Chapter 114: Control of Coagulation Reactions

Although decades have elapsed since the elaboration of the cascade model^1,2 for blood coagulation (see Chap. 113, Fig. 113–27), the basic outline of sequential conversions of protease zymogens to active serineproteases is still useful, albeit with important modifications (see Chap. 113, Fig. 113–28), to represent blood coagulation reactions. The major conceptual advances for procoagulant pathways in the past two decades emphasize both positive and negative feedback reactions affecting thrombin generation as depicted in Fig. 114–1.

In positive feedback reactions, procoagulant thrombin activates platelets and factors V, VIII, and XI (Chap. 113).^3,4,5 Small amounts of thrombin can be generated by trace amounts of tissue factor via the extrinsic pathway. Subsequently, thrombin can activate factors XI, VIII, and V, thereby stimulating each of the steps in the intrinsic pathway, thereby amplifying thrombin generation (see Fig. 114–1).

In negative feedback reactions, anticoagulant activated protein C (APC) that is generated on endothelial cell surfaces^6,7,8 (Fig. 114–2) downregulates coagulation (see Figs. 114–1 and 114–3). Furthermore, APC can exert direct cytoprotective effects on cells via reactions that involve certain receptors, including endothelial cell protein C receptor (EPCR) and protease-activated receptor-1 (PAR-1) (Fig. 114–4), PAR-3, integrin CD11b/CD18, and possibly apolipoprotein E receptor 2 (apoER2).^7,8 APC’s cytoprotective effects include antiinflammatory and antiapoptotic activities, as well as alterations of gene-expression profiles and stabilization of endothelial barriers (see “Activated Protein C Activities” below). Because inflammation, apoptosis, and vascular barrier breakdown contribute significantly to reactions that promote thrombin generation, such direct cytoprotective effects of APC on cells indirectly downregulate thrombin generation.^7,8

For APC generation by the protein C cellular pathway, binding of thrombin to thrombomodulin converts the bound thrombin from a procoagulant enzyme to an anticoagulant enzyme that converts the protein C zymogen to an anticoagulant serine protease, APC (see Figs. 114–1 and 114–2). This surface-dependent reaction is enhanced by the EPCR that binds protein C.^6,9,10 With the aid of its nonenzymatic cofactor, protein S, as well as other potential lipid and protein cofactors, APC inactivates factors Va and VIIIa by highly selective proteolysis, yielding inactive (i) cofactors, that is, factors V_i and VIII_i (see Fig. 114–3 and Chap. 113, Figs. 113–11 and 113–13). Protein S also can directly inhibit factors VIIIa, Xa, and Va.^11,12,13 Thus, APC and protein S inhibit multiple steps in the intrinsic coagulation pathway.

At each step in the coagulation pathways, each clotting protease can be inhibited by one or more plasma protease inhibitors in reactions stimulated by negatively charged glycosaminoglycans such as heparan sulfate or heparin (see “Inhibition of Coagulation Proteases by Protease Inhibitors” below). Given the highly nonlinear nature of the coagulation pathways with both positive and negative feedback reactions, synergy between the protein C pathway and plasma protease inhibitors is important for regulating thrombin generation.

There is continuous activation of coagulation factors at a basal physiologic low level. Plasma from all normal subjects contains circulating active enzymes, factor VIIa,¹⁴ and APC,¹⁵ as well as various polypeptide fragments generated by the action of clotting proteases, namely fibrinopeptides,^16,17 prothrombin fragment 1+2,¹⁸ and activation peptides for factors IX and X.^19,20 The presence of multiple clotting factors that require positive feedback activation (e.g., factors V, VIII, XI, and VII) imparts special threshold properties to the blood coagulation pathways, making the coagulant response nonlinearly responsive to stimuli. Theoretical analysis of blood coagulation as a threshold system suggests there can be an all-or-none response to various levels of stimulation, depending on the ensemble of activating and inhibitory reactions that defines upregulation and downregulation of thrombin generation.^21,22 The coagulation system is active, but idling, and is poised for extensive and explosive generation of thrombin. Because of synergy among various cellular and humoral anticoagulant mechanisms that establish a threshold system, the presence of multiple coagulation inhibitors with complementary modes of action prevents massive thrombin generation in the absence of a substantial procoagulant stimulus.

Evidence for the physiologic importance of specific factors for controlling coagulation reactions comes from clinical observations and animal model studies. Major identified genetic risk factors for venous thrombosis involve protein structural defects in factor V, protein C, protein S, and antithrombin (Chap. 130). There are also gene regulatory defects associated with thrombotic disease, such as the G20210A polymorphism in the prothrombin gene that causes elevated levels of prothrombin, and defects in protein C gene regulatory elements that decrease the expression of protein C. Deficiencies of thrombomodulin might also be associated with increased risk of arterial thrombosis. Association of hereditary abnormalities of EPCR with increased risks of thrombosis has been suggested, but this remains somewhat controversial.

Figure 114–5 is a schematic of the structures of protein C, protein S, thrombomodulin, and EPCR. These proteins contain multiple domains, each of which may mediate different molecular functions. Values for the molecular weight, normal plasma concentration, chromosomal location, and gene structures of these factors are given in Table 114–1. Factors Va and VIIIa, as substrates of APC, are also participants in the reactions of the anticoagulant protein C pathway. Moreover, factor V, but not factor V Leiden, appears to act as an APC cofactor for the inactivation of factor VIIIa (see “Factor V as Activated Protein C Cofactor” below).²³

PROTEIN C

In 1976, Stenflo designated a bovine plasma vitamin K–dependent protein that eluted in the third peak (peak C) from an anion exchange column as bovine “protein C.”²⁴ Protein C, actually previously described as the anticoagulant factor autoprothrombin II-A,²⁵ is a plasma serine protease zymogen that can be converted to an active serine protease by the action of thrombin.

Protein C is synthesized in the liver as a polypeptide precursor of 461 residues, with a prepropeptide of 42 amino acids that contains the signal for carboxylation of Glu residues by a carboxylase that forms nine γ-carboxyglutamic acid (GLA) residues and secretion of the mature protein.^26,27,28 The mature glycoprotein of Mr 62,000 contains 419 residues (see Chap. 113, Fig. 113–1 and Fig. 114–5) and N-linked carbohydrate, and the majority of the secreted protein C molecules are cleaved by a furin-like endoprotease that releases Lys156-Arg157 and generates a two-chain zymogen that circulates in plasma at 65 nM (4 mcg/mL).²⁹ The heavy and light chains of plasma protein C are covalently linked by a disulfide bond that keeps the serine protease globular domain (residues 170 to 419) covalently tethered to the N-terminal string of three domains, the GLA domain and the epidermal growth factor (EGF)-like domains EGF1 and EGF2.^{26,27,28,29,30}

The GLA domain of protein C (residues 1 to 42) and APC is important for a number of functions, including binding to phospholipid-containing membranes (see Chap. 113, Fig. 113–3), thrombomodulin, and EPCR; thus, incomplete carboxylation impairs the functional anticoagulant activity of APC.^31,32,33 The two EGF modules in the light chain may contribute to interactions of APC with protein S and of protein C with thrombomodulin.

The serine protease domain of protein C is homologous to other trypsin-like proteases, and three-dimensional modeling³⁴ and X-ray crystallographic structures³⁰ reflect the structural similarity of APC to members of the serine protease family of which chymotrypsin is the prototype (see Mather and colleagues³⁰ for conversion of protein C to chymotrypsin numbering). APC’s trypsin-like protease domain exerts its anticoagulant activity by highly specific interactions with factors Va and VIIIa followed by cleavage at only two Arg-containing peptide bonds in factors Va and VIIIa (see “Factors Va and VIIIa as Substrates for Activated Protein C” below). These stereo-specific interactions involve both the APC enzymatic active site region and a number of APC residues that are termed exosites because they are not located in the immediate vicinity of APC’s enzymatic active site. Such APC exosites are essential for specific recognition of the macromolecular substrates, factors Va and VIIIa as well as recognition of cellular APC receptors.^35–44

Protein C and Activated Protein C Therapy

Purified plasma protein C concentrate (Ceprotin) is FDA-approved for treating protein C–deficient patients.⁴⁵ Recombinant APC (Xigris) reduced all-cause 28-day mortality in severe sepsis adult patients in the PROWESS phase III trial in 2001 and was FDA-approved for this indication.⁴⁶ This successful therapy of adult severe sepsis using APC followed preclinical antithrombotic and sepsis studies in baboons.^47,48 However, a decade later, in the PROWESS-SHOCK trial, recombinant APC did not reduce mortality in adult severe sepsis patients,^49,50,51 and Xigris was withdrawn from the market. Animal injury model studies have also shed light on in vivo mechanisms for APC beneficial effects in many preclinical injury models and in many models the pharmacologic benefits of APC appear to be independent of APC’s anticoagulant actions (see “Activated Protein C Direct Cellular Activities” below).^7,8

Protein C Gene

The protein C gene, comprising nine exons and eight introns, is located on chromosome 2q14–21 and spans 11 kb (see Chap. 113, Fig. 113–10, and Table 114–1).⁵² The protein C gene is homologous to the genes for factors VII, IX, and X (Chap. 113).

Protein C Mutations

Hereditary protein C deficiency associated with thrombosis is caused by numerous mutations (see protein C mutation databases).^53,54 Based on three-dimensional structures of the protein C, the structural basis for hereditary protein C defects has been rationalized.^34,55,56 Most mutations that cause type I protein C deficiency, characterized by parallel reductions in activity and antigen, involve amino acid residues that form the hydrophobic cores of the two folded globulin-like domains that are characteristic of serine proteases. These mutations destabilize either the process or the product of protein folding, and they result in unstable molecules that are poorly secreted and/or exhibit a very short circulatory half-life. In contrast, most mutations that cause type II defects (reduced anticoagulant or enzymatic activity but normal antigen levels), that is, circulating dysfunctional molecules, involve polar surface residues that do not affect polypeptide folding or thermodynamic stability; these polar residues presumably are involved in protein–protein interactions important for expression of anticoagulant activity.

Rare variants derived from a founder’s mutation appear to be distinctive for races. Two protein C mutations, Arg147Trp and a Lys150 deletion, are significant venous thrombosis risk factors in Chinese but not in Americans of European descent or Japanese.^57,58,59

Severe protein C deficiency as a consequence of homozygous knockout of the mouse protein C gene showed a similar phenotype as severe human protein C deficiency (Chap. 130), with perinatal consumptive coagulopathy in the brain and liver and either death or massive thrombosis that occurred either intrauterine or shortly after birth.⁶⁰

PROTEIN S

Plasma “protein S,” which was named in honor of Seattle, the city of its discovery, is a vitamin K–dependent glycoprotein⁶¹ that is synthesized by hepatocytes, neuroblastoma cells, kidney cells, testis, megakaryocytes, and endothelial cells, and is also found in platelet α-granules.⁶²

Protein S is synthesized as a precursor protein of 676 amino acids, which gives rise to a mature secreted single-chain glycoprotein of 635 residues with three N-linked carbohydrate side chains (see Fig. 114–5).^63,64 Eleven GLA residues in the N-terminal region of mature protein S contribute to Ca²⁺-mediated binding of the protein to phospholipid membranes. The thrombin-sensitive region, residues 47 to 72, follows the GLA-domain (see Fig. 114–5).

The C-terminal region of protein S, residues 270 to 635, the sex hormone–binding globulin-like (SHBG) region contains binding sites for C4b-binding protein (see “Activated Protein C–Independent Anticoagulant Activity of Protein S” below)⁶⁵ and for factor V, as well as factor Va.^66,67 Protein S, like the homologous gas6, also binds to receptor tyrosine kinases, for example, Axl, and initiates cell signaling, and the SHBG region binds the receptor.⁶⁸ Thus, for the expression of its multiple activities, different domains of protein S exhibit a number of different binding sites for different proteins.

Protein S Gene

The protein S gene, comprising 15 exons and 14 introns, is located on chromosome 3p11.1–11.2 and spans 80 kb (see Fig. 113–16 and Table 114–1).^69,70 The protein S gene has limited homology with other genes for vitamin K–dependent factors in the GLA and EGF domains and notable homology of the region coding for residues 240 to 635 with genes of the SHBG family. Humans contain a protein S pseudogene that contains several stop codons and is not translated and that is located very near the normal protein S gene on chromosome 3.

Protein S Mutations

The molecular basis for hereditary protein S deficiency associated with venous thrombosis (Chap. 130) is linked to more than 100 different mutations.⁷¹ A protein S polymorphism that is strongly linked to risk for venous thrombosis in Japanese subjects is known as protein S Tokushima. It involves K155E, which ablates APC-cofactor activity.^57,72 But the K155E apparently is not present in Americans of European ancestry or Chinese populations, thus mirroring the presence of factor V Leiden and prothrombin G20210A that are risk factors in Americans of European decent but not in the Japanese, Chinese, or Americans of African descent populations.⁷³ Another single nucleotide polymorphism present in approximately 1 percent of Americans of European descent is S460P, which is designated protein S Heerlen; it results in absence of N-linked carbohydrate on Asn458 but has no accepted significant functional consequence.⁷⁴

THROMBOMODULIN

Thrombomodulin was discovered as an endothelial cell surface receptor that binds protein C and thrombin, thereby accelerating protein C activation.^75,76,77,78 Binding of thrombin to thrombomodulin converts thrombin from a procoagulant enzyme to an anticoagulant enzyme because thrombomodulin-bound thrombin loses its normal ability to clot fibrinogen or activate platelets.^79,80 Thrombomodulin is a multidomain transmembrane protein comprising an N-terminal lectin-like domain, six EGF domains, a Ser/Thr-rich region, a single membrane-spanning sequence, and an intracellular C-terminal tail (see Fig. 114–5).^{5,6,7,8,75–83} EGF domains 4, 5, and 6 are essential for activation of protein C, with the latter two domains binding thrombin and the first domain binding protein C. The mature protein has 557-amino-acid residues and variable amounts of N- and O-linked carbohydrate modifications that cause variability in molecular size. Glycosaminoglycans, notably chondroitin sulfate, covalently attached to the Ser/Thr-rich region, contribute to the functional properties of thrombomodulin by enhancing either protein C activation by thrombin or by accelerating neutralization of thrombin by protease inhibitors. Modulation of the substrate specificity of thrombin by thrombomodulin involves conformational changes in thrombin caused by binding of thrombomodulin.

Low levels of soluble thrombomodulin circulate in plasma, presumably as a result of limited proteolysis of the protein near its transmembrane cell surface anchor. The functional significance of circulating thrombomodulin is unknown, although variations in its plasma level arise in different clinical conditions.

Recombinant soluble thrombomodulin has been developed for its potential therapeutic value for disseminated intravascular coagulation and has been approved for this indication in Japan.^77,84

Thrombomodulin Gene

The thrombomodulin gene, which lacks introns, is located on chromosome 20p11.2 and spans 3.7 kb (see Chap. 113, Fig. 113–19, Fig. 114–5, and Table 114–1).^77,82 Deletion of the thrombomodulin gene in mice is embryonically lethal.⁸⁵ Downregulation of thrombomodulin gene expression is promoted by a variety of inflammatory agents, including endotoxin, interleukin-1, and tumor necrosis factor (TNF)-α, whereas its expression is upregulated by retinoic acid.^{5,6,7,8,86,87} Generally, thrombomodulin is a key member among the counterbalancing factors that contribute to inflammation, thrombin generation, and coagulation in the endothelium.

Thrombomodulin Mutations

Thrombomodulin mutations are well documented in atypical hemolytic uremic syndrome patients,^78,88 and they may also be associated with an increased risk of arterial thrombosis and myocardial infarction. In contrast, there is less supportive data for association with risk for venous thrombosis (Chap. 130).^89,90,91,92 Atypical hemolytic uremic syndrome is strongly linked to excessive complement activation, and thrombomodulin’s lectin-like domain inhibits complement activation.^77,93 Furthermore, besides promoting protein C activation by thrombin, thrombomodulin also supports activation of the carboxypeptidase, also known as thrombin-activatable fibrinolysis inhibitor (TAFI) that is a potent inactivator of bradykinin and of the activated complement components, C3a and C5a.^94,95,96

ENDOTHELIAL PROTEIN C RECEPTOR

EPCR binds both protein C and APC with similar affinities through their GLA domains and mediates multiple activities of this zymogen or its activated protease, APC.^{9,10,33,86,97–107} The mature EPCR glycoprotein contains 221-amino-acid residues and N-linked carbohydrate, giving an Mr of 46,000. EPCR is an integral membrane protein that is homologous to CD1/major histocompatibility complex class I molecules. The N-terminus is part of an extracellular domain, which is connected to a single transmembrane sequence that is followed by a short Arg-Arg-Cys-COOH cytoplasmic tail (see Fig. 114–5). The cytoplasmic tail can be palmitoylated, and this modification may help localize EPCR to certain lipid rafts or caveolae. The three-dimensional structure of EPCR determined by X-ray crystallography or inferred by molecular modeling established that the GLA domain of protein C and APC binds to EPCR.^107,108 EPCR on endothelial surfaces enhances by greater than fivefold the rate of activation of protein C by thrombin–thrombomodulin (see Fig. 114–2). EPCR is also required for the cytoprotective activities of APC by promoting the cleavage of PAR-1 by APC to induce cell-signaling pathways (see Fig. 114–4). Notably, the cytoprotective actions of APC are completely independent of its anticoagulant activity and are based on cell signaling actions (see “Activated Protein C Direct Cellular Activities” below).^{7,8,109,110,111}

The presence of functional EPCR on the cell surface is regulated by two mechanisms, namely generation of EPCR and clearance of EPCR. Inflammatory mediators induce EPCR ectodomain shedding from the endothelial cell surface by metalloproteinase TNF-α converting enzyme (known as “TACE”).¹¹² The soluble EPCR ectodomain is found in normal human plasma at 100 ng/mL; however, carriers of the H3 EPCR haplotype that includes a Ser219Gly polymorphism (rs867186) have threefold higher soluble EPCR plasma levels.¹¹³ Higher plasma levels of soluble EPCR are found in patients with disseminated intravascular coagulation or systemic lupus erythematosus, although plasma EPCR levels are not correlated with pathology-related alterations in circulating thrombomodulin levels.¹¹⁴ Soluble EPCR binds the protein C and APC via their GLA domains with an affinity similar to the membrane-bound receptor. Because binding of the APC GLA domain to negatively charged phospholipid membranes is required for its anticoagulant activity, soluble EPCR at relatively high levels in purified reaction mixtures inhibits the anticoagulant action of APC against factor Va, although it does not block the reaction of APC with protease inhibitors.^10,86,99,115

The EPCR crystal structure surprisingly revealed a single phospholipid molecule bound in a surface groove on the protein.¹⁰⁷ Secreted phospholipase A₂ group V can modify the lipid in EPCR and cause EPCR to lose its ability to bind protein C and APC.^116,117 The presence of functional EPCR on the cell surface has important implications for thrombotic and inflammatory vascular disease because EPCR inactivation in vivo increases susceptibility to thrombotic and inflammatory diseases.^10,86,97

The physiologic requirement for EPCR in mice was established by the embryonic lethality observed for knockout of the murine EPCR gene.¹¹⁸

EPCR has functionally important interactions with multiple molecules beyond protein C and APC. It binds factor VII and factor VIIa.¹⁰ Furthermore, EPCR was recently implicated to play a potentially important role in the pathogenesis of severe malaria.^{119,120,121,122}

Endothelial Protein C Receptor Gene

The EPCR gene, comprising four exons and three introns, is located on human chromosome 20q11.2 and spans 6 kb (see Table 114–1).¹²³

PROTEASE-ACTIVATED RECEPTOR-1

PAR-1, discovered as a high-affinity human platelet receptor for thrombin,¹²⁴ is the prototype of a four-member subfamily of G-protein–coupled receptors that share an unusual mechanism of activation, namely activation by proteases.^{124,125,126,127,128,129,130} Each PAR contains seven transmembrane helical domains and an extracellular N-terminal tail that is cleaved by an activating protease such that the newly generated aminoterminus is a tethered ligand that triggers activation of the coupled G-protein. Human platelets employ PAR-1 and PAR-4 for activation by thrombin whereas, curiously, murine platelets that are devoid of PAR-1 require PAR-3 and PAR-4 for thrombin’s normal effects.^125,131 PAR-1 is activated by various plasma proteases^{132,133,134,135} and is generally required for APC’s cytoprotective activities (see “Cellular Receptors for Physiologic Effects of Activated Protein C on Cells” below).^7,8,110,111

Protease-Activated Receptor-1 Gene

The PAR-1 gene contains only two introns, is located on chromosome 5q13, and spans 25 kb (see Table 114–1).¹²⁵ Much is known about many factors that can either upregulate or downregulate the PAR-1 gene.^{124,125,126,127,128,129,130}

Protein C is activated from zymogen to active protease as a result of cleavage by thrombin at the Arg169–Leu170 peptide bond in a reaction that is accelerated by thrombomodulin and EPCR (see Fig. 114–2 and “Thrombomodulin” above).^{5,7,8,76,81,86} Thrombin infusions into animals generate anticoagulant activity because of APC.^136,137 Interestingly, thrombin infusion into hyperlipidemic monkeys with atherosclerosis generates less APC and causes a poorer ex vivo response to APC compared with normolipidemic control monkeys,¹³⁸ showing that hyperlipidemia and vascular disease can affect protein C activation.

Ischemia causes protein C activation in vivo. A brief occlusion of the left anterior descending coronary artery in pigs results in APC generation.¹³⁹ During cerebral ischemia in humans undergoing routine endarterectomy, APC increases in the venous cerebral blood.¹⁴⁰ Protein C is significantly activated during cardiopulmonary bypass, mainly during the minutes immediately after aortic unclamping in the ischemic vascular beds.¹⁴¹ Streptokinase therapy for acute myocardial infarction increases circulating APC.¹⁴²

Circulating APC concentration in normal human subjects is highly correlated with circulating levels of protein C zymogen.¹⁴³ Based on protein C infusion studies in protein C-deficient subjects, the level of circulating APC is strongly determined by the concentration of protein C.¹⁴⁴ EPCR appears to be required for normal protein C activation in response to thrombin infusions in experimental animals.¹⁴⁵ EPCR and thrombomodulin must be in close proximity on cell surfaces (see Fig. 114–2), although this has yet to be experimentally demonstrated.

Thrombomodulin and EPCR appear to differ markedly in their relative distribution densities on blood vessels as the former is abundantly present in the small blood vessels but less so in large vessels, whereas the latter is more abundant in large vessels than in small vessels.^{85,86,146,147} Low levels of thrombomodulin are expressed in brain,¹⁴⁸ and brain-specific activation of protein C in humans occurs during carotid occlusion.¹⁴⁰

Proteolytic cleavage and activation of protein C can also be effected by meizothrombin, plasmin, or factor Xa.^{149,150,151,152,153} On the surface of cultured endothelial cells, negatively charged sulfated polysaccharides in the presence of phospholipid vesicles containing phosphatidylethanolamine can enhance the rate of protein C activation by factor Xa to approach the protein C activation rate of thrombin: thrombomodulin.¹⁵² No data yet indicate whether protein C activation by meizothrombin, plasmin, or factor Xa is physiologically relevant.

Protein C activation is stimulated by platelet factor 4. Both in vitro and in vivo data imply that platelet factor 4 may play a physiologic role in enhancing APC generation and influencing the activities of the protein C system.^{154,155,156,157}

The clinical phenotype of severe protein C deficiency in neonatal purpura fulminans implies that APC exerts multiple physiologically essential activities, including potent anticoagulant and antiinflammatory actions (Chap. 130). Recent advances establish that APC’s antiinflammatory actions are but one manifestation of its ability to interact directly with cell receptors to provide multiple cytoprotective activities.^7,8,110,111 These two distinct types of activities of APC—intravascular anticoagulant activity and initiation of cell signaling—are mediated by different sets of molecular interactions, and both types of activities are clinically relevant.

ACTIVATED PROTEIN C ANTICOAGULANT ACTIVITY

Mechanisms for APC’s direct anticoagulant activity involve factors V and VIII, the two homologous coagulation cofactors that circulate as inactive molecules and that are converted to active cofactors by limited proteolysis (see Chap. 113, Figs. 113–11 and 113–13). APC circulates at 40 pM (picomolars) in normal humans, and there is an inverse correlation between fibrinopeptide A, the product that is cleaved from fibrinogen by thrombin, and APC levels in healthy nonsmoking adults, suggesting APC is a significant regulator of basal thrombin activity.^15,158

Factors V and VIII are synthesized as large single-chain precursor coagulation cofactors of Mr 330,000, consisting of three homologous A domains (A1, A2, and A3) and two homologous C domains (C1 and C2) with a very large intervening, generally nonhomologous domain, designated the B domain, that connects the A2 and A3 domains (Chap. 113). Activation of the inactive precursor form of the two cofactors V and VIII involves limited proteolysis.^{23,159,160,161,162,163,164} Factor V activation involves cleavages at Arg709, Arg1018, and Arg1545 by thrombin, factor Xa, or other proteases.^{23,164,165,166,167,168} Cleavage at Arg1545 is the key step for generating factor Va activity because this proteolysis releases the B domain that blocks binding of factor Xa to factor Va.^164,169 The various forms of factor Va (see Chap. 113, Fig. 113–11) are composed of two polypeptide chains, one bearing the A1-A2 domains and the other bearing the A3-C1-C2 domains. Although generally similar to factor V activation, factor VIII activation (see Fig. 113–13) involves formation of a heterotrimer of polypeptide chains containing the A1 domain, the A2 domain, and the A3-C1-C2 domains, respectively. In contrast to heterodimeric factor Va, heterotrimeric factor VIIIa is intrinsically unstable as a consequence of spontaneous dissociation of the A2 domain.¹⁷⁰

Factors Va and VIIIa as Substrates for Activated Protein C

Irreversible proteolytic inactivation of factors Va and VIIIa by APC can be accomplished by proteolysis at Arg506 and Arg306 in factor Va and Arg562 and Arg336 in factor VIIIa (see Chap. 113, Figs. 113–11 and 113–13).^{23,171,172,173} Currently, the most common identifiable venous thrombosis risk factor involves a mutation of Arg506 to Gln in factor V that results in APC resistance (Chap. 130). The complexities of APC-dependent inactivation of factor Va and VIIIa are compounded by the number of different molecular forms of Va and VIIIa that can be generated by limited proteolysis by a variety of proteases and by their differing susceptibilities to APC and to the different APC cofactors.

Activated Protein C Resistance

APC resistance is defined as an abnormally reduced anticoagulant response of a plasma sample to APC (Chap. 130) and can be caused by many potential abnormalities in the protein C anticoagulant pathway. Such abnormalities could include defective APC cofactors, defective APC substrates, or other molecules that interfere with the normal functioning of the protein C anticoagulant pathway (e.g., autoantibodies against APC, APC cofactors, or APC substrates).

A report of familial venous thrombosis associated with APC resistance without any identifiable defect in four Swedish families¹⁷⁴ led to an intensive search for a genetic explanation that was soon found to involve replacement of G by A at nucleotide 1691 in exon 10 of the factor V gene which causes the amino acid replacement of Arg506 by Gln.^175,176,177 This factor V variant, like the prothrombin variant nt G20210A, arose in a single white founder some 18,000 to 29,000 years ago^178,179 and is known as Gln506-factor V or factor V Leiden. This mutation is currently a common, but not the only, cause of APC resistance (Chap. 130).

The molecular mechanism for APC resistance of Gln506-factor V is based on the fact that the variant molecule is inactivated 10 times slower than normal Arg506-factor Va.^{23,177,180,181,182} The variant factor Va exhibits only a partial resistance to APC because cleavage at Arg306 in factor Va also occurs, causing complete loss of factor Va activity.

Plasma and recombinant factor V can exist in two biochemically distinct forms, designated factor V1 and factor V2 that differ in N-linked carbohydrate on Asn2181, near the phospholipid binding region of the C2 domain as factor V2 has none.^183,184 Because the N-linked carbohydrate appears to decrease the apparent affinity of factor V1 or Va1 for phospholipid, it reduces the specific clotting activity and susceptibility to APC. Normal plasma contains a mixture of factors V1 and V2. Removal of the carbohydrate attached to factor V increases the rate of inactivation of factor Va by APC, although the clinical significance of this phenomenon is unknown.¹⁸⁵

APC resistance with no identifiable genetic or acquired abnormalities is well described in patients with venous and arterial thrombosis and, at least for research purposes, should be therefore examined in patients with a suspected thrombophilia. Further studies are needed to identify the causes of APC resistance in such patients.^186,187,188 One major challenge involves defining the normal range for the clotting assays that are actually used to characterize APC resistance and the multiple plasma analytes or nonplasma assay components that are present in the assays. For example, activated partial thromboplastin time-based assays are not equivalently sensitive as are dilute tissue-factor-based assays to plasma high-density lipoprotein (HDL) levels or oral contraceptive use.^189,190,191 Plasma variables, such as elevated prothrombin levels,^192,193 may affect the response to APC by inhibiting APC anticoagulant actions. Endogenous thrombin potential assays involving dilute tissue factor as the procoagulant initiator provide additional tools for defining and characterizing APC resistance and extend the tools for shedding light on the gray area of APC resistance found in some thrombosis patients that is not linked to currently known factors.

ACTIVATED PROTEIN C ANTICOAGULANT COFACTORS

APC anticoagulant activity is enhanced by a number of factors that may be termed APC anticoagulant cofactors; these include Ca²⁺ ions; certain, but not all, phospholipids; protein S; factor V; certain glycosphingolipids; and HDL.

Phospholipids as Activated Protein C Cofactors

Certain phospholipids, such as phosphatidylserine, phosphatidylethanolamine, and cardiolipin, enhance the anticoagulant activity of APC. In addition, phosphatidylethanolamine and cardiolipin stimulate the APC anticoagulant pathway activities much more than they stimulate the procoagulant pathway activities.^{194,195,196,197}

Protein S as Activated Protein C Cofactor

Protein S structure–activity relationships are informed by much biochemical work and the large number of mutations.^71,198 Protein S, as an anticoagulant APC cofactor, forms a 1:1 complex with APC and enhances by 10- to 20-fold the rate of APC’s cleavage at Arg306 in factor Va but not the Arg506 cleavage.^181,182 Part of the mechanism for this activity of protein S may be related to its ability to bring the active site of APC closer to the plane of the phospholipid membrane on which the APC–protein S complex is located when the complex is formed.^199,200 Protein S also facilitates the action of APC against factor VIIIa.²⁰¹ Protein S enhances APC’s action, in part at least, by ablating the ability of factor Xa to protect factor Va from APC.²⁰² The GLA domain, thrombin-sensitive region, and EGF1 and EGF2 domains of protein S are implicated in binding APC for expression of anticoagulant activity by the APC–protein S complex.^{198,203,204,205,206} Cleavage of the thrombin-sensitive region by thrombin abolishes normal binding of protein S to phospholipid and its normal APC-cofactor anticoagulant activity.^205,207

Factor V as Activated Protein C Cofactor

Factor V apparently can have anticoagulant as well as procoagulant properties because it enhances the anticoagulant action of APC against factors VIIIa and Va in a reaction in which protein S acts synergistically with factor V.^{23,208,209,210,211} Cleavage at Arg1545, which optimizes factor Va procoagulant activity, ablates the molecule’s anticoagulant cofactor activity. However, when factor V is cleaved at Arg506 by APC, its APC cofactor activity is increased 10-fold. This suggests that Gln506-factor V has two potential prothrombotic defects, namely, resistance of the variant factor Va to APC inactivation and resistance of the variant factor V to activation of its APC cofactor function.^{23,209,210,211}

High-Density Lipoprotein as Activated Protein C Cofactor

HDL can exert antithrombotic activity through multiple mechanisms.²¹² HDL enhances the anticoagulant activity of APC both in plasma and in purified reaction mixtures, and this APC cofactor activity requires protein S and involves, at least in part, stimulation of APC’s cleavage at Arg306 in factor Va.^189,190 HDL is heterogeneous in both protein and lipid composition, and the components responsible for this activity have not been identified, although large HDL, but not small HDL, possesses APC anticoagulant cofactor activity.¹⁹⁰ Venous thrombosis in males and in subjects experiencing venous thrombosis recurrence are associated with a pattern of dyslipoproteinemia and low HDL, consistent with the hypothesis that deficiency of large HDL is a risk factor for venous thrombosis.^213,214

Glycosphingolipids as Activated Protein C Cofactors

Although both procoagulant and anticoagulant reactions are markedly enhanced by the presence of negatively charged phospholipid surfaces in vitro, certain lipoproteins, for example, HDL,¹⁸⁹ and certain lipids, for example, glycosphingolipids and sphingosine,^{215,216,217,218} selectively enhance anticoagulant reactions in plasma. Plasma glucosylceramide deficiency is a biomarker and may be a potential risk factor for venous thrombosis.²¹⁵ Sphingosine and several of its common analogues are potent inhibitors of thrombin generation in plasma and on cell surfaces because they inhibit interactions between factors Va and Xa.²¹⁸ Further studies are needed to characterize the anticoagulant or procoagulant properties of minor abundance plasma and their significance for clinical thrombotic events.

ACTIVATED PROTEIN C DIRECT CELLULAR ACTIVITIES

As noted in Chap. 113, control of coagulation reactions does not occur in the absence of an integrated host defense system that involves a number of biologic processes involving multiple overlapping and integrated pathways. Reactions of the innate and acquired immune system including inflammatory processes, blood coagulation reactions, fibrinolysis, and thrombotic processes are intertwined in vivo via multiple molecular and cellular mechanisms.^{5,7,8,81,86,87,212,219,220} In addition to its anticoagulant activity, APC acts directly on cells to cause multiple cytoprotective effects. Cytoprotective actions of APC include antiapoptotic and antiinflammatory activities, beneficial changes in gene-expression profiles, and endothelial barrier stabilization. These cytoprotective activities of APC generally require EPCR, involve APC’s ability to activate PAR-1, and may also require additional receptors such as PAR-3, sphingosine-1-phosphate receptor 1, integrin CD11b/CD18, apoER2, EGF receptor, and/or Tie2.^{7,8,86,110,221,222,223,224,225,226,227}

Pharmacologic APC infusions showed benefits in numerous animal injury model systems, with the most informative animal studies to date being in sepsis models and in neuroprotection experiments.^{7,8,110,111,226,227} Protein engineering permitted the molecular dissection of APC’s anticoagulant activity from its cytoprotective activities^{7,8,37,40,41,42,43,44,228} and led to proof of principle that APC’s cell signaling activities are both necessary and most likely sufficient for reducing lethality in murine septic shock models^42,229 and for providing neuroprotective effects in ischemic stroke models.^{226,227,230,231,232} Notably, recombinant APC mutants that have little anticoagulant activity (<10 percent) but normal cell-signaling activity are able to convey beneficial effects in multiple injury and disease models with diminished risks for bleeding that would be anticipated with wild-type APC therapy.^{8,40,41,42,43,44}

Activated Protein C Neuroprotective Effects

Neuroprotective effects of APC have been convincingly demonstrated in rodent ischemic stroke models and N-methyl-D-aspartate (NMDA) excitotoxic injury models.^{8,223,226,227,231,233–242} APC not only provides direct cytoprotection in vitro and in vivo for brain endothelium against ischemic injury but also directly protects neurons against NMDA-induced excitotoxic injury both in vivo and in vitro. APC mutants with reduced anticoagulant activity were as neuroprotective as wild-type APC, and certain cellular receptors were required for APC’s neuroprotection, strongly implying that neuroprotection by APC involves its actions directly on the endothelium and on neurons. Remarkably, in the ischemic penumbra in a murine stroke model, APC caused neovascularization and neurogenesis.^{223,240,243,244,245} The extensive preclinical studies on APC’s neuroprotective effects paved the pathway for translation of the 3K3A-APC variant to potential neuroprotective therapy for acute ischemic stroke.^246,247 Because of the greatly reduced anticoagulant activity of 3K3A-APC (<10 percent of normal), high-dose bolus dosing in healthy volunteers can achieve circulating APC levels that are 100-fold higher than those used in the PROWESS or PROWESS-SHOCK sepsis trials without notable anticoagulant effects.^46,49,246

Cellular Receptors for Physiologic Effects of Activated Protein C on Cells

The ability of exogenously administered APC to alter gene-expression profiles of cultured endothelial cells, to stabilize endothelial barriers, to reduce lethality caused by endotoxin in murine sepsis models, to prevent apoptosis of stressed endothelial cells, and to provide neuroprotection requires EPCR and PAR-1, strongly supporting the EPCR–PAR-1 cell-signaling pathway as key for APC’s pharmacologic benefits (see Fig. 114–4).^{7,8,40,41,42,43,221,225,229,233,234,236,237,244,248,249,250,251}

Although few details are known about intracellular mechanisms for APC’s multiple cytoprotective actions, some mechanistic details for APC’s cell signaling have become clear, as depicted in Fig. 114–6.⁸ Multiple considerations help explain how PAR-1 can mediate thrombin’s disruption of endothelial barrier leading to vascular leakage while, paradoxically, the same receptor mediates APC’s endothelial barrier protection, preventing vascular leakage.^249,250 First, PAR-1–mediated APC signaling occurs in caveolae microdomains that contain EPCR whereas PAR-1–mediated thrombin signaling is not limited to caveolae (Fig. 114–6).^252,253 Second, different cleavages in the extracellular N-terminus of PAR-1, either at the canonical Arg41 thrombin-cleavage site (i.e., widely recognized as the essential thrombin cleavage site) or at the novel Arg46 APC-cleavage site, results in very different signaling initiated by different tethered N-terminal peptide sequences which begin at either residue 42 or residue 47.^124,254,255 Third, following thrombin cleavage at Arg41, PAR-1 initiates signaling involving G proteins, extracellular signal-regulated kinase (ERK)1/2 and RhoA, whereas, following APC cleavage at Arg46, PAR-1 initiates signaling involving β-arrestin-2, phosphatidylinositide 3′-kinase (PI3K)/Akt, and Rac1.²⁵⁶ Fourth, peptides mimicking the N-terminus of cleaved PAR-1 are peptide agonists with pharmacologic effects resembling those of the respective proteases that cleave PAR-1 differentially. For example, “thrombin receptor activating peptides (TRAPs)” that begin with Ser 42 promote G-protein–mediated signaling similar to thrombin. In contrast, peptides that begins with Asn 47 (TR47) promote APC-like signaling.²⁵⁴ TRAP but not TR47 promotes ERK1/2 phosphorylation on endothelial cells whereas TR47 but not TRAP promotes Akt phosphorylation.²⁵⁴ The different and opposite induction of signaling pathways is also mirrored in different and opposite functional effects, as thrombin peptide and TRAP cause endothelial barrier disruption and proinflammatory effects whereas APC and the TR47 peptide cause barrier-protective and antiinflammatory effects. Thus, PAR-1 displays biased signaling depending on the activation cleavage sites and the generated tethered-ligand with absolutely opposing outcomes for the cell, the tissue, and the host depending on which coagulation system protease, thrombin or APC, is cleaving PAR-1.

Other receptors are recognized that may also play key roles for APC’s beneficial signaling effects, including PAR-3 and sphingosine-1-phosphate receptor-1.^{249,250,257,258} ApoER2 can initiate Disabled-1-dependent pathway activation of the PI3K-Akt cell-survival pathway, which may ultimately help explain additional aspects of APC’s cytoprotection.²⁵⁹

Although most studies demonstrating the cell-signaling activities of APC have focused on pharmacologic levels of APC, several reports of murine injury models demonstrate the physiologic importance of cell signaling by endogenous APC,^260,261,262 implying that defects in APC’s endogenous cytoprotective actions might have pathophysiologic relevance. Future investigations on APC cellular receptors and on intracellular mechanisms involved in the protein C cellular pathway will likely provide novel clinical insights with diagnostic and therapeutic potential.

Blood contains circulating APC in a well-defined normal concentration range that contributes to antithrombotic surveillance mechanisms and possibly to homeostatic cell signaling.^15,142,144 Circulating APC levels are determined by the balance between countervailing mechanisms for APC generation and for APC inhibition and clearance. APC generation is influenced by protein C zymogen levels, endogenous thrombin generation, and the availability of thrombomodulin and EPCR. Clearance of circulating APC is based on inhibition of APC by protease inhibitors and clearance of APC:inhibitor complexes.^{263,264,265,266,267,268,269} The major plasma inhibitors of APC include α₁-antitrypsin, protein C inhibitor, and α₂-macroglobulin.

Because hereditary protein S deficiency^270,271 is strongly linked to increased venous thrombosis risk (Chap. 130), protein S is a significant physiologic anticoagulant factor.^71,198 In addition to its anticoagulant cofactor activity for APC, protein S can also inhibit coagulation reactions independently of APC. Several plausible mechanisms have been described for protein S’s anticoagulant activity independent of APC. First, protein S can bind directly to procoagulant factors Xa and Va and thereby inhibit directly the activity of the prothrombinase complex.^11,12,13 The thrombin-sensitive region and the EGF3 domains of protein S (see Fig. 114–5) likely bind factor Xa, contributing to APC-independent anticoagulant activity.^206,272,273 Second, protein S can also bind factor VIIIa and inhibit activation of factor X by factor IXa–factor VIIIa complexes.^274,275,276 Third, protein S binds tissue factor pathway inhibitor (TFPI) and enhances its ability to inhibit factor Xa.^277,278,279 Zn²⁺ ions might play a key role for APC-independent protein S activity.²⁸⁰ It is not easy to decipher the relative importance of each of these or other mechanisms for APC-independent anticoagulant activities of protein S or to establish their physiologic relevance, but infusions of protein S without APC are antithrombotic in baboon thrombosis models.²⁸¹

The activities of protein S can be strongly influenced by C4b-binding protein, a plasma protein that enhances inactivation of the complement cascade by binding to C4b and promoting its proteolytic inactivation by the protease, factor I. C4b-binding protein reversibly binds protein S with high affinity,^282,283,284 and formation of this complex affects some but not all of the anticoagulant activities of protein S.^{71,198,271,285} Because of the influence of C4b-binding protein on protein S activities and plasma levels, interpretation of clinical assays for protein S requires evaluation of free and bound protein S as plasma contains approximately 240 nM protein S–C4b-binding protein complexes and 120 nM free protein S.²⁸³ C4b-binding protein is a heteropolymer containing six or seven α chains that are disulfide-linked to a single β chain that binds protein S.^286,287 Residues 30 to 45 of the β chain bind with high affinity to the C-terminal SHBG domain of protein S.^65,288,289 During an acute phase reaction, the level of the C4b-binding protein α chain, but not the β chain, is increased, so that the acute phase change in total C4b-binding protein does not alter the level of free and bound protein S.²⁹⁰

Another potential mechanism for the antithrombotic actions protein S is based on its APC-independent direct interactions with cells that might contribute to its antithrombotic actions. Protein S promotes clearance of apoptotic cells,^{68,71,198,291,292,293,294} and this antiapoptotic activity of protein S might contribute to its antithrombotic activity. Protein S has direct effects on cells by activating one or more transmembrane receptor tyrosine kinases.^68,198,292 Protein S is a potent neuroprotectant as it can protect brain endothelium against ischemic injury in murine stroke models and can protect neurons against NMDA-induced excitotoxic injury, presumably acting via transmembrane receptor tyrosine kinases.^{295,296,297,298,299}

Antithrombin, initially designated antithrombin III, is clinically the best known inhibitor of clotting factor proteases. Antithrombin can neutralize all coagulation proteases in reactions that are enhanced by heparin and related glycosaminoglycans (see Chap. 113 and Fig. 113–28).³⁰⁰ However, antithrombin does not inhibit the anticoagulant protease APC. TFPI can neutralize factors VIIa and Xa, proteases of the extrinsic coagulation pathway.^{277,278,301,302,303} In addition, other plasma protease inhibitors such as α₁-antitrypsin, heparin cofactor II, protein C inhibitor, α₂-macroglobulin, or protein Z–dependent protease inhibitor, can neutralize various coagulation proteases, although the ultimate clinical significance of these reactions is less-well defined than the clinical relevance of antithrombin for thrombophilia (Chap. 130). Antithrombin is key for anticoagulant therapy based on the heparin-stimulated inhibition of thrombin and factor Xa.

ANTITHROMBIN AND HEPARINS

Antithrombin is synthesized in the liver and is present in plasma at 150 mcg/mL, and it is a typical member of the serine protease inhibitor (SERPIN) superfamily and is denoted as SERPINC1.^{300,304,305,306} Based on X-ray crystallographic studies,^{307,308,309,310,311} models of serpin–protease complexes in various reaction states have emerged and the mechanism for the effects of heparin on the reaction of thrombin with antithrombin is reasonably clear.

The neutralization of proteases by antithrombin is a result of a stable enzyme–antithrombin complex that is formed by a molecular mechanism characteristic of inhibitory serpins.^{304,305,306,307,309,310,311,312} Following binding of a protease to a “reactive site” loop in a serpin, a single peptide bond in the serpin is cleaved with formation of an acyl-enzyme intermediate via the active site Ser residue. This metastable enzyme–serpin complex can either break apart because of deacylation, or it can form a more stable covalent enzyme–serpin complex. To break apart the enzyme–serpin covalent complex, deacylation liberates the cleaved product and regenerates the active site Ser residue of the protease. However, serpins have an ability to undergo major conformational changes following cleavage at the reactive site residue that can distort that protease’s active site region and lock the enzyme into the protease–serpin complex in which both the serpin and the protease are essentially deformed.^{304,305,306,307,309,310,311,312} The dominant structural feature of native serpins is a large five-stranded β-sheet that defines the structure of an ellipsoidal protein. Following cleavage at the reactive residue in the reactive center loop by a protease, this extended loop is able to partially or completely insert itself into the five-stranded β-sheet, forming a very stable six-stranded β-sheet. If this insertion reaction proceeds before deacylation occurs, then the protease remains covalently attached to the reactive center P1 residue through the protease’s active site Ser residue, and a stable covalent protease–inhibitor complex with each protein in an altered conformation is formed.^307,308

Heparin enhancement of the rate of reaction between antithrombin and thrombin or other clotting factors is caused by two distinct effects of heparin, one involving conformational effects on antithrombin and the other involving “approximation” effects on both antithrombin and thrombin.^{300,307,308,311,312,313,314,315} For the first effect, a particular pentasaccharide sequence within heparin binds antithrombin and potently causes a conformational change that converts antithrombin from its native state of moderate reactivity to a conformation with relatively high reactivity. This pentasaccharide contains a specific sulfated sequence of glucosamine and iduronic acid residues,^{300,307,308,311,312,313,314,315} and when it is present in a large heparin molecule, in low-molecular-weight heparin, or in a synthetic pentasaccharide, it alters antithrombin conformation and greatly accelerates the reaction of antithrombin, especially with factor Xa. Synthetic pentasaccharides, such as fondaparinux, which are analogues of the naturally occurring sequence, are often termed to be indirect factor Xa inhibitors and have significant clinical utility. For the second mechanistic effect, namely the approximation effect, unfractionated heparin or low-molecular-weight heparins simultaneously bind to antithrombin and the target protease to promote frequent and geometrically productive encounters between protease and inhibitor, thus increasing the reaction rate. Heparan sulfates to some extent can also act in this manner.

The mature antithrombin polypeptide chain contains 432-amino-acid residues after cleavage of a propeptide from a 464-amino-acid-residue precursor.³¹⁶ It has four sites for N-linked carbohydrate attachment, one of which (Asn135) is variably glycosylated, giving rise to a β-isoform that has higher affinity for heparin.^317,318 Heparin binding to antithrombin is mediated by a number of positively charged Arg and Lys residues in the N-terminal region of the molecule, including Lys11, Arg13, Arg47, Lys114, Lys125, and Arg129, whereas the reactive center loop containing the scissile peptide bond at Arg393-Ser394 is near the C-terminus.³¹¹

ANTITHROMBIN GENE

The antithrombin gene comprising seven exons and six introns spans 13.4 kb and is located on chromosome 1q23–25 (see Table 114–1).^319,320

ANTITHROMBIN MUTATIONS

Hereditary deficiencies of antithrombin are risk factors for venous thrombosis (Chap. 130). More than 100 different antithrombin mutations are associated with thrombosis. An extensive database of mutations is published³²¹ and is available at http://www1.imperial.ac.uk/departmentofmedicine/divisions/experimentalmedicine/haematology/coag/antithrombin/.

Mutations that cause antithrombin deficiency are scattered throughout the gene. Molecular defects can be classified as type I, characterized by parallel decreases in antigen and activity, or type II, characterized by circulating dysfunctional molecules such that plasma has decreased functional activity but normal or near-normal antigen levels. Type II defects are further classified based on whether the dysfunction involves only reactive center defects that can be tested in the absence of heparin, only heparin-binding defects that can be tested only in the presence of heparin, or both of these defects (pleiotropic effects). Reactive center defects carry the largest risk of thrombosis, whereas heparin-binding defects are associated with less risk of venous thrombosis (Chap. 130).

TFPI, also known as lipoprotein-associated coagulation inhibitor or extrinsic pathway inhibitor, has a predicted mature protein sequence of 276 residues and a Mr of 34,000. However, TFPI is a complex protein and has at least three isoforms in blood vessels.^{277,301,302,303,322,323,324,325,326,327} There are two alternatively spliced forms of TFPI designated TFPIα and TFPIβ (Fig. 114–7).^323,324 TFPIα is the full-length mature protein that contains an acidic N-terminal sequence, three homologous but distinct Kunitz-type protease inhibitor domains (K1, K2, K3), and a C-terminal positively charged basic amino acid sequence (Fig. 114–7). TFPIβ contains K1 and K2 but an unrelated sequence replaces the K3 domain and the C-terminus. TFPIβ can be covalently modified by addition of glycosylphosphatidylinositol (GPI) that localizes TFPIβ to cell membranes (Fig. 114–7). Some TFPI in plasma is present as a disulfide-linked heterodimer of TFPI–apolipoprotein A-II,^327,328 but the functional significance of the apoA-II appendage is unknown. TFPI in its multiple forms is a significant inhibitor of the coagulation pathways that can function synergistically with the protein C pathway and antithrombin to suppress thrombin generation.

TFPI is synthesized by endothelial cells, megakaryocytes, and smooth muscle cells.^301,302,303 Free TFPI in plasma is TFPIα but it is a minor fraction of the amount of TFPI in blood vessels. More than half of TFPIα in plasma is associated with lipoproteins, especially HDL and low-density lipoprotein. TFPIα is also the main form within platelet and it is secreted by activated platelets. A substantial amount of TFPIα is released from the vessel wall when heparin is infused.³²⁹ TFPIβ is membrane bound, especially to endothelium, because of its GPI anchor.

The interaction of TFPI with lipoproteins reduces its anticoagulant activity measured in vitro though the physiologic significance of TFPI’s binding to various lipoproteins remains uncertain. In addition to binding lipoproteins, TFPIα but not TFPIβ binds to protein S and to certain forms of factor Va/factor V.^{277,278,279,330,331,332,333} Different regions of TFPIα, namely the K3 domain or the basic amino acid cluster, respectively, are responsible for binding protein S or factors Va/V (see Fig. 114–7). Inhibition of factor Xa by TFPIα is accelerated by protein S and by certain but not all forms of factor Va (see below).

TFPI neutralizes factors Xa and VIIa by multiple complicated mechanisms.^{277,301,302,303} In each mechanism, the K1 domain binds and inhibits factor VIIa while the K2 domain inhibits factor Xa (see Fig. 114–7). No protease has yet been identified as the target of the K3 protease inhibitor domain. In one mechanism, initially the K2 domain of TFPI reacts with and inhibits the enzyme activity of factor Xa. Subsequently, this binary complex reacts with factor VIIa in the tissue factor–factor VIIa complex to form a quaternary protein complex on a membrane with both proteases neutralized. In an alternative proposed scheme, TFPI first reacts with factor VIIa in a tissue factor–factor VIIa complex that has generated factor Xa, and thereafter it rapidly reacts with factor Xa before it can dissociate from the ternary tissue factor–factor VIIa–factor Xa complex. Possibly each proposal is valid. Some argue that because some kinetic studies showed that TFPI requires factor Xa for kinetically favorable reactions with factor VIIa, TFPI does not shut off the initiation of the extrinsic pathway by tissue factor until some significant though small amount of factor Xa is generated, in which case TFPI provides negative feedback inhibition of the generation of factor Xa by the factor VIIa–tissue factor complex. An additional property of TFPIα involves its inhibition of factor Xa in the absence of factor VIIa, and this reaction is accelerated by protein S and by certain forms of factor Va.^{277,278,279,330,331,332,333} In contrast to the anticoagulant factors, anti-thrombin, protein C, and protein S, for which hereditary deficiencies are linked to significantly increased risk for venous thrombosis (Chap. 130), no clear pattern for increased risk of thrombosis has been definitively established for TFPI deficiency in humans. In mice, knockout of TFPI is embryonically lethal.³³⁴ In a highly informative kindred that presented with a serious bleeding diathesis, highly elevated plasma TFPI levels were linked to increased bleeding risk, indicating that TFPI functions in man as a physiologically significant inhibitor of coagulation.^322,332 The genetic mutation causing elevated plasma TFPI levels was in the factor V gene, not the TFPI gene. The mutated factor V, named “factor V-short,” has a higher affinity for TFPIα than wild-type factor V, thereby binding more TFPIα and prolonging its half-life. Factor V-short may also enhance inhibition of factor Xa by TFPI with the effect of increasing bleeding risk. This genetic disorder, as well as previous studies showing that inhibition of TFPI reduced bleeding in preclinical hemophilia models, lends support for ongoing efforts to develop TFPI inhibitors for reducing bleeding in some hemophilia subjects, especially those with anti–factor VIII inhibitors.

TFPI GENE

The sequence of TFPI was established from cloning of its complementary DNA. The gene contains nine exons, spans 85 kb, and is located on chromosome 2q31–32.1 (see Table 114–1).^335,336

HEPARIN COFACTOR II

Heparin cofactor II, a serpin whose inhibitory activity is enhanced by dermatan sulfate, inhibits thrombin in vivo and in vitro by an approximation mechanism.^{337,338,339,340} A few reports link heparin cofactor II deficiency to venous thrombosis, but no significant clinical relevance has been established.³⁴¹ Curiously, a severe heparin cofactor II deficiency was reported in an asymptomatic subject.³⁴² Some studies imply that heparin cofactor II may play significant roles in arterial vascular wall processes, but definitive mechanisms remain to be elucidated.

PROTEIN Z–DEPENDENT PROTEASE INHIBITOR

Protein Z–dependent protease inhibitor (ZPI) is a plasma serpin that inhibits factors Xa, XIa, and IXa, but not factor XIIa or thrombin.^{343,344,345,346,347,348,349,350} Protein Z, which is a vitamin K–dependent protein that contains a GLA domain, two EGF-like domains, and a protease-like domain,³⁵¹ stimulates factor Xa inhibition by ZPI. Curiously, the protease-like domain of protein Z lacks any protease activity because it has mutations at two of the three active site triad residues. The major hypothesis for stimulation of inhibition of factor Xa by protein Z is based on a structural model in which three proteins assemble on a phospholipid membrane via the two GLA domains (see Fig. 114–7).³⁵¹ In this putative ternary complex, the protease-like domain and the second EGF-like domain of protein Z bind ZPI in an alignment that facilitates reaction of factor Xa with the reactive center loop of ZPI.

In plasma, ZPI is in slight protein molar excess over protein Z with which it associates noncovalently, and it has been speculated, but not proven, that almost all plasma protein Z is associated with ZPI.^{352,353,354,355,356,357} If the ZPI is a physiologic coagulation inhibitor, the deficiency of either protein Z or ZPI might be associated with thrombosis. Knocking out the protein Z gene in a mouse does not produce a remarkable phenotype unless protein Z deficiency coexists with factor V Leiden, in which case the mouse exhibits a hypercoagulable, prothrombotic state.³⁵³ This murine observation is mirrored by one clinical report that subnormal levels of protein Z are associated with venous thrombosis in subjects who are heterozygous for factor V Leiden.³⁵⁴ Some associations between venous thrombosis and defects in protein Z or ZPI have been reported but not uniformly confirmed.^{352,354,355,356,357} An association with peripheral arterial disease was reported.³⁵⁸ However, to date no convincing pattern between thrombosis and defects in either protein Z or ZPI has been firmly established.

OTHER MINOR PROTEASE INHIBITORS

Thrombin in plasma can be inhibited not only by antithrombin but also by α₂-macroglobulin, an acute-phase reactant. No association between defects in bleeding or thrombosis has been confirmed for this inhibitor. In purified reaction mixtures, protein C inhibitor also efficiently neutralizes thrombin in the presence of thrombomodulin,³⁵⁹ although no studies show that this is a physiologic reaction or that it is associated with thrombosis.