 hydrolysis and its linear dependence on known concentrations of thrombin established with each experiment. Initial velocities of proteinase formation were then determined from the linear appearance of proteinase with time established with at least 4 of the 6 quenched samples. For experiments with alternate substrates or inhibitors, initial velocities with increasing concentrations of dG-II Q271 were determined in the presence of the indicated fixed concentrations of dG-II QQ or dG-II Q320 .
Initial velocities for the conversion of the dG-F12/P2 mixture to thrombin (cleavage at Arg 320 Ј) were determined in the same way with varying concentrations of P2 and the dG-F12 concentration maintained at either 1.2 or 1.5 molar equivalents of P2.
The kinetics of action of prothrombinase on dG-mIIa i (cleavage at Arg 271 Ј) was inferred from the fluorescence increase seen upon conversion of dG-mIIa Alexa488 to thrombin, analogous to the change seen in fluorescein-modified but fully carboxylated mIIa (15). Reaction mixtures (200 l), prepared in black 96-well plates (Corning Inc.), contained 0.1 M dG-mIIa Alexa488 and increasing concentrations of dG-mIIa i to achieve the indicated total concentration of substrate, 30 M PCPS, and 30 nM Va in Assay Buffer. Reactions were initiated by the addition of 1 nM Xa, and product formation was monitored at 25°C in a Spectramax Gemini (Molecular Devices) using EX ϭ 488 nm and monitoring broadband fluorescence with a 510-nm-long pass filter in the emission beam. Initial, steady state velocities of the fluorescence increase were converted to concentration terms using the limits of the progress curves as described (15,29).
Platelets-Platelets were isolated from blood freshly drawn by venipuncture from aspirin-free, volunteer donors following written consent using a protocol approved by the institutional review board. Blood (45 ml) was drawn into 5 ml of anticoagulant composed of 65 mM citric acid, 85 mM Na 3 -citrate, 2% (w/w) dextrose. Platelet-rich plasma was obtained by conservative aspiration of part of the upper layer after centrifugation (1,000 ϫ g, 18 min, 25°C) in a swinging bucket rotor. Aliquots of platelet-rich plasma (5 ml) were applied in parallel to columns of Sepharose 2B-CL300 (2.5 ϫ 10.5 cm, 50 ml) equilibrated in 3.8 mM HEPES, 0.38 mM Na 2 HPO 4 , 137 mM NaCl, 2.68 mM KCl, 0.98 mM MgCl 2 , 5.55 mM dextrose, 0.2% (w/v) BSA, pH 7.3 (Hepes/Tyrode's/BSA). Platelets eluting in the void volume, well separated from the plasma fraction, were pooled and counted using a Hemavet 1500FS (CDC Technologies). Yields were typically ϳ300,000 platelets/l of blood and were essentially free of other cell types. Platelets were maintained at room temperature and used in experiments within ϳ2 h of isolation.
Endothelial Cells-Human umbilical vein endothelial cells (HUVECs; 1-7 passages) were a generous gift of Dr. Long Zheng, Children's Hospital of Philadelphia. The cells were grown to near confluence in 6-well plates using EBM-2 medium (Lonza). Prior to the experiment, HUVECs were washed 3 times with 3 ml of HEPES/Tyrode's/BSA supplemented with 5 mM CaCl 2 . Following washing, 0.5 ml of the same buffer was added to the wells in preparation for thrombin activation.
Prothrombin Cleavage on Cell Membranes-Freshly purified platelets, adjusted to 2 ϫ 10 8 /ml, were supplemented with 1 M stocks of Tris, pH 7.5, and CaCl 2 to achieve final concentrations of 20 and 5 mM, respectively. Platelets were activated by thrombin (10 nM, 3 min) followed by the addition of 12 nM hirudin. Reaction mixtures (800 l, 25°C) were prepared by mixing equal volumes of the activated platelet preparation and a solution prepared in Assay Buffer containing 2.8 M prothrombin variant, 60 M DAPA, and 60 nM Va. Cleavage reactions were initiated by the addition of 0.5 nM Xa, subsampled at the indicated times for SDS-PAGE analysis as above, and analyzed by quantitative Western blotting (34). Washed HUVECs were activated with thrombin (20 nM, 3 min) followed by the addition of 25 nM hirudin. Reaction mixtures (1 ml) were prepared in the wells of the 6-well plate by the addition of 2.8 M prothrombin variant, 60 M DAPA, and 60 nM Va in Assay Buffer (0.5 ml). The activation reaction was initiated by the addition of 0.2 nM Xa and allowed to proceed with rotary shaking (400 rpm, Thermomixer R, Eppendorf) at 25°C. Reaction mixtures were sampled at the indicated times for SDS-PAGE analysis as above and analyzed by quantitative Western blotting (34).
Data Analysis-Concentrations of PCPS and Va were chosen to saturate Xa based on the measured equilibrium constants for prothrombinase assembly (4). The concentration of enzyme was considered equal to the limiting concentration of Xa in each experiment and used to normalize initial velocities.
Quantitative densitometry of stained gels or from quantitative Western blotting was performed as previously validated and described in detail (12,15,16,34). Estimates of the initial velocity from progress curves constructed in this way were obtained by analysis according to the logarithmic approximation (36). Observed steady state kinetic constants were determined by non-linear least squares analysis according to the Henri-Michaelis-Menten equation. Global analysis according to Scheme 1 to obtain intrinsic constants was done with the rapid equilibrium assumption using Dynafit (Biokin) (37). Provided this assumption holds, the kinetics for cleavage at the individual sites in dG-II Q271 and dG-II Q320 can be described by the Henri-Michaelis-Menten equation (18,35). Intrinsic constants for either of the two cleavage events are related to the observed kinetic constant by the following, where K EXO is the cleavage site-independent equilibrium dissociation constant for exosite binding. K m,obs , (V/E) obs , K S* , and k cat would be subscripted with either 271 or 320, depending on the single cleavage site substrate variant used. These relationships were used to compute terms otherwise inaccessible from global analysis according to Scheme 1. All fitted constants are listed Ϯ 95% confidence limits. Errors from fitted constants were propagated for the calculated terms (38). With the exception of the data set obtained for the cleavage of dG-II Q271 or dG-II Q320 by Xa partially saturated by micromolar concentrations of Va in solution, the reported data are representative of two or more experiments performed at a comparable level of detail, frequently with different protein preparations.

RESULTS
Prothrombin Variants-As in previous work, we have employed recombinant variants of prothrombin in which the individual cleavage sites at Arg 271 and Arg 320 have been rendered uncleavable, either singly or in combination, by substitution with Gln (15). Culture of the stable cell lines expressing these variants in the absence of vitamin K yielded fully uncarboxylated protein. Quantitative analysis of ␥-carboxyglutamic acid (Gla) content following base hydrolysis yielded the expected 10 mol of Gla/mol of protein for the proteins expressed in the presence of vitamin K (15). In contrast, a Gla peak was undetectable for the dG variants, providing an upper limit estimate of Յ0.3 Gla/mol of protein. N-terminal sequencing verified correct processing of the propeptide during secretion, and mass spectrometry yielded the expected molecular weight for full-length prothrombin (not shown). A desGla variant (dG-II WT ) produced minimal changes in light scattering when titrated with increasing concentrations of PCPS. In a parallel experiment, the fully carboxylated protein (II WT ) produced a large and saturable change in light scattering intensity, signifying binding (not shown). In accordance with the literature (20), we estimate that the desGla variants of prothrombin bind to PCPS membranes with at least 500-fold weaker affinity than the fully carboxylated protein.
Activation of Prothrombin-The contribution of prothrombin ␥-carboxylation to its function as a substrate for prothrombinase was first assessed by progress curves of thrombin formation (Fig. 1). As expected, prothrombinase yielded thrombin at a lower rate from dG-II WT in comparison with II WT (Fig. 1). However, in contrast to the profound reduction in rate expected based on the literature, we were surprised to find that the initial rate of thrombin formation from dG-II WT was only modestly lower, by a factor of ϳ5 (10).
Progress curves were further analyzed by SDS-PAGE and protein staining (Fig. 2, A and B). The cleavage patterns observed for II WT (Fig. 2A) were consistent with sequential cleavage at Arg 320 to yield mIIa followed by cleavage at Arg 271 to produce thrombin as reported previously (15). The signature features of this cleavage pathway are the transient accumulation of the F12-A fragment, the delayed appearance of F12 following the conversion of mIIa to thrombin and the lack of any detectable P2. In contrast, consumption of dG-II WT , only modestly slower than that of II WT , yielded no obvious evidence for F12-A formation but instead produced P2 in a transient way (Fig. 2B). Thus, the modestly decreased rate of thrombin production from dG-II WT (Fig. 1) obscures major differences in the way the substrate is cleaved by prothrombinase. Nevertheless, proteinase formation, as judged by peptidyl substrate cleavage from either carboxylated or uncarboxylated prothrombin, matches well with the appearance of the thrombin B chain determined by quantitative densitometry (Fig. 1).
Cleavage of the Individual Bonds within Prothrombin-Initial evidence implying altered selectivity for bond cleavage within dG-II WT by prothrombinase was further pursued by studies with carboxylated and uncarboxylated versions of the cleavage site mutants (Fig. 2, C-F). II Q271 was rapidly consumed by prothrombinase from cleavage at Arg 320 to produce mIIa (Fig. 2C). The use of dG-II Q271 yielded much slower cleavage at Arg 320 (Fig. 2D). Cleavage at Arg 271 in II Q320 proceeded very slowly (Fig. 2E), reflecting the preference of prothrombinase for cleavage at Arg 320 in intact prothrombin (15). We were surprised to find that dG-II Q320 was cleaved at a greater rate than II Q320 (Fig. 2F). Although this observation has bearing on the altered cleavage patterns seen with dG-II WT , it also illustrates that loss of membrane binding by the substrate is not uniformly deleterious and unexpectedly produces a gain in a selected aspect of function.
Because F12 from the desGla variants migrates as an indistinct smear, initial studies were conducted at 5 M substrate to facilitate reliable interpretation of SDS-PAGE analyses (Fig. 2). To rule out the possibility that our unexpected findings may reflect a peculiarity of this choice in concentration, we pursued further work at the physiological concentration of prothrombin (1.4 M) using infrared detection of stained gels or by quantitative Western blotting with near-IR fluorescence detection (34). The findings were equivalent, borne out by the quantitative analysis of prothrombin consumption from the two types of measurements (Fig. 3). Loss of membrane binding in II WT only yielded a modest ϳ2.5-fold decrease in the initial rate of prothrombin consumption (Fig. 3A). This modest decrease arose from opposing effects producing a larger decrease in the rate (ϳ15-fold) for cleavage at Arg 320 in dG-II Q271 (Fig. 3B) but an increase in the rate (ϳ4-fold) for cleavage at Arg 271 in dG-II Q320 (Fig. 3C).
Initial rates of prothrombin consumption (Table 1) illustrate that prothrombinase consumes carboxylated prothrombin by preferential cleavage at Arg 320 with a minor contribution from cleavage at the alternate site. In contrast, prothrombinase acts on dG-II WT at a slightly reduced rate but by preferential cleavage at Arg 271 , whereas cleavage at Arg 320 also proceeds at a significant rate. This carboxylation-dependent switch in selectivity, thus far interpreted to reflect the contribution of membrane binding by prothrombin, accounts for the change in  cleavage patterns seen in the action of prothrombinase on II WT and dG-II WT (Fig. 2, A and B).
Altered Cleavage Site Selectivity in desGla Prothrombin-Further kinetic analyses using the initial velocity and rapid equilibrium assumptions were pursued using a model developed in previous studies with carboxylated prothrombin (Scheme 1). This model is rooted in kinetic and equilibrium binding measurements illustrating that equivalent exosite binding interactions are responsible for tethering the substrate to the enzyme regardless of cleavage site (16,39). Active site docking by Arg 271 in exosite-bound II Q320 or Arg 320 in enzymebound II Q271 then occurs in a mutually exclusive way for catalysis at the individual sites (Scheme 1). II QQ is uncleavable because it cannot engage the active site of prothrombinase (39). Alternate substrate studies measured the rate of dG-mIIa formation from varying concentrations of dG-II Q271 in the presence of different fixed concentrations of dG-II Q320 or dG-II QQ (Fig. 4). Global analysis of the data according to Scheme 1 yielded adequate fits and permitted meaningful assessment of K EXO , K S*,320 , K S*,271 , and k cat,320 (Fig. 4). Because the product of dG-II Q320 cleavage is not measured here, k cat,271 was estimated in combination with the initial rates determined from densitometry measurements (Table 1). These parameters reflect the intrinsic constants governing the ability of prothrombinase to discriminate between the two sites within intact but desGla prothrombin ( Table 2).
Comparisons with values previously determined for fully carboxylated prothrombin (Table 2) reveal how impaired membrane binding by desGla prothrombin affects its recognition by prothrombinase. For the desGla variants, exosite binding is ϳ70-fold weaker than for the carboxylated substrate. Membrane binding and possibly other Gla-related functions play a major role in the exosite-dependent tethering of prothrombin to prothrombinase. In fully carboxylated prothrombin, the unimolecular binding constant for active site docking by Arg 320 (K S*,320 ) is ϳ200-fold more favorable than for Arg 271 (K S*,271 ) ( Table 2). Such discrimination is lost in desGla prothrombin, wherein active site docking by Arg 320 and Arg 271  Table 1.

TABLE 1 Normalized rates of prothrombin consumption by prothrombinase
Initial rates of prothrombin consumption were measured from progress curves determined by quantitative densitometry following SDS-PAGE at either 1.4 or 5 M substrate. Initial rates were normalized for differences in the concentration of enzyme (v/E) and are expressed in multiples of the rate observed at 5 M II WT .   Table 2. and/or calculated for the action of prothrombinase at Arg 271 and Arg 320 in otherwise uncleaved desGla prothrombin ( Table  2). Initial velocity studies were also performed to examine the action of prothrombinase at the remaining site in singly cleaved desGla variants (Fig. 5). Steady state kinetic constants for the cleavage at Arg 271 Ј following initial cleavage at Arg 320 were obtained using dG-mIIa as substrate (Fig. 5A). Cleavage at Arg 320 Ј following initial cleavage at Arg 271 was assessed using P2 reconstituted with dG-F12 (Fig. 5B). Equivalent initial velocities obtained at two ratios of dG-F12/P2 support the contention that P2 was essentially saturated with dG-F12 at all concentrations used (Fig. 5B). Observed steady state kinetic constants ( Table 2) with representative values listed in Scheme 2 provide the basis for the further consideration of the action of prothrombinase on desGla prothrombin. Initial rates calculated at the physiological concentration of prothrombin reveal that desGla substrate forms are cleaved more slowly (by at least a factor of 10) than the equivalent carboxylated species for three of the four possible half-reactions (Scheme 2). The exception is cleavage at Arg 271 in intact prothrombin that is increased for desGla substrate in accordance with the rates detailed for dG-II Q320 (Table 1). Thus, asymme-try in the action of prothrombinase at Arg 271 in intact prothrombin versus mIIa in the carboxylated forms is altered for the desGla substrate (15). Consequently, ϳ80% of the rate of consumption of desGla prothrombin is expected to result in the formation of dG-F12/P2 and ϳ20% from the formation of dG-mIIa. In accordance with observations ( Fig. 2), minor amounts of dG-mIIa are predicted to accumulate as an intermediate, whereas abundant amounts of P2 are expected as a long lasting intermediate, given its slower conversion to thrombin. Simulations with these steady state kinetic constants (not shown) indicate that the initial rate of proteinase formation (mIIa ϩ IIa) through the initial cleavage of desGla prothrombin at Arg 320 would be ϳ6-fold greater than proteinase formation via the P2 pathway. This point highlights the dangers inherent in inferring the predominant pathway for product formation in such systems solely on the basis of the amounts of intermediate observed.
Correlation with Membrane Binding-A diagnostic difference in the action of prothrombinase on desGla versus fully carboxylated substrate lies in the relative rates for cleavage at the two sites within intact prothrombin (Scheme 2). In the case of carboxylated substrate, cleavage at Arg 320 is ϳ30-fold greater than cleavage at Arg 271 . For desGla prothrombin, cleavage at Arg 271 is ϳ4-fold greater than cleavage at Arg 320 . At issue is whether such effects can be wholly ascribed to a loss in membrane binding by the desGla variants rather than other ␥-carboxyglutamate-dependent effects on the substrate.
To address this uncertainty, we determined the relative rates of cleavage at the two sites within fully carboxylated prothrombin but in the absence of membranes. Progress curves for the cleavage of II Q320 and II Q271 in the absence of added membranes were constructed by quantitative blotting following the addition of 1 nM Xa partially saturated (ϳ50%) with 2 M Va (Fig. 6A). Comparable initial rates were obtained for cleavage at the individual sites in otherwise fully carboxylated intact prothrombin (Fig. 6A). Although the ϳ30-fold discrimination seen for cleavage at Arg 320 relative to Arg 271 seen for the action of membrane-bound prothrombinase on carboxylated substrate was abolished in the absence of membranes, this reaction system only partly replicated the findings seen with the desGla prothrombin variants (Table 1). Cleavage at Arg 271 in II Q320 SCHEME 2. Kinetic constants for the four half-reactions of prothrombin activation. Observed steady state kinetic constants for the action of prothrombinase (E) on desGla substrate species are listed for each half-reaction. Relative rates for the carboxylated (Gla) versus uncarboxylated (desGla) substrate forms are those calculated at 1.4 M substrate.  Table 2.
yielded v/E equivalent to that observed for the action of membrane-bound prothrombinase on dG-II Q320 . Therefore, the gain in function seen for cleavage at Arg 271 in intact desGla prothrombin and only a fraction of the loss in function seen for cleavage at Arg 320 can be replicated with fully carboxylated prothrombin but in the absence of membranes. Because essentially equivalent results were obtained using dG-II Q320 and dG-II 271 (Fig. 6A), the data implicate subtle differences between membrane-assembled prothrombinase and Xa saturated with Va in solution as the reason for the discrepancy. The findings are consistent with the conclusion that membrane binding by the substrate contributes in a major way to its constrained presentation to prothrombinase and the subsequent ability of the enzyme to discriminate between the two cleavage sites.
Cleavage Site Selectivity on Physiological Membranes-Relative rates of cleavage of fully carboxylated II Q271 and II Q320 were employed to investigate how the substrate-membrane interaction may control prothrombinase function on physiological surfaces. This comparative approach allows secure conclusions without knowledge of the concentration of productively bound prothrombinase. Studies with activated platelets revealed that cleavage at either Arg 320 or Arg 271 in intact prothrombin proceeded at equivalent rates (Fig. 6B). This indicates that prothrombinase assembled on the platelet surface does not discriminate between the two cleavage sites in prothrombin, much as Xa bound to Va in the absence of membranes (Fig. 6A). Thus, on the platelet surface, prothrombin consumption is expected to proceed approximately equally through the P2 and mIIa pathways. In contrast, when assembled on activated HUVECs, prothrombinase acted on Arg 320 in II Q271 ϳ2-fold faster than it cleaved Arg 271 in II Q320 (Fig. 6C). For these cells, cleavage of prothrombin would partition between the formation of mIIa and P2 in a ratio of 2:1. The findings are consistent with a greater contribution from membrane binding by prothrombin on HUVECs in comparison with platelets. The exact contribution of the two possible pathways to thrombin formation would require knowledge of the kinetic constants for all four cleavage reactions. However, the implications are that variable contributions of prothrombin cleavage via both possible pathways will determine the amount of intermediates observed in a cell type-dependent fashion. Such observations might be expected to correlate with the variable ability of the membrane surfaces to support prothrombin binding.
These predictions were tested by quantitative Western blotting to analyze II WT cleavage by prothrombinase assembled on platelets or HUVECs previously activated with thrombin. With platelets, bands corresponding to both P2 and mIIa (F12-A) accumulated transiently (Fig. 7A). In agreement with prior work, only a trace band arising from mIIa (F12-A) was evident (Fig. 7A) (40). In contrast, bands corresponding to mIIa and P2 were more prominent, with HUVECs yielding approximately equal peak concentrations of F12-A and P2 (Fig. 7B). In order to relate these findings to the limited ability of these activated cells to support prothrombin binding, we pursued studies with II WT and synthetic vesicles containing 2.5% (w/w) phosphatidylserine. The intent was to replicate the low amounts of phosphatidylserine expected in the outer leaflet of these cells, which would facilitate the high affinity interactions required for prothrombinase assembly but significantly impact prothrombin binding and the density of bound substrate (7). Accordingly, bands corresponding to the transient formation of both P2 and mIIa were observed with these synthetic membranes approximating the pattern of cleavage seen with the activated cells (Fig.  7C). Taken together with the predominant cleavage of II WT via the formation of mIIa seen with vesicles containing 25% (w/w) phosphatidylserine ( Fig. 2A), the data present a consistent picture implicating a limited but variable ability of physiological membranes to support substrate binding as a major determinant of the pathway for prothrombin activation.

DISCUSSION
Our strategy of using appropriate recombinant prothrombin derivatives to probe the details of the newly expanded model for the action of prothrombinase on prothrombin now sheds new light on a longstanding problem in coagulation enzymology. Prevailing dogma predicts a catastrophic decrease in the rate of action of prothrombinase on prothrombin with impaired membrane binding. Instead, by the use of full-length prothrombin variants lacking Gla modifications, we find only modest changes in rate attributable to a loss in membrane binding by the substrate at its physiological concentration. These modest decreases belie major mechanistic changes in the way prothrombin is recognized by prothrombinase. The findings reveal unexpected insights into how the interaction between the substrate and membrane in the vicinity of prothrombinase is a major determinant of the constrained presentation of prothrombin to the membrane-assembled enzyme complex.
Membrane binding by the substrate, lost in the desGla variants, contributes in a prominent way to the initial exositedriven tethering of the substrate to prothrombinase (Scheme 1). This point is evident from a ϳ70-fold increase in K EXO for the desGla variants. High selectivity (ϳ200-fold) for subsequent active site docking by Arg 320 over that of Arg 271 seen with carboxylated prothrombin is lost in the desGla variants. For these species with impaired membrane binding, the unimolecular binding constants for active site docking indicate that two cleavage sites can engage the active site of Xa within prothrombinase with approximately equal probability. Scrambling of selectivity for the two cleavage sites is consistent with the loss of a subset of exosite binding interactions that otherwise position bound substrate in a constrained way for preferential active site docking by Arg 320 . Coupled with the increased intrinsic k cat,271 , these altered constraints in exosite tethering of the substrate provide the mechanistic basis for the qualitative change in cleavage pattern observed in the action of prothrombinase on II WT versus dG-II WT .
We intentionally qualify inferences of the contribution of the substrate-membrane interaction to function from studies with the desGla variants. Altered function of these variants could arise from a loss of additional Gla-dependent functions beyond simply a loss in membrane binding. In support of this possibility are the findings with fully carboxylated prothrombin variants and Xa partially saturated with Va in solution that do not fully replicate the observations made with desGla variants and membrane-assembled prothrombinase. However, the fact that equivalent rates of consumption were observed with II Q271 , II Q320 , dG-II Q271 , and dG-II Q320 in the absence of membranes suggests that this small discrepancy most likely lies in differences in the properties of Xa saturated with Va in solution relative to the membrane-assembled enzyme. A second concern is reflected by the previously documented interaction between a peptide derived from the prothrombin Gla domain and factor Va (41). Again, compromised interactions between desGla prothrombin variants and Va are unlikely to be a major source of the present findings based on the equivalent rates of consumption of the fully carboxylated and desGla variants in the absence of membranes (Fig. 7A). However, the need for cautious interpretation is suggested by the obvious effects of the loss of Gla on the intrinsic k cat for cleavage at Arg 271 . It remains uncertain whether this solely arises from alterations in the constrained way that the exosite-tethered substrate is presented to prothrombinase.
Qualifications notwithstanding, our findings, ranging from studies with desGla prothrombin variants to studies with fully carboxylated forms in the absence of membranes and with prothrombinase assembled on physiological surfaces or on synthetic membranes with limiting amounts of phosphatidylserine, are internally consistent. They portray a unifying picture wherein membrane binding plays an essential role in dictating the presentation of substrate to prothrombinase. Altered interactions that affect these constraints are evident as changes in the intermediates observed and an apparent change in the pathway for prothrombin cleavage. These ideas highlight, both empirically and conceptually, the dangers in interpreting the major pathway for thrombin formation from the relative abundance, or lack thereof, of the two intermediate species (40,42). They also probably lie at the heart of some of the variable results and recent controversies in the field (43)(44)(45).
Recent studies have proposed a fundamental difference in the molecular architecture of prothrombinase assembled on platelets relative to synthetic membranes based on the formation of prethrombin 2 as an intermediate and the lack of observable intermediates in the fluid phase (42). An adequate kinetic explanation for these findings as well as justification for the major pathway for thrombin formation will need to await the determination of steady state kinetic constants of each of the four possible cleavage reactions. However, our results point to the inefficient binding of prothrombin to activated platelet membranes as a parsimonious mechanistic explanation. Parallels in the selectivity of prothrombinase for prothrombin in the absence of membranes and the ability of low phosphatidylserine-containing vesicles to replicate the findings of prothrombin cleavage seen on physiological membranes lend support to this contention. Mechanistic issues aside, the findings with HUVECs and platelets illustrate that different cell types may variably support the formation of mIIa during thrombin formation, perhaps reflecting a differential ability to support prothrombin binding (46 -48). Given its zymogen-like character and skewed sub-strate specificity of mIIa for the anticoagulant activities of thrombin, this may have bearing on the spatial regulation of coagulation (34).
Our findings contrast with prevailing dogma associating impaired membrane binding by prothrombin with a profound loss in the rate of thrombin formation (10,11). Indeed, when considered in the context of the four possible reactions of prothrombin activation, the desGla substrate variants exhibit a very modest decrease in rate for three of the four cleavage reactions. The increased rate observed for one of the steps even more surprisingly illustrates that loss of Gla-dependent functions is not uniformly deleterious to the function of prothrombin as a substrate for prothrombinase. Clearly, bleeding associated with incorrect dosing with warfarin cannot be ascribed to the faulty function of desGla prothrombin as a substrate. However, the reduced amounts of anticoagulant-specific mIIa produced as an intermediate during the activation of desGla prothrombin could have bearing on warfarin-induced thrombosis (9).
It could also be argued that prothrombin completely devoid of Gla is not a good facsimile of partially carboxylated zymogen forms expected in the blood of patients being therapeutically treated with warfarin. This argument is weak because an average Gla content of 3 mol/mol of protein seen in the non-membrane binding fraction of prothrombin from treated individuals reveals nothing of the fractional distribution of the various prothrombin isoforms with a Gla content varying between 0 and 10 (49). It should be noted that prothrombin devoid of Gla has been isolated from the blood of warfarin-treated cows (20). In keeping with the cooperative nature of Ca 2ϩ binding by the Gla residues, mixtures of bovine prothrombin isoforms lacking 3-4 of the full Gla complement of 10 were seen to exhibit greatly impaired membrane binding (20).
A recent structure of a variant of prothrombin lacking residues 1-44 encompassing the Gla domain from the Di Cera laboratory has provided evidence for flexibility in the linker between the fragment 1 and fragment 2 domains as well as the disorder in the region surrounding the Arg 271 cleavage site (50). Such flexibility may provide an explanation for how the desGla substrate tethered to prothrombinase, in the absence of additional constraints imposed by membrane binding, may allow active site docking of two distant sites with approximately equal probability. However, the associated claim that Arg 271 is solvent-accessible while Arg 320 is buried has yielded the major conclusion that prothrombinase must cleave first at Arg 271 before the Arg 320 site becomes exposed (50). This conclusion regarding the molecular mechanism of prothrombin activation cannot be reconciled with a large body of evidence regarding the cleavage of II WT by prothrombinase on synthetic PCPS membranes (15,17,43,51,52). The present work further rules out the relevance of such a conclusion for prothrombinase function on natural membranes (platelets or HUVECs), in solution, or even on its action on desGla-II WT . We are nonplussed by the obvious disparity between the structure-based proposal and the existing literature, particularly because the authors cite a paper from 1872 yet fail to acknowledge a broad swath of contemporary work in the field that runs counter to their claim (15,17,43,(51)(52)(53)(54).
In summary, our studies with desGla prothrombin variants provide surprising insights into a longstanding problem in coagulation enzymology. The findings point to a prominent role for the substrate-membrane interaction in mediating exosite-dependent binding to prothrombinase and the constrained presentation of the substrate for cleavage. Our findings bear on how the varied ability of physiological membranes to affect such constrained presentation of prothrombin might underlie the regulation of the pathway for prothrombin cleavage and the intermediate produced. Whether such differential proportioning via the formation of mIIa relative to the zymogen P2 has a significant regulatory role remains to be established.