Positive-sense single-stranded RNA [(+)ssRNA, see Table 1 for glossary of abbreviations] viruses represent the largest group of RNA viruses, spanning 30 divergent viral families that include important human pathogens in Flaviviridae, Picornaviridae, and Coronaviridae such as dengue virus, poliovirus, and SARS-CoV-2 (1). Despite their diversity, many viruses in this group share a common replication strategy: their (+)ssRNA viral genomes are delivered to host cells as a translation-ready mRNA that encodes a multidomain viral polyprotein. Following translation of the viral polyprotein by host ribosomes, one or more embedded viral proteases cleave the polyprotein into individual, functional proteins at numerous sequence-specific positions ( Figure 1A ). Polyprotein cleavage at these specific sites is necessary for sustained virus replication and propagation, making viral proteases an attractive target for development of antiviral therapeutics (2, 3).

In addition to their essential role in the viral life cycle, (+)ssRNA viral proteases also cleave host proteins to manipulate host processes, including the host innate antiviral immune response (4). Importantly, host targets are cleaved with the same sequence specificity as sites within the viral polyprotein ( Figure 1B ). These dual roles place viral proteases at the intersection of two opposing selective pressures. On one side, the virus and its polyprotein site targets are under strong pressure to be conserved, as any changes to the protease sequence specificity or protease sites without concomitant changes to the other would be deleterious for viral fitness. On the other side, viral fitness may be expected to benefit from a protease’s ability to adapt to and cleave new host targets, newly evolved sequences in the same host, or divergent sequences in a different host to facilitate cross-species transmission. This type of direct engagement between viral proteases and host factors thus generates an evolutionary conflict where both sides may be driven to adapt in a type of escalating molecular ‘arms race’ (5–8).

Molecular arms races exist as a result of the competing evolutionary interests of viruses and their hosts. Such competing interests establish an evolutionary equilibrium that is characterized by cyclical adaptations that exemplify a so-called ‘Red Queen’ genetic conflict (9). In these cases, viral adaptations that allow for successful infection of a host will provide a temporary advantage to the virus. However, host adaptations may restore the advantage to the host, applying selection pressure back to the virus ( Figure 2A ). Thus, molecular interactions between viruses and their hosts, particularly those interactions that contribute to potentiation or inhibition of virus replication, are shaped by immense evolutionary pressure on both parties: hosts are driven to both maintain interactions that activate or carry out antiviral defenses and evade virus interactions that prevent these responses, and viruses are driven to do the opposite. The result is recurrent adaptation of both virus and host to promote either virus replication or host antiviral mechanisms, respectively ( Figure 2A ) (5–8). Due to the fact that the direct molecular host-virus interfaces are those that are being remodeled during such molecular arms races, single amino acid changes can change the outcome of these conflicts (5, 7). Indeed, traces of these host-virus conflicts can be detected in host genomes by identifying gene codons that show evolutionary signatures of recurrent diversifying (positive) selection (5, 7, 10). Similarly, viruses are known to adapt during or following cross-species transmission to a novel host, and such adaptations can also be characterized by signatures of positive selection in viral genomes (11–14). Importantly, whether the host has evolved to the virus or the virus has evolved to the host, the resulting genetic and molecular changes determine the host range and pathogenesis of viruses, including influencing the ability of viruses to zoonotically transmit into the human population (7, 8, 13).

Due to the importance of sequence specificity to protease-host interactions, evolutionary arms races at the interfaces of proteases and their targets would be expected to exist. For instance, a single amino acid change in a targeted host protein at a position that is important for sequence-specific protease cleavage could completely reverse cleavage susceptibility ( Figure 2B ). As a result, single lineage-specific changes at any number of positions in the cleavage motif would be expected to alter cleavage susceptibility even among closely related hosts, establishing species-specific host-virus interactions that could drive viral host range ( Figure 2C ). Indeed, while a great deal of research on viral proteases has focused on conserved elements of protease function, emerging evidence suggests that both hosts and viruses are evolving in ways that can impact the host- and virus-specificity of cleavage. Here, we review the host-viral molecular conflicts engaged by the main proteases of flaviviruses, picornaviruses, and coronaviruses to emphasize how proteases of (+)ssRNA viruses act as evolutionary drivers of host innate immunity, and how viral proteases are being shaped by these same molecular conflicts. This evolutionary perspective highlights the importance of viral proteases and their host targets as being an important determinant of viral host range, tissue tropism and pathogenesis, and zoonotic potential of (+)ssRNA viruses.

Virus-encoded proteases are essential to the life cycle of numerous (+)ssRNA viruses. Newly synthesized viral polyproteins mature into individual, functional proteins via a series of cleavage events carried out by virus-encoded and host proteases. For Picornaviridae and Coronaviridae, the viral cysteine proteases 3C and 3C-Like (3CL) respectively, are responsible for the majority of polyprotein processing events ( Figures 3A, B ) (4, 16–19). Most picornaviruses have six or more 3C cleavage sites throughout the polyprotein ( Figure 1A ), and there is a preference to cleave between a glutamine (Q) in the P1 position and a small residue [e.g. glycine (G) or serine (S)] in the P1’ position ( Figure 3A ) (15, 16, 20). Likewise, coronaviruses (CoVs) have ten or more cleavage sites for the 3CL protease (also known as MPro or nsp5 in several CoVs including SARS-CoV-2) ( Figure 3B ) (17, 18, 21). Numerous other viral families, including members of Caliciviridae (e.g. norovirus) (22) and Potyviridae (e.g. tobacco etch virus) (23) encode a cysteine protease with a similar specificity for cleavage between a Q and a small residue, whereas members of Togaviridae (e.g. Chikungunya virus) use a cysteine protease with different cleavage specificity (24). Other viral families use a serine protease, including Flaviviridae, where the serine protease NS3 processes at least four polyprotein cleavage sites ( Figure 3C ) (25, 26). Here and in subsequent sections, we will predominantly discuss activities of the 3C, 3CL, and NS3 proteases of Picornaviridae, Coronaviridae, and Flaviviridae, respectively, due to their known roles in cleaving mammalian host factors. It is important to point out that because the polyprotein is sequentially processed, and not always to completion, protease activity may also be carried out when 3C, 3CL, or NS3 remains fused or associated with additional viral proteins (27, 28). This is especially true in the Flaviviridae, where the NS3 protease usually functions in association with NS2B (in the case of flaviviruses such as dengue and Zika viruses) or NS4A [in the case of hepatitis C virus (HCV)] (26, 29). However, for the sake of clarity, we will subsequently only refer to the protease domains of 3C, 3CL, or NS3. Moreover, many (+)ssRNA viruses encode additional proteases involved in both polyprotein processing and host antagonism, including the 2A protease in some picornaviruses and the papain-like protease (PLP) in coronaviruses (18, 30). Finally, viruses other than (+)ssRNA viruses can encode proteases that are important for polyprotein processing, most notably, the retrovirally-encoded aspartyl protease (31). While all of these additional proteases from (+)ssRNA viruses and retroviruses play important host antagonism roles, and likely shape host and viral evolution, they will not be extensively explored here.

The functions of the (+)ssRNA viral proteases described above are, by definition of being required for completion of the viral life cycle, well conserved. In addition to homology between the proteases themselves, the positions and sequences of the polyprotein cleavage motifs are often similar between members of the same viral family. Indeed, this conservation of polyprotein cleavage motifs has made it possible to compile sequences surrounding the cleavage site from genome sequences alone to generate a consensus motif for the viral protease that can be used to predict host and viral targets (15, 20, 32, 33) ( Figure 3A ). These consensus motifs are often generated using many diverse viruses, relying on the assumption that protease sequence specificity is well conserved among virus species. Interestingly, despite the evolutionary constraint to maintain cleavage across multiple sites in the polyprotein, virus-encoded proteases are substantially divergent across viruses ( Figure 3 ). For instance, picornavirus 3C proteases can share less than 20% amino acid sequence identity, despite sharing an overall similar fold and many homologous cleavage sites (15). Similar evolutionary distances are observed with other families of proteases, including Coronaviridae 3CL and Flaviviridae NS3 ( Figure 3 ).

Even with the divergence of protease sequences, protease sequence specificity is expected to be well conserved within closely related viruses given the essentiality of cleaving multiple site-specific polyprotein sites. Surprisingly, there is mounting evidence that this is not the case. For instance, among closely related serotypes of dengue virus (DENV), biochemical substrate profiling has revealed a subtle but clear shift in the NS3 protease cleavage sequence specificity profile (34). This type of in-depth comparative biochemical analysis of other (+)ssRNA proteases has not been conducted, but assays on model substrates have revealed differences in cleavage specificity even among 3C proteases within the Enterovirus genus of Picornaviridae (35). Some of the best evidence that protease sequence specificity is changing between related viruses has come from studies using chimeric viruses in which the protease of one virus species is inserted into the backbone of another virus. If such protease swaps result in insufficient or improper cleavage of the polyprotein and reduced viral replication, it would suggest divergence in protease sequence specificity between the parental viruses. For example, among enteroviruses ( Figure 3A ), replacing the poliovirus (PV) 3C protease with 3C proteases from human rhinovirus 14 or coxsackievirus B3 (CVB3) resulted in reduced, changed, or loss of cleavage products (36). Likewise, within the Flavivirus genus of Flaviviridae ( Figure 3C ), swapping the protease domain of DENV NS3 for the protease domain of yellow fever virus (YFV) ablates processing of polyproteins containing DENV cleavage sites (37). Additionally, West Nile virus (WNV) NS3 can cleave a polyprotein site in only one of two closely related DENV2 strains, where the only difference is in the residue in the P1’ position (38). While some of these differences may be attributed to the requirement for NS3 proteases to bind to lineage-specific activating cofactors to further augment cleavage specificity (39–47), it is also likely that these changes in cleavage specificity are dependent on non-conserved residues in the binding pocket of the NS3 protease (48). Similarly, within Coronaviridae, replication competent chimeric murine hepatitis virus (MHV) could not be recovered when the 3CL protease was replaced with one of many related alpha- or beta-coronavirus proteases including SARS-CoV, hCoV-229E and bat CoV-HKU4 ( Figure 3B ). Only when the MHV 3CL was replaced with the two most closely related beta-coronaviruses, hCoV-OC43 and hCoV-HKU1, could virus be recovered, but with a substantial fitness cost (49). Altogether, these biochemical and chimeric virus studies illustrate that (+)ssRNA viruses have undergone lineage-specific evolution in both their protease sequence specificity as well as their many polyprotein cleavage sites.

The above-described changes in protease sequence specificity do not require invocation of adaptation. Indeed, evolutionary drift could result in changes to the viral protease and its cleavage sites, including those that result in loss of fitness for chimeric viruses. However, there is another selective pressure that likely shapes viral protease evolution: the advantage that viruses gain by cleaving host targets. Proteins in multiple cellular processes have been identified as targets of viral proteases, many of which are involved in the host antiviral immune response (4). Many of these host targets are divergent between species, potentially establishing molecular barriers to cross-species transmission. Although the ability to cleave the viral polyprotein is an invariant function of viral proteases, we posit that cleavage of specific host proteins may be selected for during viral evolution, especially during or following cross-species transmission. Indeed, pathogenicity of a mouse-adapted SARS coronavirus required two mutations in 3CL to facilitate rapid, robust virus replication (50). Although it has not been established whether these 3CL changes result in changes in host target cleavage, these data indicate that protease evolution may be required for successful adaptation to a novel host species.

Several excellent reviews have been written describing the diverse host targets that are cleaved by viral proteases (4, 16, 17, 19, 51). In many cases, the described host-virus interaction has focused on a single or a small number of related viral proteases and only a single host species, often humans. Thus, the importance of host and virus diversity in these interactions is often poorly understood. However, evidence is accumulating that viral proteases and their host targets are engaged in species-specific interactions. Below, we highlight such cases in which host and viral diversity alter the outcome of the interaction between host pathways and proteases of picornaviruses, flaviviruses and coronaviruses, illustrating this ongoing molecular arms race.

Cleavage of host proteins by viral proteases often inactivates the host protein and results in a fitness advantage for the virus. In these cases, host evolutionary signatures reveal adaptations that are presumed to evade cleavage. However, another possibility exists, in which the host protein can sense the presence of the viral protease in the cytoplasm through an evolved sequence that mimics the viral polyprotein cleavage site. Sensing of pathogen-encoded activities such as toxins and effector enzymes, known as effector-triggered immunity (ETI), is well-described in plants but is also emerging as an important immune mechanism in animals (103–106). Three such signaling pathways, described below, are known to detect the main protease activity of human viruses.

The proteases of (+)ssRNA viruses have multiple roles in establishing and maintaining virus infection within a host. First and foremost, virally-encoded proteases cleave numerous sequence-specific sites within the viral polyprotein, which is essential for completion of the viral replication cycle. As a consequence of this essential activity, the ability of proteases to evolve novel sequence specificity is highly constrained. However, viral proteases also serve to manipulate numerous host processes in the infected cell through site-specific cleavage of host targets. In this context, changes in protease sequence specificity would allow the virus to cleave new host targets that might benefit the virus, or avoid cleaving host targets that are detrimental to the virus. It is at this intersection that viral proteases are engaged in evolutionary ‘arms races’ with the host, resulting in varied interactions across viral and host species and across evolutionary time. Several examples, including virus-specific cleavage of essential mRNA translation machinery and host-specific evasion of cleavage of innate antiviral immune components, highlight the consequences of these evolutionary conflicts. More recently, the discovery of host-encoded effector-triggered immunity (ETI) sensors such as NLRP1 and CARD8 suggest that host mimicry of viral protease cleavage sites is an efficient strategy to detect the cellular activity of viral proteases.

The extent to which viral protease evolution, and host target diversity, shape viral host range and pathogenesis remains unknown and is an exciting area of future research. The majority of characterized protease-host interactions have been described for a single virus against a single host, leaving open the opportunity for more detailed exploration of the evolutionary dynamics of these interactions. Indeed, examples such as cleavage of host proteins such as MAVS, STING, and NLRP1 highlight the insights that can be gained from additional analyses of host and viral diversity in these interactions. Likewise, future studies aiming to discover additional host targets of viral proteases, especially those that may be cleaved in a virus-specific manner, will advance our knowledge of the ways that protease-host interactions shape viral phenotypes. Finally, ETI sensors such as NLRP1 and CARD8 may represent just the start of host proteins that mimic viral protease cleavage sites to induce an immune response. Further studies aimed to identify ETI mechanisms against both viral and other pathogen-encoded proteases will likely continue to reveal novel mechanisms and evolutionary principles of the host innate immune response.

All authors discussed relevant literature. BT, EF, and MD wrote the first draft. All authors contributed to figure generation and editing the manuscript. All authors contributed to the article and approved the submitted version.

This work was supported by grants from the National Institutes of Health (R35 GM133633), Pew Biomedical Scholars Program, Hellman Fellows Program, and Burroughs Wellcome Investigators in the Pathogenesis of Infectious Disease Program to MD and an NIH T32 grant (GM007240) to BT.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.