A mutation in the kringle domain of human factor XII that causes autoinflammation, disturbs zymogen quiescence, and accelerates activation

Coagulation factor XII (FXII) drives production of the inflammatory peptide bradykinin. Pathological mutations in the F12 gene, which encodes FXII, provoke acute tissue swelling in hereditary angioedema (HAE). Interestingly, a recently identified F12 mutation, causing a W268R substitution, is not associated with HAE. Instead, FXII-W268R carriers experience cold-inducible urticarial rash, arthralgia, fever, and fatigue. Here, we aimed to investigate the molecular characteristics of the FXII-W268R variant. We expressed wild type FXII (FXII-WT), FXII-W268R, and FXII-T309R (which causes HAE), as well as other FXII variants in HEK293 freestyle cells. Using chromogenic substrate assays, immunoblotting, and ELISA, we analyzed expression media, cell lysates, and purified proteins for FXII activation. Recombinant FXII-W268R forms increased amounts of intracellular cleavage products that are also present in expression medium and display enzymatic activity. The active site–incapacitated variant FXII-W268R/S544A reveals that intracellular fragmentation is largely dependent on autoactivation. Purified FXII-W268R is highly sensitive to activation by plasma kallikrein and plasmin, compared with FXII-WT or FXII-T309R. Furthermore, binding studies indicated that the FXII-W268R variant leads to the exposure of a plasminogen-binding site that is cryptic in FXII-WT. In plasma, recombinant FXII-W268R spontaneously triggers high-molecular-weight kininogen cleavage. Our findings suggest that the W268R substitution influences FXII protein conformation and exposure of the activation loop, which is concealed in FXII-WT. This results in intracellular autoactivation and constitutive low-grade secretion of activated FXII. These findings help to explain the chronically increased contact activation in carriers of the FXII-W268R variant.

INH), the main plasma inhibitor of FXIIa and PKa. In addition, bradykinin-mediated HAE can also be caused by mutations in the F12 gene (encodes FXII). So far, five different mutations in FXII are associated with HAE (6 -9). All of these are located in the proline-rich region, a 53-amino acid sequence that is unique to FXII. Three of these mutations alter the enzymatic processing of FXII during activation through introduction of cleavage sites that are sensitive to plasmin (10) and thrombin (11). The resulting truncated FXII products display an enhanced sensitivity for activation (10 -12).
Recently, a new FXII mutation (FXII-W268R) was reported in the kringle domain of FXII (13,14). Interestingly, FXII-W268R carriers do not present with acute symptoms of HAE. All four identified FXII-W268R carriers are first-degree family members. They suffer from an autoinflammatory syndrome characterized by cold-inducible urticaria, arthralgia, fever, and fatigue. Plasma from these subjects contains markedly increased biomarkers of contact activation. Furthermore, clinical symptoms improve upon administration of a bradykinin receptor antagonist. Strikingly, this pathogenic FXII mutation (W268R) is located only 9 amino acids before the proline-rich region begins, which harbors mutations that cause HAE. This raises the question how FXII-W268R influences protein function and how this differs from mutations that cause HAE. We here report the investigation on the molecular characteristics of FXII with the mutation W268R.

FXII-W268R is spontaneously active
Western blotting analyses of expression medium containing FXII-W268R showed that it is partly fragmented (Fig. 1A). This fragmentation is predominantly seen under reducing conditions (nonreduced Western blotting in Fig. S1) and is much less pronounced during expression of FXII-WT or FXII-T309R (that causes HAE). The smallest fragment that FXII-W268R forms is ϳ28 kDa, resembling the light chain of FXIIa (which contains the protease domain). Chromogenic substrate assays showed that expression medium containing FXII-W268R exhibits spontaneous enzyme activity that is equivalent to 21 Ϯ 2.4 nM ␣FXIIa (progress curves in Fig. 1B, quantitation of FXIIa activity in Fig. 1C). Furthermore, incubation of FXII-W268R expression medium with plasma prekallikrein results in an accelerated development of enzyme activity, compared with expression medium containing FXII-WT or FXII-T309R (Fig.  1D). In a similar manner, prekallikrein activator activity is seen in expression medium containing FXII-W268R that was cloned without the purification tag (tagless FXII-W268R) but not in expression medium containing tagless FXII-WT (Fig. S2A). We subsequently found that the spontaneous enzymatic activity in FXII-W268R expression medium is sensitive to inhibition by the chemical serine protease inhibitor Phe-Pro-Arg-chloromethylketone (PPACK), C1-INH, and the FXIIa-inhibiting mAb

FXII-W268R is prone to intracellular autoactivation
The spontaneous enzymatic activity in expression medium that is displayed by FXII-W268R led us to investigate the possibility of intracellular FXII-W268R activation. Hereto, we analyzed lysates of cells that express FXII-W268R by Western blotting. Increased levels of ϳ28 kDa FXII fragments were found in lysates of cells that express FXII-W268R compared with those expressing FXII-WT or FXII-T309R ( Fig. 2A). This suggests that FXII-W268R has an increased propensity towards intracellular cleavage and activation. To evaluate the contribution of autoactivation of FXII to the intracellular activation of FXII-W268R, we expressed an active site mutant, FXII-W268R/ S544A. In this "silent" mutant, the active site serine is replaced by alanine. FXII-W268R/S544A is unable to autoactivate (i.e. cleave and activate other FXII molecules) and cannot activate plasma prekallikrein (Fig. S4). This FXII variant showed markedly less intracellular fragmentation ( Fig. 2A), indicating that intracellular FXII-W268R cleavage predominantly reflects autoactivation.
Alternatively, we investigated whether other intracellular enzymes are involved in the intracellular misprocessing of FXII-W268R. Proprotein convertases are abundant in most cell types; nine subtypes are important for intracellular processing of secretory proteins, with furin as the archetypical example (16). In silico predictions suggest that the W268R mutation introduces an RXRXXR consensus motif for proprotein convertases in FXII (residues 263-270; Table S1). To question its functionality, we developed four FXII-W268R variants (S267R/W268R and R265A/W268R), respectively improving or disrupting this consensus motif and their active siteincapacitated counterparts (S267R/W268R/S544A and R265A/ W268R/S544A; Table S2). Western blotting analyses of expression medium containing these FXII-W268R variants showed that FXII fragmentation occurs irrespective of disruption or improvement of the consensus motif for proprotein convertases (Fig. 2B). Disabling FXII autoactivation (S267R/W268R/ S544A and R265A/W268R/S544A) prevents this FXII fragmentation to an extent that is comparable with that of FXII-WT and FXII-W268R/S544A (Fig. 2B). Furthermore, disruption or improvement of this motif did not influence plasma prekallikrein-activator activity (Fig. 2C), whereas plasma prekallikrein-activator potential was completely lost upon

FXII-W268R disturbs zymogen quiescence
additional disabling of FXIIa enzymatic activity. It is therefore unlikely that the W268R mutation introduces a new functional proprotein convertase cleavage motif.
In addition, coexpression of FXII-W268R with ␣1-antitrypsin-Portland (␣1AT-Portland; inhibits furin (17)) does not reduce plasma prekallikrein-activator activity (Fig. 2D), ruling out a role for furin. Together, these findings show that where FXII-WT is resistant to intracellular autoactivation during protein expression, FXII-W268R is not. We hypothesized that this is the result of inappropriate activation loop exposure by this pathogenic mutant.
We next performed a similar experiment with short dextran sulfate polymers (DXS-15k; M r ϭ ϳ15,000). Because of their short chain length (3), FXII autoactivation is limited (progress curves in Fig. 3G and quantification of FXIIa activity in Fig. 3H). However, the presence of DXS-15k does enhance activation of FXII-WT by PKa ϳ4-fold (3.9 Ϯ 0.74 nM ␣FXIIa without DXS; 16.9 Ϯ 3.9 nM ␣FXIIa with DXS). Under the same conditions, FXII-W268R already becomes activated by PKa in the absence of DXS-15k (activity equivalent to 98.6 Ϯ 8.1 nM ␣FXIIa; Fig. 3, G and H). Taken together, FXII-W268R behaves as if it is already in a surface-bound, unfolded conformation, exposing its activation loop where FXII-WT does not.

Mutation W268R enhances activating cleavage of FXII
We next investigated FXII-W268R with an ELISA that recognizes two-chain FXIIa, which forms when FXII is cleaved after Arg-353. This ELISA does not require an operational active site for target recognition (18). In good agreement with our chromogenic substrate activity assays, purified FXII-W268R in buffer generates much more FXIIa in response to PKa than FXII-WT or FXII-T309R (Fig. 4A). Similarly, more FXIIa forms when this mutant is exposed to plasmin (Fig. 4B). To exclude a contribution of autoactivation, we next performed these experiment with active-site incapacitated mutants (S544A). As before, the W268R mutation accelerates formation of a two-chain form of FXII by PKa or plasmin (Fig. 4, C and D). These combined experiments suggest that cleavage site Arg-353 is cryptic in FXII-WT but uncovered by the W268R mutation, making it sensitive to activation in solution.
In further experiments, we examined whether the W268R mutation influences the interaction of FXII with its natural binding partners. We were unable to detect a stable binding interaction between plasma prekallikrein and immobilized FXII-WT or FXII-W268R in direct binding studies (by ELISA; data not shown). In contrast, plasminogen binds to immobilized FXII-WT, but only when it is immobilized in the presence of DXS (Fig. 4E). This suggests that FXII contains a cryptic binding site for plasminogen, which is exposed by surface binding. Remarkably, plasminogen binds robustly to immobilized FXII-W268R without DXS (Fig. 4E), suggesting an open conformation. This behavior is specific for this mutation, because it was not the case for several FXII mutations that cause HAE (Fig.  S6). Together, our results suggest that the conformation of FXII determines its activation rate, as well as the interaction with binding partners.

FXII-W268R activity in plasma overrides C1-INH function
In physiology, FXIIa and PKa are inhibited by C1-INH. We next investigated how FXII-W268R interacts with the plasma contact system by Western blotting. We first examined FXII fragmentation in buffer in the absence or presence of PKa. Over a 5-min time span, 15 g/ml FXII-WT, FXII-W268R, or a 1:1 combination of both (7.5 g/ml each, to mimic heterozygous FXII-W268R carriers) did not display fragmentation (Fig. 5, A and C). In contrast, progressive FXII fragmentation was observed in the presence of PKa where FXII-W268R was present (Fig. 5, B and D).
Similarly, when we resupplemented FXII-immunodepleted plasma with FXII-W268R, we observed spontaneous formation of ϳ130 kDa bands that correspond to FXIIa-C1-INH complexes (Fig. 5E, indicated by open arrows, and densitometric quantification in Fig. 5G), as well as FXIIa light chain (Fig. 5E, indicated by closed arrows). This was more apparent in the presence of PKa (Fig. 5, F and H).
Finally, we analyzed HK cleavage as a reflection of bradykinin production. Where resupplementation of FXII-immunodepleted plasma with FXII-WT did not provoke any HK cleavage, substantial HK cleavage was seen in the presence of FXII-W268R (Fig. 6, I and K). As expected, the direct addition of PKa to plasma leads to HK cleavage (Fig. 6, J and L). These combined observations suggest that the accelerated activation rate of FXII-W268R overwhelms the inhibitory capacity of C1-INH, triggering HK cleavage in plasma, and are consistent with observations in FXII-W268R carriers (13,14).

Discussion
In the present study, we investigated the biochemical features of the FXII mutant FXII-W268R that was recently discov-FXII-W268R disturbs zymogen quiescence FXII-W268R disturbs zymogen quiescence ered in a family with autoinflammation (13,14). We found that this pathogenic variant has a generally increased susceptibility for enzyme activation, unlike mutant FXII-T309R, a biochemically comparable mutation that causes HAE. This remarkable behavior of FXII-W268R is not restricted to the extracellular environment; activation already begins intracellularly. The finding that a fraction of FXII-W268R is secreted as an active enzyme and has a lowered threshold for activation enables it to override physiological C1-INH activity in plasma as a result of accelerated activation. This matches the clinical observations of excessive contact activation in plasma of FXII-W268R carriers.
Based on our findings, we hypothesize that the increased susceptibility for activation of FXII-W268R is related to its confor-mation (graphically summarized in Fig. 6). However, this might not be the only explanation for the observed clinical phenotype. It is becoming increasingly clear that leukocytes contain a distinct pool of FXII (19). The presence of intracellular FXIIa in these cells might trigger inflammasome activation, in turn leading to uncontrolled interleukin-1␤ production. Alternatively, this mutation might influence the nonproteolytic contributions of FXII to inflammation (20).
At the same time, there is a growing body of evidence that the conformation of FXII is paramount to its function. The binding of FXII to a negatively charged surface changes its conformation. This mediates (auto)activation and causes FXII to expose its activation loop (2,3). In the present study, we show that the W268R mutation influences FXII in a strikingly similar way.

FXII-W268R disturbs zymogen quiescence
We previously proposed that FXII zymogen quiescence is determined by a "closed" protein conformation, in which the N terminus of FXII shields its activation loop (1). This concept is not completely novel; it is also described for the serine proteases plasminogen (21), Factor IX (22), prothrombin (23), and the metalloprotease ADAMTS13 (a disintegrin and metalloproteinase with a thrombospondin type 1 motif, member 13) (24,25). Each of these enzymes displays increased functional properties in an open conformation, which mostly is the result of binding to phospholipid surfaces, protein cofactors, or polymeric target proteins.
Although our data are consistent with the hypothesis that FXII adopts an open conformation as a result of the W268R mutation, we did not directly demonstrate this. Currently available data on FXII structure is limited to the protease domain (26), ␤FXIIa (27, 28), and the FnI-EGF-2 tandem domain (29).
The W268R mutation is located in the kringle domain of FXII, distant from the protease domain. The functional role of this domain in FXII has not been thoroughly investigated. Kringle domains are among others conserved in members of the plasminogen activation system, hepatocyte growth factors, and prothrombin (30). However, the kringle domain peptide sequence of FXII has low similarity to those of plasminogen, which are important for lysine-dependent binding to fibrin. Although FXII has fibrin-binding properties and is implicated in fibrinolysis, it remains unclear whether this depends on lysine-mediated binding of its kringle domain (31,32). Interestingly, the two kringle domains of prothrombin are not only important for interactions with factor Va and factor Xa. Kringle 1 is involved in ensuring a closed prothrombin conformation through intramolecular binding to its serine protease domain (23). Disruption or reinforcement of this interaction directly

FXII-W268R disturbs zymogen quiescence
influences the open/closed conformational equilibrium, which impacts prothrombin activation. Based on our current findings, one can speculate that the kringle domain of FXII has a role in regulation of the open/closed equilibrium of this enzyme.
Although our findings point toward the kringle domain as a regulator of zymogen quiescence, based on the identification of a human pathogenic mutation in this domain, it cannot be ruled out that other domains are involved. Indeed, although a mAb against the kringle domain enhances FXII activation by PKa (33), the same holds true for monoclonal antibodies against the adjacent EGF-2 domain (11) and the N-terminally positioned FnII domain (34).
This current work reveals the biomolecular differences between FXII-W268R and previously studied FXII-HAE mutations. Throughout our studies, we learned that FXII-W268R was extremely sensitive to activating cleavage but not as the result of inappropriate truncation. Moreover, FXII-W268R already activates intracellularly where FXII-T309R does not. Our binding studies show that FXII-W268R exposes a binding site for plasminogen that is cryptic in FXII-WT or FXII with HAE mutations. Furthermore, studies with active site-incapacitated FXII mutants and ELISAs show us that this mutant exposes its activation loop, whereas incapacitated FXII-WT does not. Together, these findings strongly suggest that mutation W268R forces FXII into an open conformation. This contrasts the FXII-HAE mutants that we studied earlier: these are in a quiescent state until they are truncated, which requires a specific enzymatic environment. We hypothesize that the continuously increased potential for activation matches the chronic, diffuse clinical picture of autoinflammatory urticaria, whereas the accelerated activation of the FXII-HAE mutations after truncation matches the acute clinical picture of HAE.
Our work presented here on the FXII-W268R mutation supports a pivotal role for the contact system and bradykinin in pathology beyond HAE. To our knowledge, this is the first study to suggest that a properly controlled conformation of FXII is decisive for the regulation of the contact system in human physiology.
Mutants FXII-W268R and FXII-S544A were developed through two-step assembly. First, the insert encoding the FXII-WT cDNA sequence was amplified by PCR in three segments (annotated as N, M, and C, respectively; primer sequences in Table S3). Segments with the desired mutations were ordered as dsDNA fragments. First, the N and M segments were digested with XmaI. Subsequently, the fragments were fused by ligation. The fused product (NM) was amplified via PCR. Hereafter, the NM and C segments were digested with BsaI and fused together by ligation. For FXII mutants R265A/ W268R, R265A/W268R/S544A, S267R/W268R, S267R/ W268R/S544A, and W268R/S544A, the complete cDNA sequences were ordered as dsDNA fragments that were codonoptimized for expression in human cell lines. Fragments were first ligated into the pJET1.2/blunt vector according to the manufacturer's instructions. Inserts were next digested with EcoRI and NotI and ligated into the mammalian expression vector pSM2 encoding an N-terminal IgK secretion signal and 2ϫ Strep-tags for affinity purification (based on pcDNA6/ V5-His A (10)). For FXII-WT and W268R expressed without the N-terminal 2ϫ Strep-tags (tagless FXII), cDNA sequences were amplified by PCR from the FXII-WT or W268R cDNA sequence using the primers tagless_FXII_For and FXII_Rev Primer sequences in Table S3). The resulting cDNA fragment contains the N-terminal IgK secretion signal but lacks the 2ϫ Strep-tags. The cDNA product was ligated into the pJET1.2/ blunt vector according to the manufacturer's instructions. Inserts were digested with HindIII and NotI and ligated into the HindIII and NotI restricted mammalian expression vector pcDNA6/V5-His A. Sequences were verified by Sanger sequencing prior to transfection.
HEK293F cells at a concentration of 1.1 ϫ 10 6 cells/ml were transfected with 293fectin and cultured in Gibco FreeStyle 293 expression medium with 2-ml cell suspensions in 6-well culture plates. After 4 days, cell suspensions were collected and centrifuged at 500 ϫ g for analysis of expression medium and cell pellets. Stable cell lines were selected for resistance to 5 g/ml blasticidin in the presence of penicillin and streptomycin. Resistant cells were expanded and grown in 2-liter Erlenmeyer flasks for cell culture, cells were centrifuged at 500 ϫ g, and supernatant medium was harvested twice weekly. FXII inhibitor mix consisting of soy bean trypsin inhibitor (0.024 g/liter), benzamidine (0.174 g/liter), and Polybrene (0.056 g/liter) was added to the supernatant prior to storage at Ϫ20°C. Supernatants were thawed and concentrated, and the buffer was exchanged (100 mM Tris-HCl, 150 mM NaCl, and 1 mM EDTA, pH 8.0, containing FXII inhibitor mix) on a 10-kDa dialysis membrane in a Quixstand benchtop system (GE Life Sciences). Strep-Tactin-Sepharose beads were used for purification of recombinant FXII via Strep-tag, and 12.5 M PPACK was added to washing and elution buffers. Purified recombinant FXII was dialyzed against 4 mM sodium acetate-HCl and 150 mM NaCl (pH 5.5). Protein concentrations were routinely determined by absorption spectroscopy at 280 nm, Coomassie protein staining, or Western blotting (accompanied by densitometric analyses of bands) to ensure equal protein amounts in experiments. Table S2 displays an overview of the constructs used.

FXII-W268R disturbs zymogen quiescence Purified FXII in plasma and buffer
For activation experiments in plasma, 10 l of recombinant FXII-WT, FXII-W268R, or a 1:1 combination of both (final FXII concentration, 15 g/ml) were added to 15 l of FXII immunodepleted plasma or 15 l of HBS containing 0.05% BSA (w/v). Directly after mixing, samples were placed at 37°C for 2 min, and activation was started with 5 l of PKa (final concentration, 12.5 g/ml) or 5 l of HBS as a control. Samples were taken after 1 and 5 min and diluted 20 times in 1ϫ sample buffer containing 8.3 mM DTT. Prior to activation, t ϭ 0 samples were created.
All samples were heated for 10 min at 95°C and centrifuged for 5 min at 2000 ϫ g. Samples were separated on 4 -12% Bis-Tris gels at 165 V for 60 min in MOPS buffer, and proteins were transferred onto Immobilon-FL membranes at 125 V for 55 min in blotting buffer (14.4 g/liter glycine, 3.03 g/liter Tris-HCl, 20% ethanol). The membranes were blocked for 1-2 h at room temperature using Odyssey blocking reagent diluted 1:1 with Tris-buffered saline (TBS; 50 mM Tris-HCl and 150 mM NaCl, pH 7.4). Factor XII or HK were detected with affinity-purified polyclonal antibodies (1:2000) by overnight incubation at 4°C. Next, primary antibodies were detected with Alexa Fluor 680conjugated donkey anti-sheep IgG (1:5000). GAPDH in lane loading controls was detected with a mAb (1:4000) and a IRdye 800-conjugated donkey anti-mouse antibody (1:10000). The membranes were washed with TBSt and distilled water and analyzed on a near-IR Odyssey scanner (LI-COR).

FXIIa-C1-INH complex ELISA
Methods on VhH selection and production for this assay are described before (18). VhH 1B12 (5 g/ml), a monoclonal VhH directed against complexed C1-INH, was coated overnight at 4°C on 96-well MaxiSorp plates. The plates were tapped dry and blocked with HBS containing 0.05% (v/v) Tween 20 (HBSt) and 1% (w/v) skimmed milk powder (mHBSt). Plasma purified C1-INH was added to expression medium at a final concentration of 5 g/ml. As a reference standard, ␤FXIIa was added to expression medium from untransfected HEK293F cells and incubated with 5 g/ml C1-INH. After 10 min of incubation with purified C1-INH, 100 l/sample was pipetted on the plate and incubated for 1 h at room temperature while shaking. The plates were washed with HBSt. The captured FXIIa-C1-INH complexes were detected by 5 g/ml biotinylated polyclonal VhH targeted against ␤FXIIa in mHBSt and after rinsing by streptavidin-poly-HRP (1:2000) in mHBSt. Finally, plates were rinsed, and 50 l/well TMB substrate was added. Substrate conversion was stopped by adding 25 l/well H 2 SO 4 (0.3 M), and absorbance was measured at 450 nm. GraphPad Prism 7.02 software was used to interpolate FXIIa-C1-INH concentrations, using a sigmoidal 4PL fit model to the ␤FXIIa-C1-INH reference standard.

FXIIa ELISA
VhH B7, a monoclonal VhH directed against FXIIa (18) was coated overnight at 4°C on a 96-well MaxiSorp plates. The plates were tapped dry and blocked with mHBSt. 10 g/ml recombinant FXII in HBSt was incubated with 1 g/ml PKa or 25 g/ml plasmin. At the indicated time points, the reactions were stopped by diluting samples five times in mHBSt, supplemented with 250 M PPACK. Fifty l per sample was pipetted on the blocked plates and incubated for 1 h at room temperature while shaking. A reference range of purified ␣FXIIa in mHBSt containing 200 M PPACK was included as a standard. Captured FXIIa was detected by the biotinylated monoclonal VhH B2 targeted against ␤FXIIa (5 g/ml). Further detection steps were performed as described above (FXIIa-C1-INH complex ELISA).

Protein-binding assays
To study binding interactions between FXII and plasminogen, 5 g/ml recombinant FXII in HBS was coated on 96-well PolySorp plates (overnight at 4°C) in the presence of 100 M PPACK and in the presence of 1 g/ml DXS, where indicated. The plates were tapped dry and blocked with 2% BSA in HBS (w/v). A concentration series of plasminogen was diluted in HBS containing 1% (w/v) BSA in the presence of 10 M PPACK and incubated on plates for 1 h. After rinsing, bound plasminogen was detected using a polyclonal goat anti-human plasminogen antibody (1:2000) followed by a rabbit anti-sheep IgG-HRP antibody (1:4000) in the presence of 10 M PPACK. Finally, the plates were rinsed, and 50 l/well TMB substrate was added. Substrate conversion was stopped by adding 25 l of H 2 SO 4 (0.3 M), and absorbance was measured at 450 nm.

Chromogenic substrate assays
Chromogenic substrate H-D-Pro-Phe-Arg-pNA (0.5 mM) was used to detect FXIIa and/or PKa activity. 96-well Costar plates were blocked with HBS containing 1% BSA (w/v; filtered on 0.20-micron filters) for at least 20 min. All sample incubations were performed 37°C. Substrate conversion was readout at 405 nm at 37°C.
Spontaneous activity in medium was determined in 40 l of expression medium containing FXII, to which 10 l of chromogenic substrate H-D-Pro-Phe-Arg-pNA was added (final concentration, 0.5 mM). Amidolytic FXIIa activity was quantified by determining the initial (linear) slopes of substrate conversion for 20 min, which were fitted to a standard curve of ␣FXIIa in expression medium (0 -25 nM). Where indicated, inhibitors PPACK (200 M), C1-INH (25 g/ml), or mAb OT2 (5 g/ml) were added for 15 min prior to addition of chromogenic substrate H-D-Pro-Phe-Arg-pNA.
Plasma prekallikrein activator activity in 40 l of expression medium containing FXII was measured by addition of 10 l of chromogenic substrate H-D-Pro-Phe-Arg-pNA (final concentration, 0.5 mM) and plasma prekallikrein (5 g/ml final concentration), after which PKa-like activity was measured. The effect of inhibitors PPACK (200 M), C1-INH (25 g/ml), and mouse mAb OT2 (5 g/ml) was investigated by preincubation for 15 min, after which 10 l of chromogenic substrate H-D-Pro-Phe-Arg-pNA (final concentration, 0.5 mM) and plasma prekallikrein (5 g/ml) were added, and PKa-like activity was measured.
FXII activation of purified recombinant FXII variants by PKa or plasmin was investigated by preincubating FXII (10 g/ml in HBS, 0.2% m/v BSA) with PKa (1 g/ml, in HBS 0.2% BSA), plasmin (25 g/ml in HBS 0.2% BSA), or buffer (HBS, 0.2% BSA) for 15 min at 37°C, followed by a 5-min incubation with aprotinin for all conditions (100 KIU/ml in HBS 0.2% BSA) to inhibit PKa or plasmin. FXIIa activity was detected by conversion chromogenic substrate H-D-Pro-Phe-Arg-pNA (0.5 mM) and quantified by determining the initial (linear) slopes of substrate conversion for 5 min, which were fitted to a standard curve of ␣FXIIa (0 -125 nM) in HBS, 0.2% BSA. Residual PKa activity (monitored as a control without FXII) was routinely subtracted where FXII activation by PKa was investigated.
FXII activation of purified recombinant FXII variants by PKa in the presence of DXS-15 was measured by preincubating FXII (10 g/ml in HBS, 0.2% BSA) with DXS-15 (10 g/ml in HBS) or buffer (HBS) for 15 min at 37°C after which PKa (1 g/ml, in HBS 0.2% BSA) or buffer (HBS, 0.2% BSA) was added and incubated at 37°C for 15 min, after which aprotinin (100 KIU/ml in HBS 0.2% BSA) was added and incubated at 37°C for 5 min. Chromogenic substrate H-D-Pro-Phe-Arg-pNA (0.5 mM) was added, and FXIIa activity was quantified by determining the initial (linear) slopes of substrate conversion for 5 min, which were fitted to a standard curve of ␣FXIIa (0 -156 nM) in HBS, 0.2% BSA. Residual PKa activity (monitored as a control without FXII) was routinely subtracted where FXII activation by PKa was investigated.

Statistical analysis
GraphPad Prism 7.02 software was used for statistical analysis were indicated. Two-way analysis of variance with Dunn's correction of multiple testing was used for comparison of repeated experiments with biological samples.