Characterization of the enzymatic activity of the serine protease domain of Factor VII activating protease (FSAP)

Factor VII (FVII) activating protease (FSAP) is a circulating serine protease. Human genetic studies, based on the Marburg I (MI) (Gly221Glu, chymotrypsin numbering system) polymorphism, implicate FSAP in the pathogenesis of many diseases. Here, we describe the molecular and functional changes caused by the Gly221Glu substitution in the 220 loop using recombinant proteins expressed in E. coli. The serine protease domain (SPD) of wild type (WT) FSAP displayed auto-catalytic activation whereas the MI isoform displayed very low autocatalytic activation and low proteolytic activity against the chromogenic substrate S-2288, Factor VII, tissue factor pathway inhibitor as well as pro-urokinase. Introduction of a thermolysin cleavage site in the activation position (Arg15Gln) led to cleavage of both WT- and MI-SPD and the resulting WT-SPD, but not the MI-SPD, was active. Mutating the Gly221 position to Asp, Gln and Leu led to a loss of activity whereas the Ala substitution was partially active. These results suggest a disturbance of the active site, or non-accessibility of the substrate to the active site in MI-SPD. With respect to regulation with metal ions, calcium, more than sodium, increased the enzymatic activity of WT-SPD. Thus, we describe a novel method for the production of recombinant FSAP-SPD to understand the role of the MI-single nucleotide polymorphism (SNP) in the regulation of its activity.

introducing point mutations that inactivated FSAP 18 . Hasumi and co-workers used a similar strategy to produce catalytically inactive Ser195Ala mutant for use in binding studies 20 , while Mertens and co-workers solved this problem by introducing a mutation at the activation site (Arg15Gln) making it resistant to auto-activation. The Arg15Gln mutant remained proteolytically inactive until cleaved by exogenously added thermolysin 19 . They suggest that the Gly221Glu polymorphism disturbs the transition of the zymogen form into a catalytically active protease 7 . Thermolysin is a common zinc dependent protease that cleaves at positions followed by hydrophobic residues. The problem with thermolysin is that it can potentially cleave FSAP at >100 different sites, introducing potential artefacts in downstream experiments. Thus, there is a need to establish a robust expression system for recombinant active FSAP.
Here we describe a novel strategy to express the serine protease domain (SPD) of FSAP in E. coli and perform a biochemical characterization of the wild type (WT)-SPD and the MI-SPD isoform against natural and chromogenic substrates and study their regulation by inhibitors and metal ions. Additional mutations were introduced in the MI position to understand the reasons for the loss of activity. Binding of active site ligands and homology modelling was performed to characterize these isoforms.

Materials and Methods
Expression of recombinant SPD-FSAP. Full length FSAP, excluding the signal peptide, (amino acids 24-560) (numbering system in italics refers to complete FSAP including the signal peptide), SPD-FSAP (amino acids 292-560) and various other mutants were cloned into the pASK-IBA33plus vector (IBA-Lifesciences, Goettingen, Germany) including a C-terminal 6 X His tag. SPD-FSAP contains 22 amino acids from the C-terminal of the heavy chain in addition to the complete light chain, protease domain, including the interchain disulphide bond. Additionally, a Met residue was added at the N-terminal end and Ser-Ala-Arg-Gly-Ser sequence was added before the His tag. All mutants were prepared by site directed mutagenesis (Stratagene, La Jolla, CA) (Fig. 1B). Positively transfected BL21-Gold cells (DE3) (Agilent, Santa Clara, CA) were selected on 100 µg/ml ampicillin agar plates. Anhydro-tetracycline 300 µg/L was used for the induction of expression and inclusion bodies were prepared. The inclusion bodies were washed by centrifugation and sonicated in 50 mM Tris (pH 8.0), 5% (vol/vol) glycerol, 100 mM NaCl, 1 mM 2-mercaptoethanol, 100 µg/ml of lysozyme and EDTAfree protease inhibitor tablets (Roche, Mannheim, Germany) and pelleted by centrifugation. Inclusion bodies were resuspended in TNU buffer (100 mM Tris pH 8.0, 300 mM NaCl, 5 mM 2-Mercaptoethanol, 8 M Urea,) for 1 h. The soluble fraction was loaded onto pre-equilibrated Ni-Agarose column (Qiagen, Hilden, Germany) and washed with TNU buffer with 15 mM imidazole, and eluted with TNU buffer containing 400 mM imidazole.

Refolding.
Refolding was monitored and optimized by determining the hydrolysis of the chromogenic substrate S-2288 as described later. The unfolded/purified FSAP was added dropwise to 100 mM Tris (pH 7.5), 125 mM MgCl 2 , 100 mM KCl, 2 mM reduced Glutathione (GSH), 1 mM oxidized Glutathione (GSSG), 1 M NDSB-201 (3-(1-Pyridinio)-1-propanesulfonate), 5% glycerol at 21 °C at a ratio of 20 mg protein per 100 ml buffer. The preparation was mixed slowly for 48 h followed by centrifugation at 12,000 × g for 10 min to clear the precipitate. FSAP-SPD was further concentrated using spin concentrator and re-purified on a Superdex 75 column (GE Healthcare, Oslo, Norway) in gel filtration buffer (10 mM Tris pH 8.0, 150 mM NaCl) using an ÄKTA Purifier (GE Healthcare). Recombinant proteins were characterized by N-terminal sequencing using the Edman procedure (Guenther Lochnit, University of Giessen, Germany). After gel-filtration the preparation of WT-SPD was prone to auto-proteolysis upon storage and it was used immediately in experiments.   21 . The enzyme was buffer-exchanged into 5 mM Tris (pH 8.0), 150 mM NaCl, 2 mM CaCl 2 . 50 µM of p-nitrophenyl p'-guanidinobenzoate (PNPGB) in methanol was used as substrate and the burst was measured at 402 nm using 18300 M −1 cm −1 as extinction coefficient. FSAP activity assays were performed as described previously 22 . In brief, 96 well microtiter plates were used and the standard assay system consisted of 25 mM Tris (pH 7.4), 137 mM NaCl, with CaCl 2 (2 mM) and 0.1% (wt/vol) Tween-20 and 0.25 mM of the chromogenic substrate S-2288 (H-D-isoleucyl-L-prolyl-L-arginine-p-nitroanilinedihydro-chloride) (Chromogenix, Mölndal, Sweden) and was followed over a period of 60 min at 37 °C at 405 nm in a microplate reader. The data was fitted to the Michaelis-Menten equation using Graphpad Prism software, Version 7.02 (San Diego, CA).
Activation of pro-upA and fVii. FSAP-SPD was incubated with pro-uPA (Grünenthal, Stolberg, Germany) and its activation was measured by adding 0.2 mM of the chromogenic substrate S-2444 (L-pyroglutamyl-glycyl-L-arginine-p-nitroanilinedihydro-chloride) (Chromogenix). Absorbance was followed for 60 min at 37 °C at 405 nm in a microplate reader. FVIIa clotting activity was determined in the presence of a recombinant soluble mutant tissue factor (TF) 23 . Briefly, FVIIa solutions were mixed with equal volumes of FVII-deficient plasma (Roche Diagnostics, Mannheim, Germany), a mixture of TF and phospholipids (PTT reagent, Diagnostica Stago, Asnieres, France) and 25 mM CaCl 2 . Clotting and turbidity changes were monitored at 37 °C and 405 nm in a microplate reader. In each well the final concentration of total FVII/FVIIa was set to 5 nM and TF to 60 nM. The level of FVIIa generated was quantified relative to the WHO 1 st International Standard FVIIa concentrate (1 IU/ml FVIIa corresponds to 20 ng/ml or 0.5 nM). The reaction was stopped by transferring 10 µl of the reaction mixture to 1 µl of 160 µM aprotinin. The residual TFPI activity was measured as described previously with minor modifications 24 . In brief, 90 µl of 0.44 nM FXa (Haematologic Technologies Inc, Essex Junction, VT) diluted in HBS-BC + 0.1% wt/vol Tween-20) was added to the aprotinin-treated reaction mixture and was incubated for 15 min at 37 °C to allow complex formation between TFPI and FXa. After addition of 40 µl of pre-warmed phospholipids (PTT reagent) and 20 µl of 2 mM chromogenic substrate S-2765 (N-a-Benzyloxycarbonyl-D-arginyl-L-glycyl-L-arginine-p-nitroaniline-dihydrochloride) (Chromogenix), the residual FXa activity was measured at 405 nm and 37 °C in a plate reader. Residual TFPI activity was quantified relative to a dilution series of TFPI incubated with FXa as described above.

Binding of biotinylated D-phenylalanyl-prolyl-arginyl chloromethyl ketone (Biot-ppAcK).
High binding ELISA plates (Costar, Cambridge, MA) were coated with polyclonal rabbit FSAP antibody (5 ug/ ml) and after blocking with BSA (3% wt/vol), WT-or MI-SPD was captured. Biot-PPACK, Biot-aprotinin, Biot-Mab570 was added in a buffer with 25 mM Tris (pH 7.4), 137 mM NaCl, with 2 mM CaCl 2 , 0.3% (wt/vol) BSA and 0.1% (wt/vol) Tween-20. After extensive washing the plates were developed with peroxidase coupled streptavidin and developed with a 3,3′,5,5′-tetramethylbenzidine (TMB) substrate. Binding to plates without the addition of WT-or MI-SPD was used to obtain blank values, which were subtracted from the binding in the presence of the SPD's to obtain specific binding. In competition experiments the binding of Biot-PPACK was determined in the presence of increasing concentrations of aprotinin or PPACK.
Homology modelling of Wt-and Mi-SpD. The homology models of the FSAP-SPD were generated using the protein structure prediction programme I-TASSER 25 . The 247 residues at the C-terminus (Ile314-Phe560) were used as the input sequence, and the models were built using uPA (PDB ID: 2VNT) as a template. All models yielded a C-score (confidence score) in the range of 0.83-0.85 and a TM-score (structural similarity) of 0.83 ± 0.08, indicating the correct topology and similarity of the models 26 . For the analysis of the metal ion binding sites, the WT-SPD model was superimposed on FXa structure (PDB ID: 2JKH), where the Na + and Ca 2+ binding sites are described. The models were visualized and the figures were prepared by using PyMOL Molecular Graphics System; Version 1.7.2.0 (Schrödinger, LLC).

Results
Expression and characterization of FSAP-SPD. Full-length FSAP (amino acids 24-560) was expressed in E. coli but could not be refolded from inclusion bodies. A SPD construct spanning a 22 amino acids of the heavy chain (amino acids 292-560) was produced effectively in inclusion bodies and could be refolded into the active protease (see Fig. 1B for details of the constructs used in this study) and used for all the experiments in this study. The inter-chain disulphide bond between the heavy chain and Cys122 in the protease domain is retained in this construct. The refolding conditions were optimized by measuring the hydrolysis of the chromogenic substrate S-2288. The activity status of this preparation indicated that the protein underwent classical autocatalytic activation with cleavage at Arg15-Ile16 during the refolding process (see later). The WT-SPD was very stable in refolding buffer. Based on the amount of input protein, the overall yield from the refolding process was 10-15% due to extensive precipitation. Active site titration showed that 59 ± 4% (n = 3 different preparations) of the final soluble WT-SPD preparation was active. The homogeneous size of the preparation would suggest that the inactive fraction was falsely folded but this was not investigated further. In preliminary experiments, affinity selection on benzamidine-sepharose increased the active site concentration significantly but this also enhanced auto-proteolysis. Thus, we have established a robust method for preparation of SPD of FSAP by producing it as an unfolded protein in inclusion bodies and generating the active form by controlled refolding. (2019) 9:18990 | https://doi.org/10.1038/s41598-019-55531-x www.nature.com/scientificreports www.nature.com/scientificreports/ We performed extensive tests to check if the protein could be expressed in a soluble form under different conditions in different bacteria. This included inducing the expression at low temperatures such as 15 °C, 22 °C and 30 °C. The cell pellets were extracted with eight different solubility buffers with varying pH, chaotropic agents, and ion compositions but yielded no functional soluble protein. Supplementary Fig. S1 shows FSAP expression in different bacterial strains at different time points up to 4 h post induction.

Autocatalytic activation of Wt and Mi-SpD.
Auto-activation of WT-SPD could be followed by SDS-PAGE under reducing conditions since there would be a loss of 22 amino acids of the heavy chain and a shift in MW from 30.6 to 28.1 kDa. N-terminal sequencing of the 30.6 kDa band showed an expected sequence of STKLP and the 28.1 kDa band started with Ile16 (IYGGF) in WT-SPD ( Fig. 2A). The N-terminal Met residue is removed, presumably by E. coli aminopeptidases. The kinetics of auto-activation of WT-SPD showed a maximal activation at 48 h at 4 °C (Fig. 2B). MI-SPD showed no activation for up to 3 days (Fig. 2B), and in further experiments for up to 1 week, but prolonged storage at 4 °C for months lead to auto-activation in some preparations (Fig. 2C). Preparations of MI-SPD showing auto-activation, by a shift in MW, had very low enzymatic activity against the chromogenic substrate S-2288. In the refolding step the recovery of MI-SPD, with respect to protein amount, was about 2-fold higher than WT-SPD. Because of its very low catalytic activity it was not possible to perform active site titration on MI-SPD.
We also compared auto-activation of WT-and MI-SPD after mutating the activation site from Arg15 to Gln to prevent auto-activation and enable controlled activation by thermolysin. The Arg15Gln mutants of both, WT and MI, isoforms showed no auto-activation, as was expected. Both could be activated by thermolysin, as confirmed by a shift in the MW of the bands as well as N-terminal sequencing (Fig. 2D). Thermolysin also cleaved the SPDs non-specifically as seen by the generation of many low MW bands starting with the original N-terminal sequence (STKLP) (Fig. 2D). The Arg15Gln mutant activated with thermolysin showed robust activity against S-2288, whereas similarly activated MI isoform showed no detectable activity (Fig. 2E). The fact that the Arg15Gln mutant folded correctly in its zymogen form suggests that the activation is not a prerequisite for the correct folding of SPD's. Thus, the low enzymatic activity was an intrinsic property of MI-SPD and not attributed to the lack of refolding. comparison of Wt-and Mi-SpD against physiological macromolecular substrates. Since plasma-purified FSAP has been shown to activate pro-uPA and Factor VII (FVII) as well as inactivate TFPI, we tested these natural substrates with the activated forms of WT-SPD and MI-SPD. WT-SPD was effective in activating pro-uPA as well as FVII, whereas MI-SPD had no such activity (Fig. 3A,B). FVII activation required approximately 100-fold higher concentrations of WT-SPD than pro-uPA activation, which is similar to the earlier observations with plasma-purified proteins 9 . Inactivation of TFPI was also observed with WT-but not MI-SPD (Fig. 3C) as was the case with plasma-purified FSAP 27 . Thus, the comparison of properties of WT-and MI-SPD against physiological substrates showed the expected pattern of activities. Thus, the recombinant SPD's, even though they lack the regulatory domains, can phenocopy some of the known functions of full-length FSAP.
comparison of mutants at the position of Mi-Snp. To elucidate the basis of the low catalytic activity of the MI-isoform (Gly221Glu) and thus test the significance of Glu in position 221, this residue was mutated to the following amino acids; Ala, Gln, Leu or Asp. Ala would represent a small amino acid that does not have the flexibility around the psi/phi torsion angles as compared to Gly. Gln has a similar size as Glu but is uncharged, however is still able to form hydrogen bonds, while Leu is similar in size, but hydrophobic and does not form hydrogen bonds. Asp is smaller than Glu but carries the same charge. These studies were performed under optimized concentrations of Ca 2+ and Na + as described later. The kinetic properties were investigated against a standard chromogenic substrate S-2288. For WT-SPD a K m 107 ± 7 µM and k cat of 650 ± 28 min −1 (n = 3) was measured. The Ala mutant had a 50-fold and the MI-mutant had a 2000-fold lower catalytic efficiency than WT-SPD. The Gln-, Leu-and Asp-mutants exhibited negligible activity (Fig. 4A, Table 1). These three mutants that exhibited negligible activity, most likely, also did not undergo auto-activation.
All mutants were also incubated with active WT-SPD to reinforce activation but, this did not lead to any additional change in their enzymatic activity (Fig. 4B). Thus only the replacement of the original Gly with a similarly sized Ala led to a partial retention of enzymatic activity, whereas all other replacements resulted in the same properties as MI-SPD.

Binding of active site ligands to activated Wt-and Mi-SpD.
An earlier study has suggested that after activation of the zymogen form of MI-FSAP the new N-terminus (Ile16) is unable to form a salt-bridge with Asp194 to enable a conformational change into an active protease 7 . We sought further evidence for a change in the conformation of the active site of MI-SPD by performing binding studies with biot-PPACK. These studies were performed under optimized concentrations of Ca 2+ and Na + as described later. To maintain their correct conformations, WT-and MI-SPD were captured using immobilized polyclonal FSAP antibody. Biot-PPACK bound to WT-SPD in a concentration-dependent manner but there was low binding of MI-SPD (Fig. 5A,B). An inhibitory antibody directed against the SPD of FSAP, Mab570, bound to both isoforms equally well indicating that both isoforms were captured equally well on the plate and have conserved epitopes. Biot-aprotinin showed negligible binding to both isoforms indicating that its affinity for WT-SPD is very low (Fig. 5A,B). The binding of biot-PPACK to FSAP-SPD could be competed by excess PPACK as well as aprotinin but this was not the case with MI-SPD (Fig. 5C). Thus, the low binding of biotinylated PPACK to MI-SPD is non-specific in nature. The enzymatic activity of WT-SPD was inhibited equally well by biotin-inhibitors in a comparable manner to non-biotin-inhibitors indicating that biotinylation did not change their properties (Fig. 5D) www.nature.com/scientificreports www.nature.com/scientificreports/ by biotinylated and non-biotinylated Mab570 at higher concentrations was identical but is not shown. Thus, WT-SPD binds to PPACK but MI-SPD does not. 2+ and na + on the activity of Wt-SpD. FSAP activity has been reported 7 to be altered through binding to the metal ions Ca 2+ or Na + . We tested if physiological levels of Na + and Ca 2+ were able to modulate the activity of our WT-SPD preparation. The presence of Ca 2+ caused a 12-fold decrease in K m and a 50-fold increase in catalytic efficiency, whereas Na + caused a 4-fold increase in catalytic efficiency ( Fig. 6 and Table 2). www.nature.com/scientificreports www.nature.com/scientificreports/ To identify the possible metal ion binding sites in the SPD of FSAP, we superimposed the WT-SPD model with the FXa (PDB ID: 2JKH) 6 . The superimposed structures suggest the presence of both Ca 2+ (magenta sphere) and Na + (blue sphere) binding sites in WT-SPD (Fig. 6). In the WT-SPD model, compared to FXa, the amino acids Asp64, Asp66, Glu70 and Glu71 could form potential coordinates with Ca 2+ . These functional and modeling studies predict the presence of a Ca 2+ -binding site in the SPD of FSAP as described recently 7 . With regard to Na + , potentially the binding occurs between the residues in the 220 and 180 loops, as observed in FXa 6 . Thus, we can confirm earlier findings that Ca 2+ is a strong activator of FSAP enzymatic activity.

Discussion
To date the expression of recombinant active FSAP has been difficult because enzymatically active FSAP is not well-tolerated by mammalian cells 18 . One solution to this problem was to prepare a mutant that, initially, was inactive but could be activated by an exogenous protease such as thermolysin 7 . The disadvantage of this method is that the broad substrate specificity of thermolysin, which can lead to unwanted cleavage at other extraneous www.nature.com/scientificreports www.nature.com/scientificreports/ sites giving rise to potential artefacts. In our experiments the FSAP-SPD's were very sensitive to multiple cleavages by thermolysin, other than at the activation site. To overcome these issues, we have developed a method to prepare recombinant SPD of FSAP in inclusion bodies of E. coli. The protein remains stable and inactive until it is refolded. The protein was very stable in refolding buffer at −20 °C for 1 year, or at 4 °C for several months. Most likely this buffer allows auto-activation but prevents auto-degradation due to a combination of the components of the refolding buffer and a lack of Na + and Ca 2+ that is required for full enzymatic activity. These methods will enable expression of large amounts of FSAP-SPD that is crucial for detailed biochemical, cellular and in vivo studies in the future.
The fact that MI-SPD showed no auto-activation for up to 1 week underscores its resistance to auto-activation. Some auto-activation was observed after months but we could not pinpoint the factors required for this in a consistent manner. Full length constructs expressed in eukaryotic cells also exhibited slower kinetics of auto-activation in MI-FSAP 7 . Mutating WT-and MI-SPD to Arg15Gln prevented auto-activation and enabled targeted cleavage by thermolysin as determined by sequencing and activity studies. Although this procedure was effective in auto-activating both WT-and MI-SPD, the activated MI-SPD remained virtually inactive. There were additional cleavages by thermolysin as seen by the appearance of a number of smaller MW bands which limits the practicality of this approach.
The MI-SPD has a reduced proteolytic activity against chromogenic substrate S-2288, FVII, pro-uPA and TFPI. Of all the amino acid replacements at the MI-position tested, only the Ala-mutant retained substantial enzymatic activity, indicating a central role of this position for the activity of the whole protein. These results WT-SPD 650 ± 28 107 ± 1 6. are very similar to a recent report that the Ser and Ala mutants are partially active 7 but that the MI-isoform is inactive. In concordance with the lack of enzymatic activity the binding of PPACK was also lost in MI-SPD. It has been reported that aprotinin is capable of binding to the zymogen and the active forms of some serine proteases such as trypsinogen 28 . Although aprotinin could compete with biot-PPACK for binding to the active site of WT-SPD, its affinity was too low to perform binding studies with biotinylated aprotinin under the given conditions. However, the monoclonal antibody Mab570, that inhibits the activity of WT-SPD, bound to MI-SPD equally well suggesting an overall conservation of certain epitopes in both isoforms.
In an earlier study, it has been suggested that MI-isoform is inactive because the formation of the ion-pairs between Ile16 and Asp194, which is important for achieving the zymogen-to-active conformation, is disturbed in this mutant 7 . Considering that thermolysin-cleaved WT-FSAP was compared with auto-activated MI-FSAP in these experiments there is a high potential for artefacts. Another possible explanation is that the altered flexibility in the 220-loop, when Gly is replaced by the bulkier (Leu, Gln) and negatively charged residues (Glu and Asp), interferes with substrate binding in the S1 pocket that is defined by residues 189-192, 214-212, and 224-228. Gara et al. used static homology models of WT-and MI-SPD and suggest differences in the conformation of the active sites between the two isoforms 12 . Further detailed crystallographic studies are needed to dissect the exact reason for the loss of enzymatic activity in MI-FSAP.
Stavenuiter et al. 7 demonstrate a role for both Ca 2+ and Na + in increasing the enzymatic activity of FSAP. We found similar effects of Ca 2+ on K m and k cat for WT-SPD using the substrate S-2288. Homology modeling showed www.nature.com/scientificreports www.nature.com/scientificreports/ a possible binding site for Ca 2+ in the 70-and Na + in the 220-loop/180-loop respectively which are in-line with the results of Stavenuiter et al. 7 . However, the effects of Na + in our studies were much smaller. The proposed Na + binding site is between 2 loops that have low sequence identity and insertions/deletions and are very flexible thus their positional assignment is uncertain.
The major advances presented in this study represent the first demonstration of the large-scale preparation of recombinant FSAP-SPD. The results obtained with the SPD-isoforms in this study largely match the earlier observations made with plasma-derived proteins 9,27 as well as proteins obtained from mammalian expression system 7,18 . During the course of the preparation of this manuscript we have been able to develop a novel substrate, an active-site probe as well as a specific and selective inhibitor for FSAP 29 . Similarly, we also found that this preparation of FSAP can recapitulate the effects of plasma FSAP on gene expression in cells 30 . This cross-validates and further strengthens the reliability of the current approach and opens the possibility for further research in this field.  with cartoon presentation. The Ca 2+ and the Na + ions, taken from FX, are shown in magenta and blue spheres, respectively. The Ca 2+ binding loop (upper inset) and the Na + binding loop (lower inset) of the FX shows the corresponding amino acid residues with which they coordinate with the corresponding metal ions. The yellow dotted lines represent the hydrogen bonding.  Table 2. Michaels-Menten kinetic constants for WT-SPD in the absence and presence of 150 mM NaCl and/or 5 mM CaCl 2 using the chromogenic substrate S-2288. The buffer used for these experiments was Tris (50 mM, pH 7.4) containing 2 mM EDTA and Tween-20 (0.1% wt/vol). All values are mean ± SD. Each value is obtained from 4-5 independent experiments.