A serine loop in tissue factor mediates substrate selectivity by the tissue factor – factor VIIa complex

Abstract

Background:

The tissue factor – factor VIIa (TF-FVIIa) complex is the physiologic activator of blood clotting and plays a major role in many thrombotic diseases. TF-FVIIa drives clotting through proteolytic cleavage of its major protein substrates, factor IX (FIX) and factor X (FX). However, it remains unclear how TF-FVIIa exhibits selectivity between these substrates. We previously showed that TF residues adjacent to the putative substrate binding site of TF (“exosite”) facilitate FX activation, but the role of these residues in substrate selectivity had not been tested.

Objectives:

We hypothesized that a TF serine loop (residues S160-S163) mediates substrate selectivity by the TF-FVIIa complex.

Methods:

We generated TF serine loop and exosite mutants. The mutants were tested in FIX and FX enzyme activation assays as well as thrombin generation assays.

Results:

Changes in the length of the serine loop affected rates of FIX and FX activation very differently. FX activation was decreased by up to 200-fold when the loop length was changed by just one residue. In contrast, FIX activation was largely unaffected. Substrate selectivity was also detected in thrombin generation assays. Activation assays with TF serine loop and exosite double mutants revealed that the serine loop has no effect on the exosite during FIX activation. In contrast, the serine loop regulates the exosite during FX activation.

Conclusions:

Our results provide new insights into how the TF-FVIIa complex actively selects between its major protein substrates, which is mediated by a TF serine loop.

Introduction

The complex of tissue factor (TF) and factor VIIa (FVIIa) triggers blood clotting in hemostasis and many thrombotic diseases.^1,2 TF is highly expressed in atherosclerotic plaques³ and plays a major role in thrombosis associated with myocardial infarction⁴ and ischemic stroke.⁵ TF is an integral membrane glycoprotein that is present on the surface of many extravascular cells.⁶ Upon vascular damage, TF is exposed to plasma and binds the serine protease FVIIa (Fig. 1A), yielding a two-subunit enzyme (TF-FVIIa) in which TF is the regulatory subunit and FVIIa is the catalytic subunit. In blood clotting, TF-FVIIa initiates clotting by proteolytically activating two substrates, the zymogens factor IX (FIX) and factor X (FX), which then propagate the clotting cascade (Fig. 1A). Although FX is the preferred substrate for TF-FVIIa under most conditions, it is unclear how TF-FVIIa selects between FIX and FX.⁷

Figure 1. TF pathway of blood clotting and the structure of TF-FVIIa.

(A) Upon vascular damage, TF is exposed to blood after which the TF-FVIIa complex activates its major substrates, FIX and FX by limited proteolysis. Downstream proteolytic reactions lead to thrombin generation and fibrin clot formation. (B) The crystal structure (PDB entry: 3TH2³³) of the isolated TF ectodomain (teal) in complex with FVIIa (green), arranged on a cartoon of the phospholipid bilayer (grey). Bound divalent metal ions are Ca²⁺ (yellow spheres) and Mg²⁺ (orange spheres). The side chains of the TF exosite region are rendered in space-filling (red) and are located near the C-terminus of the TF ectodomain. The TF serine loop S160-S163, which is also rendered in space filling (dark blue), is adjacent to both the exosite and the proposed location of the phospholipid membrane.

Activation of both FIX and FX by TF-FVIIa is thought to be promoted by a solvent-exposed surface (“exosite”) near the C-terminus of TF.^7–9 Defined by mutagenesis studies, the exosite appears to function as a putative substrate-binding site, located some 60 Å away from the active site of FVIIa (Fig. 1B). The exosite comprises TF residues Y157, K159, S163, G164, K165, K166 and Y185.⁷ Mutation of TF residues in the exosite, especially K165 and K166, decreases FX and FIX activation by up to 100-fold, while FVIIa binding and allosteric activation are not impaired.^7,8,10 Although there are quantitative differences in the impact of specific exosite mutations on FIX versus FX activation, the magnitude of these differences is relatively small.⁷

The TF exosite partially overlaps a stretch of four consecutive serine residues, S160-S161-S162-S163 (Fig. 1B), that form a solvent-exposed loop which is unresolved in most soluble TF crystal structures.^11,12 However, this loop connects to a region on the TF ectodomain that we previously proposed to interact with phosphatidylserine (PS) headgroups.^13,14 In the present study, we hypothesized that this flexible S160-S163 loop might function as a linker between the TF exosite and the nearby region that interacts with PS residues, and thus might mediate allosteric activation of the TF exosite when this protein engages PS. Accordingly, we hypothesized that changing the length of this loop could disrupt this allosteric linkage.

To test this hypothesis, we generated an extensive set of TF exosite and serine loop mutants, including multiple deletions and insertions within the S160-S163 loop. We tested the activity of these mutants in substrate activation assays and thrombin generation assays. We now report the surprising result that, while increasing or decreasing the length of the S160-S163 loop had little to no effect on the rate of FIX activation, inserting or deleting even a single Ser residue within this loop decreased FX activation by up to 200-fold. These findings suggest that the precise length of the S160-S163 loop strongly contributes to substrate selectivity by TF-FVIIa.

Materials and Methods

Materials

Materials were from the following sources: 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) and 1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-L-serine (POPS), Avanti Polar Lipids (Alabaster, AL); Bio-Beads SM-2 absorbent, Bio-Rad Laboratories (Hercules, CA); recombinant human FVIIa, American Diagnostica (now Sekisui Diagnostics, Lexington, MA); FIX-immunodepleted plasma and purified human FIX, FIXa and FX, Haematologic Technologies (Essex Junction, VT); Pefachrome FIXa, DSM Nutritional Products Ltd., Branch Pentapharm (Parsippany, NJ); methoxycarbonyl-D-Nle-Gly-Arg-pNA acetate salt (FXa substrate), Bachem (Bubendorf, Switzerland); pooled normal plasma (PNP) and FVIII-deficient plasma (< 1% activity), George King Bio-Medical (Overland Park, KS); Thrombin Calibrator and FluCa-Kit (including Fluo-Buffer and Fluo-Substrate), Diagnostica Stago (Parsippany, NJ); and medium-binding microplates, Corning (Tewksbury, MA).

Production and relipidation of recombinant TF

Recombinant human membrane-anchored TF (residues 3–244) was expressed in Escherichia coli and purified essentially as described.¹⁵ Expression constructs for TF mutants with serine loop insertions and deletions had an N-terminal HPC4-epitope tag¹⁶ and were generated by GenScript Biotech (Piscataway, NJ). They were purified by affinity chromatography using an HPC4-antibody column. TF mutants with serine loop substitutions had a N-terminal 6×His-tag and were generated using the Q5 site-directed mutagenesis kit from New England Biolabs (Ipswich, MA). These mutants were purified using nickel-NTA affinity chromatography. TF-liposomes were prepared by incorporating TF into phospholipid vesicles of varying composition (POPS and POPC) as described,¹⁷ using 15 mM deoxycholate as the detergent.

Activation of FX and FIX

For most assays, initial rates of FX activation by TF-FVIIa were quantified using a continuous (one-stage) FX activation assay^18,19 in 96-well plates with the following modifications. TF-liposomes (0.6 nM TF, with 10 μM total phospholipid) and FVIIa (3 to 150 pM) were incubated in HBSA buffer (20 mM HEPES pH 7.4, 100 mM NaCl, 0.1% bovine serum albumin, 0.02% NaN₃) with 5 mM CaCl₂. Reactions were initiated by addition of 0.5 mM FXa substrate and 30 nM FX. The rate of change in A₄₀₅ was measured at ambient temperature using a Spectramax 96-well spectrophotometer (Molecular Devices, San Jose, CA). Initial rates of FX activation in the presence of TF mutants were normalized to those measured in the presence of wild-type (WT) TF with the same phospholipid composition. K_M and k_cat values for FX activation were measured in discontinuous (two-stage) assays in 96-well plates. TF-liposomes and FVIIa (same concentrations as for the one-stage assay) were incubated in HBSA with 5 mM CaCl₂ at 37° C for 5 min, after which we initiated reactions by adding 25 to 400 nM FX. Timed, 10 μL aliquots were taken over 20 min and quenched on ice in a microwells containing 80 μL stop buffer (made by mixing, in each well, 50 μL of 2× concentrated HBSA plus 20 mM EDTA, and 30 μL water). For the second stage, plates were warmed to ambient temperature and 10 μL of 5 mM FXa substrate (0.5 mM final) was added per well, after which the rate of change in A₄₀₅ was measured. Rates of chromogenic substrate hydrolysis were then converted to FXa concentrations by reference to a standard curve. To extract K_M and k_cat values, initial rates of FX activation were plotted versus FX concentration, to which the Michaelis-Menten equation was fitted by nonlinear regression.

Initial rates of FIX activation by TF-FVIIa were quantified using a discontinuous (two-stage) FIX activation assay in 96-well plates. TF-liposomes (10 nM TF, with 175 μM total phospholipid) and FVIIa (1 nM) were incubated in HBSA with 5 mM CaCl₂ at 37° C for 5 min. Reactions were initiated by addition of 2 μM FIX. Timed, 10 μL aliquots were taken over the course of 5 min and quenched on ice in a microwell containing 80 μL stop buffer (made by mixing, in each well, 50 μL of 2× concentrated HBSA plus 20 mM EDTA, and 30 μL of 75% (v/v) ethylene glycol). For the second stage, the plates were warmed to ambient temperature and 10 μL 5 mM Pefachrome FIXa (0.5 mM final) was added per well, after which the rate of change in A₄₀₅ was measured. Rates of chromogenic substrate hydrolysis were then converted to FIXa concentrations by reference to a standard curve. Initial rates of FIX activation in the presence of TF mutants were normalized to those measured in the presence of WT TF with the same phospholipid composition. Measurement of rates of FIX activation for determining k_cat and K_M values used the same two-stage assay configuration, with 0.1 to 2 μM FIX and with 10 μL aliquots taken over the course of 10 min. Initial rates of FIX activation were then plotted versus FIX concentration, to which the Michaelis-Menten equation was fitted by nonlinear regression.

Thrombin generation

Thrombin generation in clotting plasma was measured using the calibrated automated thrombogram (CAT) assay (using Thrombinoscope reagents and software; Diagnostica Stago).^20,21 TF-liposomes (made with either WT TF or the 3S2T mutant) were diluted in HBSA to yield TF concentrations that generated a similar lag time. For 20% POPS/80% POPC liposomes, this was 60 pM WT TF or 15 nM 3S2T TF as well as 6 pM WT TF or 1.5 nM 3S2T TF. For 10% POPS/90% POPC liposomes, this was 30 pM WT TF or 30 nM 3S2T TF. Final phospholipid concentrations in all cases were 60 μM (made by supplementing with TF-free liposomes of the same phospholipid composition). CAT assays were performed in 96-well plates by mixing 20 μL TF and 80 μL plasma (PNP, FVIII-deficient or FIX-immunodepleted plasma) at 37° C for 10 min. The reaction was then started by adding pre-warmed 20 μL FluCa buffer. The change in fluorescence was measured using a Fluoroskan Microplate Fluorometer (ThermoFisher Scientific, Waltham, MA), and converted to thrombin concentrations using Thrombinoscope software according to the manufacturer’s recommendations. Measured parameters (lag time, peak thrombin, and time to peak (ttPeak) for deficient plasmas were normalized to PNP for WT and mutant TF to generate a fold-change.

Production of XK1

XK1, a slow, tight-binding inhibitor of the TF-FVIIa complex, is an engineered hybrid protein consisting of the human FX light chain linked to the first Kunitz (K1) domain of human tissue factor pathway inhibitor.²² The XK1 coding sequence was prepared in the pcDNA3.1(+) expression vector by GenScript Biotech (Piscataway, NJ) and transfected into HEK293/VKOR cells (HEK293 cells that overexpress vitamin K 2,3-epoxide reductase C1, the expression vector for which was a kind gift of Darrell Stafford). These cells were cultured in cell factories (Corning HYPERFlask M Cell Culture Vessel) in a 1:1 mixture of DMEM and F12 media supplemented with L-Glutamine, 15 mM HEPES (Corning), 10 μg/mL vitamin K1 (Phytonadione, Henry Schein Medical, Melville, NY), 2 μg/mL puromycin and 400 μg/mL G-418. XK1 levels in culture supernatants were measured using the FX ELISA from Enzyme Research Laboratories (South Bend, IN). XK1 was affinity-purified from cell supernatants using immobilized 4G3, a Ca²⁺-dependent mouse monoclonal antibody targeting the human FX light chain.²³ Fully carboxylated XK1 was subsequently separated from any undercarboxylated protein via anion-exchange chromatography as previously described for recombinant factor VII.²⁴

Inhibition of TF-FVIIa by XK1

The ability of XK1 to inhibit the TF-FVIIa complex was assessed by quantifying its effects on the rate of FX activation, using a modification of the continuous FX activation assay described above. TF-liposomes (0.6 nM TF, with 10 μM total phospholipid) and FVIIa (3 – 150 pM) were incubated in HBSA with 5 mM CaCl₂. Varying concentrations (0 – 100 nM) of XK1 were added, and after 5 min, reactions were initiated by addition of 0.5 mM FXa substrate and 30 nM FX. The rate of change in A₄₀₅ was measured at ambient temperature. FX activation rates for WT or mutant TF in the presence of different XK1 concentrations were normalized to activation rates in the absence of XK1.

Results

Precise length of the TF serine loop is essential for activation of FX but not FIX

We performed mutagenesis studies to investigate the role of the TF serine loop S160-S163 in substrate activation. Multiple sequence alignment showed that the TF serine loop region is highly conserved, especially among mammals (Fig. 2A). Thus, S161, S162 and S163 are almost always serine or the similar amino acid, threonine, while S160 is less well conserved. Notably, the length of the solvent exposed loop is always exactly four amino acid residues. Based on the alignment, we generated a total of fifteen TF serine loop mutants (Table 1).

Figure 2. Changes in the length of the TF serine loop differentially effect rates of FIX vs. FX activation.

(A) Multiple sequence alignment of TF serine loop region was performed using the online tool T-Coffee (https://www.ebi.ac.uk/Tools/msa/). The alignment shows TF residues 156–168 (numbering according to human TF). Serine loop residues S160-S163T are marked by red box. (B-E) Initial rates of FIX and FX activation by TF-FVIIa (using WT or mutant TF in 20% POPS/80% POPC liposomes) were measured, and the rates obtained with TF mutants were normalized to those with WT TF. Relative rates that were more than twofold lower than WT are indicated with a dagger. (B) Relative rates of FX activation with TF mutants S160T, S161T, S162T and S163T. (C) Relative rates of FIX activation with TF mutants S160T, S161T, S162T and S163T. (D) Relative rates of FX activation with TF mutants of different serine loop length or composition. (E) Relative rates of FIX activation with TF mutants of different serine loop length or composition. (F, G) Kinetics of the activation of FX (panel F) or FIX (panel G) supported by WT TF or the 3S2T mutant. Initial rates of substrate activation were plotted versus substrate concentration, to which the Michaelis-Menten equation was fitted. All data are mean ± SE (n ≥ 3).

Table 1.

TF serine loop mutations

TF mutant	Amino acid sequence*	Change in loop length	FIX/FX ratio**
WT (4S)	KSSSSG	0	1
S160T	KTSSSG	0	1.6
S161T	KSTSSG	0	1
S162T	KSSTSG	0	2.4
S163T	KSSSTG	0	3.9
2S	KSS - - G	−2	28
3S	KSSS - G	−1	24
5S	KSSSSSG	+1	111
3S2T	KSSSTTG	+1	147
4ST	KSSSSTG	+1	121
6S	KSSSSSSG	+2	63
7S	KSSSSSSSG	+3	49
K159S	SSSSSG	0	31
G164S	KSSSSS	0	3

^*Amino acid sequences of human TF from positions 159–164 are given. Substitutions are in bold, insertions are in bold underline, and deletions are indicated with dashes.

^**FIX/FX ratios were calculated for each TF mutant by dividing the relative rate of FIX activation by the relative rate of FX activation, from the values plotted in Fig. 2B–E.

The first set of mutants tested how replacing each of the four serine residues individually with threonine affected substrate activation rates. The S161T substitution was without effect on FX activation, while the S160T, S162T and S163T substitutions decreased the rate of FX activation by up to 3-fold (Fig. 2B). On the other hand, replacing any of these four serine residues with threonine had little effect on the rate of FIX activation (Fig. 2C).

We next tested how changing the length of this loop affected substrate activation rates. We therefore generated mutants with deletions of one or two serine residues (termed 2S or 3S), insertion of one, two or three serines (termed 5S, 6S or 7S), insertion of one threonine residue after serine 163 (termed 4ST) or replacement of serine 163 with threonine together with insertion of an additional threonine (termed 3S2T). To create longer runs of serine residues without insertions, we also replaced the adjacent residues K159 and G164 with serine (K159S and G164S, respectively). We also attempted to delete 3 serines or all 4 serines, but the resulting constructs did not express detectable protein and are therefore not listed in Table 1. The remaining TF mutants were successfully expressed, purified, and tested for their ability to support FIX and FX activation by TF-FVIIa. We found that any insertions or deletions within this serine loop profoundly diminished the rate of FX activation, with reductions ranging from 62.5-fold to 200-fold (Fig. 2D). On the other hand, changing the length of the serine loop had much more modest effects on the rate of FIX activation (Fig 2E). In particular, lengthening the serine loop was relatively benign toward FIX activation, as the 5S mutant supported essentially wild-type levels of FIX activation while the 6S and 7S mutants supported FIX activation at levels that were reduced less than 2-fold relative to wild type. On the other hand, shortening the serine loop resulted in somewhat diminished FIX activation rates (2.3-fold and 4.6-fold for the 2S and 3S mutants, respectively), although these reductions were far less than when FX was the substrate (Fig. 2D and 2E).

Increasing the length of the serine loop by one residue via inserting a threonine residue, with or without replacing Ser163 with threonine (the 4ST or 3S2T mutants, respectively), resulted in only a 1.3-fold reduction in the rate of FIX activation but caused the most profound reductions in the rate of FX activation (158-fold and 200-fold, respectively).

Finally, changing the run of serine residues from four to five in a row via substituting the adjacent lysine or glycine residues with serine (K159S or G164S) caused less than a 2-fold reduction in the rate of FIX activation but a 55-fold or 6-fold reduction in the rate of FX activation, respectively.

When we directly compared the relative reductions in the rate of FIX activation versus factor X activation (last column in Table 1), we found the largest differential in rates when just one serine residue was inserted in the loop and one serine was substituted to threonine (3S2T). This TF mutant exhibited almost a 150-fold greater reduction in the rate FX activation relative to its impact on FIX activation. Together, these results show that the precise length of the serine loop is crucial for TF-FVIIa-mediated FX activation, but far less important for FIX activation.

We investigated the role of the TF serine loop in greater detail by determining K_M and k_cat values for FX and FIX activation supported by WT TF versus the 3S2T mutant (Fig. 2F and 2G). For FX activation, the K_M was increased 3.4-fold for the 3S2T mutant relative to WT, while the k_cat was decreased over 50-fold (Table 2). In contrast, the K_M and k_cat values for FIX activation varied by only 1.4- and 1.2-fold when comparing the 3S2T mutant to WT TF (Table 2). Thus, the 3S2T mutation decreased the k_cat/K_M for FX by 181-fold while decreasing the k_cat/K_M for FIX by only 1.76-fold. These results demonstrate that FX activation supported by 3S2T is significantly reduced due to changes in substrate binding and turnover.

Table 2.

Kinetics of FX and FIX activation by TF-FVIIa

TF variant	Substrate	KM [nM]*	Fold increase in KM [3S2T/WT]	kcat [sec−1]*	Fold decrease in kcat [WT/3S2T]	kcat/KM [M −1sec−1]
WT	FX	48 ± 10	3.4	4.07 ± 0.23	51.3	8.5 × 107
3S2T	FX	163 ± 40		0.076 ± 0.008		4.7 × 105
WT	FIX	700 ± 233	1.4	0.21 ± 0.03	1.2	3.0 × 105
3S2T	FIX	974 ± 297		0.17 ± 0.03		1.7 × 105

^*K_M and k_cat values are derived from plots in Fig. 2F & 2G.

TF serine loop provides selectivity for FX over FIX

We further tested if the TF serine loop provides substrate selectivity in human plasma using calibrated automated thrombogram (CAT) assays.²⁰ In this study, we compared thrombin generation triggered by WT TF versus the 3S2T mutant (incorporated into liposomes with either 20% POPS/80% POPC or 10% POPS/90% POPC), using PNP, FIX-immunodepleted and FVIII-deficient human plasmas. Fig. 3 reports three parameters from the thrombin generation curves: lag time, time to peak (ttPeak), and peak thrombin. For each combination of TF and lipids, the parameters were normalized to those obtained with PNP. The absolute values of these parameters are available in Supplemental Fig. S1.

Figure 3. Substrate selectivity of TF serine loop mutants detected via thrombin generation assay.

WT or mutant TF were incorporated in liposomes containing 20% POPS/80% POPC or 10% POPS/90% POPC. TF-liposomes were incubated with PNP, FIX-deficient plasma or FVIII-deficient plasma and thrombin generation was measured over 60 min in CAT assays. Lag time, ttPeak and peak thrombin obtained with deficient plasmas were normalized to the values obtained with PNP for each type of TF. Changes in lag time, ttPeak and peak thrombin for deficient plasmas that were significantly different from PNP are indicated with asterisk (t-test, p < 0.05; ns = not significant). (A) Fold-change in lag time for WT versus 3S2T TF in 20% POPS/80% POPC. (B) Fold-change in lag time for WT versus 3S2T TF in 10% POPS/90% POPC. (C) Fold-change in ttPeak for WT versus 3S2T TF in 20% POPS/80% POPC. (D) Fold-change in ttPeak for WT versus 3S2T TF in 10% POPS/90% POPC. (E) Fold-change in peak thrombin for WT versus 3S2T TF in 20% POPS/80% POPC. (F) Fold-change in peak thrombin for WT versus 3S2T TF in 10% POPS/90% POPC. All data are mean ± SE (n ≥ 3).

Thrombin generation by the 3S2T mutant was significantly altered in FIX- and FVIII-deficient human plasmas relative to WT TF. Thus, the lag time was prolonged up to 1.5-fold in FIX- and FVIII-deficient plasmas relative to PNP for 3S2T in both lipid compositions, while the only significant prolongation in lag time with WT TF (relative to PNP) was seen with FIX-deficient plasma using TF in 10% POPS/90% POPC liposomes (Fig. 3A and 3B). A larger difference was encountered with the ttPeak parameter, which was prolonged up to 2-fold in FIX- and FVIII-deficient plasmas relative to PNP with the 3S2T mutant in both lipid compositions (Fig. 3C and D). In contrast, WT TF showed a significant prolongation only with FVIII-deficient plasma, and the magnitude of the prolongation was smaller than that observed with the 3S2T mutant. Similarly, the peak thrombin level was decreased 2- to 3-fold in FIX- and FVIII-deficient plasmas relative to PNP triggered with the 3S2T mutant in both lipid compositions (Fig. 3E and 3F). When triggered by WT TF, however, the peak thrombin level was only significantly reduced in FVIII-deficient plasma using TF in 20% POPS/80% POPC liposomes, or in FIX-deficient plasma using TF in 10% POPS/90% POPC liposomes. To investigate if thrombin generation was more FIX-dependent at lower TF concentrations, we reduced the TF concentrations another tenfold (to 1 pM WT and 0.25 nM 3S2T). Fig. S2 shows that both TF concentrations followed the same trends for lag time, ttPeak and peak thrombin in WT and 3S2T. The difference in ttPeak between WT and 3S2T was somewhat more amplified with lower TF concentrations, however.

These results demonstrate that triggering of the plasma clotting cascade by the 3S2T mutant exhibits a significantly increased dependence on FIX and FVIII for thrombin generation relative to when the clotting is triggered by WT TF.

Serine loop provides substrate selectivity through differential regulation of the TF exosite

The serine loop of TF is immediately adjacent to, and partially overlaps, the putative exosite region defined by prior mutagenesis studies. It also overlaps a region that we previously identified as a putative PS-binding region on the TF ectodomain.^13,14 We therefore hypothesized that the serine loop could act as a “connector” that regulates the activity of the adjacent substrate-binding exosite. To test this hypothesis, we generated TF constructs with mutations in both the serine loop (3S2T) and the exosite (either K165A, K166A, or the K165A- K166A double mutant). These TF mutants were incorporated into 20% POPS/80% POPC liposomes and tested in FIX and FX activation assays, normalized to the rates with WT TF (Fig. 4A and 4B). By themselves, the exosite mutations resulted in 11- to 104-fold reductions in the rate of FX activation but more modest (1.2- to 4.5-fold) reductions in FIX activation. The 3S2T mutant exhibited a 103-fold reduction in FX activation by itself, with some further reduction observed in the rate of FX activation in conjunction with the exosite mutants. On the other hand, the 3S2T mutation resulted in only a 1.2-fold reduction in FIX activation by itself, with further reductions when combined with the exosite mutations.

Figure 4. TF serine loop regulates the TF exosite for activation of FX but not FIX.

Initial rates of FIX and FX activation by TF-FVIIa were measured using WT or mutant TF in 20% POPS/80% POPC liposomes. (A) Relative rates of FX activation and (B) FIX activation with TF mutations in the exosite (K165A, K166A and K165A-K166A), serine loop (3S2T), or both the exosite and serine loop (3S2T-K165A, 3S2T-K166A and 3S2T-K165A-K166A). Activation rates in panels A and B with mutant TF were normalized to those obtained with WT TF. Numbers above the bars are fold-reduction in rate, relative to WT. (C,D) Replotting of the data from panels A and B to more clearly visualize the combination of exosite and serine loop mutations. In each panel, rates obtained with exosite mutants (K165A, K166A and K165A-K166A) are normalized to WT TF, while the rates obtained with combination exosite and serine loop mutants (3S2T-K165A, 3S2T-K166A and 3S2T-K165A-K166A) are normalized to the 3S2T mutant. Normalization to 3S2T in panels C and D is indicated by striped bars. Accordingly, numbers above the bars in panels C and D are the fold-reduction in rate, relative to WT for the exosite mutants (K165A, K166A and K165A-K166A), or relative to 3S2T for combination mutants (3S2T-K165A, 3S2T-K166A and 3S2T-K165A-K166A). (E) Ability of XK1 to inhibit TF-FVIIa. Relative rates of FX activation by TF-FVIIa were measured in the presence of increasing XK1 concentrations, using TF with mutations in exosite (K165A, K166A and K165A-K166A), serine loop (3S2T), or both (3S2T-K165A, 3S2T-K166A and 3S2T-K165A-K166A). All data are mean ± SE (n ≥ 3).

In order to clarify the combined effect of these mutations, we replotted the data from Fig. 4A and 4B in Fig. 4C and 4D, respectively. In this case, the FX or FIX activation data using the combined 3S2T and exosite mutations were normalized to the rates obtained with the 3S2T mutant. With FX as substrate (Fig. 4C), one can see that the exosite mutations that resulted in 11-, 18-, or 104-fold reductions in the activation rate instead resulted in a much more modest relative reduction of 1.8-, 2.1- or 2.8-fold when combined with the 3S2T mutation (as normalized to the rate with the 3S2T mutation alone). On the other hand, when FIX was substrate, the exosite mutations resulted in very similar fold reductions in activation rates with or without the 3S2T mutation (Fig. 4D). In summary, when FX was the substrate, the exosite mutations were epistatic to the serine loop mutations, whereas when FIX was the substrate, the effects of the exosite and serine loop mutations appeared to be independent. Overall, these results are consistent with the idea that the serine loop regulates the TF exosite during FX activation but not FIX activation.

On the other hand, we do note that the already substantial reductions in FX activation brought about the exosite mutations might mask the effects of the loop mutations. To address this question in another way, we performed FX activation assays in the presence of a TF-FVIIa inhibitor, XK1 (Fig. 4E). XK1 is a hybrid protein consisting of the FX light chain linked to the first Kunitz (K1) domain of tissue factor pathway inhibitor.²² It is a slow, tight-binding inhibitor that was designed to be recognized by TF-FVIIa as a pseudosubstrate.

In the presence of TF exosite mutants K165A, K166A and K165A-K166A, it required 7- to 16-fold higher XK1 concentrations to inhibit FX activation relative to WT TF (WT IC₅₀ = 44±2 pM, K165A IC₅₀ = 305±18 pM, K166A IC₅₀ = 394±88 pM, and K165A-K166A IC₅₀ = 693±34 pM). In contrast, in the presence of TF serine loop mutant 3S2T, 24-fold higher XK1 concentrations were needed to inhibit TF-FVIIa relative to WT TF (3S2T IC₅₀ = 1063±64 pM). Somewhat surprisingly, for TF with mutations in both the serine loop and exosite (3S2T-K165A, 3S2T-K166A and 3S2T-K165A-K166A) the IC₅₀ values for XK1 were actually somewhat lower than those observed with the 3S2T mutant alone (3S2T-K165A IC₅₀ = 717±35 pM, 3S2T-K166A IC₅₀ = 710±34 pM, and 3S2T-K165A-K166A IC₅₀ = 757±46 pM). These results are consistent with the idea that the exosite mutations are epistatic to the TF serine loop mutations regarding the recognition of FX as a TF-FVIIa substrate.

Discussion

This study probed how TF-FVIIa selects between its major protein substrates, FIX and FX. We now report that substrate selectivity is strongly influenced by the TF serine loop, S160-S163, with the precise length of this loop being crucial for FX activation but relatively unimportant for FIX activation. FIX activation was especially tolerant of insertions within this loop. To our knowledge, no prior study has reported TF mutations that differentially decrease substrate activation (FIX versus FX) to the extent observed here with serine loop mutations. Numerous studies have previously identified the importance of TF exosite residues in substrate activation,^7–10,25 with some exosite mutations having a bigger effect on activation of FX than FIX.^7,26 However, none of those studies reported such a strong influence on one substrate over another as we observed with the serine loop mutations in this study.

Previous mutational studies defined the putative substate-binding exosite of TF as including residues Y157, K159, S163, G164, K165, K166,^7–9 a sequence which partially overlaps the serine loop. We also note that nearby residues, like the more N-terminal parts of the serine loop and immediately adjacent residues are postulated to interact with PS in the phospholipid membrane.^13,14 Whether our findings indicate that the serine loop is actually part of the TF exosite or if it modulates the adjacent exosite is not entirely clear and will require further study. Curiously, we found that exosite mutations were epistatic to the serine loop mutations when FX was the substrate, but not when FIX was the substate. Elucidating the physical basis for this epistasis will take further study, but this finding is consistent with the notion that the serine loop mediates the regulation of the adjacent TF exosite when FX is the substrate, but not FIX. The activity of TF-FVIIa is tightly regulated by the presence PS in the membrane.^27,28 Recently, our lab showed that exosite-adjacent TF residues, which are predicted to interact with the phospholipid membrane,²⁹ are also involved in regulating substrate activation.^13,14 Thus, the serine loop may play a role in positioning of the exosite for optimal interaction with FX, via transducing information from interaction of the TF ectodomain with anionic phospholipid headgroups.

The TF-FVIIa complex is known to have additional protein substates, including FVII³⁰ and proteinase-activated receptors (PARs)—in particular, PAR2.^31,32 In further studies, it would be interesting to examine the role of the TF serine loop in recognizing these alternative macromolecular substrates.

Overall, our study provides insights into the mechanisms of substrate selectivity by TF-FVIIa, showing an unexpected role for the serine loop of TF.

Supplementary Material

sup 01

Figure S1. Absolute values for peak thrombin, ttPeak, lag time and endogenous thrombin potential in thrombin generation assays.

Figure S2. Effect of different TF concentrations (WT and 3S2T mutant) on thrombin generation.

Essentials.

• How the tissue factor – factor VIIa complex selects between different substrates is not well understood

• We investigated a serine loop in tissue factor and its role in substrate selectivity

• The tissue factor serine loop is selective for factor X over factor IX

• Substrate selectivity is facilitated by differential regulation of the nearby tissue factor exosite