An Extended Primer Grip of Picornavirus Polymerase Facilitates Sexual RNA Replication Mechanisms

Picornaviruses have both asexual and sexual RNA replication mechanisms. Sexual RNA replication shapes picornavirus species groups, contributes to the emergence of vaccine-derived polioviruses, and counteracts error catastrophe. Can viruses distinguish between homologous and nonhomologous partners during sexual RNA replication? We implicate an extended primer grip of the viral polymerase in sexual RNA replication mechanisms. By sensing RNA sequence complementarity near the active site, the extended primer grip of the polymerase has the potential to distinguish between homologous and nonhomologous RNA templates during sexual RNA replication.

ABSTRACT

Picornaviruses have both asexual and sexual RNA replication mechanisms. Asexual RNA replication mechanisms involve one parental template, whereas sexual RNA replication mechanisms involve two or more parental templates. Because sexual RNA replication mechanisms counteract ribavirin-induced error catastrophe, we selected for ribavirin-resistant poliovirus to identify polymerase residues that facilitate sexual RNA replication mechanisms. We used serial passage in ribavirin, beginning with a variety of ribavirin-sensitive and ribavirin-resistant parental viruses. Ribavirin-sensitive virus contained an L420A polymerase mutation, while ribavirin-resistant virus contained a G64S polymerase mutation. A G64 codon mutation (G64^Fix) was used to inhibit emergence of G64S-mediated ribavirin resistance. Revertants (L420) or pseudorevertants (L420V and L420I) were selected from all independent lineages of L420A, G64^Fix L420A, and G64S L420A parental viruses. Ribavirin resistance G64S mutations were selected in two independent lineages, and novel ribavirin resistance mutations were selected in the polymerase in other lineages (M299I, M323I, M392V, and T353I). The structural orientation of M392, immediately adjacent to L420 and the polymerase primer grip region, led us to engineer additional polymerase mutations into poliovirus (M392A, M392L, M392V, K375R, and R376K). L420A revertants and pseudorevertants (L420V and L420I) restored efficient viral RNA recombination, confirming that ribavirin-induced error catastrophe coincides with defects in sexual RNA replication mechanisms. Viruses containing M392 mutations (M392A, M392L, and M392V) and primer grip mutations (K375R and R376K) exhibited divergent RNA recombination, ribavirin sensitivity, and biochemical phenotypes, consistent with changes in the fidelity of RNA synthesis. We conclude that an extended primer grip of the polymerase, including L420, M392, K375, and R376, contributes to the fidelity of RNA synthesis and to efficient sexual RNA replication mechanisms.

IMPORTANCE Picornaviruses have both asexual and sexual RNA replication mechanisms. Sexual RNA replication shapes picornavirus species groups, contributes to the emergence of vaccine-derived polioviruses, and counteracts error catastrophe. Can viruses distinguish between homologous and nonhomologous partners during sexual RNA replication? We implicate an extended primer grip of the viral polymerase in sexual RNA replication mechanisms. By sensing RNA sequence complementarity near the active site, the extended primer grip of the polymerase has the potential to distinguish between homologous and nonhomologous RNA templates during sexual RNA replication.

INTRODUCTION

RNA viruses arose billions of years ago, becoming ubiquitous parasites of cellular life (1). The RNA-dependent RNA polymerase of RNA viruses is monophyletic, providing a means to consider the evolutionary history of all RNA viruses, to compare distinct groups of RNA viruses, and to classify RNA viruses (2, 3). Picornaviruses, whose origins predate the radiation of eukaryotic supergroups (4), have been studied extensively. Poliovirus, in particular, has been studied intensively for 100 years (5). The poliovirus RNA-dependent RNA polymerase was initially purified and biochemically characterized in 1977 (6). Since then, labs across the globe have studied the poliovirus RNA-dependent RNA polymerase in great detail (reviewed in reference 7). Picornavirus RNA-dependent RNA polymerases are multifunctional, catalyzing distinct steps of viral RNA replication from making primers (VPg-uridylylation) (8–12) to replicating viral RNA (13, 14) and polyadenylation of progeny RNA genomes (15). Solving the atomic structure of poliovirus polymerase (16), and its elongation complex (17), led to unprecedented opportunities, including our ongoing work to understand viral RNA replication mechanisms and error catastrophe (18).

Picornaviruses have both asexual and sexual RNA replication mechanisms, both of which are catalyzed by the viral RNA-dependent RNA polymerase (Fig. 1). Asexual RNA replication mechanisms involve one parental template, whereas sexual RNA replication mechanisms involve two or more parental templates. Previously, we established that an L420A polymerase mutation exacerbates ribavirin-induced error catastrophe coincident with defects in sexual RNA replication mechanisms (Fig. 1) (18, 19). Asexual RNA replication mechanisms are advantageous because vast amounts of progeny can be produced very quickly from one parental RNA template. However, asexual RNA replication mechanisms, in conjunction with error-prone RNA-dependent RNA polymerases (20, 21), can be disadvantageous, contributing to a loss of fitness due to Muller’s ratchet (22, 23). Errors introduced into viral RNA during asexual RNA replication cannot be easily removed in the absence of viral RNA recombination, except by reversion or negative selection (24). In conjunction with genetic bottlenecks, reiterative asexual RNA replication leads to error catastrophe, an overwhelming accumulation of mutations in viral RNA. Negative-strand RNA viruses such as vesicular stomatitis virus, which do not recombine, are especially susceptible to Muller’s ratchet (23). In contrast, picornaviruses, which have relatively high levels of replicative RNA recombination (25, 26), are somewhat resistant to Muller’s ratchet (27–29). Nonetheless, ribavirin, a mutagenic antiviral drug, can restrict picornavirus replication via error catastrophe (30, 31). Replicative RNA recombination, a form of sexual RNA replication (32, 33), can counteract ribavirin-induced error catastrophe (18, 19), presumably by purging lethal mutations from viral RNA genomes (34, 35).

FIG 1.

Picornaviruses have both asexual and sexual RNA replication mechanisms. (A) Asexual RNA replication involves one parental template, whereas sexual RNA replication involves two parental templates. Mutations (asterisks) introduced during asexual RNA replication can be removed by sexual RNA replication. (B) Ribavirin-induced error catastrophe. The fidelity of RNA synthesis (54, 55) and the frequency of RNA recombination (18, 19) influence ribavirin-induced error catastrophe. Because an L420A mutation exacerbates ribavirin-induced error catastrophe coincident with defects in sexual RNA replication (18), we hypothesized that ribavirin-resistant poliovirus selected from an L420A background would acquire novel mutations that restore efficient sexual RNA replication. RDRP, RNA-dependent RNA polymerase. (Adapted from reference 18.)

A growing body of evidence indicates that sexual RNA replication mechanisms play an essential role in picornavirus speciation and ongoing genetic exchange between viruses within defined species groups (36–42). Mucosal surfaces in the respiratory (43) and gastrointestinal tracts (42) are important ecological environments for ongoing genetic exchange between related picornaviruses. Gastrointestinal bacteria may even enhance virus coinfection to promote recombination (44). RNA sequence complementarity between genomes undergoing recombination (26, 45), among other factors (46), influence the frequency of recombination, with higher rates of recombination when sequences are more alike. When two related viruses coinfect a cell, sexual RNA replication mechanisms lead to the generation of chimeric RNA genomes, which are often fit, culminating in sustained host-to-host transmission. Circulating vaccine-derived polioviruses, obstacles to poliovirus eradication, are products of sexual RNA replication mechanisms—formed when unfit portions of oral polio vaccine (OPV) RNA genomes recombine with more fit regions of nonpolio group C enteroviruses (47–49). When recombinant progeny are fit, as occurs in the case of circulating vaccine-derived polioviruses, the recombinant strains circulate from host to host (50, 51). Thus, an important biological consequence of sexual RNA replication mechanisms is the frequent exchange of genetic material among related picornaviruses. Picornavirus species groups are sustained over time by repeated ongoing genetic exchange between related viruses within the species group. A leucine 420 residue in the polymerase thumb domain, conserved across picornavirus species groups, is required for efficient RNA recombination/sexual RNA replication (18, 19).

The interplay between asexual and sexual RNA replication (Fig. 1) is central to what Eigen deemed a grand challenge for the 21st century, elucidating the complex mechanisms of error catastrophe, which are variable from one type of virus to another (52). Error catastrophe and lethal mutagenesis are commonly used in the literature to describe decreased virus titers associated with replication in the presence of ribavirin (30, 52, 53). In 2007, Bull et al. (53) set out to develop the theory of lethal mutagenesis, espousing that it be distinct from the theory of error catastrophe. In their theory of lethal mutagenesis, Bull et al. assume that viral RNA recombination is absent (53). By assuming that viral RNA recombination is absent, lethal mutagenesis theory fails to appreciate the important interplay between asexual and sexual RNA replication mechanisms. Asexual RNA replication mechanisms are efficient; however, they render viruses susceptible to error catastrophe/lethal mutagenesis (30, 32, 33). Sexual RNA replication (also known as recombination), while inefficient, counteracts the primary disadvantage of asexual RNA replication, namely, error catastrophe/lethal mutagenesis (32, 33). Overall, we find that the theory of error catastrophe is more realistic because it better emphasizes the distinctions and interplay between asexual and sexual RNA replication strategies. Using this theoretical framework, we propose that viral RNA recombination counteracts ribavirin-induced error catastrophe by purging ribavirin-induced mutations from viral RNA genomes (Fig. 1). Viral RNA recombination can reassemble mutant-free genomes to refresh the pool of RNA undergoing asexual RNA replication (Fig. 1). Thus, consistent with theoretical underpinnings (32, 33), genetic exchange during sexual RNA replication is advantageous because it provides a mechanism to counteract error catastrophe.

Building upon previous studies showing an L420A polymerase mutation exacerbates ribavirin-induced error catastrophe coincident with defects in sexual RNA replication mechanisms (18, 19), we predicted that ribavirin-resistant poliovirus selected from L420A parental strains will acquire polymerase mutations that facilitate sexual RNA replication mechanisms (Fig. 1).

In the work presented here, we used several ribavirin-sensitive and ribavirin-resistant parental viruses, in conjunction with serial passage in ribavirin, to select for ribavirin-resistant poliovirus. Novel ribavirin resistance mutations, along with L420A revertants and pseudorevertants, were identified and characterized. These data implicate an extended primer grip of the viral polymerase in sexual RNA replication mechanisms.

RESULTS

Serial passage of poliovirus in ribavirin.

We adapted the methods of Pfeiffer and Kirkegaard (54) to select for ribavirin resistance in poliovirus-infected cells (Fig. 2). We used serial passage in escalating concentrations of ribavirin (Fig. 2A), beginning with several ribavirin-sensitive and ribavirin-resistant parental viruses (Fig. 2B): ribavirin-sensitive L420A virus, ribavirin-resistant G64S virus, and virus with a G64 codon mutation (G64^Fix) designed to inhibit emergence of G64S-mediated resistance. Altogether, this study contained eight parental strains with three independent lineages per strain: wild type (WT), G64S, G64^Fix, D79H, L419A, L420A, G64^Fix L420A, and G64S L420A. Polioviruses with these mutations have normal one-step growth phenotypes in HeLa cells, and all of the mutations are stably maintained in virus populations in the absence of ribavirin (18). A G64S mutation in the poliovirus polymerase mediates resistance to ribavirin by increasing the fidelity of RNA synthesis (54, 55), while an L420A mutation in the poliovirus polymerase increases sensitivity to ribavirin by inhibiting replicative RNA recombination/sexual RNA replication (18, 19). Virus with a D79H polymerase mutation was included due to conflicting reports: one suggesting the mutation inhibits replicative RNA recombination (34) and another refuting these conclusions (18). Virus with an L419A polymerase mutation was included because the L419 residue is immediately adjacent to the L420 residue, and alanine substitution mutations at these sites have been functionally characterized in biochemical and virological assays (15, 18, 19). We used a multiplicity of infection (MOI) of 0.1 PFU per cell for each serial passage, harvesting virus by freeze-thaw at 24 h postinfection (hpi). We used an MOI of 0.1 PFU per cell to avoid genetic bottlenecks so that irrelevant (random) mutations that have nothing to do with ribavirin resistance would be less likely to accumulate and fix in the populations. At an MOI of 0.1 (∼10⁵ PFU per 35-mm well), advantageous ribavirin resistance mutations need to outcompete parental strains via positive selection in order to become fixed in the population.

FIG 2.

Selection of ribavirin-resistant poliovirus. (A) Serial passage of poliovirus in escalating doses of ribavirin. Methods adapted from Pfeiffer and Kirkegaard (54). (B) Diagram showing the eight parental and 24 progeny virus strains used in this study.

In order to maintain an MOI of 0.1 PFU per cell for each serial passage, we monitored the titer of poliovirus recovered from each infection (Fig. 3). Passage zero (P0) titers ∼10⁹ PFU per ml were obtained in the absence of ribavirin (Fig. 3). Under most circumstances, the titers for each lineage of virus were similar from passage to passage, with lower titers recovered as the concentrations of ribavirin increased: titers of 10⁸ to 10⁹ PFU per ml for passage 1 to 4 in 100 μM ribavirin (Rb), a wider range of titers from 10⁵ to 10⁸ PFU per ml for passages 5 to 9 in 400 μM Rb, and titers of ∼10⁶ to 10⁷ PFU per ml for passage 10 at 800 μM Rb. In some cases, the titers of poliovirus recovered for independent lineages varied considerably from one passage to another: L420A lineage 2 (L420A2) titers dropped precipitously at passage 5 (∼10⁵ PFU per ml) and remained low through passage 8, whereas L420A lineages 1 and 3 dropped at passage 5 but then increased incrementally from passage 5 through passage 9. Similar divergence was apparent in G64^Fix L420A lineages: G64^Fix L420A lineage 2 (G64^Fix L420A2) titers did not drop at passage 5, whereas G64^Fix L420A lineages 1 and 3 dropped precipitously at passage 5, with an incremental recovery by passage 9. In contrast, titers for G64S lineages 1 to 3 remained relatively high at all passages compared those of other parental strains. Altogether, these data show that WT, ribavirin-resistant, and ribavirin-sensitive parental strains behave differently during serial passage in escalating concentrations of ribavirin and that individual lineages of ribavirin-sensitive parental viruses become more resistant to ribavirin at different points during passage.

FIG 3.

Titers of poliovirus recovered during serial passage in ribavirin. HeLa cells were infected with WT or mutant parental strains of poliovirus at an MOI of 0.1 PFU per cell, with three independent lineages per virus, and passaged 10 times (P1 to P10). Virus was incubated in 0 μM (P0), 100 μM (P1 to P4), 400 μM (P5 to P9), or 800 μM ribavirin (P10). Virus was harvested at 24 hpi, and the titer was determined by plaque assay at each passage. Dashed lines provide reference points for 100-fold decrease in titers compared to no drug at P0.

Polymerase mutations selected by serial passage in ribavirin.

We expanded the population of poliovirus for each lineage after passage 10 and sequenced viral cDNA to identify polymerase mutations fixed in the virus populations (Table 1). No polymerase mutations were selected in 12 of 24 lineages, including all lineages of WT and ribavirin-resistant G64S parental viruses. These data indicate that genetic bottlenecks were avoided—no mutations were detected in expanded passage 10 (P10^e) populations from WT or G64S parental viruses. In contrast, we identified revertants, pseudorevertants, or second site mutations fixed in expanded passage 10 (P10^e) virus populations in 9/9 lineages of ribavirin-sensitive parental viruses (Table 1). Ribavirin-sensitive parental viruses include L420A, G64^Fix L420A, and G64S L420A (18). Polymerase mutations were also detected in 1/3 lineages of G64^Fix, D79H, and L419A parental viruses, parental viruses with ribavirin sensitivity like that of the WT virus (Table 1). The precise nature of polymerase mutations varied from one ribavirin-sensitive parental virus to another. A variety of polymerase mutations were selected from L420A and G64^Fix L420A parental viruses (lineages 1 to 3) (Table 1), while G64S L420A parental viruses (lineages 1 to 3) had one singular outcome—reversion to L420 while maintaining the G64S resistance mutation (Table 1). These data indicate that distinct genetic backgrounds and codon potential played a role in the nature of selected ribavirin resistance mutations. When G64S was present alone (G64S parental virus) or within an otherwise ribavirin-sensitive parental population (G64S L420A parental virus), the only outcome was L420A reversion to WT while maintaining G64S alleles.

TABLE 1.

Polymerase mutations selected by serial passage in ribavirin

Parental virus phenotype	Lineage	P10e mutation(s)a
WT	1	None
	2	None
	3	None
G64S	1	None
	2	None
	3	None
G64Fix	1	None
	2	None
	3	3Dpol M392V and C→U6163
D79H	1	None
	2	3Dpol T353I
	3	None
L419A	1	None
	2	None
	3	3Dpol G64S
L420A	1	3Dpol L420V and M299I
	2	3Dpol L420V and 3Cpro A172S
	3	3Dpol L420I and M323I
G64Fix L420A	1	3Dpol L420 reversion, M323I, and C→U6163
	2	3Dpol L420 reversion
	3	3Dpol L420 reversion and G64S
G64S L420A	1	3Dpol L420 reversion
	2	3Dpol L420 reversion
	3	3Dpol L420 reversion

^aPoliovirus recovered from passage 10 (Fig. 2 and 3) was expanded by infecting a T150 flask of HeLa cells. The 3D^pol region of expanded P10 virus (P10^e) was converted to cDNA and sequenced.

In contrast, when G64S was absent from the parental virus or a codon mutation was used to inhibit G64S emergence, a variety of novel polymerase mutations were selected during serial passage in escalating concentrations of ribavirin. Thus, by using a variety of parental viruses, a variety of polymerase mutations were selected during serial passage in escalating concentrations of ribavirin. Altogether, polymerase mutations were selected in 12/24 lineages, including L420A revertants (L420), L420A pseudorevertants (L420V and L420I), G64S, and several new putative ribavirin resistance mutations (M299I, M323I, M392V, and T353I) (Table 1). It was surprising to find that G64S was selected infrequently and even more surprising to see it selected from a G64^Fix parental strain. As noted above, by passaging virus at an MOI of 0.1 PFU per cell, ribavirin-resistant mutants that arise in the population must outcompete parental strains to become fixed in the population. Under the conditions of our experiment, G64S mutations became fixed in the populations on occasion, but infrequently overall.

The polymerase mutations selected during serial passage in escalating concentrations of ribavirin were located at several sites within polymerase elongation complexes, but all were in close proximity to RNA or the active site (Fig. 4). G64S, a well-studied ribavirin resistance mutation (54, 55), is distal from the active site, near the back of the palm domain (Fig. 4). L420A revertants and pseudorevertants are present in an alpha helix in the thumb domain that packs into the minor groove of the double-stranded RNA (dsRNA) product as it exits the polymerase (Fig. 4). L420, and presumably pseudorevertants thereof (L420V and L420I), interact with the ribose of viral RNA products 3 bases from the active site (19). M392V is found between L420 and the primer grip in polymerase elongation complexes (Fig. 4). The location of the M392V ribavirin resistance mutation underpins several important insights from this investigation. T353I is in motif D at the back of the palm domain and near the nucleoside triphosphate (NTP) entry tunnel and G64 (Fig. 4). M299I is on the motif B helix and buried in the palm domain, not far from M323I that is found in the motif C beta sheet containing the active site YGDD motif (Fig. 4).

FIG 4.

Location of polymerase mutations within elongation complexes. The atomic structure of poliovirus polymerase elongation complexes (17) was used to highlight the locations of mutations. Polymerase is shown in gray, RNA template in cyan, RNA products in green, and locations of mutations in purple (G64, D79, Y275, and L419), dark green (M299, M323, and T353), orange (M392), and yellow (L420). Classic primer grip residues are shown in red (K375 and R376). L419 (purple) and L420 (yellow) residues are adjacent to one another at the base of a thumb alpha helix (blue) that interacts with the minor groove of the RNA product helix. Poliovirus polymerase PDB entry 3OL6 was rendered using PyMOL molecular graphics (Schrodinger, LLC).

Ribavirin sensitivity of parental and progeny virus.

We compared the titers of parental and progeny virus populations grown in the presence and absence of 600 μM ribavirin (Fig. 5). Expanded passage 10 (P10^e) viruses from each lineage were compared to their respective parental viruses. Titers of ∼10⁹ PFU per ml were obtained for all viruses when they were grown in the absence of ribavirin (Fig. 5). WT poliovirus, and the P10^e viruses derived from WT after serial passage in escalating concentrations of ribavirin, remained sensitive to inhibition by ribavirin, with titers below 10⁷ PFU per ml when grown in 600 μM ribavirin (Fig. 5, WT panel). G64S parental virus, and the P10^e viruses derived therefrom, retained resistance to ribavirin, with titers well above 10⁷ PFU per ml when grown in the presence of 600 μM ribavirin (Fig. 5, G64S panel). Expanded passage 10 (P10^e) virus from other parental strains exhibited a spectrum of sensitivities to ribavirin compared to those of WT and G64S viruses (Fig. 5). G64^Fix parental virus was similar to WT poliovirus; however, G64^Fix lineage 3 was more resistant to ribavirin (Fig. 5, G64^Fix panel), presumably due to the selected M392V mutation (Table 1). D79H parental virus was similar to WT poliovirus, as reported (34), yet D79H lineage 2 was modestly resistant to ribavirin (Fig. 5, D79H panel), presumably due to the selected T353I mutation (Table 1). L419A parental virus was similar to WT poliovirus, as reported (18, 19), yet L419A lineage 3 was more resistant to ribavirin (Fig. 5, L419A panel), no doubt due to the selected G64S mutation (Table 1). L420A parental virus was more sensitive to ribavirin than WT virus, as reported (18, 19); however, L420A lineage 1, 2, and 3 viruses grew to higher titers in the presence of ribavirin (Fig. 5, L420A panel), presumably due to the selected polymerase mutations in each lineage (Table 1). G64^Fix L420A parental virus was more sensitive to ribavirin than WT virus, as reported (18, 19); however, G64^Fix L420A lineage 1, 2 and 3 viruses grew to higher titers in the presence of ribavirin (Fig. 5, G64^Fix L420A panel), presumably due to the selected polymerase mutations in each lineage (Table 1). Similarly, ribavirin-sensitive G64S L420A parental virus and its selected progeny exhibited ribavirin phenotypes consistent with the genotypes of the respective virus populations (Fig. 5 and Table 1). Altogether, these data suggest that the polymerase mutations selected during serial passage in escalating concentrations of ribavirin (Table 1) correlate with measurable resistance to ribavirin (Fig. 5).

FIG 5.

Ribavirin sensitivity of parental and progeny viruses. Titers of parental and progeny viruses grown in the absence and presence of 600 μM ribavirin. Expanded passage 10 (P10^e) viruses from each lineage were used to infect HeLa cells at an MOI of 0.1 PFU per cell. Infected cells were incubated for 24 h in the absence or presence of 600 μM ribavirin. Virus was harvested by freeze-thaw, and the titer was determined by plaque assay on HeLa cells. The names of parental strains and lineage 1, 2, and 3 strains are indicated on the x axes. Polymerase mutations selected during serial passage and identified in Table 1 are annotated on the graphs accordingly (in red).

Virus derived from infectious cDNA clones.

The polymerase mutations identified during serial passage in escalating concentrations of ribavirin were reverse engineered into infectious cDNA clones of poliovirus to derive genetically defined virus populations (Table 2, ribavirin selected mutations). In addition, several engineered variants originating from our work, and that of others, were similarly generated (Table 2, engineered variants) (18, 19, 54, 56). Many of these mutations lie within an extended primer grip region of the polymerase, consisting of L420, M392, and primer grip residues K375 and R376. All the mutations listed here were stably maintained in infectious virus, except for alanine substitutions of the two charged primer grip residues (K375A and R376A) that reverted back to wild-type residues, precluding further evaluation. However, the charge-retaining K375R and R376K mutants at these sites were stable. Viruses containing polymerase mutations were assessed in one-step growth assays (Fig. 6), ribavirin dose-response assays (Fig. 7A), viral RNA recombination assays (Fig. 7B), and replication controls (Fig. 7C).

TABLE 2.

Virus derived from infectious cDNA

Polymerasea	Ribavirin selected	Engineered variant	Stably maintained	Reference
WT			Yes	74
G64S	✓		Yes	54
L420A		✓	Yes	19
G64S L420A		✓	Yes	18
L420V	✓		Yes	This study
L420I	✓		Yes	This study
M392A		✓	Yes	This study
M392L		✓	Yes	This study
M392V	✓		Yes	This study
K375R		✓	Yes	This study
R376K		✓	Yes	This study
K375A		✓	No	This study
R376A		✓	No	This study
M299I	✓		Yes	This study
M323I	✓		Yes	This study
T353I	✓		Yes	This study
Y275H		✓	Yes	56

^aPolymerase mutations identified during ribavirin selection, along with engineered variants, were cloned into infectious cDNA. Poliovirus was recovered from HeLa cells transfected with RNA derived from each clone. cDNA from the 3D^pol region was sequenced to determine whether mutations were stably maintained in virus.

FIG 6.

One-step growth of wild-type and mutant polioviruses. Polymerase mutations selected by serial passage in ribavirin were engineered into infectious cDNA clones. Viruses derived from infectious cDNA clones were compared under one-step growth conditions: HeLa cells were infected with wild-type or mutant poliovirus at an MOI of 10 PFU per cell. Virus was harvested at the indicated times by freeze-thawing cells. Titers were determined by plaque assay and plotted versus time (hpi, hours postinfection). Mean titers from triplicates were plotted with standard deviation error bars, although some error bars are too small to show.

FIG 7.

Polymerase mutations influence ribavirin-induced error catastrophe and sexual RNA replication mechanisms. (A) Polymerase mutations influence ribavirin sensitivity and resistance. HeLa cells were infected with wild-type or mutant poliovirus at an MOI of 0.1 PFU per cell and incubated for 24 h with the indicated concentrations of ribavirin. Each condition was performed in triplicates. After three freeze-thaw cycles, viral titers were determined by plaque assay and plotted versus ribavirin concentration. Three ribavirin-responsive phenotypic clusters were observed: a ribavirin-resistant cluster (blue), a wild-type cluster (gray and black), and a ribavirin-sensitive cluster (red). M392V (green) congregated with viruses in the WT cluster. Mean titers from triplicates were plotted with standard deviation error bars, although some error bars are too small to show. Holm-Sidak P values ranged from 0.01 (200 μM) to 0.000002 (1,000 μM) when comparing viruses in the WT cluster to those in the G64S or L420A clusters. (B) Polymerase mutations influence the frequency of viral RNA recombination mechanisms, i.e., sexual RNA replication. The Δcapsid “donor” and ΔGDD “recipient” RNAs were cotransfected into murine cells as previously established (18, 19). The Δcapsid donor contained wild-type or mutant polymerases as indicated on the x axis. The titer of virus produced in cotransfected murine cells was determined by plaque assays in HeLa cells. Mean titers from triplicates were plotted with standard deviation error bars. *, P < 0.05 compared to the wild type. (C) Replication controls showing titers from poliovirus RNA containing polymerase mutations (x axis) after transfection into murine cells. At 72 h posttransfection (hpt), the amount of poliovirus produced within the murine cells was determined by plaque assay. Mean titers from triplicates were plotted with standard deviation error bars.

One-step growth of wild-type and mutant polioviruses.

Before assessing ribavirin sensitivity and resistance, we measured one-step growth of wild-type and mutant polioviruses in HeLa cells (Fig. 6). HeLa cells were infected at an MOI of 10 PFU per cell, and virus yields were determined at 0, 1, 2, 3, 4, 6, 8, and 24 hpi (Fig. 6). Each of the viruses grew by 4 orders of magnitude between 3 and 24 hpi, reaching titers of ∼10⁹ PFU per ml. While some mutant viruses had noticeably lower titers than the wild type at 6 or 8 hpi, overall differences between WT and mutant viruses were not statistically significant. A Holm-Sidak comparison of each mutant to the wild type yielded the following P values for Y275H (0.5368), M299I (0.3686), M323I (0.6709), T353I (0.3425), K375R (0.1099), R376K (0.4372), M392A (0.7341), M392L (0.4467), M392V (0.4391), L420I (0.4590), and L420V (0.9703). These Holm-Sidak comparisons included all time points, effectively comparing one virus growth curve to another. In another statistical analysis, we compared WT and mutant virus titers at individual time points using unpaired t tests. Statistically significant differences (P values < 0.05) were not obtained in unpaired t tests, except for some spurious differences at 0, 1, and 2 hpi. The error bar on the WT virus titer at 8 hpi is larger than error bars at other time points, because one of three WT virus titers was unusually higher at 8 hpi, making this data point shift higher, with a larger error bar (Fig. 6). Because of this, differences in WT and mutant virus titers are noticeable at 8 hpi but are not statistically significant. These one-step growth curves do not rule out the possibility of subtle differences in replication rates between WT and mutant viruses; however, the wild-type and mutant viruses exhibited similar one-step growth phenotypes in highly permissive HeLa cells.

Ribavirin sensitivity and resistance.

Poliovirus-infected HeLa cells were used to assess ribavirin sensitivity and resistance (Fig. 7A). Wild-type poliovirus titers decreased incrementally as the dose of ribavirin increased, from titers of ∼10⁹ PFU per ml in the absence of ribavirin to titers of ∼2 × 10⁶ PFU per ml in 1,000 μM ribavirin. Consistent with other studies (57), we normalized the ribavirin dose-response data using the titers of each virus at 0 μM ribavirin (Fig. 7A). Wild-type poliovirus titers decreased by ∼1,000-fold as ribavirin treatment increased from 0 to 1,000 μM (Fig. 7A). Poliovirus with a G64S polymerase mutation resisted inhibition by ribavirin, as previously reported (18, 19, 54), with titers decreasing by ∼100-fold in 1,000 μM ribavirin (Fig. 7A). In contrast, poliovirus with an L420A polymerase mutation was more sensitive to inhibition by ribavirin, with titers decreasing by ∼10,000-fold in 1,000 μM ribavirin (Fig. 7A) (18, 19). Likewise, poliovirus containing both G64S and L420A mutations was more sensitive to ribavirin, with titers decreasing by ∼10,000-fold in 1,000 μM ribavirin (G64S L420A) (Fig. 7A). Thus, the ribavirin-resistant (G64S) and ribavirin-sensitive (L420A) controls behaved as previously reported (18, 19).

The viruses used in this investigation segregated into three ribavirin-responsive phenotypic groups (Fig. 7A): a ribavirin-resistant cluster (blue), a wild-type cluster (black), and a ribavirin-sensitive cluster (red). Holm-Sidak P values were significant at all ribavirin concentrations when comparing viruses in the WT cluster to those in the G64S or L420A clusters: P values ranging from 0.01 (200 μM) to 0.000002 (1,000 μM). Ribavirin-resistant polioviruses included those with G64S, M299I, M323I, and T353I (Fig. 7A, in blue). Although an M392V mutation was selected during serial passage in ribavirin, poliovirus with an M392V mutation was not as resistant to ribavirin as G64S: rather, it grouped with other viruses in the WT cluster (Fig. 7A, in green). Furthermore, M392V resistance to ribavirin was not statistically significant compared to that of WT poliovirus: P values of 0.262 (200 μM), 0.173 (400 μM), 0.393 (600 μM), and 0.173 (1,000 μM). Ribavirin-sensitive polioviruses included those with L420A, G64S L420A, R376K, and M392A polymerase mutations. Polioviruses in the wild-type cluster included WT and Y275H, K375R, M392V, M392L, L420I, and L420V mutations. These results indicate that the polymerase mutations selected during serial passage in escalating concentrations of ribavirin (G64S, L420V, L420I, M392V, M299I, M323I, and T353I) are indeed ribavirin resistance mutations, although one mutant (M392V) was not statistically different from WT. By comparison to the selected M392V mutation, an M392A variant was ribavirin sensitive and an M392L variant had wild-type sensitivity to ribavirin. L420A and G64S L420A parental strains were ribavirin-sensitive viruses, whereas L420V- and L420I-containing viruses exhibited wild-type sensitivity to ribavirin. Conservative substitutions in the primer grip exhibited divergent ribavirin phenotypes: poliovirus containing a K375R mutation had wild-type sensitivity to ribavirin, whereas poliovirus containing an R376K mutation was ribavirin sensitive. These data indicate that polymerase mutations influence ribavirin sensitivity and resistance, with some mutations rendering virus more resistant to ribavirin while others render virus more sensitive to ribavirin.

Viral RNA recombination.

Based on our previous investigation (18), we predicted that ribavirin-resistant poliovirus might acquire polymerase mutations that facilitate sexual RNA replication mechanisms as a means to avoid ribavirin-induced error catastrophe. Consequently, we assayed the impact of polymerase mutations on the frequency of replicative RNA recombination (Fig. 7B). In replicative RNA recombination assays, we cotransfect murine cells with two viral RNAs, each of which contains a lethal mutation: (i) a subgenomic replicon with an in-frame capsid deletion carrying the mutant polymerase, and (ii) a full-length poliovirus RNA with a lethal deletion of the active site GDD motif in the polymerase (18, 19). Wild-type or mutant polymerase is produced from the subgenomic replicon in the cotransfected cells. The amount of infectious poliovirus produced in the cotransfected murine cells corresponds with the frequency of viral RNA recombination. By comparing the titers of poliovirus, we assess the impact of polymerase mutations on the frequency of viral RNA recombination (Fig. 7B).

WT polymerase established wild-type levels of replicative RNA recombination in our experiments, with titers of poliovirus of ∼10⁵ PFU per ml (Fig. 7B). While Lowry et al. (45) reported that a G64S mutation inhibits viral RNA recombination, we consistently find that a G64S mutation has no impact on the frequency of viral RNA recombination under the conditions of our experiments (Fig. 7B). Technical differences in viral RNA recombination assays between the Barton and Evans labs may be responsible for these divergent outcomes, as previously reported (18, 19). In contrast, an L420A mutation inhibited replicative recombination by 2 orders of magnitude, with poliovirus titers of ∼10³ PFU per ml (Fig. 7B). Likewise, a G64S L420A polymerase supported decreased levels of replicative RNA recombination, with titers near 10² PFU per ml (Fig. 7B). These outcomes for WT, G64S, and L420A polymerases are similar to those previous reported (18): WT and G64S polymerases support wild-type magnitudes of viral RNA recombination, whereas an L420A polymerase mutation significantly inhibits replicative RNA recombination.

L420A pseudorevertants (L420V and L420I), selected during serial passage in escalating concentrations of ribavirin, restored replicative RNA recombination frequencies to wild-type levels (Fig. 7B). All the other polymerase mutations selected during serial passage in escalating concentrations of ribavirin maintained replicative RNA recombination frequencies at near-wild-type levels, albeit with some variation in titers both higher and lower than 10⁵ PFU per ml (M392V, M299I, M323I, M392V, and T353I) (Fig. 7B).

Engineered polymerase mutations had divergent ribavirin and recombination phenotypes: the M392A variant was ribavirin sensitive with wild-type levels of recombination, while the M392L variant exhibited wild-type ribavirin sensitivity and wild-type levels of recombination (Fig. 7). Likewise, the K375R variant exhibited wild-type ribavirin sensitivity and wild-type levels of recombination, whereas the R376K variant was ribavirin sensitive with wild-type levels of recombination (Fig. 7). Finally, consistent with data reported by Acevedo et al. (56), a Y275H mutation inhibited viral RNA recombination without impacting ribavirin sensitivity or resistance (Fig. 7A and B).

As a control for the viral RNA recombination assay, we examined the impact of polymerase mutations on virus replication in murine cells (Fig. 7C). Full-length infectious RNA containing each of the polymerase mutations was transfected into murine cells. Infectious poliovirus was recovered at 72 hpi by freeze-thaw and the titer was determined by plaque assay (Fig. 7C). These data show that the polymerase mutations do not inhibit virus replication within murine cells. Consequently, the decreased titers of virus recovered from viral RNA recombination assays of Y275H, L420A, and G64S L420A polymerases are due to defects in sexual RNA replication mechanisms (i.e., viral RNA recombination) rather than defects in asexual RNA replication mechanisms.

Altogether, these results indicate that ribavirin-resistant poliovirus selected during serial passage in escalating concentrations of ribavirin restored (or maintained) efficient viral RNA recombination mechanisms. Furthermore, polymerase with engineered amino acid substitution mutations at M392 and primer grip residues (K375 and R376) exhibited divergent ribavirin and recombination phenotypes, with some mutations increasing sensitivity to ribavirin without negatively impacting viral RNA recombination frequencies.

Biochemical phenotypes of wild-type and mutant polymerases.

Having examined viral RNA recombination and ribavirin sensitivity, we next examined the biochemical phenotypes of wild-type and mutant polymerases (Table 3). We used purified polymerase to assess the effects of individual mutations on various biochemical parameters, including RNA synthesis initiation, elongation complex stability, processive elongation rate, and single nucleotide addition cycle rate. The fidelity of nucleotide addition was indirectly assayed via a nucleotide discrimination factor (DF) that is based on the relative catalytic efficiency of rNTP versus 2′-dNTP incorporation (58). Previous studies established that the speed of polymerase elongation tends to correlate with the fidelity of RNA synthesis: as the rate of elongation increases, the fidelity of RNA synthesis decreases (7, 55, 58).

TABLE 3.

Biochemical phenotypes of purified polymerases

Polymerasea	Initiation (min)	ECb stability (min)	Processive rate (nt/s)	Processive Km (μM)	Single NTP rate (per s)	Single NTP Km (μM)	Discrimination factor
From Kempf et al.c
WT	6 ± 1	88 ± 9	22 ± 1	67 ± 2	30 ± 1	37 ± 2	120 ± 10
G64S	8 ± 1	24 ± 7	12 ± 1	46 ± 4	23 ± 1	39 ± 5	220 ± 40
L420A	8 ± 1	28 ± 4	30 ± 1	75 ± 3	40 ± 1	39 ± 3	100 ± 10
G64S L420A	6 ± 3	40 ± 8	26 ± 1	84 ± 9	28 ± 1	36 ± 6	220 ± 50
This study
WT	4 ± 1	130 ± 20	26 ± 1	72 ± 4	40 ± 1	46 ± 3	100 ± 10
L420V	4 ± 1	170 ± 10	29 ± 1	82 ± 5	41 ± 1	40 ± 3	99 ± 9
L420I	4 ± 1	60 ± 2	33 ± 1	90 ± 5	45 ± 1	41 ± 2	107 ± 8
M392A	3 ± 1	20 ± 1	30 ± 1	90 ± 8	33 ± 1	38 ± 3	120 ± 10
M392L	4 ± 1	39 ± 4	25 ± 1	74 ± 9	38 ± 1	34 ± 1	102 ± 6
M392V	4 ± 1	11 ± 1	22 ± 1	74 ± 7	42 ± 1	51 ± 3	160 ± 10
K375R	5 ± 1	51 ± 2	20 ± 1	96 ± 8	36 ± 1	44 ± 3	200 ± 20
R376K	2 ± 1	30 ± 3	65 ± 2	160 ± 10	48 ± 1	46 ± 3	17 ± 1
M299I	4 ± 1	180 ± 30	18 ± 1	66 ± 4	35 ± 1	42 ± 2	120 ± 10
M323I	4 ± 1	150 ± 20	18 ± 1	64 ± 4	37 ± 1	50 ± 3	130 ± 10
T353I	5 ± 1	90 ± 20	18 ± 1	73 ± 2	38 ± 1	49 ± 3	160 ± 10
Y275H	13 ± 6	95 ± 3	27 ± 1	95 ± 8	35 ± 1	42 ± 2	140 ± 10

^aPolymerases were expressed and purified, and biochemical phenotypes were determined as previously reported (59, 76).

^bEC, elongation complex.

^cFrom Kempf et al. (18), with permission.

In a previous study (18), we compared the biochemical phenotypes of WT, G64S, L420A, and G64S L420A polymerases (Table 3). A G64S mutation reduced the rate of RNA elongation in both single nucleotide (23 nucleotides [nt] per s for G64S versus 30 per s for WT) and processive elongation assays (12 nt/s for G64S versus 22 nt/s for WT), with a corresponding increase in the fidelity of RNA synthesis (DF of 220 ± 40 for G64S polymerase versus 120 ± 10 for WT). Consequently, G64S is considered a high-fidelity polymerase (54, 55). An L420A mutation decreased elongation complex stability, increased RNA elongation rates, and exhibited a wild-type discrimination factor of 100 ± 10. Thus, L420A polymerase is considered to have normal (wild-type) fidelity. Polymerase containing both G64S and L420A mutations had hybrid phenotypes, with faster elongation rates like those of L420A polymerase and an increased discrimination factor of 220 ± 50. Consequently, G64S L420A is considered a high-fidelity polymerase, like G64S polymerase. Consistent with our previous report (18), an L420A mutation inhibited viral RNA recombination (Fig. 7B) and rendered virus more susceptible to ribavirin-induced error catastrophe (Fig. 7A). Furthermore, the high-fidelity G64S L420A polymerase failed to render virus resistant to ribavirin when viral RNA recombination was substantially reduced (Fig. 7). These observations for WT, G64S, L420A, and G64S L420A polymerases provide context to interpret the biochemical data for the new polymerase mutants identified in this study (Table 3).

Compared to that of WT polymerase, the L420V, M299I, and T353I polymerases form more stable elongation complexes (Table 3). In contrast, mutations in the extended primer grip of the polymerase significantly reduced elongation complex stability (L420A, L420I, M392A, M392L, M392V, K375R, and R376K) (Table 3). The M392V mutation reduced elongation complex stability by 10-fold to 11 ± 1 min versus 130 ± 20 min for WT.

Primer grip mutations (K375R and R376K) exhibited divergent biochemical phenotypes (Table 3). The K375R mutation reduced the rate of RNA elongation in both single nucleotide (36 nt per s versus 40 nt per s for WT) and processive elongation assays (20 nt per s versus 26 nt per s for WT), with a corresponding increase in nucleotide discrimination (DF of 200 ± 20). In contrast, the R376K mutation increased the rate of RNA elongation in both single nucleotide (48 nt per s) and processive elongation assays (65 nt per s), with a dramatic decrease in the fidelity of RNA synthesis (DF of 17 ± 1 for R376K polymerase). The low-fidelity R376K polymerase rendered poliovirus more susceptible to ribavirin (Fig. 7A), as one might expect; however, the high-fidelity K375R polymerase exhibited normal (wild-type) sensitivity to ribavirin (Fig. 7A). The R376K polymerase also initiated RNA synthesis 2-fold faster than WT polymerase. Together, these data show that conservative substitution mutations in the primer grip (K375R and R376K) result in divergent biochemical phenotypes (Table 3) and divergent ribavirin sensitivity (Fig. 7A).

Polymerase mutations selected during serial passage in ribavirin tend to have reduced rates of RNA elongation and increased discrimination factors (G64S, M392V, M299I, M323I and T353I) (Table 3). The M299I, M323I, and T353I mutations reduced the rate of RNA elongation in processive elongation assays (18 nt per s for M299I, M323I, and T353I versus 26 nt per s for WT), with corresponding increases in nucleotide discrimination (discrimination factors of 120 ± 10 for M299I polymerase, 130 ± 10 for M323I polymerase, and 160 ± 10 for T353I polymerase). The M392V mutation reduced the rate of RNA elongation in processive elongation assays (22 nt per s for M392V versus 26 nt per s for WT) and increased the fidelity of RNA synthesis (discrimination factor of 160 ± 10 for M392V polymerase). Increases in the fidelity of RNA synthesis are consistent with increased resistance to ribavirin (G64S, M392V, M299I, M323I, and T353I) (Fig. 7A). Altogether, these data indicate that the polymerase mutations selected during serial passage in escalating concentrations of ribavirin tend to increase the fidelity of RNA synthesis and decrease rates of RNA elongation.

We included a Y275H mutation in this study based on the report from Acevedo et al. (56). Consistent with their data, we find that the Y275H mutation inhibits viral RNA recombination (Fig. 7B) without increasing sensitivity to ribavirin (Fig. 7A). These phenotypes are distinct from those associated with an L420A polymerase mutation, where decreased viral RNA recombination is mechanistically linked with increased sensitivity to ribavirin (L420A) (Fig. 7A and B). The biochemical phenotypes of Y275H are important in this regard (Table 3). The Y275H mutation inhibits the initiation of RNA synthesis by 3-fold (13 ± 6 min for Y275H versus 4 ± 1 min for WT). None of the other mutations in our panel exhibit this phenotype. The proximity of Y275 to the template entry channel of the polymerase suggests this defect in the initiation of RNA synthesis may arise from defects in template RNA binding (59, 60). In the discussion, we elaborate on the structural and functional distinctions of Y275 and L420 residues in the polymerase, especially as they relate to viral RNA recombination and ribavirin-induced error catastrophe.

DISCUSSION

In this study, we investigated the interplay between sexual RNA replication mechanisms and ribavirin-induced error catastrophe. Sexual RNA replication involves two parental templates, wherein mutant-free genomes can be reassembled from viral RNAs containing mutations (Fig. 1). An L420A polymerase mutation inhibits sexual RNA replication by ∼50-fold (Fig. 7B) coincident with increased susceptibility to ribavirin-induced error catastrophe (Fig. 7A). During serial passage in ribavirin, L420A revertants and pseudorevertants (L420I and L420V) regained efficient viral RNA recombination coincident with ribavirin resistance (Fig. 7). These data reinforce previous studies suggesting that viral RNA recombination counteracts error catastrophe (18). We conclude that sexual RNA replication mechanisms counteract ribavirin-induced error catastrophe by purging ribavirin-induced mutations from viral RNA genomes (Fig. 1).

The structural orientation of L420 interacting with primer-template duplexes provides insights into its role in sexual RNA replication mechanisms (Fig. 8). Within this region of the structure, residues K375 and R376 constitute the classic primer grip electrostatic interaction that brackets the phosphodiester bonds of the nascent RNA product one base from the active site. L420 is located on the thumb domain and expands these RNA interactions via a hydrophobic contact with the product strand ribose group three bases from the active site. M392, which mutated during our serial passage experiments, is found immediately between L420 and the classic primer grip. Altogether, residues L420, M392, and K375/R376 form an extended surface for direct interactions with the RNA product strand at the third, second, and first bases from the active site, respectively (Fig. 8). We refer to this structural element as an “extended primer grip.” These protein-RNA interactions facilitate both asexual and sexual RNA replication mechanisms by modulating the orientation and dynamics of the primer (i.e., product) RNA strand in the active site. Notably, the L420A mutation specifically disrupts sexual RNA replication mechanisms without inhibiting asexual RNA replication mechanisms (Fig. 1).

FIG 8.

An extended primer grip in the viral polymerase mediates sexual RNA replication mechanisms. Poliovirus polymerase L420 (yellow), M392 (orange), and classic primer grip residues K375 and R376 (red) constitute an “extended primer grip.” Altogether, these residues work coordinately to hold nascent RNA products on homologous RNA templates near the catalytic site of the polymerase. (A) Back view of extended primer grip with the active site YGDD residues (fuchsia), template (cyan), and product (green) RNAs. (B) Surface representation showing how extended primer grip residues form a continuous surface in direct contact with the 3′ terminal nucleotides of the RNA primer strand (in green). The template strand was omitted for clarity. Poliovirus polymerase PDB entry 3OL6 rendered using the PyMOL molecular graphics system (Schrodinger, LLC).

By using several ribavirin-sensitive and ribavirin-resistant parental viruses, serial passage in escalating doses of ribavirin led to the selection of both known and novel ribavirin resistance mutations (Table 1). These data reinforce and build upon prior studies regarding ribavirin-induced error catastrophe (18, 30, 52, 54). A G64S mutation was selected in two instances, reinforcing its well-established role as a modulator of RNA synthesis fidelity (54, 55). In addition, three novel ribavirin resistance mutations were identified: M299I, M323I, and T353I. Two important phenotypes were shared by all polymerases containing these ribavirin-resistant mutations; efficient viral RNA recombination and increased fidelity of RNA synthesis. Thus, both asexual (fidelity and nucleotide discrimination) and sexual (viral RNA recombination) RNA replication mechanisms influence ribavirin sensitivity and resistance.

L420A revertants and pseudorevertants.

Because an L420A polymerase mutation exacerbates ribavirin-induced error catastrophe coincident with defects in sexual RNA replication mechanisms (18), we predicted that ribavirin-resistant poliovirus selected from L420A parental strains would acquire polymerase mutations that facilitate sexual RNA replication mechanisms. Consistent with this prediction, L420A revertants or pseudorevertants (L420I and L420V) were selected in every lineage of virus containing an L420A mutation: 3/3 lineages of L420A parental virus, 3/3 lineages of G64^Fix L420A parental virus, and 3/3 lineages of G64S L420A parental virus (Table 1). Importantly, L420A revertants and the L420I and L420V pseudorevertants restored efficient sexual RNA replication mechanisms coincident with ribavirin resistance (Fig. 7). These data reinforce the correlation between efficient viral RNA recombination and resistance to ribavirin-induced error catastrophe (18).

Novel ribavirin resistance mutations.

Serial passage in escalating concentrations of ribavirin leads to the selection of ribavirin resistance polymerase mutations in both poliovirus and foot-and-mouth-disease virus (FMDV): a G64S mutation in the poliovirus polymerase (54) and an M296I mutation in the FMDV polymerase (61, 62). By using several ribavirin-sensitive and ribavirin-resistant parental viruses, we obtained both known and novel ribavirin resistance polymerase mutations following serial passage of poliovirus in escalating concentrations of ribavirin (Fig. 2 and Table 1). G64S mutations were selected twice, M323I was selected twice, and M299I, T353I, and M392V were each selected once. The M299I mutation in poliovirus is distinct from the ribavirin resistance M296I mutation selected in FMDV; the poliovirus M299 residue corresponds to FMDV residue I309 (61, 62). Our biochemical data indicate that polymerase mutations selected during serial passage in ribavirin tend to have reduced rates of RNA elongation and increased discrimination factors (Table 3). M299I, M323I, and T353I mutations reduced the rate of RNA elongation in processive elongation assays from 26 to ≈18 nt per s, with slight increases in the fidelity of RNA synthesis with DF values of 120, 130, and 160 for M299I, M323I, and T353I polymerases, respectively.

The M392V polymerase mutation selected during serial passage in ribavirin was not statistically resistant to ribavirin in poliovirus rederived from an infectious cDNA clone, although it was perhaps trending toward slight resistance (green in Fig. 7A). This mutation increased the polymerase discrimination factor (Table 3) without changing the frequency of viral RNA recombination (Fig. 7B), consistent with an impact on the fidelity of viral RNA synthesis. Curiously, a mutation at an analogous site in the EV-71 polymerase (M393L) is reported to resist the antiviral drug NITD008 (63). NITD008 is an adenine analogue, whereas ribavirin is a general purine analogue. Both NITD008 and ribavirin are prodrugs converted into triphosphate forms in vivo, where they function as NTP substrates in the active site of the polymerase. Deng and colleagues noted that the EV-71 M393 side chain interacts with the primer grip of the polymerase (63). They suspected that the M393L mutation influences the polymerase active site in a way that inhibits NITD008 antiviral activity. We find that an M392V mutation decreased the rate of RNA elongation coincident with an increase in the poliovirus polymerase discrimination factor (Table 3) without changing the frequency of viral RNA recombination (Fig. 7B). These data suggest that the M392V mutation can influence the rate of catalysis in the active site of the polymerase, consistent with the conclusions of Deng and colleagues (63).

Two important phenotypes were shared by all of the ribavirin resistance polymerase mutations in poliovirus: efficient viral RNA recombination (Fig. 7) and increased fidelity of RNA synthesis (Table 3). A G64S mutation renders poliovirus resistant to ribavirin-induced error catastrophe by increasing the fidelity of RNA synthesis (54, 55), but a G64S mutation is insufficient for ribavirin resistance when viral RNA recombination is disabled by an L420A mutation (18). Consequently, poliovirus with both G64S and L420A polymerase mutations is highly sensitive to inhibition by ribavirin (Fig. 7). Furthermore, when subjected to serial passage in ribavirin, the L420A mutation virus reverted in 3/3 lineages (Table 1), resulting in the classic ribavirin-resistant G64S poliovirus. While polymerase speed, fidelity, and efficiency of recombination are intricately intertwined (64, 65), our data suggest that polymerase residue L420 is required for efficient viral RNA recombination and for counteracting ribavirin-induced error catastrophe (18). Furthermore, even though polymerase speed and fidelity are intricately intertwined, recent studies suggest that polymerase speed is more important than fidelity for virulence (66). Thus, it is possible to attribute specific biochemical functions of the polymerase to important biological outcomes.

Remarkably, despite substantial selective pressure, neither WT parental strains nor G64S parental strains acquired any new polymerase mutations during serial passage in escalating concentrations of ribavirin (Fig. 2 and Table 1). We used an MOI of 0.1 PFU per cell for each serial passage in ribavirin to avoid genetic bottlenecks. Under these conditions, ribavirin-resistant variants only become fixed in the population if they outcompete the parental strain from which they arise. Serial passage of poliovirus at lower MOIs might increase the selective pressure of ribavirin, as decreased virus populations are more sensitive to inhibition by ribavirin (57). Because sexual RNA replication via viral RNA recombination requires two parental templates, decreased populations of virus should reduce the frequency of viral RNA recombination, effectively mimicking the effects of an L420A mutation.

Population size and genetic diversity among homologous partners in recombination are important to effectively overcome ribavirin-induced error catastrophe. It is intuitive that small populations (which are inherently less diverse) are likely to be more susceptible to ribavirin-induced error catastrophe. When the population size and genetic diversity decrease coordinately, it becomes less probable that fit recombinants can be reassembled from a limited and genetically debilitated (mutated) population of RNA templates. Graci et al. (57) showed that coxsackievirus B3 was more susceptible to ribavirin at a low MOI versus a high MOI, implicating population size as an important variable. Larger and more diverse populations of otherwise homologous virus within cells make it more probable that RNA recombination can reassemble fit genomes from ribavirin-mutated genomes.

An extended primer grip in the viral polymerase.

By using a codon mutation to inhibit the rapid emergence of the classic G64S mutation, we identified M392V as a novel ribavirin resistance mutation, although it provided statistically insignificant ribavirin resistance compared to that found in the G64S virus cluster (G64S, M299I, M323I, and T353I) (Fig. 7A). Structurally, M392 is found between L420 and the primer grip, and L420, M392, and K375/R376 interact directly with viral RNA product strand at the third, second, and first bases from the active site, respectively (Fig. 8). These protein-RNA interactions at the core of the polymerase facilitate both asexual and sexual RNA replication mechanisms. An L420A mutation specifically disrupts sexual RNA replication mechanisms without inhibiting asexual RNA replication mechanisms (Fig. 1) (18, 19), while K375A and R376A mutations impaired asexual RNA replication mechanisms to a sufficient degree that both mutations were unstable (Table 2). These results are consistent with the lethal effects of clustered charge-to-alanine substitution mutations at these locations in the polymerase (67). More conservative substitution mutations, K375R and R376K, were stably maintained in virus, and neither of these mutations disabled viral RNA recombination; however, they had divergent effects on ribavirin sensitivity, which the R376K mutation increased whereas the K375R mutation had little impact.

Due to the interesting location of the M392V mutation within the elongation complex, we engineered two additional mutations, M392A and M392L, at this residue. Both were stably maintained in virus, but divergent phenotypic effects were observed. Virus with the M392A mutation was highly susceptible to inhibition by ribavirin, virus with the M392L mutation had no impact on ribavirin sensitivity, and virus with the M392V mutation had statistically insignificant resistance to ribavirin. None of these M392 mutations had significant impacts on the magnitudes of viral RNA recombination. However, biochemical data provided key information: the M392V mutation decreased elongation rates coincident with increased nucleotide discrimination (Table 3). These data suggest that the M392V mutation impacts asexual RNA replication mechanisms by slowing the rates of elongation coincident with very slight increases in the fidelity of RNA synthesis.

Together, the data from the mutations at L420, M392, and the primer grip suggest that these residues form an extended primer-grip surface on the polymerase that senses the complementarity of the product duplex in the immediate vicinity of the active site (Fig. 8). Recombination involves resumption of elongation with a new primer-template pairing in the active site, and this extended primer grip would energetically favor a properly structured dsRNA helix in the priming region and thus facilitate recombination by selecting for proper base pairing at the critical step when elongation resumes on the new template. Consistent with this, the recombination mutations generally show reduced elongation complex stability that is indicative of impaired polymerase-RNA contacts.

Reconciling distinct phenotypes of Y275H and L420A mutations.

An L420A mutation renders poliovirus susceptible to ribavirin-induced error catastrophe coincident with defects in viral RNA recombination (18, 19). Consequently, we conclude that viral RNA recombination counteracts ribavirin-induced error catastrophe (18). In this study, we find that L420A revertants and pseudorevertants (L420I and L420V) restored efficient sexual RNA replication mechanisms coincident with increased resistance to ribavirin compared to that of L420A parental strains (Fig. 7). These data reinforce the correlation between efficient viral RNA recombination and resistance to ribavirin-induced error catastrophe (18). However, data from a Y275H mutation seem to be inconsistent with the conclusion that viral RNA recombination counteracts ribavirin-induced error catastrophe (Fig. 7). A Y275H mutation inhibited viral RNA recombination without rendering poliovirus more susceptible to ribavirin (Fig. 7A and B). These results are similar to those reported by Acevedo and colleagues (56). Importantly, the Y275 and L420 residues are structurally and functionally distinct, providing insights to help reconcile their alternate effects on ribavirin-induced error catastrophe. The Y275 residue is located in the fingers domain (Fig. 4), and the Y275H mutation is 4-fold slower for template binding and initiation (Table 3), suggesting a role in template binding; the Y275H elongation complex could not be purified in sufficient yield for crystallization studies. In addition, a recent enterovirus 71 elongation complex structure shows a downstream template strand base inserted into a surface pocket with Y275 at its bottom (Y276 in EV71 polymerase) (60). Importantly, this RNA binding role for Y275 is mechanistically upstream of L420’s role in discriminating between homologous and nonhomologous base pairing of the primer-template duplex near the active site. Thus, Y275 helps recruit RNA templates in a presumably sequence-independent manner, whereas L420 enforces sequence-dependent steps of homologous recombination. Further work will be required to better understand these two distinct steps of viral RNA recombination and their divergent effects on ribavirin sensitivity.

Picornavirus species groups and sexual RNA replication mechanisms.

The ancient origin of picornavirus RNA-dependent RNA polymerase suggests that viral RNA recombination has been a characteristic feature of viruses for billions of years (1, 3, 4). Thus, both asexual and sexual RNA replication mechanisms likely arose coordinately very early during virus evolution. Genetic exchange is a common event among modern picornaviruses and inevitable when two related viruses with shared RNA sequence homology coinfect cells at mucosal surfaces in the respiratory or gastrointestinal tracts (42, 43), a process facilitated by bacteria (44). The extended primer grip of the poliovirus polymerase highlighted by the work presented herein is conserved across picornavirus species groups and likely mediates genetic exchange between viruses in nature. Intraspecies recombination is well documented for enterovirus species A (40, 68), species B (39, 69), species C (37), and species D viruses (70, 71), as well as rhinovirus species groups (38, 72), parechoviruses (36), and nonhuman enteroviruses (42). Circulating vaccine-derived polioviruses (cVDPVs) arise, in part, due to improved fitness from genetic exchange with nonpolio group C enteroviruses in the field (47–51). Now that we appreciate the molecular and ecological conditions that contribute to genetic exchange between picornaviruses, we may be able to exploit recombination to combat human disease. Recombination-deficient oral poliovirus vaccines are one option to consider (73); however, new approaches to exploit viral RNA recombination deserve consideration.

Summary.

An extended primer grip in picornavirus polymerases contributes to the fidelity of RNA synthesis and to efficient sexual RNA replication mechanisms. This region of the polymerase interacts with nascent RNA products near the active site, effectively sensing the degree of RNA sequence complementarity between RNA products and RNA templates. With the potential to discriminate between homologous and nonhomologous RNA templates, this region of the polymerase mediates sexual RNA replication mechanisms, thereby counteracting error catastrophe and contributing to genetic exchange between related picornaviruses.

MATERIALS AND METHODS

Poliovirus and infectious cDNA clones.

Poliovirus type 1 (Mahoney) and mutant derivatives thereof were derived from an infectious cDNA clone (74). Poliovirus RNAs were produced by T7 transcription of MluI-linearized cDNA clones (Ampliscribe T7; Cellscript Inc.) and transfected into HeLa cells to make infectious virus as previously described (15, 18, 19).

Serial passage of poliovirus in escalating concentrations of ribavirin.

Poliovirus was grown in HeLa cells in the presence of escalating concentrations of ribavirin (Sigma-Aldrich) using methods modified from those of Pfeiffer and Kirkegaard (54). HeLa cells were plated in 35-mm 6-well dishes 24 h before infection with WT or mutant parental strains of poliovirus at an MOI of 0.1 PFU per cell and incubated in 0 μM (P0), 100 μM (P1 to P4), 400 μM (P5 to P9) or 800 μM ribavirin (P10). Virus was harvested at 24 hpi by three freeze-thaw cycles, and the titer was determined by plaque assay in HeLa cells. Three independent lineages of each parental strain were subjected to 10 serial passages in escalating concentrations of ribavirin. We expanded the population of poliovirus for each lineage after passage 10 by infecting a T150 flask of HeLa cells.

Viral cDNA was prepared from expanded passage 10 (P10^e) viruses from each lineage and sequenced to identify polymerase mutations fixed in the virus populations.

One-step growth of poliovirus.

HeLa cells were plated in 35 mm 6-well dishes 24 h before infection with wild-type and mutant polioviruses. An MOI of 10 PFU per cell was used for one-step growth conditions. After 1 h for virus adsorption, the inoculum was removed and the cells were incubated with 2 ml of culture medium at 37°C. Total virus was harvested by three freeze-thaw cycles at the designated times postinfection. Titers were determined by plaque assays.

Mean titers from triplicates were plotted versus time postinfection with standard deviation error bars. Statistical significance was assessed using the Holm-Sidak method of pairwise comparisons from GraphPad Prism (La Jolla, CA).

Ribavirin dose-response assays.

HeLa cells were plated in 35-mm 6-well dishes 24 h before infection with wild-type and mutant polioviruses. An MOI of 0.1 PFU per cell was used for ribavirin-dose response assays. After 1 h for virus adsorption, the inoculum was removed and the cells were incubated with 2 ml of culture medium at 37°C. Total virus was harvested by three freeze-thaw cycles at 24 h postinfection. Titers were determined by plaque assays.

Mean titers from triplicates were plotted with standard deviation error bars. Statistical significance was determined using the Holm-Sidak method of pairwise comparisons from GraphPad Prism (La Jolla, CA).

Viral RNA recombination and replication controls.

Viral RNA recombination assays and replication controls were performed in L929 cells as previously described (18, 19). For viral RNA recombination, L929 murine cells were cotransfected with two viral RNAs, each of which contained a lethal mutation: Δcapsid donor and 3D^pol ΔGDD recipient. The Δcapsid donor is a subgenomic replicon containing an in-frame deletion of VP2 and VP3 capsid gene sequences (Δ nucleotide positions 1175 to 2956) (19, 74). The 3D^pol ΔGDD recipient is a full-length poliovirus RNA containing a 9-base deletion in 3D^pol (Δ ₆₉₆₅GGU GAU GAU₆₉₇₃). Deleting three catalytic residues from the viral polymerase (ΔGDD) results in a noninfectious RNA replication-incompetent derivative of poliovirus (19). Wild-type and mutant derivatives of Δcapsid donor RNA were used, with the following 3D^pol substitution mutations: G64S, Y275H, G64S L420A, L420A, L420I, L420V, M392A, K375R, R476K, M299I, M323I, T353I, and ΔGDD. Mutations were engineered into Δcapsid donor cDNA clones as previously described (18, 19).

L929 cells were plated in 35-mm 6-well dishes ∼24 h before transfection, with ∼10⁶ cells per well. Two micrograms of viral RNA (1 μg each of Δcapsid donor and 3D^pol ΔGDD recipient) was transfected into each well in triplicates (i.e., three independent samples for every experimental condition) (TransMessenger transfection reagent; Qiagen). Following transfection, 2 ml of culture medium (Dulbecco modified Eagle medium containing 100 units of penicillin and 100 μg per ml of streptomycin, 10% fetal bovine serum, and 10 mM MgCl₂) was added to each well, and the cells were incubated at 37°C in 5% CO₂. Virus was harvested at 72 h posttransfection, recovered after three rounds of freezing and thawing, cleared of cellular debris by centrifugation at 3,000 rpm, and quantified by plaque assay.

RNA recombination is evident when infectious virus is recovered from cells cotransfected with two noninfectious RNAs. L929 murine cells, which naturally lack the poliovirus receptor, prevent multiple rounds of virus amplification beyond that within cotransfected cells (19, 45). Virus was harvested from cotransfected cells at 72 h posttransfection, an endpoint when cotransfected cells have produced as much recombinant virus as possible.

For replication controls, L929 cells were transfected with full-length infectious poliovirus RNAs containing wild-type or mutant polymerase, as indicated in the text and figures. The titer of infectious poliovirus recovered from the transfected cells was determined by plaque assay in HeLa cells.

Mean titers from triplicates were plotted with standard deviation error bars. Statistical significance was determined using Tukey’s test for single-step multiple comparisons from GraphPad Prism (La Jolla, CA).

Biochemical characterization of purified polymerase.

Biochemical characteristics of purified polymerase were examined as previously described (59, 75–77). Briefly, proteins were expressed in Escherichia coli and purified through metal affinity, ion exchange, and gel filtration chromatography. Initiation rates are based on the time needed to form a +1 product after mixing 5 μM polymerase with 0.5 μM “10 + 1–12 RNA” and 40 μM GTP in reaction buffer containing 50 mM NaCl, 4 mM MgCl₂, 25 mM HEPES (pH 6.5), and 2 mM Tris(2-carboxyethyl)phosphine hydrochloride (TCEP), all at room temperature. Elongation complex stability measurements are based on diluting a 15-min initiation reaction mixture 10-fold in the same buffer with 300 mM NaCl and then testing the amount of elongation-competent complex present at time points up to 4 h and fitting the resulting data to a single exponential decay function. Kinetics assays were performed using rapid stopped-flow fluorescence methods in 75 mM NaCl, 50 mM HEPES (pH 7), and MgCl₂ in 4 mM excess over total NTP concentration. Processive elongation rates were determined with 26-nt-long template RNA bearing a 5′-fluorescein end label (75), and the single cycle data were obtained with an RNA whose fluorescence reports on translocation of a template strand 2-aminopurine base from the +2 to +1 site following incorporation of a single CMP or 2′-dCMP (77). The discrimination factor is the ratio of catalytic efficiencies of CTP and dCTP incorporation reactions, i.e., (k_pol/K_m)_CTP/(k_pol/K_m)_dCTP.