The Mechanism Underlying Activation of Factor IX by Factor XIa

Abstract

Factor XI (fXI) is the zymogen of a plasma protease, factor XIa (fXIa), that contributes to thrombin generation during blood coagulation by proteolytic conversion of factor IX (fIX) to factor IXaβ (fIXaβ). There is considerable interest in fXIa as a therapeutic target because it contributes to thrombosis, while serving a relatively minor role in hemostasis. FXI/XIa has a distinctly different structure than other plasma coagulation proteases. Specifically, the protein lacks a phospholipid-binding Gla-domain, and is a homodimer. Each subunit of a fXIa dimer contains four apple domains (A1 to A4) and one trypsin-like protease domain. The A3 domain contains a binding site (exosite) that largely determines affinity and specificity for the substrate fIX. After binding to fXIa, fIX undergoes a single cleavage to form the intermediate fIXα. FIXα then rebinds to the A3 domain to undergo a second cleavage, generating fIXaβ. The catalytic efficiency for the second cleavage is ~7-fold greater than that of the first cleavage, limiting fIXα accumulation. Residues at the N-terminus and C-terminus of the fXIa A3 domain likely form the fIX binding site. The dimeric conformation of fXIa is not required for normal fIX activation in solution. However, monomeric forms of fXI do not reconstitute fXI-deficient mice in arterial thrombosis models, indicating the dimer is required for normal function in vivo. FXI must be a dimer to be activated normal by the protease fXIIa. It is also possible that the dimeric structure is an adaptation that allows fXI/XIa to bind to a surface through one subunit, while binding to its substrate fIX through the other.

INTRODUCTION

The serine protease factor XIa (fXIa) contributes to blood clot formation at a wound site primarily by converting factor IX (fIX) to the protease fIXaβ [1]. Two structural features distinguish fXIa from other plasma proteases required for hemostasis. First, fXIa lacks an N-terminal calcium-binding Gla-domain [1–3]. As a result, fXIa-mediated activation of fIX is not a phospholipid-dependent processes. Second, FXIa is a dimeric protein comprised of identical 80 kDa subunits [1,4]. As this unusual structure suggests, factor XI (fXI, the zymogen of fXIa) is not closely related to the vitamin K-dependent serine proteases that form the core of the vertebrate blood coagulation mechanism [5]. FXI is found only in mammals, and is the result of a duplication of the gene for plasma prekallikrein (PK), the zymogen of the kininogenase α-kallikrein [6]. In this review we will discuss structural features of fXIa that are required for fIX activation, and the importance of the protease’s dimeric structure.

FACTOR XI STRUCTURE

Each fXIa subunit contains four apple domains (A1 to A4) and a trypsin-like serine protease domain (Figure 1A) [2,3]. Apple domains are a type of PAN (plasminogen, apple, nematode) module, with homology to the N-terminal domains of plasminogen and hepatocyte growth factor [7]. In the fXI structure reported by Papagrigoriou et al. [4], the apple domains form a 60Å wide planar platform on which the catalytic domain rests (Figure 1B). The apple domain platforms of two fXI subunits join at a 70° angle through a hydrophobic interface on the A4 domains to form a dimer (Figure 1C) [4]. There is a disulfide bond between Cys321 residues on each subunit in most species, but this is not required for dimer formation [8–10]. Binding sites on the fXI A1 domain for thrombin [11,12]; on the A3 domain for factor IX [13,14], platelet glycoprotein 1b [15] and polyanions such as heparin [16] and polyphosphate [17,18]; and on the A4 domain for factor XIIa [19] have been described. A distinct polyanion-binding site has been identified on the catalytic domain [18]. Factor XI circulates in plasma in a complex with the glycoprotein high molecular weight kininogen (HK) [1,20]. The D6 domain of HK appears to bind to a groove formed by residues from A2, with contributions from A1 and A4, on the face of the apple domain platform opposite the catalytic domain [1,21].

Figure 1. Structures for factor XI and factor IX.

(A) Schematic diagram of the primary and disulfide bond structures of a human fXI subunit (after McMullen et al. [3]) showing the apple domains (A1 to A4) and the trypsin-like catalytic domain. Residues comprising the catalytic triad are shown in red. Cys321 forms an inter-chain disulfide bond in the dimer. Cleavage after Arg369 converts the zymogen subunit to the active protease. (B and C) Models of the fXI monomer (B) and dimer (C) showing the relative positions of the apple domains (numbered 1 through 4) and the catalytic domains (CD). Hydrophobic residues in the A4 domain form the dimer interface. Models based on the crystal structure of fXI [4] (D) Schematic diagram of factor IX. The protein is comprised of a N-terminal calcium-binding Gla-domain, two epidermal growth factor (EGF) domains, and activation peptide (blue) and a catalytic domain (CD). Factor IX is cleaved initially by factor XIa after Arg145, forming the intermediate factor IXα, followed by cleavage after Arg180, forming the protease factor IXaβ. (E) Surface representation of the human factor XI A3 domain, showing positions of positively (blue) and negatively (red) charged side-chains. Positions of Arg184, Asp185, and an adjacent hydrophobic pocket not present on the PK A3 domain are indicated. (F) Structure of the factor XI monomer showing A3 as a topological model; apple domains A1, A2, and A4 as blue ribbon drawings; and the catalytic domain as a black ribbon drawing. Note that the area comprised of Arg184 (indicated in deep blue), Asp185 (red), and the hydrophobic pocket (orange) are covered by components of the catalytic domain in the zymogen structure.

Activation of a fXI subunit involves proteolysis of the Arg369-Ile370 bond [1–3]. The reaction is catalyzed by fXIIa [1,22], various forms of thrombin (meizo-, α-, β-, and γ-thrombin) [23,24], or by fXIa in the presence of polyanions (autoactivation) [22]. Forms of fXI with one (1/2-fXIa) or two (fXIa) activated subunits have been described [25]. While a structure for full-length fXIa has not been reported, major conformational changes appear to accompany activation that result in expression of a binding site for the fXIa substrate fIX [26].

FACTOR XIa ACTIVATION OF FACTOR IX

Conversion of fIX to fIXaβ requires proteolytic cleavages after Arg145 and Arg180, releasing an activation peptide (Ala146-Arg180, Figure 1D) [1,13,27]. This calcium-dependent reaction is catalyzed by factor VIIa (fVIIa) in the presence of tissue factor and phosphatidylserine-rich phospholipid [27,28], or by fXIa in a phospholipid-independent process [1,9,13,14,25,29]. FIX bond cleavage is ordered, with the Arg145-Ala146 bond cleaved first, forming the inactive intermediate fIXα [32–35]. FIXα accumulates during fIX activation by fVIIa, indicating the second cleavage after Arg180 is rate limiting [27,28]. In contrast, fIXα does not accumulate during fIX activation by fXIa [13,25,29], a phenomenon that cannot be accounted for by the dimeric structure of fXIa, as fIXα does not accumulate during activation catalyzed by monomeric fXIa [10] or 1/2-fXIa [25].

Recent work has shed light on details of the mechanism by which fXIa activates fIX. FIX first binds to the fXIa A3 domain followed by engagement at the protease active site and cleavage of the Arg145-Ala146 bond to form fIXα [13,14]. Cleavage after Arg145 facilitates cleavage of the Arg180-Val181 bond, forming fIXaβ. Catalytic efficiency for the second cleavage is 7-fold greater than for the first cleavage, explaining the low accumulation of fIXα. A significant portion of fIXα is released from fXIa, and must rebind to the A3 domain for the second cleavage to occur. The mechanism depends on substrate (fIX and fIXα) binding to an exosite on the fXIa A3 domain. Replacing the A3 domain with the A3 domain from the homolog PK (fXIa/PKA3) results in a marked reduction in catalytic efficiency for cleavage of both fIX bonds, but with a disproportionately greater deleterious effect on cleavage after Arg180, leading to significant fIXα accumulation. A similar defect is observed during fIX activation by the isolated fXIa catalytic domain, or when fIX is activated by fXIa in the absence of calcium ions. Taken as a whole, the kinetic data show that calcium-dependent fIX activation requires fIX and fIXα to bind to the fXIa A3 domain before catalysis of the appropriate bonds. This hypothesis is supported by binding studies showing that neither fIX nor fIXα bind to fXIa in the absence of calcium, or to the chimera fXIa/PKA3 even when calcium is present [13].

THE FACTOR IX BINDING SITE ON FACTOR XIa

A fIX binding site is not available on zymogen fXI, therefore, the binding site must be formed or unmasked by structural changes that occur when fXI is converted to fXIa [25]. Replacement of Ile183, Arg184 and Asp185 at the N-terminus of the fXIa A3 domain with alanine residues results in a protease with a similar defect in fIX activation to fXIa/PKA3 [14]. These residues form a charged ridge adjacent to hydrophobic patches on the surface of the A3 domain that are not present in PK (Figure 1E). This area is not exposed to the solvent phase in the zymogen fXI structure (Figure 1F) [4,14], consistent with the observation that fIX does not bind to fXI. Arg184 forms salt bridges with residues on the catalytic domain in fXI, holding the catalytic domain like a cover over the putative fIX binding site (Figure 1F) [4,14]. While a structure is not available for fXIa, the working hypothesis is that Arg184 releases the catalytic domain during conversion of fXI to fXIa exposing the underlying fIX binding site.

The fIX-fXIa interaction requires the calcium-binding fIX-Gla domain [14]. Results from recent mutagenesis studies suggest that residues in the phospholipid-binding region of the Gla-domain (the Ωloop, residues 4 to 11) are required for binding [14]. The Ω-loops of vitamin K-dependent proteases facilitate binding of the protein to phospholipid membranes through charged and hydrophobic interactions [27]. In this regard, it is interesting to note that the topology of the putative fIX binding site on the fXIa A3 domain is consistent with a binding interaction involving charged and hydrophobic components. If the fXIa A3 domain engages the fIX Ω-loop, then fXIa should compete with phospholipid for binding to fIX. Using surface plasmon resonance techniques, we recently demonstrated that fXIa, but not fXIa/PKA3, effectively prevents fIX from binding to a phosphatidylserine-rich surface (Messer, Bajaj, Geng and Gailani, unpublished observation).

THE FACTOR XIa DIMER

The dimeric structure of fXIa is unique among coagulation proteases, and is unusual for a trypsin-like enzyme. The hydrophobic residues on the A4 domain of human fXI that form the dimer interface [1,4,10] are conserved across species, indicating functional importance. However, a role for the dimer in protease function has been difficult to define. Monomeric forms of fXIa activate fIX similarly to dimeric fXIa [10,13,25]. Recent work indicates that fXI must be a dimer to be activated normally by fXIIa [10,22], but not by thrombin or autoactivation [22]. The physiologic significance of this is not clear, as fXIIa is not required for hemostasis. Dimeric and monomeric forms of fXI were tested in fXI-deficient mice in an arterial thrombosis model that requires fXI activation by fXIIa [22]. Dimeric fXI reconstitutes the wild type phenotype in this model, but fXI monomers failed to do so, perhaps because of the defect in fXI activation by fXIIa. Alternatively, fXIa may bind to a platelet through one of its subunits, while interacting with fIX through the other subunit [30]. This could also explain the defect in mice reconstituted with monomeric fXI in the thrombosis model. While fXIa lacks a phospholipid-binding Gla-domains, it does bind to platelets through an interaction involving the A3 domain and platelet glycoprotein 1b [15]. This raises the possibility that one A3 domain of the dimer may bind to the platelet, while the other engages fIX. Additional work is required to verify or disprove this hypothesis.