Zafirlukast is a broad‐spectrum thiol isomerase inhibitor that inhibits thrombosis without altering bleeding times

Background and Purpose Multiple members of the thiol isomerase (TI) family of enzymes are present in and released by platelets. Inhibition of these enzymes results in diminished platelet responses, aggregation, adhesion and thrombus formation. Recently, the therapeutic potential of TI inhibition has been recognised and drug‐development technologies were used to identify selective small molecule inhibitors. To date, few pan‐TI inhibitors have been characterised and the most studied, bacitracin, is known to be nephrotoxic, which prohibits its systemic therapeutic usage. Experimental Approach We therefore sought to identify novel broad‐spectrum inhibitors of these enzymes and test their effects in vivo. A total of 3,641 compounds were screened for inhibitory effects on the redox activity of ERp5, protein disulphide isomerase (PDI), ERp57, ERp72 and thioredoxin in an insulin turbidity assay. Of the lead compounds identified, zafirlukast was selected for further investigation. Key Results When applied to platelets, zafirlukast diminished platelet responses in vitro. Zafirlukast was antithrombotic in murine models of thrombosis but did not impair responses in a model of haemostasis. Since TIs are known to modulate adhesion receptor function, we explored the effects of zafirlukast on cell migration. This was inhibited independently of cysteinyl LT receptor expression and was associated with modulation of cell‐surface free thiol levels consistent with alterations in redox activity on the cell surface. Conclusion and Implications We identify zafirlukast to be a novel, potent, broad‐spectrum TI inhibitor, with wide‐ranging effects on platelet function, thrombosis and integrin‐mediated cell migration. Zafirlukast is antithrombotic but does not cause bleeding.


| INTRODUCTION
Thrombosis-related pathologies such as myocardial infarction, stroke, pulmonary embolism or deep vein thrombosis are the primary cause of mortality for much of the Western world , highlighting the importance of identifying new therapeutic targets to treat these diseases. One such interesting target is protein disulphide isomerase (PDI), which, when inhibited, attenuates both arterial and venous thrombosis, unlike current clinically used therapies (Cho et al., 2012;Jasuja et al., 2012). Protein disulphide isomerase is the most well-studied member of a family of thiol isomerases (TIs) that includes ERp5 (Jordan et al., 2005), ERp57 (Holbrook et al., 2012), ERp72 (Holbrook et al., 2018), ERp44, TMX3 and thioredoxin (Holbrook et al., 2010), which are found on the surface of platelets and involved in the propagation of platelet activation and subsequent thrombus formation. Using inhibitory antibodies that are selective for cell-surface protein disulphide isomerase (Essex & Li, 1999;Lahav et al., 2002), ERp5 (Jordan et al., 2005), ERp57 (Holbrook et al., 2012) or ERp72 (Holbrook et al., 2018), thiol isomerase (TI) inhibition has been shown to result in decreased platelet aggregation, granule secretion, integrin activation, integrin-mediated adhesion and thrombus formation.
We have previously shown that while inhibition of specific platelet thiol isomerase results in broadly similar impact on platelet activation and thrombosis, these effects are not identical suggesting different substrates and/or mechanisms of action and therefore each thiol isomerase could be an independent target (Holbrook et al., 2012;Holbrook et al., 2018). Importantly, the effects of platelet thiol isomerases are non-redundant, where addition of one family member is unable to rescue the absence of another (J. Zhou et al., 2017), again suggesting distinctive modes of action. Given the non-redundant roles of these enzymes, it is important to determine if broad-spectrum inhibition is an effective therapeutic strategy and to determine its effects in vivo with respect to prevention of thrombosis without compromising haemostasis. To date, the only known broad-spectrum thiol isomerase inhibitor is the antibiotic bacitracin, which is limited to topical use due to nephrotoxicity (Michie, Zintel, Ma, Ravdin, & Ragni, 1949). Thus, in this study, we sought to identify new family-wide (pan) inhibitors that may show promise for human disease prophylaxis. We report that the structurally related cysteinyl LT (CysLT ) receptor antagonists zafirlukast and montelukast are powerful inhibitors of thiol isomerase activity.
Zafirlukast, the more potent inhibitor, was found to modulate platelet function and integrin function in vitro and inhibit thrombosis in mice while bleeding was unaffected.

What is already known
• Platelets contain multiple TIs which are key to the regulation of platelet activation and thrombosis.
• No pan-TI inhibitors that are safe and tolerated in vivo are therapeutically available.

What this study adds
• Zafirlukast inhibited multiple TI and platelet function in vitro and in vivo.
• Zafirlukast inhibits cell migration through redox modifications of cell-surface proteins.

What is the clinical significance
• Zafirlukast shows antithrombotic effects without affecting haemostasis.

| Insulin-based turbidometric assay
Initial compound library screening was performed in an insulin turbidity assay adapted for high-throughput screening (Holmgren, 1979;Smith et al., 2004). Lead compounds were diluted in a 6-point dose curve (concentrations depending on the observed effects in the initial screen) and insulin turbidity was measured kinetically each minute for 75 min using a Spectramax M3 microplate reader (Molecular Devices, Sunnyvale, CA).

| Platelet preparation and stimulation
Washed human platelets from consenting, drug-free donors were prepared by differential centrifugation as described previously (Holbrook et al., 2018) and suspended to a density of 4 × 10 8 cellsÁml −1 in 2.3 | Platelet granule secretion and calcium mobilisation measurements P-selectin exposure in human and mouse platelets was measured by flow cytometry and calcium mobilisation in human platelets assayed as described previously using Fura-2 AM (Holbrook et al., 2018). ATP secretion from dense granules was measured using lumiaggregometry (Holbrook et al., 2012).

| Proliferation assays
MDA-MB-231 or HEK293T cells were seeded at a density of 2.5 × 10 5 cells per well in a six-well cell culture plate and grown for 24 hr. Vehicle (DMSO, 0.1% v/v) or zafirlukast was then added for a further 72 hr. Following drug treatment, wells were washed and cells manually counted by haemocytometer or automatically using a Coulter counter (Beckman Coulter, CA, USA). Cell growth inhibition was calculated by comparing the counts of drug treated cells with untreated (control) cells as described previously (Kennedy et al., 2007).

| Zafirlukast is a potent, broad-spectrum thiol isomerase inhibitor
A total of 3,641 compounds were assayed for their ability to inhibit ERp57 activity by insulin turbidity assay. Initial screen data were cross-checked against databases of compounds with known false positive readouts, and consequently, 11 compounds selected as potential leads. These compounds were assayed for time-dependent enzymeinhibitory activity of the other known platelet thiol isomerases PDI, ERp5, ERp72 and thioredoxin. A compound of particular interest, zafirlukast, from the CysLT receptor antagonist family was found to exert inhibition on all enzymes tested ( Figure 1). Additionally, the effects of the related CysLT receptor antagonist montelukast were explored ( Figure S1). The insulin turbidity assay was also repeated to measure the effects of zafirlukast in this assay in the absence of enzyme. Zafirlukast did not cleave insulin or alter its optical properties in solution ( Figure 1a).
Despite zafirlukast (Figure 1b-f) and montelukast ( Figure S1) exhibiting differing potencies, both compounds (at the highest concentration tested of 300 μM) were able to almost entirely ablate the enzyme activity of PDI, ERp57 and thioredoxin.
Both zafirlukast and montelukast are FDA-approved treatments for the prevention of asthma (marketed as Accolate and Singulair, respectively). Both drugs are clinically tolerated well, can be administered orally for long-term therapy and show little toxicity. In addition to their role in CysLT receptor function, we demonstrate a novel role for zafirlukast and montelukast as potent thiol isomerase inhibitors. Since zafirlukast showed higher potency than montelukast in the insulin turbidity assay, we focused our functional studies on the use of zafirlukast. F I G U R E 1 Zafirlukast (ZFL) is a broad-spectrum thiol isomerase inhibitor. A total of 3,641 compounds were screened for inhibitory effects on the redox activity of ERp5, protein disulphide isomerase (PDI), ERp57, ERp72 and thioredoxin (TRX), all assayed at 10 μgÁml −1 except ERp5 which was assayed at 30 μgÁml −1 . ZFL was selected for further investigation. The ability of zafirlukast to inhibit thiol isomerase activity and its specificity was measured by insulin turbidity assay in which insulin chain precipitation through disulphide bond reduction was measured at a wavelength of 650 nm in a spectrophotometer at 37 C.  . Figure 4c shows maximum fluorescence intensity measurements for vehicle (n = 18 thrombi, black circles) or zafirlukast-treated mice (n = 12 thrombi, red squares); thrombus size was significantly impaired by zafirlukast treatment. Since significant and consistent effects were observed with 12 thrombi in the zafirlukast-treated mice, no further mice were killed. Therefore this group had fewer replicates than the control but was still adequately powered. We assessed the effects of zafirlukast on bleeding time, measured by murine tail bleeding assay. Figure 4d shows that zafirlukast treatment (bleeding time of 243 s ± 31 s, n = 10, red squares) did not significantly prolong or F I G U R E 3 Platelet granule secretion and calcium flux in human and mouse platelets is diminished by zafirlukast (ZFL). Washed human platelets (4 × 10 8 cellsÁml −1 ) were incubated with ZFL (0.1-10 μM) or vehicle for 5 min prior to stimulation with collagen (1 μgÁml −1 ). Chronolume reagent was added for 2 min prior to stimulation, and luminescence recorded. (a) Representative traces of ATP secretion, and (b) is log dense granule secretion values, n = 5. (c) For α-granule secretion, human platelet P-selectin exposure was measured by flow cytometry. Platelets (2 × 10 8 cellsÁml −1 ) were incubated with vehicle or zafirlukast (0.1-20 μM) for 5 min prior to stimulation with 0.25 μgÁml −1 CRP-XL. Anti-human P-selectin PE conjugate was added for 20 min (1:500 dilution), and samples were fixed with 0.2% (v/v) paraformaldehyde, n = 6. (d) Following incubation with vehicle or zafirlukast (0.6-20 μM), mouse platelets were stained with anti-mouse P-selectin FITC conjugate and stimulated with thrombin (0.1 UÁml −1 ), n = 12. For Ca 2+ mobilisation assays, platelets (4 × 10 8 cellsÁml −1 ) were loaded with Fura-2 AM and incubated with either vehicle or zafirlukast (2.5-40 μM) for 5 min before addition of CRP-XL (0.5 μgÁml −1 ), (e) mean Ca 2+ trace. (f) peak Ca 2+ concentration normalised to vehicle values, n = 6. Graphs represent mean ± SEM; data were analysed by one-way ANOVA, *P < 0.05 shorten bleeding times compared to vehicle treatment (bleeding time of 225 s ± 28 s, n = 10, black circles).

| Zafirlukast inhibits cell migration by mediation of cell-surface thiols in MDA-MB-231 and HEK293T cells
In addition to the effects of thiol isomerases on the modulation of platelet integrin activation, thiol isomerases are known to be impor- Expression of the thiol isomerases PDI, ERp57, ERp72 and ERp5 was detected on "the surface of" non-permeabilised ( Figure S3) and permeabilised MDA-MB-231 cells ( Figure  S4). Using

F I G U R E 4 Thrombus formation in vivo is inhibited by zafirlukast with no impact on bleeding times. The effects of zafirlukast (ZFL) on
thrombus formation in mice were determined following laser injury of cremaster muscle arterioles and observed by intravital microscopy. Male C57/BL6 mouse platelets were labelled with DyLight 649-conjugated anti-GPIb antibody (0.2 μgÁg −1 body weight) and either vehicle (a) or zafirlukast infused (at a volume required to achieve a circulating concentration of 20 μM) (b). Following laser injury, images were recorded for 5 min. The vessel wall is outlined in each pre-injury panel, and the red arrow indicates direction of blood flow. (c) Maximum fluorescence intensity of each thrombus formed in vehicle-treated mice (n = 18 thrombi, black circles) or zafirlukast-treated mice (n = 12 thrombi, red squares). (d) The effects of zafirlukast on bleeding were determined by tail bleeding assay. Vehicle or zafirlukast (at a volume required to achieve a circulating concentration of 20 μM) was infused into the femoral veins of C57/BL6 mice, 5 min prior to tail biopsy; 0.5 cm of tail tip was excised, and blood collected in PBS, and time to cessation of bleeding was recorded. Graphs represent mean ± SEM, n = 10 per treatment; data were analysed by Student's t-test, *P < 0.05 immunocytochemistry, staining for all enzymes was observed. Secondary antibody controls displayed no signal in the Alexa-Fluor 488 channel. With permeabilization, some differences in staining were observed; PDI and ERp57 staining appeared to increase suggesting that although a small pool of cell-surface thiol isomerase exists, a greater pool is retained within the cells.
The effects of zafirlukast in integrin-mediated cell migration were measured by scratch assay in which a scratch was inflicted on a monolayer of cells treated with either vehicle (Figure 5a) or zafirlukast (20 μM, Figure 5b), and images recorded until successful closure of vehicle-treated scratch wells was achieved. Figure 5c, was also able to impair proliferation. At the concentration used in the scratch assay (20 μM), zafirlukast had a minimal effect on cell proliferation and at a concentration of 30 μM, zafirlukast only modestly reduced proliferation by around 20% (Figure 5f).
In order to confirm that this effect was associated with decreased cell-surface redox activity, labelling of surface free thiols was performed using the cell-impermeant label MPB. At assay endpoint, cells were washed and labelled with MPB and following SDS-PAGE separation, visualisation of the biotin moiety was performed using a streptavidin conjugate (Figure 5g). In vehicle-treated samples, many bands were detected corresponding to multiple proteins containing free thiols. This was reduced by 42% in the zafirlukast-treated lane consistent with changes in the redox state of cell-surface proteins, specifically free thiols, implicating decreased thiol isomerase reductive function as a potential mode of action (Figure 5h). (c,d) At 37 hr post-scratch, cells were fixed and permeabilised, then stained with phalloidin 568 for 1 hr, and following washes, counterstained with DAPI for 20 min. Cells were washed in PBS, and Vectashield H-1000 was added prior to imaging. (e) Images were analysed using Zeiss Axiovision software, and closure areas calculated using Fiji ImageJ, n = 9. (f) Log proliferation data after 72-hr treatment with zafirlukast of 1-30 μM (n = 6).

MDA-MB-231 cells express
(g) For thiol labelling, MDA-MB-231 seeded into wells of a 12-well cell culture dish were grown to confluence, and vehicle or zafirlukast was added to each well in the presence of DMEM supplemented with 1% (v/v) FBS. Cells were washed 4 hr post-treatment with PBS and free thiols labelled by incubation with MPB (100 μM) for 20 min. Free label was removed by multiple PBS washes and cells lysed in reducing sample treatment buffer for SDS-PAGE analysis. Cell-surface thiol labelling was detected using streptavidin-HRP conjugate (1:10,000 dilution in 5% [w/v] BSA/TBS-T), and chemiluminescence captured following the addition of ECL substrate. (h) Quantification of band intensities for surface thiol labelling (n = 9). Graphs represent mean ± SEM. Data were analysed by Student's t-test, *P < 0.05 Mackie, & Straiker, 2011;Figueroa, Kramer, Strange, & Denton, 2019).
Similar to data presented for MDA-MB-231, expression of thiol isomerases was evident in both non-permeabilised ( Figure S5) and permeabilised cells ( Figure S6). In permeabilised HEK293T, staining of PDI and ERp57 appear to be increased and change distribution compared to non-permeabilised cells whereas ERp72 and ERp5 staining appear to decrease suggesting that a greater pool of these proteins is associated with the cell surface.
In order to explore the effects of zafirlukast on thiol isomerase- Graphs represent mean ± SEM. Data were analysed by Student's t-test, *P < 0.05 demonstrated that hepatic clearance of zafirlukast from the plasma is reduced in patients with cirrhosis of the liver and in elderly patients, so treatment of these patients with zafirlukast is not advised. Zafirlukast is metabolised exclusively in the liver by the cytochrome P450 isoenzyme CYP2C9. Liver clearance of zafirlukast can be altered by co-administration with the drugs erythromycin, theophylline or terfenadine, which reduce plasma zafirlukast levels while aspirin enhances its plasma levels, Therefore these drugs are not recommended in patients receiving zafirlukast treatment (Dekhuijzen & Koopmans, 2002).
Clinical efficacy is observed when oral doses of 20 mg are given twice daily and higher doses are tolerated. This dose has been shown to result in a peak plasma concentration (cMax) of 697 ngÁml −1 at 3 hr, which is equivalent to a plasma concentration of 1.2 μM (Fischer et al., 2000).
Platelets do not produce CysLT as they lack 5-LOX and therefore are unable to convert arachidonic acid to CysLTs unless they are interacting with immune cells that can generate the lipid precursor (Laidlaw & Boyce, 2012). Healthy human donor platelets have been shown to express the CysLT receptors CysLT 1 and CysLT 2 , a small pool of which are found on the cell surface (Hasegawa et al., 2010). Zafirlukast treatment diminished platelet aggregatory responses and granule secretion in a concentration-dependant manner. In addition to the effects on granule secretion, zafirlukast impaired calcium mobilisation and fibrinogen ligation to integrin α IIb β 3 . Using a murine thrombosis model, we were able to demonstrate that zafirlukast diminished thrombosis in vivo but did not impair or enhance bleeding suggesting that pan-thiol isomerase inhibition is tolerated and is a viable therapeutic option, worthy of further exploration as many currently prescribed anti-thrombotic therapies inhibit platelet function but also cause bleeding.
Differential regulation effects of inhibition of multiple thiol isomerases between different functional assays were evident with dose response profiles not being identical in the effects observed. IC 50 values obtained for platelet aggregation were around fourfold lower (1.66 μM) than required to elicit the same level of inhibition of dense granule secretion. Similarly, α-granule secretion required concentrations of 5-μM zafirlukast or more to cause significant inhibition. Previous studies have highlighted differential effects of this nature; antibody-mediated inhibition of ERp5 or ERp72 results in dysfunctional α-granule secretion (Holbrook et al., 2018;Jordan et al., 2005), whereas little or no effect is observed with PDI or ERp57 blockade (Holbrook et al., 2012;Kim et al., 2013). Likewise, calcium flux is regulated by ERp57 and ERp72 but not PDI or ERp5. In in vivo assays, all thiol isomerases have been shown to be critical for initial thrombus formation and propagation; however, curiously not all thiol isomerases are critical for haemostasis in mice. ERp57 and ERp72 ablation results in prolonged tail bleeding times (Wang et al., 2013;J. Zhou et al., 2017), while, as for zafirlukast treatment of mice, PDI-deficient mice do not show a haemostasis defect (Kim et al., 2013). We propose that this may be a consequence of different effects of each enzyme on individual aspects of platelet activation balanced with different enzymes being inhibited at different zafirlukast concentrations. Therefore, non-overlapping aspects of platelet regulation may require greater levels of inhibitor to counterbalance the sensitive to inhibition and insensitive to inhibition effects of each enzyme in the multiple pathways of platelet activation.
Thiol isomerases are also important for mediating integrin conformational changes associated with cell adhesion and migration. We therefore explored the effects of zafirlukast on these processes using a  et al., 2005). Similarly, recent findings suggest that zafirlukast can directly activate PPARγ signalling in adipocytes (Gobel et al., 2019). In platelets, PPARγ ligands negatively regulate platelet responses (Moraes et al., 2010). The potential effects of zafirlukast on these pathways in platelets remain to be explored and cannot be ruled out as additional mechanisms of zafirlukast action on platelet function.
Since the onset of this study, recent publications report that zafirlukast may also be beneficial for the treatment of endothelial dysfunction, inflammation (X. Zhou, Cai, Liu, Wu, & Gao, 2019), glioblastoma (Piromkraipak et al., 2018) and bladder cancer (Nguyen et al., 2018). It would be interesting to establish if these effects are related to the inhibition of thiol isomerase activity.
Most inhibitors of thiol isomerases characterised to date act by competitive binding to the substrate binding domains of the enzyme; therefore, future studies may entail a detailed kinetic analysis of each enzyme to determine how zafirlukast binds and modifies its activity.
Since zafirlukast is able to inhibit at least five enzymes from a family of at least 20 thiol isomerase proteins, such elaborate studies were beyond the scope of this study.
In summary, we provide evidence that the anti-asthma compound zafirlukast is a potent pan-thiol isomerase inhibitor which modulates platelet function, cell migration, and thrombosis but does not show detrimental effects on haemostasis.

CONFLICT OF INTEREST
The authors declare no conflicts of interest.

RIGOUR
This Declaration acknowledges that this paper adheres to the principles for transparent reporting and scientific rigour of preclinical research as stated in the BJP guidelines for Natural Product Research, Design & Analysis, Immunoblotting and Immunochemistry and Animal Experimentation, and as recommended by funding agencies, publishers, and other organisations engaged with supporting research.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request. Some data may not be made available because of privacy or ethical restrictions.