Control of the coagulation system by serpins

Getting by with a little help from glycosaminoglycans


  • Robert N. Pike,

    1. Department of Biochemistry & Molecular Biology and Co-operative Research Centres for Vaccine Technology and Oral Health Sciences, Monash University, Clayton, Victoria, Australia
    2. The Victorian Bioinformatics Consortium, Monash University, Clayton, Victoria, Australia
    Search for more papers by this author
  • Ashley M. Buckle,

    1. Department of Biochemistry & Molecular Biology and Co-operative Research Centres for Vaccine Technology and Oral Health Sciences, Monash University, Clayton, Victoria, Australia
    2. The Victorian Bioinformatics Consortium, Monash University, Clayton, Victoria, Australia
    Search for more papers by this author
  • Bernard F. le Bonniec,

    1. INSERM U428, Faculté de Pharmacie, Université Paris V, Paris, France
    Search for more papers by this author
  • Frank C. Church

    1. Departments of Pathology and Laboratory Medicine, Pharmacology, and Medicine, Carolina Cardiovascular Biology Center, The University of North Carolina at Chapel Hill, School of Medicine, NC, USA
    Search for more papers by this author

R. N. Pike, Department of Biochemistry & Molecular Biology, Monash University, Clayton, Victoria 3800, Australia
Fax: +61 3 99054699
Tel: +61 3 99053923


Members of the serine protease inhibitor (serpin) superfamily play important roles in the inhibition of serine proteases involved in complex systems. This is evident in the regulation of coagulation serine proteases, especially the central enzyme in this system, thrombin. This review focuses on three serpins which are known to be key players in the regulation of thrombin: antithrombin and heparin cofactor II, which inhibit thrombin in its procoagulant role, and protein C inhibitor, which primarily inhibits the thrombin/thrombomodulin complex, where thrombin plays an anticoagulant role. Several structures have been published in the past few years which have given great insight into the mechanism of action of these serpins and have significantly added to a wealth of biochemical and biophysical studies carried out previously. A major feature of these serpins is that they are under the control of glycosaminoglycans, which play a key role in accelerating and localizing their action. While further work is clearly required to understand the mechanism of action of the glycosaminoglycans, the biological mechanisms whereby cognate glycosaminoglycans for each serpin come into contact with the inhibitors also requires much further work in this important field.






heparin cofactor II


protein C inhibitor


serine protease inhibitor


Efficient functioning of the coagulation system is vital to human health [1]. However, control of this system, in particular its regulation to prevent inappropriate, excessive or mislocalized clotting of blood, is also vital to prevent cardiovascular diseases such as deep vein thrombosis.

Because many of the principal procoagulant components of the system are serine proteases, regulation of the system is principally by the action of serine protease inhibitors. One major class of serine protease inhibitors regulating procoagulant enzymes is the serpin superfamily [2]. The principal inhibitor of procoagulant enzymes such as thrombin and factor Xa is the serpin antithrombin (AT). There are, however, other serpins that act to control coagulation enzymes, such as heparin cofactor II (HC-II), protease nexin I and C1-inhibitor. Some serpins, such as protein C inhibitor (PCI), act to control the action of anticoagulant enzymes, such as activated protein C.

A feature of many of the serpins that control enzymes in the coagulation system is that they themselves are under the control of glycosaminoglycans (GAGs) [3]. Glycosaminoglycans such as heparin, heparan sulfate and dermatan sulfate have been found to significantly accelerate the interaction between serpins and coagulation proteases, usually increasing the reaction rates from values that are not relevant under physiological conditions to rates that are relevant. This control over the action of serpins that have the particular role of regulating procoagulant enzymes is probably vital in that it allows the enzymes to act, as they must, to clot blood. It follows that the serpins mostly act to localize the clotting process, which is likely to be the crucial element in the regulation of clotting, and also to the eventual shutting off of the clotting process, although this latter element is most probably a complex multifactorial process also involving the platelets and endothelium.

This review will examine the basic elements of the structure and function of the serpins involved in controlling the central coagulation enzyme, thrombin, and their control by GAGs. Control over other enzymes, such as factor Xa, will also be mentioned where relevant. We will focus on antithrombin, heparin cofactor II and protein C inhibitor.

General mechanism of serpin action

Serpins are a highly evolved family of proteins, which have a mechanism of action that appears to be common to most members of the family [4]. The mechanism, hotly debated for many years, involves the attack of the protease on the P1-P1′ bond in the reactive centre loop (RCL) of the serpin [5]. The catalysis of the peptide bond cleavage appears to be arrested at the acyl intermediate by the unique action of the serpin, whereby the RCL of the serpin inserts into the major A β-sheet causing the protease to be rapidly translocated from the top to the bottom of the inhibitor [6]. In the process, the structure of the protease appears to be deformed by being ‘crushed’ against the bottom of the inhibitor [7]. In particular, this deformation of the protease affects the geometry of the catalytic triad, preventing the completion of catalysis beyond the acyl intermediate and therefore trapping the protease in a covalent bond with the serpin. The mechanism is a suicide substrate mechanism, irreversibly inactivating the serpin. The serpin–enzyme complex is later removed from the circulation by the action of receptors which specifically recognize the inhibited conformation of the serpin (reviewed in [4]).

The structure and mechanism of serpins is highly amenable to control via binding of molecules such as GAGs, but the same level of conformational mobility which aids in the function of serpins also renders them susceptible to mutations which cause the A β-sheet in particular to become susceptible to insertion of the serpin's own RCL. This results in either a so-called latent state, or in polymers of serpins, where the insertion of another molecule's RCL takes place [8]. Both of these result in the irreversible inactivation (generally) of the serpins, and, in the case of the anticoagulant serpins, a lowering of the effective concentration of the serpins and therefore diseases such as thrombosis [8].


Antithrombin is arguably the major anticoagulant serpin. It is a 58 kDa glycoprotein, which circulates in blood at a concentration of ≈ 125 µg·mL−1 (2.3 µm) [9]. AT inhibits a large number of serine proteases of the coagulation system including thrombin (factor IIa) and factors IXa, Xa, XIa and XIIa. The principal targets of the serpin are usually regarded as being thrombin and factor Xa, although it is likely that inhibition of the other enzymes by this serpin is also important.

The serpin has a structure (Fig. 1) which is highly similar to that of other serpins, with a few important features. In its native state the RCL is partially inserted into the top of the A β-sheet of the molecule [10]. Upon the addition of heparin, the RCL is expelled from the A β-sheet by a closing of the sheet caused by a conformational transition in the molecule following the binding of a specific heparin pentasaccharide sequence to a highly positively charged cluster located at the D-helix of the serpin [11,12]. The pentasaccharide sequence of heparin on its own is able to induce the conformational change in AT [13,14] and this change in the structure of the serpin is apparently able to substantially accelerate the interaction with serine proteases such as factor Xa, but not enzymes such as thrombin [15]. It is thought that the overall increase in the rate of interaction with factor Xa brought about by the heparin pentasaccharide-mediated conformational change occurs through a combination of the changes in the structure of the RCL, allowing the interaction of residues on the RCL with subsites in the active site of fXa [16,17], and the exposure of a new exosite on the serpin for interaction with the protease [18,19]. Given the plasma concentration of AT and the rates of interaction in the presence and absence of heparin pentasaccharide (Table 1), one can calculate that the half-life of enzyme activity in the absence of heparin pentasaccharide would be 133 s (full lifetime, ≈ 22 min), and this would be ≈ 1.33 s in the presence of heparin pentasaccharide (full lifetime, 0.22 min). This action of the synthetic heparin pentasaccharide is apparently effective enough, and has allowed its introduction as a new antithrombotic drug [20].

Figure 1.

The structures of serpins controlling thrombin. The structures of AT with and without the heparin pentasaccharide bound, heparin cofactor II and cleaved protein C inhibitor are shown as indicated. In all of the structures, the A β-sheet is shown in red, the B β-sheet in green and the C β-sheet in yellow. The reactive centre loop is shown in magenta and the D-helix in dark blue. The H-helix is shown in dark pink in the protein C inhibitor structure. Positively charged residues on the D- and H-helices and the heparin pentasaccharide are shown in ball and stick format.

Table 1.  Second order rates of association (kass) values for the reaction of serpins with proteases in the presence and absence of a range of GAGs (values are representative of those reported in a range of publications cited in this article).
SerpinEnzymeGlycosaminoglycankass (m−1·s−1)
AntithrombinThrombin1 × 104
Heparin4 × 107
Heparan2 × 107
High affinity heparin4 × 107
Heparin pentasaccharide2 × 104
Factor Xa2 × 103
Heparin4 × 107
Heparan2 × 107
High affinity heparin4 × 107
Heparin pentasaccharide5 × 105
HCIIThrombin7 × 102
Dermatan sulfate1 × 107
Heparin1 × 107
PCIThrombinHeparin5 × 105
fXaHeparin3 × 103
aPC1 × 104
 Heparin2 × 104
 Heparin/calcium3 × 106

Heparin pentasaccharide on its own does not substantially increase the rate of inhibition of some coagulation enzymes, such as thrombin, indicating that the conformational change in AT alone does not cause much acceleration in the rate of interaction [15]. For full acceleration of the rate of inhibition of enzymes such as thrombin, full-length heparin (> 26 saccharide units in length) is required. The longer chains of heparin appear to accelerate the interaction between AT and thrombin by ‘templating’ the serpin and enzyme, binding to both molecules (via an exosite on the protease) and facilitating their diffusion towards each other in solution [21]. This accelerates the interaction of AT with thrombin 1000-fold and with fXa 10 000-fold [22]. With regard to the latter interaction, it is clear that calcium ions are required to overcome the negative effects of the Gla-domain of factor Xa on the templating interaction mediated by heparin. For thrombin, this means that AT controls the enzyme in 0.27 s in the presence of heparin, compared to 4.4 min in the absence of heparin. Clearly this is important, as the impairment of heparin binding on mutants of AT has disease-causing consequences [23]. Recently, the structure of AT templated to a genetically modified form of thrombin by a synthetic heparin has been solved [24,25]. These structures have supported much of what has been published before in terms of the mechanism by which templating occurs and have provided additional insights into the conformation of the reactive centre loop of AT when it is in complex with a target protease.

It is interesting to note that clot bound thrombin and factor Xa are protected from inactivation by AT [26,27]. This is consistent with the role of AT in localizing the clot and preventing it from spreading too far, rather than actually shutting down clotting. It would appear that AT might localize to clots due to the exposure of heparan sulfate chains on the endothelium following vascular disruption or the localized release of heparin from the granules of mast cells which are found lining the vasculature [28,29]. Thus the AT may act as a sentinel to prevent escape of active procoagulant enzymes from their site of action, allowing clotting to proceed where it is required, but not allowing it to spread.

Antithrombin is clearly critical to survival. Homozygous null mutants of AT die in utero[30] and heterozygous mutants which have about 50% of the normal concentrations of AT are predisposed towards disease [31]. The experiments using the genetically manipulated mice have confirmed a host of studies which reveal that mutations of AT which impair its normal function predispose patients to thrombotic disorders [8], particularly when found in combination with other predisposing factors [32].

Heparin cofactor II

Heparin cofactor II mRNA has been detected only in human liver, and the normal concentration of HC-II in blood plasma is ≈ 1.2 µm and the mature protein is 65.6 kDa [33]. The HC-II reactive site peptide bond is Leu444-Ser445 [34,35]. Intriguingly, HC-II is a very specific inhibitor of thrombin, but no other serine protease in blood coagulation; however, it does exhibit some inhibitory action to the chymotrypsin-like proteases, cathepsin G and chymotrypsin [36,37].

Heparin cofactor II rapidly inhibits thrombin following binding to GAGs (Table 1). However, the GAG specificity of HC-II is much less discriminating than that of AT. While both heparin/heparan sulfate and dermatan sulfate GAGs are physiological activators of HC-II, many different polyanions, including polyphosphates, polysulfates and polycarboxylates, are able to accelerate HC-II inhibition of thrombin [38,39]. The GAG binding site of HC-II has been identified as the D-helix region [40–52]. The effects of mutagenesis of thrombin anion-binding exosites-1 and -2 on GAG acceleration of the HC-II–thrombin reaction suggest that the template mechanism makes only a minor contribution to heparin acceleration and no contribution to dermatan sulfate acceleration [42,45,49,50,53]. Instead, the major mechanism of GAG enhancement appears to be allosteric and uses conformational activation of the serpin. Heparin cofactor II possesses a unique amino-terminal extension that contains two tandem repeats rich in acidic amino acids with two sulfated tyrosines (contained in the region encompassed by residues 54–75). The acidic region repeats of HC-II are significantly homologous to the carboxyl-terminal sequence of hirudin (the thrombin inhibitor from the medicinal leech), which binds to thrombin anion-binding exosite-1 [54,55]. Glycosaminoglycan binding to HC-II is thought to allosterically activate the serpin by displacing the acidic amino terminus from an intramolecular interaction with the basic GAG binding site and freeing it for binding to the thrombin anion-binding exosite-1 [44,46,47,51]. An alternative allosteric mechanism has been suggested based on the recently described crystal structures of both native HC-II (Fig. 1) and HC-II complexed with catalytically inactive S195A thrombin [40]. In a surprising revelation, the native HC-II structure showed that the hinge of the reactive centre loop is partially inserted into the A β-sheet, similar to the situation seen in native AT, and the short segment of the amino terminus that was visible suggested that this region might be interacting with an alternative basic site on the serpin near the reactive loop [40]. Thus, GAG activation of HC-II was proposed to resemble AT, where GAG binding to the D-helix causes the expulsion of the buried reactive centre loop hinge from the A β-sheet, which in turn alters the amino-terminal tail interaction to promote binding to the thrombin anion-binding exosite-1. Regardless of which allosteric mechanism turns out to be more correct, it is obvious that release of the amino-terminal portion of HC-II to bind to thrombin anion-binding exosite-1 is a primary part of the allosteric activation mechanism.

For many years, the physiological activator of HC-II has been assumed to be extravascular dermatan sulfate [56–64], which would complement the intravascular effect of heparan sulfate binding to AT. Maimone and Tollefsen [60] described the structure of a high affinity dermatan sulfate hexasaccharide that bound to HC-II. Furthermore, dermatan sulfate proteoglycans on the surface of cultured fibroblasts and vascular smooth muscle cells and purified biglycan and decorin dermatan sulfate proteoglycans accelerate the rate of thrombin inhibition by HC-II [58,62]. Dermatan sulfate proteoglycans in the extracellular matrices and on certain cell surfaces may localize HC-II to sites appropriate for inhibiting thrombin. The murine knock-out studies of HC-II revealed a role for this serpin in regulating thrombin formation, especially in the arterial circulation [65]. Recent studies using HC-II deficient mice confirmed that the antithrombotic effect of exogenously added dermatan sulfate is due to its interaction with HC-II [66]. Collectively, these findings imply that HC-II has a major role in thrombin regulation at extravascular tissue sites following vessel injury.

Protein C inhibitor (also named PAI-3)

Protein C inhibitor antigen is found in human blood plasma (3.6–6.8 µg·mL−1 or ≈ 90 nm) [67], numerous other human tissues, in urine and several other body fluids (e.g. tears, saliva, cerebral spinal fluid, amniotic fluid), and in seminal fluid at ≈ 200 µg·mL−1, which is almost 40 times the amount in plasma [68–76]. The mature protein is 57 kDa [77,78]. The PCI reactive site peptide bond is Arg354-Ser355, and PCI displays a protease inhibition profile for numerous ‘argininespecific’ serine proteases, including trypsin, thrombin (in the absence and presence of thrombomodulin), activated protein C, acrosin, kallikrein, urokinase, tissue plasminogen activator and factor XIa [69–71,74,77–91]. The GAG binding site in protein C inhibitor appears to be localized not to the D-helix as in AT and HC-II, but to the H-helix region, with possible contributions from the N-terminal A+-helix region [92–96]. Both regions have sequences of basic residues consistent with a general heparin-binding consensus sequence motif. Mutagenesis of four basic residues in the H-helix, Lys266, Arg269, Lys270 and Lys273, in recombinant PCI has shown that all of these residues are important for heparin binding. With the recent report of the crystal structure of cleaved-PCI (Fig. 1) [97], there are clearly other basic residues near the primary H-helix GAG binding site that probably contribute to GAG/polyanion binding (including Arg26, Arg27, Arg213, Arg234, Arg229, Lys255 and Arg362).

In contrast to both AT and HC-II, there is no evidence for an allosteric activation mechanism and instead the mechanism appears to involve only a ternary complex with heparin bridging the serpin and protease [61,98]. As found for other serine proteases with γ-carboxyglutamic acid domains, heparin bridging of PCI and activated protein C is only modest unless calcium ions are present to bind the acidic domain and prevent its interaction with the heparin-binding site of the protease [89,91]. Thrombin is also inhibited by PCI and the inhibition is accelerated by heparin, but the heparin-enhanced rate does not appear to be physiologically relevant when compared to thrombin inhibition rates by both AT-heparin/heparan sulfate and HC-II-heparin/dermatan sulfate. A more physiologically significant rate of thrombin inhibition by PCI results when thrombin binds to thrombomodulin, the endothelial cell receptor/proteoglycan [84,86,90,99]. This is consistent with PCI regulating the anticoagulant protein C pathway, because the thrombin-thrombomodulin complex initiates this pathway by the activation of zymogen protein C to an anticoagulant serine protease. Interestingly, the increased rate of thrombin inhibition when bound to thrombomodulin that is measured with PCI does not involve the chondroitin sulfate moiety of thrombomodulin, but rather is apparently promoted by the epidermal growth factor-like domains of thrombomodulin.

Like HC-II, PCI has a broad GAG/polyanion specificity for acceleration of protease inhibition reactions [61,98]. A variety of GAGs and polyanions [including heparin, low molecular weight heparin, heparan sulfate, fucoidan, and other polyanions (phosvitin)] accelerate both thrombin and activated protein C inhibition by PCI; this is consistent with a relatively nonspecific heparin-binding site in protein C inhibitor. In contrast to AT, there is no evidence for any sequence-specific binding of heparin/heparan sulfate to PCI. In a cell-derived series of studies, heparan sulfate-containing proteoglycans were involved in binding to PCI using cultured human epithelial kidney tumor cells (TCL-598); furthermore, dermatan sulfate-containing proteoglycans were implicated in binding PCI in the extracellular matrix [70,100]. However, identification of the physiological proteoglycan responsible for acceleration of PCI's activity in vivo has not been clearly identified.

Overall conclusions and future directions

It is readily apparent that the action of the three serpins discussed here is highly controlled by interactions with GAGs. There are differences in the way which each of the three serpins bind the GAGs, but common to each is that GAGs increase the rate of interaction with target proteases. The interaction of AT with heparin is clearly the most understood in structural terms, although a number of elements of the conformational change brought about in AT by heparin remain a little unclear. Further structural studies are obviously required to fully understand the interaction of HC-II and PCI with cognate GAGs.

The action of GAGs, in particular that of heparin on AT, has been very successfully exploited in clinical practice and this has been brought to even greater sophistication by the introduction of synthetic analogues of heparin. There is still a great need to fully understand the situation in the physiological setting, however. It is not completely clear when each serpin comes into contact with the GAGs that modulate its activity and how this leads to the vital regulation which evidently occurs. This is a major area of basic research for the immediate and medium term future.


This work was supported by the National Health & Medical Research Council of Australia, the Australian Research Council, the National Heart Foundation of Australia (to RNP and AMB), Research Grants HL-06350 and HL-32656 from the National Institutes of Health (to FCC), the Institut National de la Santé et de la Recherche Médicale of France and the Foundation pour la Recherche Médicale of France (to BFLB).