Introduction

Viruses, whether known or yet to emerge (‘virus X’), remain a major threat to global health. Vaccines, monoclonal antibodies (mAbs) and antiviral drugs are each important and complementary in controlling viral infections. Vaccination has led to the eradication of smallpox, brought polio close to eradication and massively reduced disease and mortality associated with numerous viral infections, such as measles virus, mumps virus, rubella virus, influenza virus, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), hepatitis B virus (HBV) and human papillomavirus1,2,3. For several important human pathogens, such as HIV, hepatitis C virus (HCV), herpes simplex viruses (HSV), cytomegalovirus (CMV), rhinoviruses and many others, the development of highly effective vaccines has proved difficult or remains elusive, often despite decades of intensive effort. Therefore, it cannot be guaranteed that efficacious vaccines can always be developed when new threatening viruses emerge4,5.

mAbs can offer immediate protection and are transforming the prophylaxis or treatment of certain acute viral diseases, such as those caused by the rabies virus6 and respiratory syncytial virus (RSV)7. They generally have an excellent safety profile and can be highly effective, although they require injections, are virus specific or variant specific and can rapidly select for escape variants8,9. Like vaccines, mAbs can be developed only once a new pathogen has been identified and characterized.

By contrast, small molecule antiviral drugs offer several distinct advantages. They can be administered orally and have the potential to cover a broader spectrum of viruses (Box 1) than vaccines and mAbs. Antiviral drugs have revolutionized the treatment of certain (chronic) viral infections such as those caused by either HIV, HBV, HCV, HSV-1 and HSV-2, and CMV. More recently, antiviral drugs have also become available against some other viruses including influenza viruses, SARS-CoV-2 and hepatitis delta virus. However, for the majority of the more than 200 known human pathogenic viruses, no antiviral drugs are available. Most of these viruses are RNA viruses — the focus of this Review — including many emerging and clinically important pathogens10,11,12.

Alongside vaccine development, sustained investment in antiviral drug discovery is therefore essential. Antiviral drug development has so far largely been focused on viral infections that offer large and predictable commercial markets (for example, those caused by HIV and HCV). It is obviously unrealistic and economically unsustainable to develop specific antiviral drugs for each known viral pathogen. Newly emerging viruses will most likely belong to viral families already known (Table 1) today. Advances in metagenomics and high-throughput sequencing greatly expanded our understanding of viral diversity, and although new viral families may still be discovered, such events are becoming increasingly rare. Together, this strongly supports a strategy in which broader-acting small molecule antiviral (BSA) drugs are developed. Such drugs can be used for the treatment and prophylaxis of infections with known viruses for which no antiviral treatment is available and, in the case of epidemics or pandemics, as a rapid response option while other countermeasures, such as vaccines and mAbs, are being developed12,13.

Table 1 RNA virus families covered in this Review
Full size table

In this Review, we discuss the rationale for the development of BSAs and provide a non-exhaustive overview of approved or experimental antiviral drugs active against viruses with an RNA genome. We also outline strategies for target discovery and accelerated clinical development.

One drug to treat them all?

The term ‘broad-spectrum’ antiviral has been widely used but lacks, to the best of our knowledge, a clear definition. Here, we propose instead to use the term ‘broader-spectrum’ antivirals and distinguish three categories: (i) cross-family BSAs with activity across unrelated viral families within or across a certain genome architecture, that is, negative-sense single-stranded RNA ((−)ssRNA), positive-sense single-stranded RNA ((+)ssRNA) or double-stranded RNA (dsRNA); (ii) pan-family BSAs that act across the different genera of a given viral family (for example, pan-paramyxovirus drugs); and (iii) intra-family BSAs, which are active against multiple serotypes and genotypes of a particular virus (for example, pan-serotype dengue virus (DENV) inhibitors) or active against one or more (but not all) genera of a viral family (Box 1). The need for broader-acting antivirals and their desired spectrum of activity are mostly defined by the particular (clinical) context (Boxes 2 and 3). For viruses that cause a high number of annual cases (for example, RSV), there might be a sufficiently large market, and thus incentives, to develop virus-specific drugs. Conversely, there might be no or less financial incentive to develop virus-specific drugs for viruses that cause a rather limited number of cases but that collectively, within their genus or family, are responsible for a substantial clinical burden (for example, coxsackieviruses and echoviruses, EV-D68 and EV-A71 that all belong to the genus Enterovirus). In this case, the development of antivirals that are broadly active across the entire Enterovirus genus (pan-Enterovirus; category intra-family BSAs) represents a more realistic approach. In the context of epidemic or pandemic preparedness, access to pan-family and cross-family BSAs may be particularly important, as it cannot be predicted which viral family a newly emerging pathogen (‘virus X’) will belong to. Given the extensive genetic diversity and distinct replication strategies of viral families, the development of a single cross-family drug with potent activity against all RNA virus families is likely to be unrealistic. Rather, a more feasible strategy may be to establish a portfolio of different types of BSAs that collectively provide, either alone or in combination, coverage across the RNA virus landscape. BSAs may also be valuable in the context of viral co-infections, in which a single drug could simultaneously target two (or more) viruses14,15.

Beyond the spectrum of activity, BSA development must be guided by clearly defined target product profiles16. To have a meaningful clinical impact, BSAs (or combinations thereof) should ideally be highly efficacious, be orally available (preferably once daily dosing), be safe for a broad patient population (preferably including pregnant women and children), possess a (relatively) high barrier to resistance and exhibit minimal drug–drug interactions16. Virus tissue tropism further influences desirable drug properties. For example, infections involving the central nervous system require efficient blood–brain barrier penetration, whereas limited penetration may be preferable for peripheral viral infections. Consequently, distinct target product profiles may be required, even within a single viral genus. Within Orthoflavivirus, for instance, West Nile virus and Japanese encephalitis virus typically invade the central nerve system, whereas yellow fever virus and DENV do not17.

The importance of preparedness against known viruses (such as EV-D6818, Ebola virus19, monkeypox virus20 and Andes hantavirus21) as well as against emerging viral threats (for example, SARS-CoV-2) has been repeatedly illustrated during outbreaks caused by these diverse viral pathogens. During the coronavirus disease 2019 (COVID-19 pandemic), no antiviral drugs were available at the onset. A stockpile of an oral, pan-family coronavirus BSA before the start of the COVID-19 pandemic would likely have saved countless lives22,23. Remdesivir, originally discovered as an RNA-dependent RNA polymerase (RdRp) inhibitor of RSV24 and Ebola viruses25,26, was the only drug for which antiviral activity against coronaviruses had been systematically demonstrated in preclinical models before the pandemic27.

In the absence of readily available and effective antiviral treatments, drug repurposing may need to be considered. Repurposing can provide a rapid and relatively cost-effective response by leveraging de-risked drugs with existing preclinical and clinical safety data28,29. However, drugs developed or used for non-viral indications are mostly unlikely to possess sufficient antiviral potency to curb outbreaks caused by newly emerging and (potentially highly) pathogenic viruses. At the start of the COVID-19 pandemic, extensive repurposing initiatives studied whether drugs approved for other indications could inhibit SARS-CoV-2 replication. Drugs such as azithromycin (antibiotic), itraconazole (antifungal), hydroxychloroquine (antimalarial, antirheumatic), niclosamide (anthelmintic), ivermectin (anthelmintic) and nitazoxanide (antiparasitic) rapidly entered numerous clinical trials, many of them small and underpowered, and even became, for some time, standard treatment in certain settings despite limited preclinical evidence30. Overall, drug repurposing efforts during the SARS-CoV-2 pandemic were largely unsuccessful, highlighting that such strategies should not be initiated only at the time a new virus emerges. A more realistic approach is to systematically assess existing drugs during epidemic or pandemic ‘peace-time’ against panels of viruses that are representative of high-risk families. Drugs with ‘reasonable’ in vitro activity can be further tested, alone or in combinations, in preclinical models13.

Studies in SARS-CoV-2-infected hamsters revealed that combinations of two or three antivirals, each used at suboptimal or even inactive doses, can result in important antiviral effects31. To rationally develop combination strategies for BSAs, systematic preclinical testing of combinations across viral families is required. Extensive preclinical studies of the most promising combinations may help to guide strategies for stockpiling for epidemic or pandemic preparedness. The use of fixed-dose combinations (such as those used in the treatment of HIV infections) would be ideal. Yet, in outbreak situation with new pathogens, there will be insufficient time to develop such fixed-dose combinations; co-administration of individual drugs would therefore represent a more practical approach. Maximizing broader-spectrum activity while minimizing the risk of cross-resistance requires combining molecules with distinct modes of action and complementary safety and pharmacokinetic profiles.

Direct-acting antivirals versus host-targeting antivirals

Viral replication is a complex, multistep process involving steps such as attachment to and entry into host cells, uncoating, viral protein synthesis (including, for some viral families, polyprotein processing by viral proteases), genome replication, particle assembly and egress. Many of these steps also rely on extensive interactions with processes in the host cell. Despite this complexity, antiviral drug development has historically focused on a limited subset of these steps. For RNA viruses in particular, most approved therapies act on the viral RdRp or viral proteases32. Although RNA viruses are genetically very diverse, many viral families rely on conserved biological processes that can be exploited for antiviral drug development. These include viral enzymes with structurally and functionally conserved catalytic domains, such as RdRps, proteases, endonucleases, guanylyltransferases, methyltransferases, neuraminidases and helicases, as well as host factors and pathways that are essential for replication of multiple virus families. In addition, viruses also encode non-structural or structural proteins that are typically not conserved across families but play central roles within specific genera or families and may therefore represent attractive targets for intra-family or pan-family BSAs. Examples include (i) the orthoflavivirus NS4B33 and the HCV NS5A34, which are essential for the formation of the replication complex, and (ii) the coronavirus membrane protein (M-protein), a key orchestrator of virion assembly35. Given the complexity of viral replication cycles, it is most likely that, beyond the known druggable targets, additional yet-unidentified targets exist. These may include, for example, new functional sites within viral proteins, previously unrecognized viral protein–protein or protein–RNA interactions that are critical for replication, as well as co-opted host factors involved in lipid metabolism36, membrane trafficking37 and nucleotide metabolism38.

Antiviral strategies can be divided into the use of direct-acting antivirals (DAAs) and host-targeting antivirals (HTAs). DAAs inhibit viral proteins or interactions between viral proteins that are essential for replication, including enzymes as well as regulatory or structural proteins. HTAs target host factors or pathways that viruses co-opt to support their replication cycle. Each approach offers distinct advantages and limitations with respect to antiviral spectrum of activity, resistance development and safety.

In the following sections, we provide a selected, non-exhaustive overview of druggable antiviral targets and representative molecules (DAAs and HTAs) with broader-spectrum antiviral activity. We first focus on viral targets for which cross-family DAAs already exist (mostly viral RdRp and viral proteases), followed by a discussion on additional viral and host targets that may support BSA development.

The viral RdRp

The viral RdRp catalytic core represents one of the most attractive targets for BSA development because of its essential role in genome replication and transcription, the absence of a cellular equivalent39 and its relative conservation across (certain) virus families40. This catalytic core has a canonical ‘right-hand’ architecture composed of palm, fingers and thumb domains, which contain conserved structural motifs that coordinate multiple steps of RNA synthesis41. RdRps have a wide range of architectural organizations with varying complexity. Some (+)ssRNA viruses, such as members of the Picornaviridae, encode relatively compact, single-subunit RdRps, whereas others rely on more complex multi-domain or multisubunit assemblies42,43,44. For example, coronaviruses use a multi-domain RdRp (nsp12) that functions together with accessory cofactors (nsp7 and nsp8) to support large-genome replication44. Segmented (−)ssRNA viruses such as influenza viruses use heterotrimeric polymerase complexes consisting of the RdRp subunit PB1, the cap-binding subunit PB2 and the endonuclease subunit PA or P3 (ref. 45). By contrast, non-segmented (−)ssRNA virus families (such as Paramyxoviridae, Rhabdoviridae and Filoviridae) encode large multifunctional L proteins that integrate polymerase, capping and methyltransferase activities45. Although these architectural differences result in differences in replication mechanisms, they do not reflect major differences in the catalytic core of the polymerase42,46.

The RdRp can be targeted by nucleoside analogues (NAs) or by non-nucleoside inhibitors. The close structural similarity of NAs to natural nucleosides allows some of these molecules to achieve broader-spectrum antiviral activity. Historically, antiviral NAs have been primarily developed against DNA viruses and retroviruses, including herpesviruses (such as acyclovir and cidofovir), HIV (for example, emtricitabine and tenofovir) and HBV (such as lamivudine and tenofovir). These NAs functionally resemble deoxynucleosides, and their discovery has largely relied on empirical approaches. This also largely applies to antiviral NAs against RNA viruses, which instead resemble ribose nucleosides. The therapeutic activity of NAs depends on their intracellular conversion into the active triphosphate form. The first phosphorylation is often a critical bottleneck, as nucleoside kinase expression and catalytic efficiency vary substantially between cell types and tissues, limiting antiviral efficacy to anatomical sites where activating enzymes are most abundant47,48 (Fig. 1). To bypass this limitation, several NAs are designed as monophosphate prodrugs using strategies such as the ProTide approach, in which lipophilic masking groups shield the phosphate moiety to facilitate cellular uptake49. In the cell, these masking groups are enzymatically cleaved off to release the nucleoside monophosphate. It should be noted that the use of a nucleobase alone can also yield a nucleoside monophosphate via intracellular phosphoribosylation, a pathway used by a drug such as favipiravir (Fig. 1). Once in the phosphorylated forms, NAs are incorporated by the viral polymerase in the nascent viral RNA strand and inhibit further elongation either through RNA chain termination or by acting as viral RNA mutagens50. Distinct mechanisms of chain termination are dictated by both the chemical architecture of the NA and the site at which arrest occurs. Obligate terminators lack a 3′-hydroxyl group, resulting in immediate termination of RNA synthesis upon incorporation. By contrast, non-obligate terminators possess a 3′-hydroxyl group and induce termination via steric hindrance. Some non-obligate terminators act as immediate terminators (for example, sofosbuvir), with RNA synthesis stalling directly after incorporation (at the i + 1 position) because bulky substitutions — often at the 2′ position — obstruct binding of the next incoming nucleotide or prevent proper closure of the active site. Others act as delayed terminators (for example, remdesivir), allowing incorporation of several additional nucleotides before a structural clash between the inhibitor — typically involving a bulky substitution at the 1′ position — and the RdRp triggers termination, commonly at the i + 3 or i + 5 position. NAs that act as mutagens, such as molnupiravir, introduce base-pairing ambiguity that results in C-to-U and G-to-A transition mutations, pushing the virus beyond its error threshold, leading to nonviable progeny, also known as lethal mutagenesis47,48,51. Although some viruses, such as coronaviruses, possess 3′-to-5′ exonuclease proofreading activity capable of excising mis-incorporated NAs, they remain susceptible to the antiviral activity of mutagenic nucleoside analogues such as molnupiravir51,52 and favipiravir53.

Fig. 1: Nucleoside-analogue activation and their antiviral mechanisms of action.
Fig. 1: Nucleoside-analogue activation and their antiviral mechanisms of action.
Full size image

Examples of the activation of nucleoside analogues (or nucleobase-like favipiravir) and their prodrugs to their 5′-triphosphate metabolite, together with an indication of the different mechanisms by which the resulting nucleotides block viral replication. CatA, cathepsin A; CES1, carboxylesterase 1; HINT1, histidine triad nucleotide-binding protein 1; HGPRT, hypoxanthine–guanine phosphoribosyltransferase.

Selected examples of nucleoside and non-nucleoside RdRp inhibitors are discussed below, beginning with clinically approved drugs, followed by investigational compounds. These examples illustrate that some RdRp inhibitors exhibit cross-family antiviral activity against both (+)ssRNA and (−)ssRNA viruses, whereas others display a more restricted activity spectrum, limited to selected genome types or virus families (Fig. 2).

Fig. 2: Chemical structures of example nucleoside analogues and spectrum of antiviral activity against RNA viruses.
Fig. 2: Chemical structures of example nucleoside analogues and spectrum of antiviral activity against RNA viruses.
Full size image

The figure summarizes the chemical structures and mechanisms of action and antiviral activity of selected nucleoside analogues. Next to the structures, their likely mechanism is indicated as follows: A, immediate chain termination; B, delayed chain termination; C, induction of mutagenicity; and A*, under investigation. Green dots indicate at least some reported level of in vitro antiviral activity against one or more viruses belonging to the particular family. This qualitative information does not account for differences in antiviral potency; for example, PSI-6206 is markedly more potent against hepatitis C virus (HCV) (Flaviviridae) than against hepatitis E virus (HEV) (Hepeviridae) in vitro. Quantitative data (such as half-maximal effective concentration (EC50) values) are not provided as these cannot be directly compared between published studies. Note that several of these compounds are used clinically in the form of prodrugs (see Fig. 1); however, for simplicity and structural comparison, the nucleoside forms are depicted here. (+)ssRNA, positive-sense single-stranded RNA; (−)ssRNA, negative-sense single-stranded RNA; dsRNA, double-stranded RNA; RdRp, RNA-dependent RNA polymerase.

Sofosbuvir

Sofosbuvir is used for the treatment of infections with HCV. It is a prodrug of the 5′-monophosphate of PSI-6206 (2′-deoxy-2′-fluoro-2′-C-methyluridine) and acts as a chain terminator54,55. Beyond HCV (family Flaviviridae), sofosbuvir has been reported to inhibit the replication of some other (+)ssRNA viruses — albeit with less potency — including the hepatitis E virus (HEV; family Hepeviridae)56, chikungunya virus (family Togaviridae)57 and Zika virus (family Flaviviridae)58. The susceptibility of viruses to a given nucleoside analogue may be influenced by the efficiency of phosphorylation of the nucleoside analogue in different cell lines and tissues. For sofosbuvir, the in vitro antiviral effects against both HCV and HEV have been demonstrated in similar cell types, yet the activity against HCV is more pronounced than against HEV, suggesting a more efficient interaction with the active site of the HCV RdRp than with that of HEV RdRp. Sofosbuvir can be considered a BSA with cross-family activity against certain (+)ssRNA virus families.

Remdesivir

Remdesivir is approved for the treatment of SARS-CoV-2 infection59. It is a (non-oral) prodrug of the 5′-monophosphate of GS-441524 (1′-cyano-4-aza-7,9-dideazaadenosine) that acts as a chain terminator. The drug has antiviral activity against viruses belonging to different viral families (either with (−)ssRNA or (+)ssRNA genomes). Remdesivir inhibits several (−)ssRNA viruses, such as RSV (family Pneumoviridae)24, Ebola viruses (family Filoviridae)26,60, human parainfluenza 3 (HPIV3) and Nipah virus (family Paramyxoviridae)61,62, and Lassa virus and Junín virus (family Arenaviridae)26, as well as several (+)ssRNA viruses, such as other members of the family Coronaviridae besides SARS-CoV-2 (refs. 27,63), DENV, West Nile virus, yellow fever virus and Zika virus (family Flaviviridae)64, and enteroviruses (family Picornaviridae)64. Remdesivir has little or no antiviral activity against influenza A and B viruses (family Orthomyxoviridae; (−)ssRNA) and HEV (family Hepeviridae; (+)ssRNA)64,65. Yet, its relatively broad antiviral spectrum against RNA viruses makes remdesivir (and other prodrug forms such as the oral form obeldesivir) a key example of a cross-family BSA.

Favipiravir

Favipiravir (6-fluoro-3-hydroxy-2-pyrazinecarboxamide, T-705) acts as a purine nucleoside analogue and viral RNA mutagen following intracellular activation to its ribofuranosyl-5′-triphosphate form66. It is approved in Japan against new or re-emerging influenza viruses67,68, in particular those that are resistant to other treatments. Besides its activity against influenza viruses, it exhibits activity, although mostly relatively weak, against several (+)ssRNA virus families (Coronaviridae, Flaviviridae, Togaviridae and Caliciviridae) and (−)ssRNA virus families (Orthomyxoviridae, Paramyxoviridae, Filoviridae, Phenuiviridae, Nairoviridae and Arenaviridae)69. Although favipiravir seems generally safe and well tolerated70, inconclusive efficacy data for Ebola, Lassa fever and COVID-19 have so far limited its broader clinical use71,72. Thus, despite cross-family antiviral activity in vitro and in a number of animal infection models, its clinical benefit — at least in monotherapy — remains to be established.

Molnupiravir

Molnupiravir is an oral prodrug of N4-hydroxycytidine and, upon phosphorylation to its 5′-triphosphate, induces lethal viral mutagenesis51,52 but is considered not mutagenic to host cells73. It received Emergency Use Authorization in the USA for certain groups of high-risk patients with COVID-19. The drug was initially identified as an inhibitor of influenza viruses (family Orthomyxoviridae)74,75 and Venezuelan equine encephalitis virus (VEEV, family Togaviridae)76. Besides its activity against VEEV and influenza viruses, it inhibits several (+)ssRNA viruses such as SARS-CoV-2 (family Coronaviridae), orthoflaviviruses such as Zika virus (family Flaviviridae)77,78 and norovirus (family Caliciviridae)79 and several (−)ssRNA viruses such as RSV (family Pneumoviridae)75,78,80, Ebola virus (EBOV; family Filoviridae)81 and Crimean–Congo haemorrhagic fever virus (CCHFV; family Nairoviridae)82. In addition, molnupiravir is also active against rotaviruses79 (family Reoviridae), which have a fragmented double-stranded (ds)RNA genome. Concerns exist that its mutagenic mechanism may accelerate the emergence of new variants, as was demonstrated in vitro for SARS-CoV-2 (ref. 83). Molnupiravir can be considered another example of a cross-family BSA. The extent to which the preclinical findings translate into meaningful clinical outcomes for viruses belonging to other families remains uncertain.

Galidesivir

Galidesivir (BCX-4430) is an investigational drug initially developed for the treatment of HCV (Flaviviridae) that acts as a chain terminator84. In vitro activity has been demonstrated against more than 20 RNA viruses across nine families, including viruses with either an (+)ssRNA genome (Coronaviridae, Picornaviridae, Togaviridae and Flaviviridae) or an (−)ssRNA genome (Filoviridae, Phenuiviridae, Arenaviridae, Paramyxoviridae, Pneumoviridae and Orthomyxoviridae)85, although with relatively low in vitro potency. The antiviral potency against several viruses (for example, those in the families Arenaviridae and Phenuiviridae, with reported half-maximal effective concentration (EC50) values approximately 40 µM) is likely too low for potential clinical use; however, it can be considered an example of a cross-family BSA. Galidesivir is currently in clinical development for the treatment of Marburg virus infection; following completion of a phase I study, it is now proceeding with the FDA through the Animal Rule pathway (see section ‘Pathways to (temporary) licensure’).

Bemnifosbuvir

Bemnifosbuvir (AT-527) is an oral dual prodrug of the 5′-monophosphate of AT-273 (2′-fluoro-2′-C-methyl-guanosine) that acts as a chain terminator86. Bemnifosbuvir is in phase III clinical development (in combination with ruzasvir) for the treatment of HCV infections87. The drug is also active against Zika virus (Flaviviridae; (+)ssRNA)88,89, against hepatitis E virus90 and against several members of the Coronaviridae family ((+)ssRNA)91. It lacks antiviral activity against (−)ssRNA viruses such as influenza (family Orthomyxoviridae) and RSV (family Paramyxoviridae)89. The drug has some cross-family BSA activity.

4′-Fluorouridine

The investigational nucleoside analogue 4′-fluorouridine (4′-FU) acts as a viral RNA chain terminator92. It was originally developed as an RSV inhibitor (family Pneumoviridae)92 but also exhibits antiviral activity against other (−)ssRNA viruses such as measles virus, HPIV3, Cedar virus and Nipah virus (family Paramyxoviridae)92,93,94,95, influenza (family Orthomyxoviridae)96, rabies virus (family Rhabdoviridae)92, Ebola viruses (family Filoviridae)93 and in particular viruses belonging to the family of the Hantaviridae and Arenaviridae, including in infected nonhuman primates94. Antiviral activity has also been shown against (+)ssRNA viruses, such as SARS-CoV-2 (family Coronaviridae)92, chikungunya virus and VEEV (family Togaviridae)97 and DENV2 (family Flaviviridae)98. 4′FU is more potent than, for example, galidesivir or favipiravir and can be considered a remarkably broad-spectrum cross-family BSA.

JNJ-9117

JNJ-9117 is an investigational oral prodrug of the 5′-monophosphate of 2′-deoxy-2′,4′-difluoro-2′-C-methylguanosine (a 2′-C-methyl nucleoside analogue, like the parent compound of valopicitabine (see below)) that acts as a chain terminator. It was initially developed as an inhibitor of HCV (family Flaviviridae; (+)ssRNA)99. Development was halted, after phase I studies, for strategic reasons. JNJ-9117 also exhibits pan-genotype activity against hepatitis E virus (family Hepeviridae; (+)ssRNA), with a high barrier to resistance100. The molecule also has in vitro antiviral activity against several other members of the family Flaviviridae and alphaviruses (family Togaviridae; our unpublished data). It serves as an example of a cross-family BSA.

Lumicitabine

Lumicitabine (ALS-8176), an investigational prodrug of 4′-chloromethyl-2′-deoxy-2′-fluorocytidine, was originally developed for the treatment of RSV infection101. It acts as a chain terminator102 and has shown efficacy in preclinical models and early-phase clinical trials103. It is also active against other members of the family Pneumoviridae (including human metapneumovirus (HMPV)) and several members of the family Paramyxoviridae (including Nipah virus and measles virus)104. Clinical development was discontinued because of reversible neutropenia.

Valopicitabine

Valopicitabine (NM283) is an investigational prodrug of 2′-C-methylcytidine (2CMC) that acts by chain termination. It was originally developed for the treatment of HCV infections (family Flaviviridae)105. 2CMC and a number of analogues (for example, 7-deaza-2′-C-methyladenosine) are also active against other (+)ssRNA viruses (such as orthoflaviviruses (family Flaviviridae)106,107 and noroviruses and sapoviruses (family Caliciviridae)108) but generally lack activity against (−)ssRNA viruses. Valopicitabine is also active against rotaviruses (family Reoviridae)109. Clinical development was discontinued because of safety reasons. 2′-C-methyl nucleoside analogues can be considered cross-family BSAs, but their activity is limited to certain viral families with a (+)ssRNA genome.

1′-Cyanocytidine

The nucleoside analogue 1′-cyanocytidine (CNC) is an inhibitor of SARS-CoV-2, with activity in animal infection models, but does not inhibit the human coronavirus OC43 (HCoV-OC43). Mechanistically, it has been proposed to act as a competitive nucleoside inhibitor and is therefore distinct from chain terminators or mutagenic analogues110. CNC has also no relevant activity against other (+)ssRNA viruses (Flaviviridae and Togaviridae) or (−)ssRNA viruses (Orthomyxoviridae and Paramyxoviridae)110. CNC is thus (likely) not to be considered a BSA.

Towards broader-spectrum RdRp inhibitors

As shown in Fig. 2, it is obvious that most, but not all, NAs with activity against RNA viruses exhibit some level of cross-family activity, although their antiviral spectra and potencies vary substantially. A systematic side-by-side mapping of the antiviral activity of these molecules against panels of representative viruses from the various families in standardized antiviral assays and across multiple cell lines and culture systems may be needed to obtain a full picture of their antiviral spectrum. This set of data, together with information on intracellular activation and phosphorylation, kinetic data on the inhibition of purified enzymes by the corresponding nucleoside 5′-triphosphates, structural data on the active site and the 5′-nucleotide (ideally in complex with the RNA template) and advanced computational and artificial intelligence (AI)-based analyses, could ultimately support the rational design of next-generation NAs and the development of BSAs against RNA viruses. The recently developed NA-DB database, which integrates both structural and annotated biological information on reported NAs, may be a useful tool to that end111. Besides targeting the active site of RdRps, intracellular phosphorylation of nucleoside analogues in various cell types and tissues will remain another key challenge. Designing prodrug strategies to overcome this bottleneck and to allow oral dosing will also be essential.

Non-nucleoside inhibitors (NNIs) act by binding to allosteric sites on the RdRp, resulting in conformational changes that disrupt the polymerase activity. They exhibit resistance profiles distinct from those of nucleoside analogues. Their antiviral spectrum is generally narrower, as allosteric binding sites tend to be less conserved across or even within virus families. Yet, broader antiviral activity within a family may (for some genera) be achievable. Indeed, GHP-88309 (and the better-tolerated analogue GHP-88310) is an allosteric non-nucleoside RdRp inhibitor with antiviral activity against HPIV3 (and related respiroviruses) and measles virus (and related morbilliviruses) within the family Paramyxoviridae112,113. However, viruses belonging to the more distantly related genus Orthorubulavirus (HPIV2) are not inhibited112, making GHP-88309 an example of an intra-family BSA. Another example of an NNI is MRK-1, which targets the L polymerase of RSV (genus Orthopneumovirus) and HMPV (genus Metapneumovirus) within the family Pneumoviridae114. MRK-1 is an example of a pan-family BSA. The discovery of entrectinib (RAI-13), a tyrosine kinase inhibitor used for the treatment of certain cancers, as an inhibitor of both human norovirus (family Caliciviridae) and DENV (family Flaviviridae) through targeting a structurally equivalent allosteric site on the viral polymerase suggests that the development of cross-family NNI activity might be possible, although this is likely to be challenging115.

Viral proteases

Several (+)ssRNA virus families (Picornaviridae, Coronaviridae, Flaviviridae and Caliciviridae) translate their genomes as one or more polyproteins that must be proteolytically processed by viral proteases into functional viral proteins (Fig. 3). By contrast, viruses with a (−)ssRNA genome do not encode proteases. Viral proteases are highly attractive antiviral targets and constitute a cornerstone of antiviral therapy against HIV and HCV116,117. The catalytic class of viral proteases differs across families. Viruses of the family Flaviviridae encode a serine protease, whereas viruses of the families Picornaviridae, Coronaviridae and Caliciviridae encode cysteine proteases. Protease inhibitors (PIs) can be categorized into (i) peptidomimetics, which mimic the natural peptide substrate, or (ii) non-peptidomimetics, which are chemically diverse scaffolds that target either the active site or allosteric sites. A key principle in the design of effective peptidomimetic PIs is to adhere strictly to the ‘substrate envelope’, defined as the consensus 3D volume occupied by natural substrates within the active site. This constraint ensures that any resistance mutations that disrupt inhibitor binding would simultaneously impair cleavage of viral polyproteins, imposing a significant fitness penalty on the virus118.

Fig. 3: Antiviral targets in the replication cycle of (+)ssRNA and (−)ssRNA viruses and a non-exhaustive overview of druggable targets per viral family.
Fig. 3: Antiviral targets in the replication cycle of (+)ssRNA and (−)ssRNA viruses and a non-exhaustive overview of druggable targets per viral family.
Full size image

a, A generalized and simplified representation of viral replication cycles of viruses with a positive-sense single-stranded RNA ((+)ssRNA) genome (enveloped and non-enveloped) and negative-sense single-stranded RNA ((−)ssRNA) viruses, indicating the major stages, namely, viral entry and uncoating, translation and replication, and assembly and egress. b, Examples of known druggable direct-acting antiviral targets in the replication cycle of selected viral families. Grey circles indicate the absence of this particular target for the indicated viral family. Green circles indicate that inhibitors have been reported with (at least some level of) in vitro antiviral activity against the indicated family, validating the target as druggable. Open circles indicate that information on inhibitors of the indicated target is unavailable or was not discussed in this Review. RdRp, RNA-dependent RNA polymerase.

PIs have been widely used for the treatment of HIV infection (family Retroviridae). For RNA viruses, particularly HCV, boceprevir and telaprevir were approved as PIs. They are peptidomimetics with a reactive chemical ‘warhead’ that forms a stable, reversible covalent bond with the catalytic serine residue of the protease. They were eventually succeeded by a second generation of non-covalent peptidomimetic PIs (glecaprevir, grazoprevir and voxilaprevir), which are currently part of the standard of care for HCV119. The advantages and disadvantages of covalent versus non-covalent inhibitors have been discussed elsewhere120,121.

The successful development of highly potent serine PIs against HCV has not yet been achieved for orthoflaviviruses (such as DENV). The lack of antiviral activity of HCV PIs against orthoflaviviruses can be explained by the fact that the conserved catalytic triad (His–Asp–Ser) of both proteases is embedded in very different environments and that their active-site pocket architectures are structurally divergent117,122. In addition, NS3 of orthoflaviviruses forms a complex with NS2B, which influences the active site of the enzyme, whereas HCV NS3 requires NS4A as its cofactor. Substrate conservation across orthoflaviviruses displays a strong preference for dibasic residues (arginine or lysine) at the P1 and P2 positions123. This poses challenges for drug development, as such features are associated with poor drug-like properties and overlap with substrate preferences of several host proteases, including furin, thrombin and trypsin. Alternative approaches are being explored for orthoflaviviruses, including active-site inhibitors that do not closely resemble the natural substrate117,124.

The SARS-CoV-2 main protease (3CLpro/Mpro), a cysteine protease, has proved to be an excellent antiviral target. Two PIs, nirmatrelvir and ensitrelvir, have been approved for the treatment of COVID-19, with the latter approved only in Japan and Singapore. Nirmatrelvir is a reversible covalent, peptidomimetic PI, whereas ensitrelvir is a non-covalent, non-peptidomimetic PI125,126. Efforts to develop a 3CLpro-targeting coronavirus PI were launched at the time of the SARS-CoV outbreak in 2003 but were later discontinued; however, the knowledge gained enabled rapid progress in 2020, leading to the development of nirmatrelvir127. The active-site architecture and recognition networks of the 3CLpro (for example, the P1-Gln hydrogen-bonding network and the hydrophobic S2 subsite) are strongly conserved across α-coronaviruses and β-coronaviruses, enabling pan-coronaviral inhibition. For example, nirmatrelvir is antivirally active (at least as demonstrated in vitro) against the seven human coronaviruses (SARS-CoV-2 (including all variants of concern), severe acute respiratory syndrome coronavirus (SARS-CoV-1), Middle East respiratory syndrome coronavirus (MERS-CoV), human coronaviruses HCoV-229E, HCoV-NL63, HCoV-OC43 and HCoV-HKU1) and even against animal coronaviruses (for example, feline and porcine), whereas the spectrum of activity of ensitrelvir is largely confined to SARS-CoV-2 (including all variants of concern), SARS-CoV-1 and HCoV-OC43 (refs. 125,128). Resistance of SARS-CoV-2 to nirmatrelvir seems to be very rare in clinical settings129, whereas the barrier to resistance to ensitrelvir is lower130. The coronavirus 3CLpro can therefore be considered an excellent target for the development of pan-family BSAs, at least when the latter is a peptidomimetic. Ensitrelvir, administered to household contacts of individuals with COVID-19 within 72 h of symptom onset in the index case, demonstrated efficacy in preventing COVID-19 in the contacts23. New pan-coronaviral 3CLpro PIs with a high barrier to resistance, such as ibuzatrelvir131 and ALG-097558 (ref. 132), are currently in clinical development. In contrast to nirmatrelvir, these drugs do not require co-administration with the pharmacokinetic booster ritonavir, which inhibits the liver enzyme CYP3A4 to slow drug metabolism and maintain therapeutic plasma concentrations133. The structural similarity between the serine proteases of Picornaviridae, Coronaviridae and Caliciviridae may possibly create opportunities for the development of pan-family or cross-family PIs134,135,136,137,138. This was demonstrated for rupintrivir, a viral PI with pan-enterovirus and pan-rhinovirus activity that also exerts some in vitro antiviral activity against noroviruses139.

Note that the genomes of viruses within the family Togaviridae ((+)ssRNA) also encode a viral protease; however, the discovery of inhibitors is still at an early stage140,141. For the family Hepeviridae, another (+)ssRNA virus family, it remains a matter of debate whether a functional protease is encoded within the genome142.

Viral endonucleases

Endonucleases are encoded by several segmented (−)ssRNA viruses, including members of the family Orthomyxoviridae and viruses belonging to the class Bunyaviricetes, such as those in the families Nairoviridae, Peribunyaviridae, Phenuiviridae, Arenaviridae and Hantaviridae143,144 (Fig. 3). Endonucleases cleave the 5′ ends of cellular RNA transcripts (mRNA or small nuclear RNA) to generate capped RNA fragments that serve as primers for viral mRNA synthesis by the RdRp, in a process called ‘cap-snatching’45,145,146. The endonuclease active site contains a conserved acidic pocket that coordinates divalent metal ions, providing a clear opportunity for the development of inhibitors that act by chelating these catalytic metal ions147. Baloxavir marboxil (a prodrug of baloxavir acid) was developed through rational molecular design147. It is so far the only endonuclease inhibitor approved for the treatment of influenza148. Baloxavir binds tightly to the endonuclease active site in the N-terminal domain of the PA subunit of the viral RdRp and chelates the two divalent metal ions145. In addition to influenza A and B viruses (family Orthomyxoviridae), baloxavir also inhibits Bunyamwera virus and La Crosse virus (family Peribunyaviridae)149, albeit with much lower potency than that observed for influenza viruses. The endonuclease activity of viruses in the class Bunyaviricetes is encoded by the N-terminal domain of the L-protein143,150. Beyond baloxavir, other metal-chelating endonuclease inhibitors have been identified, including carbamoyl pyridone carboxylic acids, which exhibit cross-family antiviral activity. These compounds inhibit influenza viruses (family Orthomyxoviridae) as well as several members of the families Arenaviridae (lymphocytic choriomeningitis virus, Junín virus and Lassa virus) and Peribunyaviridae (La Crosse virus) and selected members of the family Phenuiviridae (severe fever with thrombocytopenia syndrome virus but not Rift Valley fever virus (RVFV)) but have no activity against members of the family Hantaviridae (Thottapalayam virus)150,151. The observed overlap in antiviral activity reflects structural homology among the active sites of endonucleases across Orthomyxoviridae and some bunyavirus families150,152 and suggests that the feasibility of developing endonuclease-targeting BSAs with cross-family activity against a spectrum of segmented (−)ssRNA viruses warrants investigation. Pimodivir (VX-787) is another inhibitor of the ‘cap-snatching’ activity of the influenza RdRp. Rather than inhibiting endonuclease activity, pimodivir prevents binding of the 7-methyl GTP cap of cellular mRNA to the cap-binding pocket of the PB2 subunit153. Clinical development has been halted owing to limited efficacy154. Pimodivir is active against influenza A viruses but has little to no activity against influenza B viruses155.

Beyond viral RdRps, proteases and endonucleases, which are known as potential targets for cross-family BSAs, other targets that are conserved within or across viral genera and/or families may also provide a basis for the development of specific classes of BSA. In the following sections, we provide a non-exhaustive overview of such targets.

Viral guanylyltransferases and methyltransferases

Certain viruses with either a (−)ssRNA or (+)ssRNA genome acquire their mRNA cap structures through mechanisms other than cap-snatching and, to this end, encode guanylyltransferases (GTases) and methyltransferases (MTases) (Fig. 3). GTases catalyse the transfer of a GMP molecule to the 5′ end of viral RNA to form GpppN, whereas MTases sequentially methylate the N7 position of this guanine and the 2′-O position of the ribose of the first nucleoside, using S-adenosylmethionine as a methyl donor, to generate the mature m7GpppNm cap structure required for efficient translation, protection from degradation and immune evasion156,157. These enzymatic activities may be encoded by a single multifunctional protein or distributed across multiple proteins, depending on the virus family156,157. For example, Paramyxoviridae, Pneumoviridae, Filoviridae and Rhabdoviridae encode both activities within the large L-protein, whereas in Togaviridae, Flaviviridae and Hepeviridae, these capping functions are embedded within individual non-structural proteins nsP1, NS5 and ORF1, respectively. Coronaviridae encodes capping functions within higher-order replication–transcription complex assemblies. Capping is initiated by the nidovirus RdRp-associated nucleotidyltransferase (NiRAN) domain of nsp12 through an unconventional reaction involving the RNAylation of nsp9, whereas nsp13 facilitates this process by unwinding nascent RNA duplexes158. Interestingly, the diphosphate form of bemnifosbuvir, an RdRp inhibitor, can also bind the NiRAN domain of nsp12 (ref. 86), suggesting a potential dual antiviral mechanism involving both inhibition of RNA synthesis and viral mRNA capping. Cap maturation is subsequently completed by the nsp14 N7-methyltransferase159 and nsp16/nsp10 2′-O-methyltransferase160. Nsp14 has recently emerged as an attractive antiviral target. Small molecule inhibitors, such as TDI-015051 and C10, show antiviral activity in cell-based assays159,161, inhibiting multiple β-coronaviruses, with TDI-015051 additionally active against α-coronavirus, without inhibiting human methyltransferases and showing efficacy in SARS-CoV-2 animal models. These findings support nsp14 inhibition as a promising pan-coronavirus antiviral strategy.

Despite these recent advances, GTases and MTases have not yet been widely explored as antiviral drug targets. A small molecule inhibitor of the chikungunya virus (family Togaviridae) capping enzyme nsP1 was identified through phenotypic antiviral screening. Biochemical studies using the enzyme from the closely related VEEV demonstrated selective inhibition of the nsP1 guanylyltransferase activity, thereby validating the capping machinery as a potential antiviral target for intra-family broad-spectrum activity162. Whether the capping machinery of different families can be exploited as a target for the development of BSAs remains an open question.

Neuraminidases

Neuraminidases of influenza viruses (family Orthomyxoviridae) are classified into 11 subtypes and catalyse the cleavage of terminal sialic acids from the surface of the host cell glycoproteins, a process essential for release of progeny virions upon viral budding (Fig. 2). However, despite this diversity, the catalytic site is highly conserved among influenza A and B viruses163, making it a validated and attractive antiviral target for pan-family inhibitors. Neuraminidase inhibitors, such as oseltamivir, zanamivir and peramivir, are sialic acid mimetics that competitively block the active site164. They are used for both the treatment and prophylaxis of influenza virus infections in high-risk patients. The emergence of resistance mutations nevertheless highlights the need for next-generation inhibitors165. Certain members of the family Paramyxoviridae (genus Respirovirus, such as HPIV1 and HPIV3, and genus Orthorubulavirus, such as HPIV2, HPIV4 and mumps virus) encode a multifunctional haemagglutinin–neuraminidase (HN) glycoprotein that combines receptor-binding (haemagglutinin) and receptor-destroying (neuraminidase) activities within a single protein. The haemagglutinin activity of HN mediates cell attachment and activation of the viral fusion protein, whereas its neuraminidase activity is required for efficient virus release and spread to uninfected cells166. Unlike the neuraminidases of influenza viruses, paramyxovirus HN proteins perform these dual functions through a single sialic acid-binding site, exploiting structural plasticity to alternate between receptor-binding and catalytic states167,168. Similar to influenza virus, several sialic acid mimetics have been designed based on the 3D structure of the HN as inhibitors of HPIV1–HPIV3 and mumps virus169,170. However, several other paramyxoviruses, such as measles (genus Morbillivirus) and Nipah virus (genus Henipavirus), do not encode a neuraminidase. Consequently, in contrast to Orthomyxoviridae, neuraminidase-targeting approaches will not allow the development of pan-family inhibitors for Paramyxoviridae. In addition, this fundamental biological divergence makes the development of neuraminidase inhibitors active against both orthomyxoviruses and paramyxoviruses challenging, if feasible all possible.

Viral helicases

Helicases use the free energy derived from NTP hydrolysis to unwind double-stranded RNA or RNA secondary structures during viral genome replication171. Helicases are essential for the replication of several RNA virus families, including (i) Togaviridae (encoded by nsP2)172 and Coronaviridae (encoded by nsp13)173, which have superfamily 1 (SF1) helicases; (ii) Flaviviridae (encoded by NS3)174, which have SF2 helicases; (iii) Picornaviridae (encoded by 2C)175; and (iv) Caliciviridae (encoded by NS3)176, which have SF3 helicases. Interestingly, some virus families encode proteins associated with helicase activity without having any canonical conserved helicase motif, such as RSV (family Pneumoviridae; P protein)177 and EBOV (family Filoviridae; VP35)178 (Fig. 2). Despite their essential role, helicases remain relatively underexplored as antiviral targets for RNA viruses. By contrast, the helicase–primase complex has already been clinically validated as an antiviral target for herpesviruses179. Examples of small molecule antivirals that target helicases of RNA viruses have also been reported. For instance, structure-based drug design has led to the development of enterovirus 2C inhibitors with broader-spectrum antiviral activity against multiple enteroviruses, including EV-D68, EV-A71 and CVB3 (family Picornaviridae), and demonstrated efficacy in animal infection models180. Despite these promising findings and the presence of conserved structural elements, including ATP-binding and nucleic acid-binding sites, clinical validation of helicase inhibitors as BSAs remains limited, largely owing to conformational flexibility and structural similarities with host helicases that complicate selective targeting181.

Selected other antiviral targets

The orthoflavivirus non-structural protein 4B

The orthoflavivirus non-structural protein 4B (NS4B) represents an example of a viral protein without enzymatic activity that serves as an excellent target for inhibition of viral replication by small molecule antivirals. Mosnodenvir is an ultrapotent pan-serotype DENV inhibitor that blocks the interaction between viral proteins NS4B and NS3, an interaction essential for viral replication33. Mosnodenvir is safe and well tolerated and has demonstrated efficacy in humans against dengue182. At higher concentrations, it also inhibits the replication in vitro of other orthoflaviviruses, including West Nile virus, Japanese encephalitis virus and Zika virus. Other NS4B-targeting compounds have been described, including NITD-688 (ref. 183) and BDAA184, which display more restricted activity profiles and selectively inhibit DENV and yellow fever virus replication, respectively. Whether further optimization of these molecules can yield pan-orthoflavivirus activity remains to be explored. Owing to its highly hydrophobic and conformationally dynamic nature and its dependence on a membrane context, high-resolution structures of NS4B have not yet been resolved. This lack of structural information is an obstacle to the target-based development of NS4B-targeted antivirals185. As NS4B is distinct to orthoflaviviruses, these compounds are primarily suited as intra-family BSAs.

Viral entry

Viral entry into host cells involves multiple sequential steps and is mediated by virus–host interactions that are often highly specific, even within a single viral genus. For example, although poliovirus, rhinovirus C and enterovirus 71 all belong to the genus Enterovirus (family Picornaviridae), they use a highly diverse set of cellular receptors (CD155, CDHR3, and SCARB2 or PSGL-1, respectively) and engage different endocytic routes. This diversity makes the development of BSAs particularly challenging186,187. Interestingly, entry inhibitors with broader antiviral activity against the genus Mammarenavirus have been reported. ARN-75039 inhibits membrane fusion and entry of Old World (such as Lassa virus) and New World mammarenaviruses (such as Junín and Machupo viruses) within the family Arenaviridae by targeting conserved structural features of the glycoprotein complex188. However, entry inhibitors targeting viral glycoproteins in other families are mostly highly selective, typically exhibiting activity restricted to a single genus or even individual viruses, as exemplified by inhibitors of the respiratory syncytial virus fusion protein and the influenza virus haemagglutinin fusion protein189,190,191. A suggested approach to achieve broader-spectrum activity is the use of synthetic carbohydrate receptors that bind conserved N-glycans on viral envelopes, thereby inhibiting entry across multiple virus families, including coronaviruses, filoviruses, and paramyxoviruses192. Another proposed strategy is targeting host cell glycans, such as sialic acid and heparan sulfate, using engineered β-cyclodextrins that display virucidal activity against a broader panel of respiratory viruses (for example, HPIV3, RSV, influenza A and B viruses and SARS-CoV-2)193. The physicochemical properties may limit potential clinical use.

Uncoating and assembly

Uncoating and assembly represent another underexploited target in the viral replication cycle (Fig. 2). The ultrapotent and long-acting HIV drug lenacapavir targets both functions, suggesting that these processes could possibly also be targeted in other viral families194. We and others recently demonstrated that the coronavirus assembly process can be efficiently inhibited by targeting the viral M-protein, a key regulator of virion assembly. The druggable pocket identified in the M-protein may allow the design of pan-Betacoronavirus genus inhibitors35,195. For measles and Nipah virus (family Paramyxoviridae), pharmacological reduction of phosphatidylinositol 4,5-bisphosphate (PI(4,5)P2) results in the mis-localization of the M-protein and inhibition of viral replication by blocking viral assembly196. The relative conservation of the structure and function of the M-protein across paramyxoviruses may possibly allow the development of broader-acting antivirals within this family.

Host-targeted antivirals

HTAs act on cellular factors or pathways that viruses exploit for replication. By targeting host functions rather than viral proteins, HTAs may offer the potential for broader-spectrum antiviral activity and a higher barrier to resistance. However, interference with cellular processes (for example, MTase as previously discussed) also increases the risk of adverse effects. A successful example of an HTA is maraviroc, a CCR5 receptor antagonist used in the treatment of HIV infection197. Type I interferons (IFNs), although not small molecules, represent another widely used host-directed antiviral strategy. Pegylated IFNα (PEG-IFN) in combination with ribavirin was the standard of care for the treatment of HCV infections until 2014, when highly potent DAAs became available198. Ribavirin, a nucleoside analogue with (at least in vitro) antiviral activity against various viruses belonging to several viral families (including Flaviviridae, Arenaviridae, Paramyxoviridae, Coronaviridae, Filoviridae, Orthomyxoviridae and Picornaviridae)199, has a complex and incompletely understood mechanism of action that involves both host-directed and virus-directed effects200. Its antiviral activity against members of the families Flaviviridae and Paramyxoviridae seems to result predominantly from depletion of intracellular GTP pools through inhibition of inosine monophosphate dehydrogenase201. Ribavirin has also been reported to restore interferon responsiveness202 and to exert direct antiviral effects, including lethal mutagenesis and inhibition of RNA capping203. Ribavirin is used in the treatment of Lassa fever (family Arenaviridae) and CCHF (family Nairoviridae), although evidence for a clinical benefit remains limited204. Signal peptide peptidase was recently identified as a target for inhibition of multiple orthoflaviviruses205. Alisporivir is a non-immunosuppressive cyclosporin A analogue that inhibits HCV replication by targeting cyclophilin A (CypA)206. Depletion of CypA by the drug seems to be well tolerated by the host cell as its function can be partially compensated by other cyclophylins207. Despite promising antiviral activity in phase II and III trials, development was halted owing to adverse effects when combined with PEG-IFN208.

Although HTAs are conceptually attractive, relatively few have succeeded in clinical development. Limited success is frequently attributed to toxicity or insufficient antiviral activity, as exemplified by inhibitors of the lipid kinase phosphatidylinositol 4-kinase III (PI4KIII), such as enviroxime. Enviroxime potently inhibits replication of HCV (family Flaviviridae) and enteroviruses (family Picornaviridae) in vitro209,210; however, safety concerns were observed in several studies211,212 and disappointing efficacy in clinical trials213 raised concerns regarding the use of PI4KIII inhibitors for the treatment of viral infections214. Future progress in host-directed antiviral therapy will require careful target selection, prioritizing host factors with a high therapeutic index (that is, large separation between antiviral efficacy and host toxicity) to minimize off-target effects.

Proteolysis-targeting chimeras (PROTACs) are rationally designed, heterobifunctional small molecules that induce selective degradation of target proteins215 and represent an emerging technology, particularly in oncology. Unlike classical inhibitors, PROTACs eliminate rather than transiently suppress protein function and may expand the druggable target space as they only require a suitable binding interface. The PROTAC approach is being explored against numerous viral proteins (such as HBV X protein216, HCV NS3/NS4A protease217 and SARS-CoV-2 Mpro218) and also against host factors essential for virus replication (for example, GSPT1 (ref. 219) and cyclophilin A220). However, poor drug-like physicochemical properties and delivery challenges may limit clinical implementation221,222.

New approaches to discover new targets for broader-acting antivirals

Despite major advances in antiviral drug development, current efforts towards the development of antiviral drugs, including BSAs, remain largely focused on a limited number of well-established targets, such as the viral RdRp and viral proteases32. These targets have yielded clinically important drugs and clearly demonstrate that broader-spectrum antiviral activity is achievable. However, continued reliance on this relatively narrow set of possible targets may not provide sufficient coverage for the diversity of existing viruses and against future viral threats. Given the complexity of viral replication cycles, it can be assumed that many more druggable antiviral targets, beyond those currently exploited, remain undiscovered or underexplored and could serve as intra-family, pan-family or cross-family antiviral targets.

Antiviral drug discovery has relied on two main strategies: target-based drug discovery (TBDD) or phenotypic screening drug discovery (PSDD). To identify antiviral host targets, druggable genome screens have been and are being used205. TBDD requires identification of a specific molecular target and detailed structural and mechanistic knowledge. This approach has been highly successful for targets such as the HIV, HCV and SARS-CoV-2 proteases223, as well as influenza virus neuraminidase224 and endonuclease147. However, TBDD inherently biases discovery towards known targets and is poorly suited to uncover new or unconventional antiviral drug targets, particularly in emerging or less well-characterized RNA viruses. PSDD offers a complementary, target-agnostic strategy by identifying molecules that inhibit viral replication in cell-based systems without making assumptions about the target. In the antiviral field, PSDD has led to the identification of several important drug classes and key targets, such as non-nucleoside HIV reverse transcriptase inhibitors225, inhibitors of the DENV NS3–NS4B interaction33, enterovirus capsid pocket inhibitors226, alphavirus capping machinery inhibitors162, coronavirus assembly inhibitors35,195 and HCV NS5A inhibitors34. These examples highlight the value of phenotypic approaches to reveal essential viral functions that lack obvious enzymatic activity or are inaccessible to classical target-based assays. Nevertheless, conventional phenotypic screens also have important limitations. Most rely on simple monoparametric or biparametric assays (such as reduction in cytopathogenic effect and cytotoxicity) and typically prioritize compounds with maximal antiviral potency and minimal toxicity. As a result, molecules that exert weaker or partial antiviral activity are often discarded, despite the possibility that such compounds could serve as chemical tools to reveal new targets or pathways and may open avenues for the development of more diverse BSAs. AI and machine learning are rapidly revolutionizing high-throughput phenotypic screening in drug discovery, offering new opportunities to overcome these limitations. In other therapeutic areas, such as oncology, neuroscience and regenerative medicine, AI-driven analysis of multiparametric imaging data has enabled detection of subtle, mechanism-linked phenotypic changes induced by disease or drug treatment227,228. In antiviral research, progress has been more limited. Recent efforts include approaches that combine (i) cell painting (a high-content imaging approach using multiplexed fluorescent dyes to generate rich morphological profiles of cells in response to genetic or chemical perturbations229) with (ii) antibody-based detection of viral infection230 and (iii) AI-powered frameworks for image analysis231, highlighting the potential of these approaches for antiviral drug discovery. However, throughput remains limited, and systematic comparisons across multiple virus families are still in their infancy232. Advanced AI-driven analysis could unlock complex phenotypic datasets and has the potential to link compound structures to cellular responses and mechanistic insights, allowing prioritization of antiviral candidates not only by potency and safety but also by shared or distinct mode of action. When applied across panels of representative viruses from selected families, such approaches could facilitate the systematic mapping of virus-specific and conserved targets. Ultimately, combining high-content phenotypic screening with AI-assisted interpretation, new strategies in virtual screening141,233, medicinal chemistry and mechanistic studies will provide a powerful approach to expand the antiviral target landscape and could substantially accelerate the development of next-generation BSAs, improving preparedness for future viral outbreaks.

Translational readiness

Further clinical development of broader-acting antivirals

Once a potential drug candidate shows favourable preclinical activity, including efficacy, tolerability and safety in relevant animal models, clinical development may be warranted. This includes scaling up manufacturing, conducting safety assessments in humans, generating preliminary (indirect) evidence of efficacy and exploring appropriate pathways for (temporary) licensure. Economic incentives may be poorly aligned with the development cost of drugs against viruses with uncertain incidence, limited markets or geographically restricted disease burden. This challenge is not limited to epidemic-prone viruses but also applies to endemic, sporadic or certain opportunistic viral infections. Public–private partnerships (PPPs) and product-development consortia, bringing together academia, industry, governmental agencies, regulators and civil society, can effectively advance the early clinical development (phase I/IIa) of drugs of limited commercial interest234. Examples of such PPPs include the Medicines for Malaria Venture, the Drugs for Neglected Diseases initiative and the Innovative Medicines Initiative235. In this context, in the USA, the Biomedical Advanced Research and Development Authority (BARDA) fulfils a key role in supporting the development of antivirals that are important for current or future epidemics but that are not necessarily commercially viable. BARDA initiatives include the Flexible and Strategic Therapeutics programme, which funds adaptable therapeutic platforms for epidemic viral diseases, and the Small Molecule Approaches for Rapid and Robust Treatment Antiviral Prize aimed at accelerating the development of scalable antivirals against pandemic-prone viruses. Earlier, the NIH-launched Antiviral Drug Discovery Centers provided a dedicated infrastructure for early discovery and preclinical development of antiviral candidates against several high-risk viral families. Also, certain European initiatives, in particular those launched by EU Horizon Europe and the European Commission’s Health Emergency Preparedness and Response Authority, aim to support the development of broader-spectrum antiviral development.

Pathways to (temporary) licensure

In the particular context of epidemic or pandemic preparedness, phase I clinical studies in healthy volunteers should be conducted during interepidemic periods to establish safety, tolerability and pharmacokinetics. Subsequent clinical development can be challenging when target viruses circulate intermittently or at low incidence. For viruses that rarely infect humans or have been eradicated (for example, SARS-CoV, Hendra virus and Variola virus), licensure may rely on alternative regulatory pathways such as the FDA Animal Rule. In contrast to an Emergency Use Authorization, which grants only temporary approval, the Animal Rule allows full market access based on efficacy in validated animal models when human efficacy studies are neither ethical nor feasible236. Complementing this, new approach methodologies, including AI-based computational models, organ-on-a-chip systems and organoids, are being explored as potential alternatives to animal testing237. The European Medicines Agency has a comparable mechanism through the ‘exceptional circumstances’ pathway, which allows marketing authorization when comprehensive clinical data cannot be obtained, subject to ongoing evidence generation and periodic reassessment.

Tecovirimat, an orthopoxvirus inhibitor, illustrates the utility and limitations of this approach. Tecovirimat progressed through efficacy studies in rodent and non-human primate models, pharmacokinetic evaluation and human safety trials, ultimately becoming an antiviral approved under the Animal Rule238. During the 2022 mpox outbreak, the drug was deployed in the USA under the CDC’s expanded access protocol. Similar provisions in Europe (European Medicines Agency) allowed authorization under ‘exceptional circumstances’. Such mechanisms enable strategic stockpiling and emergency deployment. However, regulatory approval based on animal efficacy does not guarantee clinical benefit in human disease. Indeed, subsequent randomized trials in moderately ill patients with mpox failed to demonstrate a reduction in time to lesion resolution239.

Clinical testing

The recent failure of tecovirimat for the treatment of mpox despite licensure underscores the continued need for phase II and III clinical trials when human efficacy trials become feasible, that is, when larger numbers of people are infected by a certain pathogen. To prepare for ‘virus X’, candidate BSAs should be clinically assessed against circulating prototype viruses. For example, pre-developed pan-coronavirus inhibitors could have been tested against seasonal coronaviruses before the emergence of SARS-CoV-2. In selected cases, phase II studies can be conducted using controlled human infection models, which exist for several viral pathogens, including influenza viruses (family Orthomyxoviridae), RSV (family Pneumoviridae), DENV (family Flaviviridae) and rhinoviruses (family Picornaviridae)240. Clinical trials should ideally be prepared before epidemics occur.

Experience from outbreaks of Ebola virus disease, COVID-19 and other viral epidemics has shown that ad hoc clinical trials initiated during public health emergencies often fail to reach target enrolment or generate definitive results241. This has resulted in fragmented landscapes of underpowered studies evaluating multiple therapeutic compounds in parallel, many of which were terminated prematurely once larger international platform trials had overtaken them in demonstrating futility or superiority242. Regulatory and ethical review processes often represent critical bottlenecks during outbreaks. Pre-approved, generic master or core protocols can help to overcome these hurdles. These protocols define standardized trial elements, such as eligibility criteria, randomization procedures, endpoints and analytical methods243, which can be rapidly adapted to specific interventions or settings through appendices. Adaptive platform trial designs further increase efficiency by allowing multiple therapies to be evaluated simultaneously (for example, through factorial randomization) or consecutively, with predefined rules for adding, dropping or transitioning interventions across trial phases243. In addition, more favourable interventions can be prioritized through response-adaptive randomization243. Successful examples include REMAP-CAP, which evaluated multiple interventions for the management of patients with severe community-acquired pneumonia242,244. The trial was embedded in clinical practice, increasing efficiency and decreasing the gap between research and clinical care243. During the COVID-19 pandemic, large pragmatic platform trials such as SOLIDARITY, ACTIV and RECOVERY, executed by international clinical trial networks, rapidly generated high-quality evidence at scale242.

A recent example of a pre-epidemically conceived protocol is the SOLIDARITY PARTNERS trial. This WHO-led, multi-country adaptive platform randomized trial evaluates the effect of new and repurposed therapies on filovirus diseases. Distinctive features of this protocol are that it covers multiple viruses of the same family (Ebola virus, Marburg virus and Sudan virus), includes different classes of drugs (antivirals, monoclonals and immunomodulating drugs) and can recruit participants across different outbreaks in different countries. During the 2024 Marburg outbreak in Rwanda, the study was implemented two and a half weeks after the outbreak was declared and included five participants. Similarly, for the recent Ebola outbreak in the Democratic Republic of the Congo (DRC), caused by Bundibugyo virus, initiatives are underway to establish a clinical trial evaluating two experimental treatments, namely, the mAb cocktail MBP134 and the nucleoside analogue remdesivir245. Between outbreaks, the study goes dormant, but readiness is ensured through continuous training and site-capacity development246.

Equity and global access

Equitable access must be a core consideration in the development of BSAs. Many viruses with a high public health impact are endemic or emerge disproportionately in low-income and middle-income countries (LMICs), often in remote areas (such as filoviruses, monkeypox virus, orthoflaviviruses, arenaviruses, Nipah virus, CCHFV and RVFV). Historical underinvestment in these settings has delayed the development and evaluation of effective countermeasures. Conducting clinical trials in affected regions is essential not only for pandemic preparedness but also to ensure that resulting therapies address local clinical needs. The multicentre randomized PALM-007 trial in the DRC demonstrated the feasibility of this approach by evaluating tecovirimat for mpox while generating data relevant to smallpox preparedness239. However, research conducted in LMICs should not primarily serve the interests of high-income countries. Instead, it should be embedded in long-term strategies that strengthen local research capacity, ensure affordability and guarantee sustainable access to effective antivirals247.

Conclusions

Over the past decades, remarkable progress has been made in developing highly potent and safe antiviral drugs for the treatment of infections such as those caused by HIV, HBV, HCV, herpesviruses, influenza viruses and coronaviruses. Yet, effective antiviral drugs remain unavailable for most other, often neglected, life-threatening viral infections, many of which are caused by RNA viruses. Newly emerging (highly pathogenic) viruses (‘virus X’) will most likely also belong to known viral families, further underscoring the urgent need for broader-acting antivirals that address both current unmet medical needs and future outbreaks.

The catalytic site of viral RdRps is relatively conserved across different RNA virus families and represents a promising target for the development of broader-spectrum antivirals. Several nucleoside analogues targeting the RdRp have been shown to exert pan-family or even cross-family antiviral activity. Whether nucleoside RdRp inhibitors can be developed with even broader cross RNA-virus family activity remains to be explored. Other viral enzymes, including proteases, helicases and endonucleases, are not encoded by all viral families but may nonetheless be exploited to develop inhibitors with pan-genus or pan-family activity. In addition, targeting structural viral proteins (such as the coronavirus M-protein, which orchestrates viral assembly) or essential viral protein–protein interactions (such as those involving orthoflavivirus NS4B and NS3) may yield antiviral drugs with pan-genus or pan-family antiviral activity. Host-targeted approaches have the potential for broader-spectrum antiviral activity and a higher barrier to resistance but require careful balancing of antiviral efficacy and host tolerability. The optimal characteristics of such BSAs and how to deploy them will depend on the clinical and epidemiological context. Pandemic preparedness may require potent orally available cross-family or pan-family antivirals targeting (relatively) conserved proteins (such as RdRps or proteases), whereas for known infections (such as those caused by enteroviruses), intra-family activity may suffice. Cross-family antivirals with relatively modest potency may still be relevant as interim solutions, for example, in combination regimens or as repurposed agents, whereas time and efforts are invested in the development of more potent and selective drugs. For neurological infections, efficient central nervous system penetration is essential, whereas for other infections more restricted tissue distribution may be preferable. Advances in structural virology, high-throughput phenotypic screening and AI-driven analysis offer unprecedented opportunities to uncover new druggable targets that can be exploited for the development of either pan-genus, pan-family or cross-family activity. Finally, preparedness also requires translational readiness, including pre-approved master protocols, appropriate regulatory pathways and mechanisms that ensure global access.

In conclusion, building a strategically diversified portfolio of BSAs that covers most, if not all, RNA virus families will be indispensable for enabling the treatment and prophylaxis of currently untreatable infections and for rapid deployment during outbreaks caused by newly emerging viruses. Such portfolios of BSA will be key to strengthening global resilience against viral pathogens.