In the early days of the pandemic, there were no approved antiviral compounds against SARS-CoV-2 (World Health Organization, 2020). This changed in October 2020, when remdesivir (brand name: Veklury; Gilead Sciences) was granted emergency use authorization (EUA) by the US Food and Drug Administration (FDA) for treatment of hospitalized patients (World Health Organization 2020). Remdesivir was the only approved medicine until the EUA of molnupiravir (Merck and Ridgeback) and paxlovid (Pfizer) in December 2021 (U.S. Food and Drug Administration, 2021).

The approved drugs have different mechanisms of action; remdesivir, a nucleotide analogue, acts by stalling SARS-CoV-2 RNA-dependent RNA polymerase (RdRp) (Kokic et al., 2021). Remdesivir exhibited conflicting impact in studies, showing improvement in time to recovery in the initial study cited during authorization (Beigel et al., 2020), but later studies showed either no statistically significant effect (Wang et al., 2020), or a statistically significant but clinically minor effect (Spinner et al., 2020). Concerns over renal toxicity (Gérard et al., 2021; Wu et al., 2022), as well as a cardiac safety signal (Rafaniello et al., 2021) challenge the safety of the drug. The second drug under EUA, molnupiravir, was approved based on a study showing a reduction in hospitalization and death (Jayk Bernal et al., 2022). Molnupiravir, in addition to remdesivir, targets RNA-dependent RNA polymerase and increases the frequency of mutations during SARS-CoV-2 replication (Kabinger et al., 2021). Concerningly, it has also been shown to induce mutations in mammalian cells (Zhou et al., 2021). The mechanism of action of molnupiravir is concerning as it has a possibility of driving new variants (Kabinger et al., 2021; Hashemian et al., 2022), as a result, its use is cautioned by the World Health Organization (World Health Organization, 2022). The third approved antiviral, paxlovid acts as a 3CL protease inhibitor. 3CL protease is necessary for viral replication (Marzi et al., 2022). Paxlovid displays a reasonable safety profile, although patients often report a “paxlovid rebound” where there is a resurgence of symptoms, often worse than the initial bout (Charness et al., 2022). Moreover, drug-drug interactions have been shown to cause adverse events (Burki, 2022). Drug resistance is also a concern, as mutations have been characterized which drastically reduce the effectiveness of paxlovid (Zhou et al., 2022).

Depending on the drug target, medication is tailored for different stages in infection. Different proteins can be targeted for therapy depending on the stage of infection. Compounds targeting the spike protein will inhibit entry of SARS-CoV-2 into cells, whereas compounds targeting RNA-dependent RNA polymerase will inhibit the replication process, but will not prevent entry into the cell. Therefore, depending on the clinical course, certain compounds can be used at different stages of infection. The helicase, being a replication protein, is active in unwinding the RNA secondary structure so that it can be either replicated by RNA-dependent RNA polymerase or translated by the host ribosome.

Responding to emerging and pandemic viral illnesses requires a multifaceted approach, one strategy is drug repurposing. Drug repurposing is the use of approved drugs for novel targets and diseases. First, finding a useful medication amongst already existing drugs obviates the need to create novel drugs, thus saving time in disease response. Moreover, the side-effects of marketed drugs, having undergone clinical trials and prescribed use, are extensively researched and documented. Lastly, the manufacturing process is already known, and needs only to be scaled. Drug repurposing has previously found success, for example in sildenafil, an angina medication, that was successfully repurposed for erectile dysfunction as Viagra® (Pushpakom et al., 2019).

One example of a successfully repurposed and widely available medication for treatment of COVID-19 is fluvoxamine, a well-tolerated and selective serotonin reuptake inhibitor. Fluvoxamine is commonly used as an antidepressant (Sukhatme et al., 2021). It has been shown to reduce hospitalization in a large-scale randomized control trial (Reis et al., 2022). Being a repurposed drug, fluvoxamine, which was first approved by the FDA in 1994 (trade name: Luvox), has the advantage of decades of safety data surrounding its use. Unlike molnupiravir and paxlovid where a treatment course costs approximately 700 and 500 USD, respectively (Goswami et al., 2022; Morrison Ponce et al., 2022), fluvoxamine is accessible at 4 USD per course (Wang et al., 2021). Remdesivir is also expensive at over 2000 USD per 5-day treatment course (Carta and Conversano, 2021). The price and availability of drugs is an important consideration, especially considering that developing nations have far lower vaccination rates than developed nations (Bollyky et al., 2020). As of 25 July 2022, 73.2% of EU citizens have completed a full course with an EU-approved vaccine ¹ and 55.0% have received at least one booster shot (Ritchie et al., 2020). For comparison, in Africa 42.7% of individuals have been vaccinated and only 2.5% have received at least one booster shot (Ritchie et al., 2020).

Finally, other concerns shape the adoption of a particular pharmacological compound in response to a global pandemic; these include intellectual property concerns, current and future availability, distribution, and (un)known side-effects. Ultimately, an effective treatment of COVID-19 is preferred, that is widely available, inexpensive and without significant toxicity.

Drug repurposing is mostly a phenotypic approach, meaning that protein target and mechanism of action are often unknown. In contrast, target-based approaches seek to first identify protein targets (chemical biology) and to subsequently develop small-molecule inhibitors (medicinal chemistry) for the target. In principle, every SARS-CoV-2 protein can be considered a target, but it is preferable to target essential and/or conserved proteins. A previous review has already reviewed and postulated the main drug targets for COVID-19 (Gil et al., 2020), while this report focuses on the helicase of SARS-CoV-2. The SARS-CoV-2 nsp13 gene encodes a molecular motor, which is a 5′ to 3′-translocating helicase, belonging to superfamily 1B. Helicases act on (deoxy)-ribonucleic acid substrates and are fueled by (deoxy)-nucleotide triphosphates (Figure 1A). The primary functions of helicases are in DNA repair, replication, recombination, and transcription.

Nsp13 is one of the most conserved genes in the SARS-CoV-2 genome, having one of the lowest mutation rates of any of the essential SARS-CoV-2 proteins (Martin et al., 2021; Newman et al., 2021). The SARS-CoV-2 helicase differs from the SARS-CoV-1 helicase by only one amino acid residue, i.e., V570 in SARS-CoV-2 helicase (Figure 1A, highlighted in red) compared to I570 in SARS-CoV-1 helicase, allowing drugs discovered for SARS-CoV-1 to potentially be re-used. Potential binding pockets of Nsp13 were explored via crystallographic fragment screening (Figure 1B), presenting a starting point for structure-based drug discovery (Newman et al., 2021). Moreover, the helicase plays a critical role in replication of the viral genome (Jia et al., 2019). The combination of these two argues for the functional importance of SARS-CoV-2 helicase and makes it an attractive target for the development of antivirals. This is also evidenced by an upcoming CACHE challenge ² that aims to discover new molecules that target SARS-CoV-2 helicase.

The viral helicase is not a new target in drug discovery, for example the helicases of herpes simplex virus and hepatitis C virus have been targeted, as reviewed by Shadrick et al. (Shadrick et al., 2013). More recent reports feature the helicases of polyomaviruses, Zika virus, and MERS-CoV (Bonafoux et al., 2016; Kumar et al., 2020; Zaher et al., 2020; Mehyar et al., 2021b). Additionally, human helicases have also attracted research interest, and inhibitors for DDX and BLM, among others, have been reported (Datta and Brosh, 2018). This approach aims to use small molecule inhibitors to sensitize cancer cells to chemotherapy and DNA-damaging agents and/or to utilize specific tumor backgrounds for hypersensitization of tumors to pharmacological inhibition, a concept which is known as synthetic lethality (Datta and Brosh, 2018).

As previously mentioned, SARS-CoV-2 helicase is among the most conserved proteins in the SARS-CoV-2 genome (Martin et al., 2021; Newman et al., 2021). Throughout the pandemic, it has remained largely stable. Phylogenetic evidence demonstrates increasing negative, i.e., purifying, selection over time, making it a stable target (Figure 2A). The development of drug resistance is an issue that undermines many treatments, most notably anti-biotics. Under the selection pressure of a drug treatment, the target protein can mutate such that the compound no longer binds (Richman, 1994; Menéndez-Arias and Richman, 2014). It was evaluated whether the mutations observed through genomic surveillance of COVID-19 cases (Kumari et al., 2022) altered the initial protein sequence (Newman et al., 2021). For a drug to retain effectiveness over time, the major mutations would not alter binding affinity of the drug compounds, thus maintaining drug effectiveness against mutations. Possibly, conservation of structure may enable production of pan-beta coronaviral inhibitors to guard against future zoonotic coronaviral outbreaks (Li et al., 2021; Munshi et al., 2022). This possibility is supported by the low level of nsp13 genetic variation within beta-coronaviruses, as demonstrated by the phylogenetic tree shown in Figure 2B.

The technique used here to identify mutations is exploratory, in that the predicted energetic shift was used as a proxy for conformational change. It has been assessed whether there are any changes likely to significantly impact the structural conformation of SARS-CoV-2 helicase. If a mutation was near a binding site and significantly shifted the energetic stability of the protein, it is likely that the mutation alters compound binding. Selection was determined using the toolkit made from the GISAID database ³ , for all SARS-CoV-2 genomes up to 2 January 2022. In Figure 2C, site selection in terms of fixed-effects likelihood (FEL) (Kosakovsky Pond and Frost, 2005) is displayed (blue and red stem plots), FEL is a measure of selection pressure in phylogenetic trees and is calculated by comparing the expected number of non-synonymous mutations with the actual observed rate. In short, observing a higher than expected frequency of non-synonymous mutations suggests positive selection, i.e., evolutionary pressure for the protein to change. Observing fewer than expected non-synonymous mutations is evidence of negative purifying selection, whereby mutants are not likely to survive and reproduce.

Additionally, a color plot depicts the average change in energetic stability of the protein resulting from the set of possible mutations at that site (Kwasigroch et al., 2002). Most mutations result in a slight destabilization of the helicase protein, suggesting a high level of structural optimization. While this makes it less likely that the protein will develop a drug resistant mutation, it is not certain. Residues where mutations have a destabilizing effect are more likely to alter the helicase structure, which affects the binding of compounds. The limitation of this approach is the lack of experimental data to support the generated model. Our assessment shows potentially worrisome loci for future drug resistance, where there is a confluence of positive selection and an energetically destabilizing impact (upward stems in Figure 2C). These sites should be monitored for development of drug resistance and ideally a drug will either act on a different location, or the destabilization is significant enough to render the protein non-functional.

Having established the validity of the helicase as a drug target, multiple methods can be applied for the discovery of inhibitors. In silico screening is experimentally less intense, requiring mostly computational power. This methodology requires the availability of an X-ray or cryo-EM structures. The crystal structure for SARS-CoV-1 helicase was solved in 2019 (Jia et al., 2019), whereas for SARS-CoV-2 helicase structural information was first published in 2021 (Newman et al., 2021). Earlier in silico research made use of homology models based on either SARS-CoV-1 or MERS-CoV helicase to perform molecular modelling studies. Orthogonal to in silico, is in vitro, the screening of compounds directly on the protein of interest. This methodology can be low- (1–100), medium- (100–10.000) or high- (>10.000) throughput, depending on the equipment used and assay deployed. The most common in vitro assay performed for helicases is an ATP-turnover assay, there is, however, a high risk for false positives, e.g., aggregators or DNA-binders, when running these experiments (McGovern et al., 2003; Acker and Auld, 2014). Another common in vitro assay for helicase activity is to measure the unwound fraction by using a DNA construct with a double stranded region formed by an annealed oligonucleotide. If the helicase is active, it will separate the oligonucleotide from the construct, and a lighter band will show up on the gel. Form the intensity of this band, the unwound fraction and subsequent helicase activity can be calculated (Kim and Seo, 2009).

The potential side-effects of any treatment are a concern for medical practitioners when making a choice of which therapy to implement. Certainly, drugs with minimal off-target toxicity are preferred. While toxicity information exists for some compounds in the included tables, many have limited application as treatments and therefore little associated data on side-effects. Toxicity prediction applies machine learning to chemical structures with known toxicity tests on model organisms. Based on chemical similarities, the toxicity of untested compounds can be predicted. Toxicity prediction is a useful tool for evaluating potential harmful side-effects before taking the drug through costly pre-clinical and clinical trials.

It was not possible to use the same assay or toxicity prediction for all compounds. Individual studies often use different assays and thus report different values. Additionally, the toxicity prediction software was not always successful, and therefore several different tools were used: the Quantitative Structure-Activity Relationship (QSAR) toolbox, developed by the Organization for Economic Cooperation and Development (OECD) (Dimitrov et al., 2016); the Toxicity Estimation Software Tool (TEST) software developed by the US Environmental Protection Agency (US-EPA) (Martin et al., 2008); and the lazar toxicity prediction web server (Maunz et al., 2013). For some compounds, particularly pharmaceutical drugs, toxicity data was accessible from public documents for their approval by either the FDA or the European Medicines Agency. Many of the natural products included have long histories of use in food as well as herbal medicines (Wang and Yang, 2020, 2021; Yang and Wang, 2021; Wang et al., 2022). Many are found in common foods and show strong association with positive health outcomes (Kumar and Pandey, 2013), including possible antiviral and antineoplastic (Rodriguez-García et al., 2019) properties. Since these products have been consumed for millennia, it is unlikely that they exhibit toxicity, although this may of course be different when the active compound becomes highly concentrated. The retrieved experimental toxicities and/or the predicted toxicity values for every compound are provided in Supplementary Tables S1,S2,S3 for the reader’s consideration. For most assays, acute toxicity values were reported, this certainly has its drawbacks, as compounds may exhibit toxicity at much lower doses. These toxicities should not be overly interpreted, since the effective IC₅₀ doses of compounds differ, it is more beneficial to take a selective ratio against a toxicity endpoint.