A mechanism for hereditary angioedema with normal C1 inhibitor: an inhibitory regulatory role for the factor XII heavy chain

The plasma kallikrein-kinin system (KKS) is composed of the protease precursors prekallikrein (PK) and factor XII (FXII) and the cofactor/substrate high-molecular-weight kininogen (HK).^1-4  PK and FXII undergo reciprocal conversion to the proteases kallikrein and αFXIIa by a process that is accelerated by a variety of natural and artificial “surfaces.” The surface-mediated reactions are referred to as contact activation.^1,2,5,6  Kallikrein cleaves HK to release the nanopeptide bradykinin, which contributes to inflammation, vasodilatation, and vascular permeability through interactions with specific cellular receptors.^3,4,6,7  Excessive bradykinin formation due to accelerated activation or dysregulation of the KKS contributes to a range of pathologies including angioedema (rapid soft tissue swelling).^4,7,8  In hereditary angioedema (HAE), which affects ∼1 in 50 000 individuals,^9,10  patients experience episodic swelling of the face, airway, limbs, or gastrointestinal tract. Consistent with a role for bradykinin, kallikrein inhibitors reduce the frequency and severity of angioedema in HAE patients.^11-13 

HAE is usually caused by reduced plasma activity of the serpin C1 inhibitor (C1-INH), the main regulator of kallikrein and αFXIIa.^8-10  However, HAE does occur in patients with normal C1-INH activity.^14,15  In some, Thr³⁰⁹ in FXII is replaced with lysine or arginine.^16-19  de Maat et al showed that these substitutions create a novel cleavage site for the protease plasmin, and that inducing plasmin generation in patient plasma results in cleavage of FXII-Lys/Arg³⁰⁹ that enhances kallikrein and bradykinin generation.²⁰  Angioedema in patients with FXII-Lys/Arg³⁰⁹ often follows trauma,^18,19  suggesting a relationship with thrombin generation. Here, we show that FXII-Lys/Arg³⁰⁹ is cleaved after residue 309 by proteases generated during blood coagulation, removing the protein’s noncatalytic heavy chain (HC) region. The resulting truncated FXII is activated by kallikrein more rapidly than is FXII, accelerating kallikrein and bradykinin generation in a surface-independent manner. The findings explain the predisposition of patients with FXII-Lys/Arg³⁰⁹ to develop angioedema after trauma, and support an important role for elements of the FXII HC in regulating bradykinin production.

The following materials were used (obtained from the companies shown in parentheses): normal and FXII-deficient plasma (George King Bio-Medical); FXII, αFXIIa, βFXIIa, PK, HK, thrombin, and corn trypsin inhibitor (CTI; Enzyme Research Laboratory); FXI and FXIa (Haematologic Technologies); plasmin (Athens Research & Technology); C1-INH (Sigma-Aldrich); RecombiPlasTin human tissue factor (TF; Instrumentation Laboratory); recombinant hirudin (Bayer); human rhinovirus 3C protease (3CP; ACRO Biosystems); PTT-A reagent (Diagnostica Stago); argatroban (GlaxoSmith Kline); and S-2302 (H-d-prolyl-l-phenylalanyl-l-arginine-p-nitroanilide dihydrochloride) and S-2366 (l-pyro-Glu-L-Pro-L-Arg-p-nitroanilide) (DiaPharma). Polyphosphate (200-1300 U; ILC Performance Products) was provided by James Morrissey (University of Michigan, Ann Arbor, MI).

The following antibodies were used (obtained from the companies shown in parentheses): goat horseradish peroxidase (HRP)–anti-human FXII, sheep HRP–anti-human PK, and goat HRP–anti-human C1-INH immunoglobulin Gs (IgGs) (Enzyme Research Laboratories); goat anti-human HK IgG (Nordic Immunology); and rabbit anti-plasminogen IgG 1677-1-AP (Proteintech). Rabbit anti-mouse (anti-mHK) HK IgG was raised against a peptide representing residues 607-638 of mHK. Monoclonal IgGs 15H8 (FXII HC),²¹  1B2 (FXII catalytic domain),²²  and H03 (kallikrein active site)²³  have been described.

Diagrams of human FXII and its activation products are shown in Figure 1A. FXII (FXII–wild type [WT]) and FXII with Lys, Arg, or Ala replacing Thr³⁰⁹ (FXII-Lys³⁰⁹, FXII-Arg³⁰⁹, or FXI-Ala³⁰⁹; Figure 1B) were expressed in HEK293 cells as described (Figure 1C).²⁴  FXII-Arg³⁰⁹ in which Arg³³⁴ is replaced with alanine (FXII-Arg³⁰⁹, Ala³³⁴; supplemental Figure 1A [available on the Blood Web site]), FXII containing Leu-Glu-Val-Leu-Phe-Gln-Gly inserted between FXII Gln³⁰⁷ and Pro³⁰⁸ to introduce a 3CP cleavage site (FXII-3C; Figure 1C), and FXII-3C in which Arg³³⁴, Arg³⁴³, and Arg³⁵³ are replaced with alanine (FXII-T)²⁴  were also prepared. FXII was purified by anion exchange chromatography from conditioned media,²⁴  and stored in 4 mM sodium acetate, 150 mM NaCl, pH 5.2 at −80°C.

FXII-3C (8 μM) was incubated with 3CP (25 U/mL) in 50 mM Tris-HCl, 150 mM NaCl overnight at 4°C. 3CP was removed with a glutathione column (Thermo Scientific). Cleaved FXII-3C was dialyzed into sodium acetate buffer and stored at −80°C. Activated δFXII (truncated form of FXIIa-3C [δFXIIa-3C]) was prepared by converting FXII-3C (10 μM) to αFXIIa-3C by incubation with 175 μg/mL dextran sulfate (500 kDa). αFXIIa-3C was dissociated from dextran sulfate with polybrene (500 μM), and isolated on a 15H8 IgG column.²¹  After elution with 2M NaSCN, αFXIIa-3C was cleaved with 3CP to generate δFXIIa-3C. Proteins were dialyzed into sodium acetate buffer and stored at −80°C.

Assays were performed in PEG-20000–coated microtiter plates.

Continuous assays for FXII and/or PK activation were in 100-μL standard buffer (20 mM N-2-hydroxyethylpiperazine-N′-2-ethanesulfonic acid [HEPES], pH 7.4, 100 mM NaCl, 0.1% PEG-8000, 10 μM ZnCl₂) with 200 μM S-2302. Some reactions contained polyphosphate (70 μM) or 15H8 (400 nM). Changes in optical density (OD) at 405 nm at 37°C were monitored.

Discontinuous assays for FXII, PK, and FXI activation were run in 100-μL standard buffer at 37°C. FXII cleavage by PK, kallikrein (10-20 nM), FXIa (10 nM), or plasmin (20 or 500 nM) with or without polyphosphate (70 μM) or 15H8 was assessed. For measuring FXIIa, reactions were stopped with polybrene (0.1 mg/mL) and H03 (100 nM for reactions with kallikrein) or aprotinin (10-40 μM for reactions with FXIa or plasmin) and S-2302 (500 μM) was added. For measuring kallikrein, reactions were stopped with CTI (500 nM) and polybrene (0.1 mg/mL) and S-2302 (200 μM) was added. For partial cleavage of FXII-Lys309 or FXII-Arg309 (100 nM) by thrombin (25 nM), reactions were stopped with argatroban (125 μM) and 10 μL of reactions were incubated with 60 nM PK. For measuring FXIa, reactions were stopped with CTI (500 nM) and polybrene (0.1 mg/mL), and S-2366 (250 μM) was added. For all reactions, ΔOD 405 nm was followed, and amounts of FXIIa, kallikrein, or FXIa extrapolated from standard curves prepared with pure proteases.

FXII (1.25 μM) was incubated with dextran sulfate (20 μg/mL) in standard buffer. At various times, aliquots were removed into reducing or nonreducing sample buffer, size-fractionated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE), and stained with Coomassie blue.

FXII-deficient plasma was supplemented with 400 nM FXII and 10 mM H-Gly-Pro-Arg-Pro-OH to prevent fibrin polymerization. Coagulation was initiated with 2.5 pM human TF and 6.25 mM CaCl₂ (final concentrations). At various times, aliquots were removed into nonreducing sample buffer, size-fractionated by SDS-PAGE, transferred to nitrocellulose membranes, and probed with HRP-conjugated goat anti-human FXII IgG, using chemiluminescence.

FXII (200 nM) was incubated with 10 nM thrombin, FXIa, plasmin, FVIIa/TF, FIXa, or FXa, or with 50 nM kallikrein, at 37°C in standard buffer with 5 mM CaCl₂. At various times, aliquots were mixed with nonreducing sample buffer, size-fractionated by SDS-PAGE, transferred to nitrocellulose, and probed with 15H8 and 1B2. HRP-conjugated goat anti-mouse IgG and chemiluminescence were used for detection.

FXII (200 nM) and PK (200 nM) were incubated in standard buffer at 37°C. At various times, aliquots were mixed with reducing sample buffer, size-fractionated by SDS-PAGE, transferred to nitrocellulose, and probed with HRP-conjugated goat anti-human FXII or sheep anti-human PK IgG and chemiluminescence.

FXII-deficient plasma was supplemented with FXII (400 nM), with or without PTT-A reagent (10% final volume), and incubated at 37°C. At various times, aliquots were mixed with nonreducing sample buffer, size-fractionated by SDS-PAGE, transferred to nitrocellulose, and probed with goat anti-human HK IgG or anti-human C1-INH and chemiluminescence.

Procedures with mice were approved by the Vanderbilt University Animal Care and Use Committee. FXII (0.6 mg/kg) in phosphate-buffered saline (PBS) was infused into WT, FXI-deficient (F11^−/−),²⁵  or PK-deficient (Klkb1^−/−)²⁶  C57Bl/6 mice through a tail vein. In some experiments, mice also received infusions of human TF (70 ng/kg), IgG 1B2 (1 mg/kg), IgG 15H8 (1 mg/kg), or hirudin (10 mg/kg). Blood samples were collected from tail veins into capillary tubes, and plasma was prepared by centrifugation. Samples underwent SDS-PAGE, then transfer to nitrocellulose. Blots were probed with anti-mHK IgG using HRP-conjugated goat anti-rabbit IgG, and chemiluminescence. In some experiments, animals were pretreated with antisense oligonucleotide (ASO) to mouse plasminogen messenger RNA (mRNA) (AGTGATGGTCTATTGTCACA, provided by Alexey Revenko and Gourab Bhattacharejee, Ionis Pharmaceuticals, Carlsbad, CA), 330 mg/kg/week for 2 weeks. Plasminogen in plasma was assessed by western blot.

FXII is an 80-kDa protein containing several noncatalytic domains and a C-terminal protease domain (Figure 1A).^27,28  Kallikrein cleaves FXII after Arg³⁵³ to form αFXIIa, and subsequent cleavage after Arg³³⁴ forms βFXIIa (Figure 1A; supplemental Figure 1B). FXII with Lys or Arg replacing Thr³⁰⁹ (Figure 1B) migrate slightly faster than FXII-WT on nonreducing SDS-PAGE (Figure 1C). Lys/Arg³⁰⁹ substitutions disrupt 1, and perhaps 2, glycosylation site(s) at Thr³⁰⁹ and Thr³¹⁰,^27,29  which may account for the apparent mass loss. Cleavage of FXII after Arg³⁵³ during autoactivation or activation by kallikrein generates the 50- and 30-kDa HC and light chain (LC) of αFXIIa seen on reducing SDS-PAGE (Figure 1D).^24,27,28  On nonreducing gels (Figure 1E), αFXIIa runs as a single 80-kDa band because the HC and LC are connected by a disulfide bond (Figure 1A). With FXIIa-Lys³⁰⁹ and FXIIa-Arg³⁰⁹, additional bands appear on reducing and nonreducing gels (Figure 1D-E). Cleavage after Lys/Arg³⁰⁹ would generate fragments unconnected by a disulfide bond (Figure 1A), explaining this pattern. Hereafter, the term “HC” will be used to indicate the N-terminal noncatalytic portion of FXII as well as the larger polypeptide generated when FXII is cleaved after Arg³⁵³ or Arg³⁰⁹.

FXII-WT, FXII-Lys³⁰⁹, FXII-Arg³⁰⁹ (Figure 2A), and FXII-Ala³⁰⁹ (supplemental Figure 1C) shorten the activated partial thromboplastin time (aPTT) of FXII-deficient plasma comparably, and autoactivate in the presence of polyphosphate similarly (Figure 2B), indicating that Thr³⁰⁹ substitutions do not compromise basic function. When thrombin generation was induced with TF in FXII-deficient plasma reconstituted with FXII-Lys³⁰⁹ or FXII-Arg³⁰⁹ (Figure 2C), bands appear on FXII western blots that are not present in plasma containing FXII-WT (Figure 2C). The novel bands are similar to those generated when FXII-Lys/Arg³⁰⁹ is incubated with plasmin (Figure 2D), as reported by de Maat et al,²⁰  and to those generated during autoactivation (Figure 1D-E). They do not migrate in the same position as βFXIIa (supplemental Figure 1B), indicating that they are not due to cleavage after Arg³³⁴. Consistent with this, FXII-Arg³⁰⁹ in which Arg³³⁴ was replaced with alanine (FXII-Arg³⁰⁹, Ala³³⁴) is cleaved similarly to FXII-Arg³⁰⁹ (supplemental Figure 1A,D). This indicates that 1 or more proteases generated during coagulation cleave FXII after Lys/Arg³⁰⁹. In reactions with pure proteases, thrombin and FXIa (FXIa) cleave FXII-Lys/Arg³⁰⁹ to generate patterns identical to those in plasma (Figure 2E), but do not cleave FXII-WT (Figure 2E) or FXII-Ala³⁰⁹ (supplemental Figure 1A). Thrombin appears to be more potent than plasmin as a cleaver of FXII-Lys/Arg³⁰⁹ (Figure 2E). Mass spectrometry verified that thrombin cleaves FXII-Lys³⁰⁹ after residue 309 (supplemental Figure 2). FXII-Lys/Arg³⁰⁹ was not cleaved by FVIIa/TF, IXa, or Xa (supplemental Figure 3). Kallikrein cleaves FXII-Lys/Arg³⁰⁹ after residue 309, while also cleaving after Arg³⁵³ to form FXIIa (Figure 2F).

FXII-Lys/Arg³⁰⁹ cleaved by thrombin does not hydrolyze the tripeptide S-2302 (Figure 2G), indicating that it is a form of FXII and not FXIIa (ie, it is not cleaved after Arg³⁵³). Cleavage after Lys/Arg³⁰⁹, therefore, produces a truncated FXII (Thr³¹⁰ to Ser⁵⁹⁶), which we designated δFXII, and a δHC (Ile1 to Lys/Arg³⁰⁹) (Figure 1B). Western blot analyses indicate that the ∼45-kDa band is δHC and the ∼35-kDa doublet is δFXII (supplemental Figure 4A). δFXII derived from FXII-Lys/Arg³⁰⁹ may run as a doublet due to variable glycosylation around residues 309 and 310 (supplemental Figure 4B).²⁹  These results show that FXII-Lys/Arg³⁰⁹ may be cleaved after residue 309 by plasma proteases generated during TF-induced coagulation (Figure 2C) or during contact activation (Figures 1D and 2E).

FXII and PK undergo reciprocal conversion to FXIIa and kallikrein in the absence of a surface (Figure 3A).^24,30-32  The process is supported similarly by FXII-WT, FXII-Lys³⁰⁹, FXII-Arg³⁰⁹ (Figure 3B white symbols), and FXII-Ala³⁰⁹ (supplemental Figure 5). Preincubating FXII with thrombin enhances kallikrein generation in reactions with FXII-Lys³⁰⁹ or FXII-Arg³⁰⁹, but not FXII-WT (Figure 3B black symbols) or FXII-Ala³⁰⁹ (supplemental Figure 5). As thrombin does not activate PK,³³  this shows PK activation is enhanced in reactions in which thrombin cleaves FXII-Lys³⁰⁹/Arg³⁰⁹ to form δFXII.

Isolating δFXII for detailed analyses by cleaving FXII-Lys³⁰⁹/Arg³⁰⁹ with thrombin or plasmin proved difficult because complete cleavage after Lys/Arg³⁰⁹ does not occur before cleavage at other sites causes degradation and/or activation. To address this, we prepared FXII with a unique cleavage site for human rhinovirus 3CP near residue 309 (Figure 3C-D). FXII-3C has comparable activity to FXII-WT in aPTT (Figure 2A) and autoactivation (supplemental Figure 6) assays. Cleaving FXII-3C with 3CP produces a form of δFXII (δFXII-3C) comprising FXII residues Pro³⁰⁸ to Ser⁵⁹⁶ with an additional glycine at the N terminus (Figure 3C). δFXII-3C and the accompanying δHC (Figure 3D) can be separated chromatographically (Figure 3E), but as PK activation by δFXII-3C is the same with or without δHC (Figure 3F), subsequent experiments were performed without separating the polypeptides. As expected, polyphosphate did not induce δFXII-3C autoactivation (supplemental Figure 6), and δFXII-3C did not cleave S-2302 because it is not a form of FXIIa (data not shown). When mixed with PK, δFXII-3C induced kallikrein generation more rapidly than FXII-WT or full-length FXII-3C (Figure 3F left). Polyphosphate enhanced kallikrein generation in reactions with FXII-WT or FXII-3C,^31,32  but not δFXII-3C (Figure 3F right), consistent with the importance of the FXII HC in surface-mediated reactions.^28,34,35 

Kallikrein converted δFXII-3C to δFXIIa-3C with an initial rate ∼40-fold faster than it activated FXII-WT or FXII-3C (Figure 4A), and at least 100-fold faster than plasmin activated δFXII-3C (Figure 4B).³⁶  Fitting curves for FXII activation by kallikrein to the Michaelis-Menten equation revealed a catalytic efficiency (k_cat/K_m) ∼15-fold higher for δFXII-3C than for FXII-WT or FXII-3C (65 800 M⁻¹s⁻¹ vs 4400 and 3400 M⁻¹s⁻¹, respectively; supplemental Figure 7A). k_cat/K_m values correspond to the initial slopes of the v₀ dependences on the substrate concentration. K_ms for the reactions appear ≥3.5 μM, but could not be accurately determined because reactions did not reach saturation. FXIa, a homolog of kallikrein, also converted δFXII-3C to δFXIIa-3C ∼50-fold faster than full-length FXII (Figure 4C). Previously, we showed that FXII locked into its precursor form by replacing Arg³⁵³ with alanine has intrinsic proteolytic activity that initiates reciprocal PK/FXII activation.^24,32,37  We created similar versions of FXII-3C and δFXII-3C (FXII-T and δFXII-T) that cannot be converted to αFXIIa and δFXIIa, respectively.²⁴  δFXII-T catalyzed PK conversion to kallikrein more rapidly than FXII-T (Figure 4D). k_cat/K_m for activation by δFXII-T was ∼2.5-fold higher than by FXII-T (67 M⁻¹s⁻¹ and 26 M⁻¹s⁻¹, respectively; supplemental Figure 7B). Thus, δFXII-3C has 2 properties that enhance KKS activation relative to FXII-WT. δFXII-3C initially converts PK to kallikrein more rapidly than does FXII, and then kallikrein converts δFXII-3C to δFXIIa more rapidly than FXII to αFXIIa.

δFXIIa-3C, the activated (cleaved after Arg³⁵³) form of δFXII-3C, and the truncated species βFXIIa activate PK similarly, and at slightly reduced rates compared with αFXIIa (Figure 4E). Enhanced KKS activity with δFXII, therefore, is not due to δFXIIa being a better PK activator than αFXIIa. During contact activation in plasma, αFXIIa converts FXI to FXIa.^5,6,38  δFXIIa-3C, like βFXIIa, activated FXI poorly compared with αFXIIa (Figure 4F), and rates were not enhanced by polyphosphate (data not shown), consistent with loss of the FXII HC.

In FXII-deficient plasma supplemented with FXII-WT or FXII-3C, rapid HK conversion to cleaved HK (HKa, an indicator of kallikrein-mediated proteolysis and a surrogate for bradykinin release)^39,40  was evident when contact activation was induced with a silica-based reagent, but not in its absence (Figure 4G). δFXII-3C, in contrast, induced HK cleavage without silica, and silica did not enhance cleavage. This indicates that δFXII-3C supports PK activation in plasma by a surface-independent process that overcomes regulation by C1-INH. C1-INH inhibition of δFXIIa-3C is only slightly slower than inhibition of αFXIIa (supplemental Figure 8), indicating that δFXIIa-3C is not resistant to C1-INH. Indeed, complexes between C1-INH and kallikrein or FXIIa appear in plasma with δFXII-3C in the absence of a surface as they do with full-length FXII with a surface (Figure 4G). Most likely, the enhanced reciprocal PK/δFXII activation simply exceeds the capacity of C1-INH to control the rate of the process.

Plasma HK in mice is primarily the product of the mKng1 gene.⁴¹  An antibody specific for the high-molecular-weight species encoded by mKng1 recognizes 2 prominent bands on plasma western blots that are not present in mKng1-null mice (supplemental Figure 9A). Inducing contact activation in plasma results in reductions in the apparent sizes of these bands (supplemental Figure 9B), consistent with HK conversion to HKa and bradykinin release.^39,40 

Infusing FXII-WT or FXII-3C into WT mice so that the FXII in plasma is ∼25% (140 nM) recombinant protein does not alter the baseline HK pattern (Figure 5A). In contrast, 140 nM δFXII-3C induced rapid complete HK cleavage. Significant cleavage occurs when δFXII-3C is only ∼10% (40 nM) of plasma FXII. δFXII-3C–induced HK cleavage was reduced by coinfusing an antibody to human FXII (1B2) (Figure 5B).²²  HK cleavage did not occur when δFXII-3C was infused into mice lacking PK (Klkb1^−/−) but did occur in mice lacking FXI (F11^−/−) (Figure 5B), consistent with a major role for kallikrein in converting δFXII-3C to δFXIIa-3C, and in subsequent HK cleavage. These findings support a role for δFXII in HAE, and suggest that the FXII HC serves a role in regulating bradykinin generation.

Thrombin generation was induced in mice by TF infusion. This did not result in HK cleavage in mice given FXII-WT (Figure 5C), but did in mice receiving FXII-Arg³⁰⁹ (Figure 5D). TF-induced HK cleavage was blocked by the thrombin inhibitor hirudin and did not occur when experiments were run in F11^−/− mice (Figure 5E), supporting the conclusion that FXII-Arg³⁰⁹ is cleaved after residue 309 during coagulation to form δFXII. In this model, intravascular thrombin generation could have stimulated fibrinolysis, with plasmin subsequently cleaving FXII-Arg³⁰⁹.²⁰  We analyzed HK cleavage in mice in which plasminogen was reduced to undetectable levels with an ASO (Figure 5F). TF induced HK cleavage in plasminogen-depleted mice receiving FXII-Arg³⁰⁹, but not FXII-WT (Figure 5G), indicating that plasmin is not required for HK cleavage in this model.

IgG 15H8 binds the EGF2 domain of human FXII (Figure 1A), blocking contact activation and inhibiting collagen-induced thrombosis in baboons.²¹  In the absence of surface, however, 15H8 enhanced reciprocal PK/FXII activation, and PK activation by FXII-T (Figure 6A), suggesting that it blocks an inhibitory activity intrinsic to the FXII HC. Consistent with this, 15H8 induced HK cleavage in mice supplemented with FXII-WT (Figure 6B). In vitro, 15H8 did not induce FXII autoactivation (Figure 6C), but did enhance FXII activation by kallikrein (Figure 6D) and plasmin (Figure 6E) ∼14-fold. Therefore, 15H8 appears to neutralize an activity that normally restricts PK/FXII activation in the absence of a surface, producing an effect similar to loss of the HC in δFXII.

Angioedema refers to rapid submucosal or subcutaneous soft tissue swelling that can be life-threatening if the airway or gastrointestinal tract is involved.^42-44  It is common in allergic reactions, where it is triggered by histamine release from IgE-activated mast cells. Angioedema initiated by bradykinin is less common, and is usually associated with acquired or hereditary C1-INH deficiency.^4,8-10  More rarely, HAE occurs in individuals with normal C1-INH activity, and has been linked to mutations in other proteins, including FXII, plasminogen, and angiopoietin 1.^9,10,14,15  Lysine or arginine replacements for Thr³⁰⁹ in the FXII proline-rich region have been identified in several families with HAE and normal C1-INH.^16-20  Trypsin-like enzymes including coagulation and fibrinolytic proteases cleave substrates after lysine or arginine.^45-47  Indeed, the X-Pro-Arg sequence created by Arg³⁰⁹ in FXII is present in many thrombin substrates.³³  The fibrinolytic protease plasmin activates FXII by cleavage after Arg³⁵³.^3,10,48  de Maat et al showed that plasmin also cleaves FXII-Lys/Arg³⁰⁹ after residue 309, and that inducing plasminogen activation in plasma containing FXII-Lys/Arg³⁰⁹ enhances HK cleavage.²⁰  Here, we show that FXII is cleaved after Lys/Arg³⁰⁹ by proteases generated during blood coagulation, including thrombin. This may explain trauma-induced bouts of angioedema in patients carrying these mutations.^18,19  In contrast to plasmin, which cleaves FXII after Arg³⁵³ and Lys/Arg³⁰⁹, thrombin does not activate FXII by cleavage after Arg³⁵³,³³  indicating that FXII-Lys/Arg³⁰⁹ activation requires a second protease.

Our data support a mechanism for enhanced bradykinin production in patients carrying FXII-Lys/Arg³⁰⁹ that involves a specific sequence of proteolytic reactions, as proposed recently by Jukema et al.⁴⁹  The initial step is generation of a truncated form of FXII (δFXII) and not FXII activation. This could be catalyzed by any protease that attacks the Lys/Arg³⁰⁹-Thr³¹⁰ peptide bond, including plasmin, thrombin, or FXIa. Two properties of δFXII then accelerate KKS activation. First, the intrinsic capacity of δFXII to convert PK to kallikrein is greater than that of FXII,^24,33,37  leading to more kallikrein generated early during reciprocal activation. Second, δFXII is a better kallikrein substrate than is FXII. Although αFXIIa and δFXIIa are inhibited by C1-INH at roughly similar rates, the accelerated PK/FXII activation with δFXII appears to overwhelm the regulatory function of C1-INH at normal plasma levels of the serpin. Consistent with this, C1-INH infusion to achieve supraphysiologic plasma concentrations can be effective in treating angioedema in patients with FXII-Lys/Arg³⁰⁹.⁵⁰  Our findings do not support a process in which FXII is converted first to αFXIIa, followed by truncation to δFXIIa. δFXIIa, like βFXIIa, converts PK to kallikrein, but its ability to do so is, at best, comparable to that of αFXIIa, and would not account for the increased kallikrein generation associated with FXII-Lys/Arg³⁰⁹. It is the superiority of δFXII as a kallikrein substrate, and not a property of δFXIIa as an enzyme, that appears to accelerate KKS activation with FXII-Lys/Arg³⁰⁹.

Removal of the FXII HC is central to the proposed mechanism. During contact activation, FXII, PK, and FXI bind to an activating surface.^1-3,6  HK functions as a cofactor that facilitates PK and FXI surface binding.⁵  For FXII, the comparable cofactor activity resides in the protein’s HC region,^28,34,35  and its loss greatly reduces activity in surface-dependent processes. This is illustrated by βFXIIa, a protease formed by cleaving αFXIIa after Arg³³⁴.^4,7,51  βFXIIa converts PK to kallikrein in solution, but does not support surface-dependent FXII, PK, or FXI activation. δFXIIa, the activated form of δFXII, has similarly restricted activity. For enhancing reciprocal FXII/PK activation, then, there is no advantage to removing the HC after αFXIIa is generated, and such a process would not explain enhanced KKS activation associated with FXII-Lys/Arg³⁰⁹.

Although removing the FXII HC compromises performance in surface-dependent reactions, it enhances surface-independent FXII activation. This points to a role for this part of the protein in restricting FXII activation and, consequently, PK activation in solution. Removing the HC may expose the Arg³⁵³-Val³⁵⁴ peptide bond, facilitating activation. The antibody 15H8, which binds to the FXII HC, may produce a similar effect. 15H8 inhibits contact activation-induced thrombosis in mice and primates,²¹  but in the absence of a surface it enhances PK/FXII reciprocal activation in vitro and in vivo, mimicking loss of the HC. The results are similar to those reported by Ravon et al in 1995 for an antibody to the FXII kringle domain.³⁶  Previously, Jukema et al proposed that removing the HC with neutrophil elastase produces a FXII fragment with increased susceptibility to activation by kallikrein or plasmin.⁴⁹  It seems likely that removal of the HC can be catalyzed by several proteases that use different cleavage sites, all producing an effect similar to the one observed with δFXII. Cleavage after FXII-Lys/Arg³⁰⁹ in HAE patients, then, may be an example of a general process relevant to regulating FXII and perhaps other proteases. It is interesting to note that kallikrein removes the HC from hepatocyte growth factor activator, a FXII homolog, creating a “short form” of the protein that may be the major functional species.^32,52  Similarly, removing the plasminogen HC with elastase results in “mini-plasminogen,” which is activated by urokinase faster than full-length plasminogen.^53,54  As an aside, if truncated FXIIs with similar properties are created by cleavage at multiple sites, the term δFXII may not be ideal for the group as a whole and a revised nomenclature would be required going forward.

δFXII does not participate in contact activation because it lacks an HC. A reasonable conclusion stemming from this is that angioedema in patients with FXII-Lys/Arg³⁰⁹ results from solution-phase PK activation. The broader implications, however, remain uncertain. Bradykinin is almost certainly generated at a basal rate in normal health,^4,7,30,55  and symptoms in patients with FXII-Lys/Arg³⁰⁹ may reflect loss of the regulatory activity of the FXII HC on this process. Alternatively, normal basal bradykinin generation may occur primarily on surfaces, and loss of the HC would transfer the process to the fluid phase where it may not be regulated properly, and could more easily spread systemically.

The authors thank Jon Schoenecker (Vanderbilt University Medical Center, Nashville, TN) and Alexey Revenko and Gourab Bhattacharjee (Ionis Pharmaceuticals, Carlsbad, CA) for discussions on plasminogen, and supplying plasminogen ASO.

This work was supported by awards HL81326, HL58837, and HL140025 (D.G.) and HL130081 (I.M.V.) from the National Institutes of Health, National Heart, Lung, and Blood Institute.

Contribution: I.I. designed and conducted experiments with recombinant FXII, and contributed to data interpretation and to the writing of the manuscript; A.M. contributed to experimental design and data interpretation; M.-f.S. and S.K.D. developed systems for expressing recombinant proteins and carried out initial characterizations of recombinant FXII; B.M.M. and Q.C. contributed to design and conducted animal studies; S.K. developed the antibody to murine HK and characterized its performance; I.M.V. conducted kinetic analyses; A.G. developed the anti-FXII antibody 15H8 and contributed to experimental design, data interpretation, and the writing of the manuscript; K.M. contributed to experimental design, data interpretation, and the writing of the manuscript; and D.G. oversaw experimental design and data interpretation and the writing of the manuscript.

Conflict-of-interest disclosure: D.G. is a consultant for and receives consultant’s fees from several pharmaceutical companies (Aronora, Bayer, Dyax, Ionis, Merck, Novartis, Ono) with an interest in inhibition of contact activation proteases for therapeutic purposes. A.G. and Oregon Health and Sciences University may have financial interest in the results of this study. The remaining authors declare no competing financial interests.

Correspondence: David Gailani, Department of Pathology, Microbiology, and Immunology, Vanderbilt University Medical Center, 4918 TVC, 1301 Medical Center Dr, Nashville, TN 37232-5310; e-mail: dave.gailani@vanderbilt.edu.