Visual Overview
Abstract
Nafamostat is an approved short-acting serine protease inhibitor. However, its administration is also associated with anaphylactic reactions. One mechanism to augment hypersensitivity reactions could be inhibition of diamine oxidase (DAO). The chemical structure of nafamostat is related to the potent DAO inhibitors pentamidine and diminazene. Therefore, we tested whether nafamostat is a human DAO inhibitor. Using different activity assays, nafamostat reversibly inhibited recombinant human DAO with an IC50 of 300–400 nM using 200 µM substrate concentrations. The Ki of nafamostat for the inhibition of putrescine and histamine deamination is 27 nM and 138 nM, respectively For both substrates, nafamostat is a mixed mode inhibitor with P values of <0.01 compared with other inhibition types. Using 80–90% EDTA plasma, the IC50 of nafamostat inhibition was approximately 360 nM using 20 µM cadaverine. In 90% EDTA plasma, the IC50 concentrations were 2–3 µM using 0.9 µM and 0.18 µM histamine as substrate. In silico modeling showed a high overlap compared with published diminazene crystallography data, with a preferred orientation of the guanidine group toward topaquinone. In conclusion, nafamostat is a potent human DAO inhibitor and might increase severity of anaphylactic reaction by interfering with DAO-mediated extracellular histamine degradation.
SIGNIFICANCE STATEMENT Treatment with the short-acting anticoagulant nafamostat during hemodialysis, leukocytapheresis, extracorporeal membrane oxygenator procedures, and disseminated intravascular coagulation is associated with severe anaphylaxis in humans. Histamine is a central mediator in anaphylaxis. Potent inhibition of the only extracellularly histamine-degrading enzyme diamine oxidase could augment anaphylaxis reactions during nafamostat treatment.
Introduction
Nafamostat [(6-carbamimidoylnaphthalen-2-yl) 4-(diaminomethylideneamino)benzoate; Futhan; Fut-175] is a serine-protease inhibitor of various enzymes such as thrombin, factor Xa, factor XIIa, plasmin, or kallikrein (Aoyama et al., 1984; Fujii and Hitomi, 1981; Hitomi et al., 1985; Okajima et al., 1995). For several decades, it has been used as a treatment of pancreatitis and also as a short-acting anticoagulant for hemodialysis, leukocytapheresis, extracorporeal membrane oxygenator (ECMO) procedures, and disseminated intravascular coagulation (Han et al., 2011; Sawada et al., 2016; Minakata et al., 2019). Nafamostat is additionally considered a new treatment option for coronavirus disease (COVID-19), because it inhibits transmembrane protease serine 2 (TMPSS2), which is involved in severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) entry into target cells (Hoffmann et al., 2020; Asakura and Ogawa, 2020; Hempel et al., 2021; Yamamoto et al., 2020).
At a steady infusion rate of 0.2 mg/kg/h, nafamostat plasma concentrations reached 90 and 240 nM (Mori et al., 2003). For nafamostat treatment during ECMO, a mean dosage of 0.64 mg/kg/h was used. This would correspond to concentrations between 288 and 768 nM (Han et al., 2011). Some patients received 1.15–2.19 mg/kg/h, resulting in plasma concentrations between 518 and 2628 nM. During cardiopulmonary bypass (CPB), 0.2 mg/kg/h nafamostat was administered before and after CPB and 2 mg/kg/h during CBP (Miyamoto et al., 1992; Sakamoto et al., 2014). These high-dose infusion rates result in µM nafamostat plasma concentrations. The volume of distribution of nafamostat calculated using five healthy volunteers was 0.36 L/kg or 25.2 L in a person with 70 kg bodyweight but only 0.08 L/kg in 8 hemodialysis patients (Osono et al., 1991). The lower volume of distribution in hemodialysis patients was attributed to a possible arterialvenous fistula (Osono et al., 1991).
Camostat (4-[2-[2-(dimethylamino)-2-oxoethoxy]-2-oxoethyl]phenyl] 4-(diaminomethylideneamino)benzoate; Foipan) is a related serine-protease inhibitor used for the treatment of chronic pancreatitis and postoperative reflux esophagitis. No cases of anaphylaxis have been associated with camostat, but this may be because it is used much less frequently compared with nafamostat, and also because it is not used during extracorporeal circulation treatment. Camostat is only a precursor and is rapidly degraded in the liver to the active 4-(4-guanidinobenzoyloxy)phenylacetic acid (Midgley et al., 1994).
Several reports of nafamostat-induced anaphylactic reactions have been published in patients undergoing hemodiafiltration (Maruyama et al., 1996; Ookawara et al., 2018; Kim et al., 2016; Kim et al., 2021). When compared with that demonstrated by heparin, the adverse reaction profile during leukocytapheresis showed increased rates of “typical” histamine-mediated symptoms such as headache, nausea, rash, itching, palpitations, dyspnea, and anaphylactic shock (Sawada et al., 2016). Miyamoto et al. (1992) measured increased histamine concentrations during CBP using high dose nafamostat and suggested that nafamostat might be a diamine oxidase (DAO) inhibitor.
Human DAO (E.C. 1.4.3.22) is a copper-containing amine oxidase that oxidatively deaminates histamine and various polyamines, releasing ammonia and hydrogen peroxide (Elmore et al., 2002). In humans, high DAO mRNA levels and enzymatic activity were measured in the gastrointestinal tract with increased concentrations in the duodenum and ileum/jejunum compared with the colon. In addition, high levels were found in the proximal tubular epithelial cells of the kidneys and in the extravillous trophoblast cells, fetal cells invading the decidua and the myometrium during placenta development (Elmore et al., 2002; Schwelberger et al., 1998a; Schwelberger et al., 1998a; Velicky et al., 2018). Plasma DAO concentrations increase at least 100-fold during pregnancy (Southren et al., 1964; Boehm et al., 2017). In two large animal studies involving pigs and sheep, irreversible pharmacological inhibition of DAO activity using high doses of aminoguanidine resulted in increased morbidity and mortality after exogenous histamine challenge (Sattler et al., 1988; Sjaastad, 1967). Nevertheless, the role of DAO in the elimination of endogenous histamine during anaphylaxis or during mast cell activation is not clear.
The symmetrical diamidines, pentamidine and diminazene, are potent DAO inhibitors (McGrath et al., 2009; Duch et al., 1984). Although nafamostat is not a classic diamidine and is not symmetrical, it could be a potent DAO inhibitor because it contains one terminal amidinium and one terminal guanidinium moiety, in addition to a naphthyl group. Two of the most potent DAO inhibitors, isometamidium and prothidium, contain a phenanthridine triple aromatic ring structure and two additional individual aromatic rings (Duch et al., 1984). The marginally less active antricyde is composed of a quinoline group, a double aromatic ring like naphthalene, and a benzyl ring (Duch et al., 1984). For efficient DAO inhibition, aromatic ring structures in combination with terminal amidinium/guanidinium moieties and several nitrogen atoms are clearly important. Amiloride contains seven nitrogen atoms and a pyrazine aromatic ring but is a much weaker DAO inhibitor compared with phenamil or benzamil, which carry a second aromatic ring, phenyl or benzyl group, respectively, linked to the guanidinium moiety of amiloride (Novotny et al., 1994).
If DAO is potently inhibited by nafamostat, endogenous histamine degradation might be impaired during nafamostat administration, possibly augmenting hypersensitivity reactions. In a first step toward answering this clinically relevant question, we tested whether nafamostat is a bona fide human DAO inhibitor.
Materials and Methods
References used only in the Material and Methods section are listed in the Supplemental Material. In this section only numbers are assigned. If the reference is also used in other sections, it is regularly listed with name and year.
Chemicals and reagents
Albunorm (Octapharma, Vienna, Austria), a 20% human serum albumin (HSA) solution, is authorized for human use and 96% of the protein content is HSA. Its other ingredients, 16.8 mM caprylic acid and 16.8 mM N-acetyl-tryptophan, do not inhibit DAO activity. PBS pH 7.4 without MgCl2 and CaCl2 was purchased from Gibco (Vienna, Austria). Diminazene aceturate (D7770), DMSO (D2650), putrescine (P5780), cadaverine (C8561), histamine (53300), ortho-aminobenzaldehyde (A9628), horse radish peroxidase (HRP; P6782), amiloride (A7410), benzamil (B2417), phenamil (P203), camostat (SML0057), sodium fluoride (201154), phenylhydrazine (114715), methylhydrazine (M50001), 2-hydroxyquinoline (270873), glucose oxidase (G6125), ethyl acetate (34858), DNTB (Ellman’s reagent; D8130), vanillic acid (94770), and 4-aminoantipyrine (33528) were purchased from Sigma-Aldrich (Vienna, Austria). Nafamostat was purchased from Torii Pharmaceuticals (Futhan) and Cayman (14837; Vienna, Austria). Aminoguanidine (81530) was bought from Cayman. Foipan tablets and 5% glucose solution were provided by the pharmacy at the General Hospital Vienna. Naphthalene-2-yl-benzoate (330610050) and Amplex red (12222) were purchased from Thermo Fisher Scientific (Vienna, Austria). If not otherwise indicated, compounds were dissolved in DMSO at 10 mM and stored at −32°C for no longer than 2 years or were freshly prepared and used immediately. Foipan tablets were ground and dissolved in DMSO or water at a 10 mM camostat concentration. The DMSO solution was clear and used immediately. The water solution was centrifuged at high speed for 5 minutes, and the supernatant was assumed to contain 10 mM camostat.
DAO Activity Assay Using Hydrogen Peroxide HRP Coupling with Luminol, Amplex Red or Vanillic Acid/4-Aminoantipyrine
The luminol DAO activity assay is based on a published luminescence assay (Supplemental Ref. 1) and is described in detail (Boehm et al., 2020). The expression and purification of recombinant human DAO in Chinese hamster ovary cells has been previously published (Supplemental Ref. 2). Final DAO concentrations were 0.2–1 µg/ml (1.2–6 nM based on the dimer and excluding the molecular weight of the extensive glycosylation), quantified using absorption measurements or using an in-house developed DAO ELISA (Boehm et al., 2017). For activity assays we used 0.05–0.1% HSA PBS buffer or 80–90% plasma from healthy volunteers as matrix. The pH of the luminol solution from a commercial ECL Western blotting kit (Amersham RPN2106; Vienna, Austria) was adjusted from 9.2 to 8.0. The lower pH is closer to a physiologically relevant pH level, and the quantum yield is still sufficiently high to effectively measure DAO activity.
Using 50 µM Amplex red instead of luminol allows H2O2–HRP coupling in HSA PBS buffer with higher sensitivity. It also allows for easier oxidation rate calculations because the assay can be performed using continuous measurements with accumulating stable resorufin (Boehm et al., 2020).
We also used H2O2–HRP vanillic acid/4-aminoantipyrine coupling as described (Supplemental Ref. 3). The chromophore measured at 490 nm is relatively pH insensitive between pH 6 and pH 10. This assay was also used to measure the potency of nafamostat between pH 6 and 10 using the Britton-Robinson buffer system with 0.1% HSA (Supplemental Ref. 4).
DAO Activity Assay Using P-Dimethylaminomethylbenzylamine Oxidation
The presence of antioxidants or other molecules interfering with H2O2-HRP coupling distort proper DAO activity measurements (Boehm et al., 2020). Dimethylaminomethylbenzylamine is a substrate for DAO, which can be directly quantified. However, the Michaelis constant (Km) is only 110 µM or approximately 5.5- to 39-fold higher than putrescine and histamine, respectively (Elmore et al., 2002; Boehm et al., 2020). We used a benzaldehyde extinction coefficient of 11000 M−1cm−1 at 250 nm for rate calculations. This value is about 100-fold higher than the benzylamine moiety of the parent compound. UV-compatible 96-well half-area plates (UltraCruz UV plates; SCBT; Heidelberg, Germany) were used. An HSA concentration of 0.05% reduced the protein-based signal at 250 nm to acceptable levels.
DAO Activity Assay Using Fluorescence Measurements
DAO generates δ-1-piperideine (2,3,4,5-tetrahydro-pyridine), the autocyclized reaction product, using cadaverine as substrate. The condensate between δ-1-piperideine and ortho-aminobenzaldehyde (oABA; stored at −32°C for 4 to 6 months as a 200 mM stock solution in absolute ethanol) generates 5,5a,6,7,8,9-hexahydropyrido[2,1-b]quinazoline-10-ium or abbreviated HHPQ. Absorption and fluorescence measurement procedures of HHPQ using a Synergy H1 Multi-Mode Microplate reader (BioTek; Winooski, Vermont) have been published (Boehm et al., 2020).
For the detection of HHPQ in plasma from healthy volunteers, we mixed 85 µl plasma adjusted to pH 7.4 with 11 µl 1 M HCl per ml of plasma with 5 µl 10% ethanol or 5 µl 20 mM oABA and 5 µl PBS or 5 µl 4 mM cadaverine and 5 µl 20-fold concentrated inhibitors. Only 80 µl plasma was used testing in addition the influence of esterase inhibitors. All samples were analyzed in duplicate. After incubation for 30 minutes to 1 hour at 37°C in the dark, 200 µl 7.5% TCA (99.5% trichloroacetic acid; 91228; Sigma-Aldrich; Vienna, Austria) was added and the solution incubated on ice for 20 minutes. After 10 minutes of high speed centrifugation, 150–200 µl were recovered and fluorescence was measured as described (Boehm et al., 2020).
Measurement of Histamine Concentrations
Histamine concentrations were measured using the homogeneous time-resolved fluorescence histamine dynamic kit (62HTMPEG) from Cisbio (now Perkin Elmer) according to the instructions. However, the concentration range was adjusted to the spiked histamine concentrations. In addition, a standard curve using in-house histamine was used, rather than the histamine provided in the kit. Dilutions were selected to measure less than 10 ng/ml (90 nM) histamine, because the slope of the standard curve is steeper, improving measurement precision below this concentration. The assays were performed in 20 µl using low-volume Cisbio plates (66PL96001) and measured using custom fluorescence filters (EX 330/80; EM 620/10 and EX 330/80; EM 665/8). Plasma was diluted using plasma sample diluent from Cisbio (62DLPDDD).
Irreversible DAO Inhibition Assay
DAO was immobilized onto high-protein-binding black fluorescence microtiter plates using a monoclonal antibody against human DAO. This process has previously been described for the development and characterization of a human DAO ELISA (Boehm et al., 2017). After washing, 0.1% HSA PBS containing a final concentration of 10 µM of the tested inhibitors was added and the wells incubated for 30 minutes at 37°. After this, the wells were washed as described in the DAO ELISA protocol (Boehm et al., 2017). H2O2–HRP Amplex red coupling, the most sensitive DAO activity assay, was used to measure the remaining DAO activity over 2 hours at 37°C.
Kinetic Analysis
The H2O2–HRP Amplex red coupling assay was used for kinetic analysis. Putrescine was tested between 5 and 80 µM and histamine between 0.63 and 10 µM. All samples were tested in duplicate, and the mean was used for further calculations. For both substrates, a DAO concentration of 200 ng/ml (1.2 nM) was used. DAO activity of the different nafamostat and substrate combinations was determined using the linear part of the slope of the increasing fluorescence signal. The coefficient of determination (R2) of the used part of the curve was consistently above 98%. Statistical kinetic analysis was performed using GraphPad Prism.
In Silico Prediction of Nafamostat Binding to DAO
The docking preparations and experiments were performed using Maestro 2019-4 (Supplemental Refs. 5–8). The DAO structure in complex with diminazene, which is also the best resolution structure of DAO in the Protein Data Bank (PDB 3HIG), was prepared using the Protein Preparation Wizard (Supplemental Ref. 9). Nafamostat and diminazene were prepared using LigPrep (Supplemental Ref. 9). Nafamostat and diminazene dockings were performed with Glide by defining the grid in the centroid of diminazene in the B chain of the DAO crystal structure. The performance of the docking program and the best parameters (the centroid of diminazene was used as the centroid of the grid, no constraints and flexible ligand) were tested using control docking with diminazene. The control docking resulted in 6 diminazene poses (docking scores between 9.131 and −8.267) identical to the DAO/diminazene crystal structure (PDB ID 3HIG) (Supplemental Fig. 7). It also resulted in 21 similar poses among a total of 48. The same parameters were used for nafamostat docking. The interaction maps were created using the Protein–Ligand Interaction Profiler (Adasme et al., 2021). The Molecular Mechanics Generalized Born Surface Area binding energy was calculated for the diminazene/DAO complex structure (PDB ID 3HIG) and for the two best nafamostat poses obtained in the docking study (Supplemental Refs. 10 and 11). All figures were prepared using the PyMOL Molecular Graphics System, version 2.4.1, Schrödinger, LLC.
Ethics
The study numbers for the collection of plasma samples from healthy volunteers are EC:2030/2013 and EC:1810/2015. All healthy volunteers provided their informed consent before blood samples were collected. All procedures were performed in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national) and with the Helsinki Declaration of 1975 (revised 2013).
Results
Nafamostat is a Potent Recombinant Human DAO Inhibitor
Using luminol with H2O2–HRP coupling, the IC50 of nafamostat for DAO inhibition is 407 nM. This is comparable to phenamil, a known DAO inhibitor (Fig. 1A) (Novotny et al., 1994). The IC50 of nafamostat inhibition using vanillic acid/4-aminoantipyrine H2O2–HRP coupling is 341 nM (Supplemental Fig. 2C). Nevertheless, potential DAO inhibitors with antioxidant activity might trap radicals released by HRP, interfering with the assay validity. We therefore tested interference of nafamostat using detection of H2O2 released from glucose oxidase. It is unlikely that nafamostat is a potent glucose oxidase inhibitor. Nafamostat did not block glucose oxidase activity, thus demonstrating that there is no antioxidant activity (Supplemental Fig. 2A). We also measured direct conversion of the artificial DAO substrate p-dimethylaminomethylbenzylamine to the respective aldehyde using absorption rate changes at 250 nm (Bardsley et al., 1972). The obtained data were comparable to H2O2–HRP coupling (Fig. 1B). Finally, fusion of deaminated autocyclized cadaverine, δ-1-piperideine, with ortho-aminobenzaldehyde (oABA) was quantified. This was performed both without and in the presence of different nafamostat concentrations. The IC50 was 317 nM (Fig. 1C). Camostat or Foipan tablets dissolved in water or DMSO were at least five times less potent than nafamostat, and therefore camostat was not further pursued as a DAO inhibitor (Fig. 1D). Camostat is a prodrug, and the active metabolite is likely less active compared with the parent compound.
The data from three different DAO activity assays characterize nafamostat as a potent direct DAO inhibitor, excluding relevant assay interferences, possibly explaining DAO inhibition.
Nafamostat Is a Reversible DAO Inhibitor Whose Potency Is pH Independent at Physiologic pH Levels
All known irreversible DAO inhibitors like phenylhydrazine, methylhydrazine, or aminoguanidine possess a terminal hydrazine group covalently binding to the highly reactive topaquinone in the active center of DAO (Janes et al., 1992; McGrath et al., 2010). Although the potent DAO inhibitor diminazene aceturate carries an endogenous hydrazine group, it is a reversible DAO inhibitor (McGrath et al., 2009). Nafamostat does not have a hydrazine group. Nevertheless, tryptase is slowly irreversibly inhibited by nafamostat (Aoyama et al., 1984; Fujii and Hitomi, 1981). Nafamostat (and camostat) inhibition of the transmembrane protease serine 2 (TMPRSS2), involved in the cellular entry of SARS-CoV-2, seems covalent (Hempel et al., 2021). Therefore we tested whether nafamostat is an irreversible DAO inhibitor.
After washing microtiter plates with immobilized DAO, DAO activity after preincubation with 10 µM nafamostat was equivalent to wells without the addition of an inhibitor. The two hydrazine derivatives showed strong continuous DAO inhibition after the washing step, although irreversibility is not stable and DAO activity is slowly recovering (Fig. 2A). The potent DAO inhibitor diminazene rapidly dissociated from the active center, and DAO activity was immediately restored to control values after washing the wells (Fig. 2, A and B). These results strongly argue against irreversible DAO inhibition by nafamostat.
The Chemicalize software (ChemAxon) calculated a pKa of 7.6 and 8.5 for the guanidine moiety of nafamostat and camostat, respectively (data not shown). This is unusually low for a guanidine group, because guanidine shows a pKa of 12.5, similar to amidine. We therefore tested pH-dependent DAO inhibition of nafamostat using the Britton Robinson buffer system. If a double-protonated nafamostat molecule is important for inhibition, relative DAO inhibition at pH 6.0 should be stronger when compared with pH 8.0. If a single-protonated nafamostat is a better relative DAO inhibitor, nafamostat should be more potent at higher pH values. The data presented in Fig. 2C demonstrate that the pKa of the guanidine moiety is higher than 9.0. Relative DAO inhibition is constant between pH 6 and 8 and decreases afterward. This pattern is congruent with a pKa of 9.6 but not 7.6 (Fig. 2D). The pH-dependent DAO activity pattern is equivalent to published data (Elmore et al., 2002).
Nafamostat Is a Mixed-Mode DAO Inhibitor
In the next experiments, we employed kinetic analysis using different nafamostat concentrations combined with various histamine and putrescine substrate levels three- to fourfold below and above the published Km values. The published Km values presumed to be most reliable are 20 and 2.8 µM for putrescine and histamine, respectively (Elmore et al., 2002). We obtained nafamostat inhibition constant Ki values of 27 nM and 138 nM for putrescine and histamine respectively (Fig. 3, A and B). The ratio of the Km to the Ki of putrescine is 741 and the equivalent histamine ratio just 20. The 37-fold difference between putrescine and histamine suggests that nafamostat is a more potent DAO inhibitor using the lower-affinity substrate putrescine compared than the higher-affinity substrate histamine. The Ki ratio of histamine to putrescine of 5.1 is similar to the Km ratio of 7.1. The corresponding Vmax and Km data, including the standard errors at different nafamostat concentrations, are shown in Fig. 3, C and D. At low nafamostat concentrations, the Km values for putrescine and histamine increase 7.1- and 8.4-fold per µM increase in nafamostat (data not shown). The IC50 values dependent on the substrate concentrations are shown in Fig. 3, E and F. Based on changing Km and Vmax values, nafamostat is likely a mixed-mode inhibitor. This hypothesis was substantiated by comparing different models of inhibition using the extra sum-of-squares F test, the most appropriate test to identify the mode of inhibition according to the GraphPad manual. These data and additional parameters obtained from kinetic analysis are summarized in Table 1.
Human Plasma Shifts the Inhibitory Potency of Nafamostat
The kinetic data imply that the assay substrate concentration will have a significant influence on the potency of nafamostat to inhibit DAO activity. Normal histamine concentrations are below 1 ng/ml (9 nM) (Kaliner et al., 1982). The mean histamine concentration during severe anaphylaxis following insect sting challenge was 140 ng/ml or 1.3 µM (van der Linden et al., 1992). The other known natural substrates of DAO are the polyamines putrescine, spermidine, and spermine, with Km values of 20 µM, 1100 µM, and >3000 µM, respectively (Elmore et al., 2002). Nevertheless, the plasma or serum concentrations of these three polyamines are below 1 µM combined and are unlikely to significantly interfere with DAO inhibition by nafamostat (Russell, 1983).
Giardina et al. (2018) showed that at 1 nM tryptase 10% and 20% of human plasma shifts the IC50 values of a bivalent serine protease inhibitor 1.6- and 2.6-fold, respectively. Linear extrapolation to 100% plasma would cause an eightfold shift in the IC50 (Supplemental Fig. 6). The mechanism responsible for this shift was not elucidated. We used DAO at 1.2–6 nM (0.2 to 1 µg/ml) in our experiments. Finally, stability of nafamostat could influence the potency to inhibit DAO in plasma. Nafamostat is possibly hydrolyzed by esterases with a half-life of approximately 40 minutes using 100 µM nafamostat starting concentration and a rate of 6.5 µM/min, but a low Km of 8.9 mM (Yamaori et al., 2006). At lower nafamostat concentrations, the rate is likely to be significantly reduced. Addition of 1 mM Ellman’s reagent or DNT [5,5′-dithiobis(2-nitrobenzoic acid)] to plasma inhibited degradation of 500 µM nafamostat by 83% (Yamaori et al., 2006). The responsible esterase was not identified. In the following experiments we attempted to address two main questions. First, is the inhibitory potency of nafamostat reduced using the complex matrix plasma, and second, do esterase inhibitors increase the potency of nafamostat by blocking degradation using 80–90% plasma?
The highly sensitive H2O2–HRP Amplex red coupling assay cannot be used in plasma or serum because of the high antioxidant capacity of these complex matrices (Boehm et al., 2020). Using the recently published sensitive fluorescence assay to measure DAO activity in complex matrices like plasma or tissue extracts, we were able to reduce the substrate concentration from 200 µM to 20 µM cadaverine (Supplemental Fig. 3). The IC50 in 85% plasma shifted from 2.3 µM to 386 nM, a sixfold drop, after reducing cadaverine 10-fold from 200 to 20 µM (Fig. 4A) (Boehm et al., 2020). Lower substrate concentrations cannot be used because of assay sensitivity limitations. Addition of 1 or 2 mM DNT did not influence the IC50 values of DAO inhibition by nafamostat (Fig. 4B). Camostat is also hydrolyzed by plasma esterases and sodium fluoride inhibited camostat degradation (Midgley et al., 1994), but no effect was measured using 40 mM NaF in our assay (Supplemental Fig. 5B). We also tested 2-hydroxyquinoline (2HQ), a potent arylesterase inhibitor (Khersonsky and Tawfik, 2005), but did not see any effect on DAO inhibition (Supplemental Fig. 5B). Finally, we reasoned that high concentrations of naphthalen-2-yl benzoate might occupy the responsible enzymes potentiating nafamostat inhibition. Naphthalen-2-yl benzoate corresponds to nafamostat without both the terminal amidinium and guanidinium moieties. Diamine oxidase inhibition was not increased using 1 mM naphthalen-2-yl benzoate (Supplemental Fig. 5A). The influence of the different (aryl)esterase inhibitors on DAO activity measurements was within acceptable boundaries (Supplemental Fig. 4).
Although the IC50 values shifted sixfold after reducing the substrate concentrations from 200 to 20 µM cadaverine, the IC50 values were still at least 10-fold higher compared with the Ki of 27 nM for putrescine. In general, putrescine and cadaverine behave very similarly as DAO substrate. We also tested 0.9 µM (100 ng/ml) and 0.18 µM (20 ng/ml) histamine substrate concentrations and the IC50s were approximately 2 to 3 µM, which is 15- to 22-fold above the Ki of 138 nM (Fig. 4D). Lower histamine concentrations cannot be accurately quantified, because plasma must be diluted to avoid matrix effects. No histamine degradation was measured without addition of DAO (data not shown) and endogenous histamine was below 10 nM (data not shown).
In Silico Docking Predicts Nafamostat Binding to DAO Similar to Diminazene
Crystal structures of two potent diamidine-type DAO inhibitors, pentamidine and diminazene, complexed with DAO have been published (PDB IDs 3HII, 3HIG). Both are mixed-mode inhibitors similar to nafamostat, which may use similar amino acids for tight binding. Nevertheless, nafamostat is not a strict diamidine because it carries a terminal amidinium and a terminal guanidinium moiety. Unlike diminazene and pentamidine, it is not symmetrical because it contains a single naphthalene double aromatic ring structure linked to the terminal amidine group (Supplemental Fig. 1). Therefore, it is possible that nafamostat might prefer only one orientation for DAO binding. We used in silico docking to predict amino acid interactions and the preferred orientation for nafamostat binding. Nafamostat docking resulted in six poses, and three of them had a binding mode similar to the binding mode of diminazene in the crystal complex (PDB ID 3HIG). The aromatic rings of two similar poses with the docking scores of −7.893 and −6.910 superimposed well with those of diminazene in the DAO complex structure (Fig. 5A). In the crystal complex (Fig. 5B) the buried amidinium group of diminazene interacts with the catalytic Asp373, and the vicinal phenyl group pi-stacks with Tyr371 and Trp376. The nitrogen atoms in the triazine make hydrogen bonds with Asp186 and Tyr148. The Tyr148 residue also pi-stacks with the distal phenyl ring, which is clamped between Tyr148 and Phe435 from the other chain. The terminal amidinium forms a water-mediated hydrogen bond with Thr145. The binding site is surrounded by the hydrophobic Val458 and Ala149 residues and the aromatic residues Tyr459, Phe184, and Tyr152.
Based on the interaction analysis (Fig. 5, C and D) the second-best pose for nafamostat showed more interactions than the best pose. Because it also showed a better calculated binding energy ΔG of −67.00 kcal/mol (the best post demonstrated a ΔG of −63.09 kcal/mol), the second pose was selected as a representative binding mode for nafamostat. Like the buried amidinium in the diminazene crystal complex (Fig. 5B), the buried guanidino group in the second nafamostat pose (Fig. 5C) forms a salt bridge with a catalytic Asp373, and the vicinal phenyl group pi-stacks with Tyr371 and interacts with Trp376. Compared with amidinium, the guanidinium group in nafamostat possesses an additional nitrogen, which makes both direct and water-mediated hydrogen bonds to Asn460. Like the phenyl group in diminazene, the naphtyl group in nafamostat pi-stacks with Tyr148 and interacts with Phe435 from the other chain. The central carbonyl group creates a hydrogen bond with Tyr148, and the distal amidinium interacts with Thr145 via a direct hydrogen bond and forms water-mediated hydrogen bonds with Tyr152 and Ala149. The predicted binding mode of nafamostat to DAO is highly similar to the binding mode of diminazene in the crystal complex with a preferred orientation of the guanidinium group toward the topaquinone.
Discussion
It is not surprising that nafamostat is a potent DAO inhibitor when one compares nafamostat with the structure of the known DAO inhibitors pentamidine and diminazene. In silico ligand docking studies revealed a remarkable similarity in amino acid interactions between nafamostat and diminazene. The Ki of diminazene using insect cell–derived DAO and putrescine was 14 nM and therefore quite similar to the 27 nM measured in our experiments (McGrath et al., 2009).
The fivefold increased Ki for histamine (138 nM) compared with putrescine (27 nM) is likely a reflection of the seven times lower Km of histamine (2.8 µM) compared with putrescine (20 µM) for DAO (Elmore et al., 2002). We have seen similarly higher IC50 values for histamine compared with the simple diamines putrescine or cadaverine using other inhibitors (unpublished data). Nafamostat is clearly a potent DAO inhibitor in vitro, but the key question is whether DAO inhibition could be involved in the hypersensitivity reactions during nafamostat treatment and possibly also in effects currently ascribed to protease inhibition in animal models or clinical trials. Therefore, nafamostat inhibition of histamine degradation in plasma is more important, and here we measured IC50 values of 2–3 µM using 180 nM histamine.
Normal histamine concentrations are below 0.5 ng/ml or 4.5 nM (Pollock et al., 1991). Histamine starts to induce symptoms such as flush and headache at less than 3 ng/ml, and significant hypotension develops above 5 ng/ml (Kaliner et al., 1982; Pollock et al., 1991). We could not test lower histamine concentrations in our DAO activity assays, but during severe anaphylaxis, mean histamine concentrations of 140 ng/ml or 1260 nM have been measured, and at this level, 2–3 µM nafamostat would be sufficient to inhibit plasma DAO activity (Van der Linden et al., 1992). Table 2 summarizes published IC50 and Ki data for nafamostat using frequently pure buffer matrices or more than fivefold diluted plasma samples. Hitomi et al. (1985) and Fujii and Hitomi (1981) published ratios of IC50 to Ki values of 0.68 and 0.39 respectively for prothrombin, which circulates at 1.4 µM. In our case the ratio of IC50 using 90% plasma to Ki in buffer for histamine is 22-fold. Giardina et al. (2018) published an extrapolated IC50 shift for tryptase inhibition of at least sevenfold comparing buffer with >80% plasma (Supplemental Fig. 6). Protease inhibition assays with nafamostat might show significantly higher IC50 values, indicating weaker potency of nafamostat using high plasma concentrations. Pâques and Römisch (1991) used 75% plasma to test coagulation parameters, and the Ki values seem higher compared with the other data in Table 2. Published nafamostat inhibition data are inconsistent (Table 2 and Supplemental Table 2). What might be the reason for this strong shift in the plasma IC50 values when using plasma versus buffer?
We initially assumed that nafamostat is rapidly degraded by abundant plasma arylesterases but we were not able to find any effect of several arylesterase inhibitors on DAO inhibition using nafamostat (Yamaori et al., 2006). An alternative explanation would be that plasma proteins trap nafamostat and therefore the IC50 values increase. Table 2 lists the concentrations of nafamostat target proteases. The sum of the known nafamostat-binding proteins is approximately 5 µM. Nafamostat might bind to these proteins with significant affinity before the proteases are activated.
In our assays, we used maximally 12 nM DAO concentrations, and therefore the ratio of all nafamostat-binding proteins to DAO would be 417. This indicates that nafamostat binding to DAO competes with binding to 400-fold more abundant additional target proteins. Plasma DAO might be significantly inhibited at the high nafamostat concentrations used during hemodialysis or cardiopulmonary bypass operations with saturation of protease-binding sites.
Some of the SARS-CoV-2 cell entry inhibition data have been performed with buffer or only 10% plasma. The relatively low IC50 values of 55 nM (Hempel et al., 2021) or 10 nM (Yamamoto et al., 2020) might increase significantly using higher plasma concentrations, but irreversible binding likely follows different inhibition kinetics.
Normally, human DAO is not circulating but is present at high local concentrations in the extracellular matrix of the jejunum/ileum and renal proximal tubular epithelial cells (Boehm et al., 2017). Nevertheless, exogenous high molecular/unfractionated heparin is able to rapidly release DAO from the extracellular storage sites in the gastrointestinal tract (D’Agostino et al., 1988). During cardiopulmonary bypass operations with or without nafamostat treatment, heparin was used at 300 IU/kg, and this amount of heparin very likely released endogenous DAO (Miyamoto et al., 1992; D’Agostino et al., 1988). Nafamostat, probably present at µM plasma concentrations during the cardiopulmonary bypass procedure, might have mediated inhibition of heparin-released plasma DAO, and this inhibition could have caused the more than fivefold difference in plasma histamine concentrations (Miyamoto et al., 1992).
Diamine oxidase is also released into plasma during severe anaphylaxis and mast cell activation events in both animals and humans (Rose and Leger, 1952; Code et al., 1961; Boehm et al., 2019). Therefore, nafamostat could cause anaphylaxis and mast cell activation events to deteriorate by blocking plasma DAO activity. Recombinant human DAO might also be developed as a new first-in-class biopharmaceutical for the treatment of anaphylaxis, mastocytosis, chronic urticaria, or asthma exacerbations (Gludovacz et al., 2021). Coadministration of nafamostat might interfere with the potency of recombinant human DAO.
High concentrations of DAO are bound extracellularly to interstitial heparan sulfate proteoglycans in the gastrointestinal tract and renal proximal tubular cells, but the role of DAO in the degradation of endogenous histamine released during anaphylaxis is not clear. Nevertheless, if the concentration of nafamostat in the interstitial fluid is high enough to efficiently inhibit local matrix-bound DAO, higher histamine concentrations are not only locally present after release from degranulating mast cells but will also reach the circulation. Locally and systemically elevated histamine levels will increase clinical symptom severity during anaphylaxis, mast cell activation syndrome, mastocytosis, or chronic urticaria. Of the approximately 100 mg total histamine stored in the body, 50% and 30% are located within the granules of the mast cells in the gastrointestinal tract and the skin, respectively. It is these organs that frequently demonstrate symptoms during hypersensitivity reactions (Boehm et al., 2021).
The interstitial fluid contains about 30% of the protein concentration compared with plasma (Supplemental Table 1) (Fogh-Andersen et al., 1995) and therefore, simplified, the concentration of nafamostat-binding proteins might be only about 1.5 µM, and this could increase the potency of nafamostat to inhibit local DAO activity in the gastrointestinal tract and kidneys.
What are the concentrations of nafamostat in the interstitial fluid or in the gastrointestinal tract and kidneys? The volume of distribution for nafamostat in healthy volunteers was described as 0.36 L/kg or 25 L in a person weighing 70 kg, indicating that nafamostat is present at high concentrations in the interstitial fluid (Osono et al., 1991). The interstitial fluid compartment is about three times the plasma volume. In rats, 15 minutes after intravenous infusion, nafamostat rapidly accumulated in the kidneys and duodenum/jejunum (65- and 8- to 10-fold, respectively, compared with plasma). It might then be reabsorbed in the kidneys in the proximal tubules via organic cation transporters (Supplemental Tables 3 and 4) (Nanpo et al., 1984; Li et al., 2004). Pentamidine and furamidine, two diamidines with related chemical structures compared with nafamostat, are also transported via human organic cation transporters (Ming et al., 2009). Diamine oxidase is located in the extracellular matrix of renal proximal tubular and gastrointestinal epithelial cells. If humans show a similar accumulation of nafamostat when compared with rats, DAO might be potently inhibited by the high tissue concentrations of nafamostat. Nafamostat might compete with the transport of histamine into basophils or other cells. Histamine also uses organic cation transporters, and if high plasma nafamostat concentrations interfere with histamine reabsorption into basophils, and possibly endothelial cells and other cells, the relative histamine exposure would increase (Schneider et al., 2005; Sakata et al., 2010).
Nafamostat is clearly associated with hypersensitivity reactions in humans, but is there additional in vivo evidence of DAO inhibition and consequently of elevated histamine concentrations? High-dose nafamostat administration in dogs, defined as plasma concentrations of more than 15 µM for 24 hours, caused the mean arterial pressure to drop statistically significantly after 6 hours and remain low until the end of the infusions 18 hours later (Okamoto et al., 1994). The mean arterial pressure decrease was approximately 35%, from 125 to 80 mm Hg. DAO inhibition at these high nafamostat concentrations could have increased the circulating histamine levels and at least cocontributed to hypotension in dogs, which are sensitive to histamine (Owen et al., 1982).
Ceuleers et al. (2018) tested three nafamostat concentrations (0.1, 1, and 10 mg/kg) in a rat colitis model for irritable bowel syndrome. Tryptase, a biomarker for mast cells, was increased during inflammation, indicating the involvement of mast cells. Although nafamostat is a highly potent inhibitor for tryptase, only the lower nafamostat concentration showed beneficial effects in this model. Interestingly, experimental colitis in mice seems to be driven by histamine released from mast cells via histamine 4 receptor binding (Wechsler et al., 2018). We speculate that medium and high nafamostat concentrations not only inhibited tryptase but also local DAO, thus resulting in elevated proinflammatory histamine concentrations counteracting the positive effects of tryptase inhibition.
In conclusion, nafamostat is a potent DAO inhibitor. During anaphylaxis or in general mast cell activation with massive release of histamine, concomitant nafamostat treatment might cause potent DAO inhibition leading to elevated histamine levels with possibly life-threatening consequences.
Acknowledgments
The authors thank three diploma students, Julia Henkel, Linda Thurner, and Marija Gorickic, for performing some of the experiments used in this publication. The library at the MUV was indispensable for providing literature. The authors also thank the bioinformatics (J.V. Lehtonen), translational activities and structural biology (FINStruct) infrastructure support from the Biocenter Finland and the computational infrastructure support from the CSC IT Center for Science at the Structural Bioinformatics Laboratory (SBL) of the Åbo Akademi University. The SBL is part of the NordForsk Nordic POP (Patient Oriented Products) and the Solutions for Health strategic area of the Åbo Akademi University. The authors are indebted to Sarah Ely for the final polish in the proper usage of the English language.
Authorship Contributions
Participated in research design: Boehm, Jilma.
Conducted experiments: Petroczi, Gludovacz, Alix, Vakal.
Performed data analysis: Boehm, Alix, Vakal, Salminen.
Wrote or contributed to the writing of the manuscript: Boehm, Borth, Salminen, Jilma.
Footnotes
- Received April 1, 2022.
- Accepted May 16, 2022.
This work was supported by the InFLAMES Flagship Program of the Academy of Finland [decision number: 337530] and the Sigrid Juselius foundation.
↵This article has supplemental material available at jpet.aspetjournals.org.
Abbreviations
- DAO
- diamine oxidase
- H2O2
- hydrogen peroxide
- HRP
- horseradish peroxidase
- HSA
- human serum albumin
- Ki
- inhibitory constant
- Copyright © 2022 by The American Society for Pharmacology and Experimental Therapeutics