A therapeutic combination of two small molecule toxin inhibitors provides pancontinental preclinical efficacy against viper snakebite

Snakebite is a medical emergency causing high mortality and morbidity in rural tropical communities that typically experience delayed access to unaffordable therapeutics. Viperid snakes are responsible for the majority of envenomings, but extensive interspecific variation in venom composition dictates that different antivenom treatments are used in different parts of the world, resulting in clinical and fiscal snakebite management challenges. Here, we show that a number of repurposed Phase 2-approved small molecules are capable of broadly neutralizing distinct viper venom bioactivities in vitro by inhibiting different enzymatic toxin families. Furthermore, using multiple in vivo models of envenoming, we demonstrate that a single dose of a rationally-selected dual inhibitor combination consisting of marimastat and varespladib prevents lethality caused by venom from the most medically-important vipers of Africa, South Asia and Central America. Our findings strongly support the translation of combinations of safe and affordable enzyme inhibitors as novel broad-spectrum therapeutics for snakebite.

combinations of safe and affordable enzyme inhibitors as novel broad-spectrum therapeutics for snakebite.

Introduction
Snakebite is a Neglected Tropical Disease (NTD) that causes extensive mortality (~138,000/annum) and morbidity (~400,000/annum) in the impoverished rural communities of sub-Saharan Africa, South and Southeast Asia, and Central and South America 1 . Despite annual snakebite deaths equating to one quarter of those that succumb to malaria 2 , this NTD has long been overlooked by the global health community, resulting in little investment in snakebite management, the development of new therapeutics or improving speed of access to treatment. In 2017, snakebite was reclassified as a priority NTD by the World Health Organization (WHO) and, soon after, a global roadmap was published outlining the goal of halving snakebite-related deaths and disabilities by 2030 3 . Key tasks to achieve this goal include those relating to therapeutics, specifically the necessity to improve their safety, efficacy, affordability and accessibility to those in greatest need.
Snake venoms are complex mixtures of numerous proteins and peptides and extensive interspecific variation in venom composition poses major challenges for the development of generic snakebite treatments 4,5 . Current therapies, known as antivenoms, consist of polyclonal immunoglobulins purified from the plasma/serum of large animals (e.g. equines, ovines) hyperimmunized with snake venoms. Because of the specificity of the resulting immunoglobulins towards the toxins present in the venoms used in manufacture, antivenoms typically have limited efficacy against envenoming by different snake species 6 .
Consequently, distinct antivenom products are produced (>45 manufacturers worldwide) to treat envenoming by numerous snake species found in different parts of the world, resulting in a highly fragmented drug market, issues with affordability, and a lack of sustainability 7,8 .
Other limitations with current antivenom include: (i) poor dose efficacy, as the majority (~80-90%) of their immunoglobulins do not bind venom toxins 1,9 , (ii) high incidences of adverse reactions due to the administration of large doses of foreign immunoglobulins 10 , (iii) the requirement for intravenous delivery in a hospital setting, and (iv) reliance on cold chain transport and storage. In addition, many rural snakebite victims suffer major delays in accessing healthcare facilities following a bite, if they choose to attend at all, as evidenced by estimates suggesting that 75% of snakebite deaths occur outside of a hospital setting 11 .
Cumulatively, these limitations identify an urgent and compelling need to develop crossgenerically efficacious, stable and affordable, prehospital treatments as an effective means to considerably decrease snakebite mortality and morbidity 12,13 .
Vipers represent a major group of medically important snakes that are widely distributed across the globe, ranging from the Americas to Africa and Asia, and are responsible for causing the majority of snake envenomings in these regions [14][15][16] . Treatments for systemic viper envenoming need to neutralize a number of major classes of hemotoxins, which are found in varying abundances across medically important snake species, and typically include the Zn 2+ -dependent snake venom metalloproteinases (SVMPs), phospholipase A 2 (PLA 2 s) and snake venom serine proteases (SVSPs) 17 . Collectively, these three enzymatic families typically comprise >60% of all toxins found in viper venom proteomes 5 and, in combination, are responsible for: (i) the destruction of local tissue, often resulting in necrosis, (ii) the degradation of the basement and cellular membranes resulting in extravasation, and (iii) the onset of coagulopathy via the activation and breakdown of clotting factors -with the latter two effects often culminating in life-threatening systemic hemorrhage [17][18][19][20] .
Historically, small molecule toxin inhibitors have received limited attention as potential alternatives to immunoglobulin-based snakebite therapies 12,[21][22][23][24][25] , although recent findings have suggested that a number of Phase-2 approved drugs may hold therapeutic promise 23,[26][27][28] . Perhaps the most notable of these is the PLA 2 -inhibitor, varespladib, which has been widely explored for repurposing as a snakebite therapy, and has shown substantial promise in preclinical models against a number of elapid and viper venoms 22,26,27,29 . In addition, several SVMP-inhibitors have been demonstrated to be capable of abolishing venom-induced hemorrhage or dermonecrosis, including metal ion chelators 21,24,25,28 and peptidomimetic hydroxamate inhibitors 23,24,30 . We recently reported that 2,3-dimercapto-1-propanesulfonic acid (DMPS), a Zn 2+ chelator that is a licensed oral medicine used to treat heavy metal poisoning, was particularly effective in preclinically neutralizing both the local and systemic toxicity of Zn 2+ -dependent SVMP-rich saw-scaled viper venoms (genus Echis) 28 . However, despite the great promise of both varespladib and DMPS as orally delivered prehospital therapeutics for snakebite, both are likely to be somewhat restricted in terms of their efficacy, as each predominately targets only one of the handful of major toxin families found in the venoms of medically important snakes.
To address this limitation, and cognisant of the complexity of snake venoms, herein we explored the potential of combinations of small molecule toxin inhibitors as new 'broad spectrum' snakebite therapeutics. Our goal -to rationally select and preclinically validate a therapeutic small molecule mixture capable of neutralizing distinct pathogenic toxins found in the venoms of geographically diverse, medically important, hemotoxic vipers -was achieved. Thus, we demonstrate, in a mouse model of envenoming, that a single dose of the SVMP-inhibitor marimastat, combined with the PLA 2 -inhibitor varespladib, provides in vivo protection against the lethal effects of envenoming caused by the most medically important vipers of Africa, south Asia and Central America. Our findings hold much promise for the future translation of combinations of small molecule toxin inhibitors into generic prehospital therapies for treating hemotoxic snakebites.

Venom SVMP activities are neutralized by peptidomimetic inhibitors and metal chelators
SVMPs represent a major class of enzymatic toxins responsible for causing severe snakebite pathology, including hemorrhage, coagulopathy and tissue necrosis [17][18][19] . Two classes of SVMP-inhibitors have been historically investigated in the field of snakebite: metal chelators and peptidomimetic hydroxamate inhibitors 13 . These different molecules have distinct modes of action; chelators reduce the available pool of Zn 2+ required for SVMP bioactivity, while peptidomimetic hydroxamate inhibitors directly bind the Zn 2+ ion present in the catalytic core of the metalloproteinase 31 . Here, we compared the inhibitory capabilities of the peptidomimetic inhibitors marimastat and batimastat (both Phase 2-approved) and the chelators DMPS and dimercaprol (both licensed drugs) (Fig. S1) against a variety of venoms representing highly medically important viper species from distinct geographical regions 15,[32][33][34] and with variable toxin compositions (Fig. 1); namely the West African and south Asian saw-scaled vipers (Echis ocellatus and Echis carinatus), the Central American fer-de-lance or terciopelo (Bothrops asper), the African puff adder (Bitis arietans) and the south Asian Russell's viper (Daboia russelii).
We used an in vitro kinetic fluorogenic assay 28 to assess the SVMP bioactivity of each venom and its inhibition by varying concentrations of the four SVMP-inhibitors. All venoms exhibited considerable SVMP activity when compared to the PBS control ( Fig. 2A), except for D. russelii, whose venom SVMP abundance was the lowest (6.9% of all venom proteins;  Fig. 2C), but could not be tested at the two higher concentrations due to the low water solubility of this drug and the interference of DMSO (>1%) in our assay. Conversely, dimercaprol was generally effective down to 1.5 µM (IC 50 s = 0.02-0.4 µ M), and 15 µM of DMPS was required to fully inhibit SVMP activities (IC 50 s = 0. 29-4.19 µ M) (Fig. 2B). We therefore concluded that both peptidomimetic inhibitors are equally effective (with IC 50 s <150 nM), and that both supersede the preclinically validated metal chelators 28 in neutralizing the in vitro SVMP activities of the African, Asian and American snake venoms tested here.

Procoagulant venom activities are antagonized by peptidomimetic inhibitors and metal chelators
Since SVMPs are key toxins associated with causing coagulopathy, we next investigated whether the same peptidomimetic and metal chelating inhibitors could also neutralize the procoagulant bioactivities of viper venoms. To do so, we used a validated kinetic absorbancebased assay monitoring plasma clotting 35 in the presence or absence of venoms and inhibitors The results of the SVMP and coagulation assays demonstrated that the peptidomimetic inhibitors outperformed the metal chelators and, although marimastat and batimastat are similar drugs in terms of both mechanism of action and in vitro efficacy, marimastat has a number of potential clinical advantages over batimastat, including: (i) increased solubility, (ii) excellent oral bioavailability vs. parenteral administration, and (iii) generally well tolerated vs. some reports of acute bowel toxicity 38 . Therefore, we selected marimastat as our candidate SVMP-inhibitor for use in vivo venom-neutralization experiments.

Combinations of inhibitors inhibit distinct pro-and anti-coagulant venom toxins
The coagulation assay findings described above for D. russelii provided a strong rationale for exploring combinations of small molecule toxin inhibitors as snakebite therapeutics. While the SVMP-inhibitors potently inhibited the dominant procoagulant activities of this venom, inhibition revealed a secondary, uninhibited, anticoagulant activity (Fig. 3E). To better understand these effects, we applied a validated nanofractionation approach 35,37 to D. russelii venom and reassessed the inhibition of pro-and anti-coagulant bioactivities of the resulting venom fractions (Fig. S1A). Consistent with our findings using whole venom, the resulting nanofractionated bioactivity profiles displayed procoagulant peaks that were effectively inhibited in a dose-dependent manner by marimastat, while fractions with anticoagulant activity were not neutralized by this inhibitor at any of the tested concentrations (Fig. S2A).
Prior research suggests that the anticoagulant activity of D. russelii venom is mediated by PLA 2 toxins 37 . Indeed, of the toxins found in D. russelii venom, 35% are PLA 2 s, while only 16% are SVSPs and 6.9% SVMPs 5 (Fig. 1). Consequently, we tested the well-established Since marimastat effectively neutralizes the SVMP-driven procoagulant activity of D. russelii venom, while varespladib inhibits the anticoagulant PLA 2 toxins, we next tested whether a combination of these two drugs could restore normal clotting caused by the whole venom. At the two highest doses tested (15 µ M and 150 µ M), the combination of these two inhibitors restored clotting profiles to levels similar to those observed in the control (Fig. S2B), demonstrating that a rationally designed small molecule toxin inhibitor mix is capable of simultaneously inhibiting both procoagulant and anticoagulant venom toxins.

Venom SVSP activities are abrogated by the serine protease-inhibitor nafamostat
While inhibitors against SVMP and PLA 2 toxins have been actively researched, to our knowledge no serine protease-inhibitors have been investigated as drugs against snakebite.
We selected nafamostat (Fig. S1), a serine protease-inhibitor licensed as an anticoagulant medicine in Japan 39 , as a candidate SVSP-inhibitor and tested its in vitro efficacy using a chromogenic assay. Among the tested venoms, all except D. russelii displayed detectable SVSP activity (Fig. 4A). These activities were broadly neutralized in a dose-dependent manner by nafamostat (Fig. 4B), with the highest doses (150 and 15 µM) completely inhibiting SVSP activity, irrespective of venom (IC 50s = 0.12-1.07 µ M). Although SVSP toxins can also perturb coagulation, we were unable to test the efficacy of nafamostat in the plasma assay described above due to nafamostat's inherent anticoagulant potency (Fig. S3), which is mediated via interactions with cognate serine proteases found in the blood clotting cascade, such as thrombin and factors Xa and XIIa 40 . Because of these off-target interactions, generic SVSP-inhibitors must be carefully evaluated prior to any inclusion in a human snakebite therapy, especially since SVSPs are often less abundant in venom than SVMP or PLA 2 toxins 5 ( Fig. 1). Nevertheless, the in vitro efficacy of nafamostat demonstrated here justified its evaluation in in vivo models of envenoming to select the most efficacious mixture of inhibitors.

Preclinical efficacy of small molecule toxin inhibitors as solo and combination therapies
We used an established in vivo model of envenoming 41,42 to test the efficacy of small molecule toxin inhibitors. This model consists of the preincubation of the test therapy with venom, followed by intravenous injection of the mixture into groups of five male CD-1 mice (18-20 g) via the tail vein, and is based on the gold standard method of preclinical efficacy recommended by the World Health Organization 43 . We first tested the ability of marimastat, varespladib and nafamostat as solo therapies to prevent venom-induced lethality in mice challenged with a 2.5 x median lethal dose (LD 50 ) of E. ocellatus venom (45 µg) 25 . We selected this snake venom and venom dose as our initial model based upon its medical importance and results from our recent work exploring the preclinical venom-neutralizing efficacy of metal chelators 28  ng/ml) and comparable with the two combination therapies, but those detected in the less efficacious varespladib-and nafamostat-only treatment groups displayed substantially higher TAT levels (472.7-649.9 ng/ml and 543.9-859.1 ng/ml, respectively), although these remain lower than those of the venom-only controls. In combination, these findings suggest that marimastat is likely responsible for much of the observed efficacy against the lethal effects of E. ocellatus venom, but that small molecule combinations with additional toxin inhibitors provide superior preclinical efficacy than treatment with marimastat alone.

Inhibitor mixes protect against lethality caused by a diverse range of viper venoms
We next investigated whether the two inhibitor combination therapies were equally effective against the other viper venoms tested in vitro, as these venoms exhibit highly variable toxin compositions in comparison with the SVMP-rich toxin profile of E. ocellatus (Fig. 1). We adopted the same approach as described above, and intravenously challenged groups of experimental animals with 2.5 × LD 50 doses of E. carinatus (47.5 µg) 41  (iii) the inhibition of SVMP and PLA 2 toxins appears sufficient to protect against lethality caused by a diverse array of viper venoms, we decided to proceed with the marimastat and varespladib combination as our lead candidate for testing in more therapeutically challenging preclinical models of envenoming.

Administration of the marimastat and varespladib (MV) dual therapy after venom challenge broadly protects against venom lethality
To better mimic a real-life envenoming scenario, we next tested the marimastat and varespladib inhibitor mixture in a preclinical 'challenge then treat' model of envenoming, where the venom is first administered intraperitoneally and then the test therapy is administered intraperitoneally separately after the venom challenge 28 . To this end, we injected venom from each of the five viper species in doses equivalent to at least 5 x the intravenous (iv) LD 50  'envenomed' mice. This observation was noted for both the 'intravenous preincubation' and 'intraperitoneal challenge then treat' models of envenoming ( Fig. S4A and B). These elevated levels were reduced to control levels in experimental animals treated with the inhibitor combination in both experimental approaches (Fig. S4C). These findings suggest that, in addition to protecting against the lethal effects of the various viper venoms, the marimastat and varespladib therapeutic combination is capable of preventing coagulopathy, and in the case of B. asper, inhibiting toxins acting to disrupt the endothelium.

Discussion
Snakebite is the world's most lethal NTD, resulting in ~138,000 deaths annually and primarily affecting the world's resource-poor populations of the tropics and subtropics 1 .  23 , we found both drugs to be equipotent in vitro. We selected marimastat as our candidate for in vivo efficacy experiments due to a number of desirable characteristics that make it amenable for a future field intervention for snakebite, specifically its oral vs.
intraperitoneal route of administration, and its increased solubility and tolerability compared to batimastat 38 . Indeed, these characteristics seemingly contributed to the demise of batimastat during development, although both drugs were ultimately discontinued following lack of efficacy in Phase 3 clinical trials 38  Even when considering the differences in route of administration (intravenous/intraperitoneal vs. oral), our extrapolated dose (33.6 mg per 70 kg adult) is very low compared to that welltolerated in Phase 1 trials (800 mg) and this, combined with the relatively high oral bioavailability of marimastat (70%), offers substantial scope for the development of this drug as a prehospital therapeutic for use soon after a snakebite.
The second drug in our mixture, varespladib, is a secretory PLA 2 -inhibitor previously investigated for use in the treatment of various acute coronary syndromes 55 . Both varespladib and varespladib methyl (its oral prodrug, which is rapidly converted in vivo to varespladib) have been used clinically in Phase 1 and 2 trials 55-57 , although a lack of efficacy at Phase 3 ultimately resulted in its discontinuation 58 . More recently, varespladib has been explored for re-purposing as a potential therapeutic for the treatment of snakebite. Both varespladib and its oral prodrug have been shown to exhibit promising neutralizing capabilities against a variety of different snake venoms 22,29 , but have proven to be particularly effective at mitigating the life-threatening effects of neurotoxicity caused by certain elapid venoms in animal models of envenoming 26,27,59 . Similar to marimastat, varespladib shows good oral bioavailability, and has a half-life equating to 5 hr when delivered by iv infusion 55 4). These findings, alongside evidence that nafamostat provides no protection against the lethal effects of E. ocellatus venom when used as a solo therapy (Fig. 5A), strongly suggest that nafamostat does not appear to substantially contribute to the observed preclinical efficacy (Fig. 5). Despite being a licensed anticoagulant drug in Japan since the early 1990s 39 , nafamostat has potential detrimental off-target effects for use in snake envenoming via interaction with cognate coagulation cascade serine proteases 40 , has a short half-life (~8 min) 60 , and requires intravenous administration, thereby limiting its utility and applicability as a potential prehospital snakebite therapeutic. For those various reasons, our lead candidate therapeutic mixture remained restricted to the marimastat and varespladib combination.
The administration of the marimastat and varespladib combination 15 mins after 'envenoming' resulted in the survival of experimental animals for at least 17 h after mortality was observed in the venom-only control groups (Fig. 6). In our previous work, we demonstrated that the licensed metal chelator DMPS, which shows much promise as an early intervention therapeutic against snakes with SVMP-rich venoms (e.g. the West African sawscaled viper, E. ocellatus) 28 , prevented lethality for ~8 h in the same preclinical model, but required a later dose of antivenom (1 hr after venom delivery) to extend protection to a comparable duration to that observed here with the marimastat and varespladib combination ( Fig. S5). While DMPS undoubtedly represents a highly promising future therapeutic for snakebite, not least because of its oral formulation, licensed drug status and decades of therapeutic use for other indications 61,62 , it seems unlikely to be highly efficacious as a solo therapy against a wide variety of snakes due to only targeting SVMP toxins 28   Prior to use, venoms were resuspended to 10 mg/ml in PBS (pH 7.4) and then further diluted to 1 mg/ml stock solutions (with PBS) for the described experiments.
Working stocks (tenfold dilutions from 2 mM to 2 µM) were made using deionized water, with the exception of varespladib and batimastat, for which we used DMSO due to water insolubility.

Enzymatic assays
The SVMP assay measuring metalloproteinase activity and the plasma assay measuring coagulation in the presence or absence of venoms and inhibitors were performed as previously described 28  slope values across the plate, and the processed coagulation chromatograms were plotted to visualize very fast coagulation, medium increased coagulation and anticoagulation, as previously described 37 .

In vivo experimentation
All animal experiments were conducted using protocols approved by the Animal Welfare and    Note the different y-axis scales.