|Home | About | Journals | Submit | Contact Us | Français|
RNA recombination is important in the formation of picornavirus species groups and the ongoing evolution of viruses within species groups. In this study, we examined the structure and function of poliovirus polymerase, 3Dpol, as it relates to RNA recombination. Recombination occurs when nascent RNA products exchange one viral RNA template for another during RNA replication. Because recombination is a natural aspect of picornavirus replication, we hypothesized that some features of 3Dpol may exist, in part, to facilitate RNA recombination. Furthermore, we reasoned that alanine substitution mutations that disrupt 3Dpol-RNA interactions within the polymerase elongation complex might increase and/or decrease the magnitudes of recombination. We found that an L420A mutation in 3Dpol decreased the frequency of RNA recombination, whereas alanine substitutions at other sites in 3Dpol increased the frequency of recombination. The 3Dpol Leu420 side chain interacts with a ribose in the nascent RNA product 3 nucleotides from the active site of the polymerase. Notably, the L420A mutation that reduced recombination also rendered the virus more susceptible to inhibition by ribavirin, coincident with the accumulation of ribavirin-induced G→A and C→U mutations in viral RNA. We conclude that 3Dpol Leu420 is critically important for RNA recombination and that RNA recombination contributes to ribavirin resistance.
IMPORTANCE Recombination contributes to the formation of picornavirus species groups and the emergence of circulating vaccine-derived polioviruses (cVDPVs). The recombinant viruses that arise in nature are occasionally more fit than either parental strain, especially when the two partners in recombination are closely related, i.e., members of characteristic species groups, such as enterovirus species groups A to H or rhinovirus species groups A to C. Our study shows that RNA recombination requires conserved features of the viral polymerase. Furthermore, a polymerase mutation that disables recombination renders the virus more susceptible to the antiviral drug ribavirin, suggesting that recombination contributes to ribavirin resistance. Elucidating the molecular mechanisms of RNA replication and recombination may help mankind achieve and maintain poliovirus eradication.
Picornaviruses are widespread in nature. Based on their ancient origins (1) and shared molecular features (2), picornaviruses are taxonomically organized by order, family, genus, species, and virus. Viruses in the family Picornaviridae are classified within 29 genera and 50 species groups. The genus Enterovirus, with 12 species groups (enterovirus species A to J and rhinovirus species A to C), contains the vast majority of medically important picornaviruses, including the polioviruses (PV), coxsackieviruses, enteroviruses, echoviruses, and rhinoviruses. These viruses, which possess single-stranded positive-sense RNA genomes, replicate via RNA intermediates in the cytoplasm of infected cells. A virus-encoded RNA-dependent RNA polymerase, 3Dpol, is responsible for RNA replication (3).
Viral RNA recombination was important in picornavirus speciation and remains important in the ongoing evolution of viruses within species groups (4, 5). Enterovirus and rhinovirus species groups are dependent upon, and defined by, ongoing recombination among viruses within each species group (6,–12). The capsid genes and nonstructural genes of viruses within each species group evolve separately due to the distinct functions and selective pressures associated with these gene cassettes (8). Recombinant viruses arise most frequently from intraspecies recombination (13), although interspecies recombination occurs, as well (14). Moreover, recombination dynamics reveal subgroups within species (15). Recombinant enterovirus strains arise frequently in nature, so individual serotypes of virus circulating at any one time are often genetically distinct (16, 17). Distinct lineages of recombinant circulating vaccine-derived polioviruses (cVDPVs) arise annually (18,–20), creating an obstacle to the eradication of polioviruses.
The mechanisms by which recombinant picornavirus RNA genomes arise are only partially understood. Viral RNA recombination can occur via two distinct mechanisms, replicative (21, 22) and nonreplicative (23, 24) recombination. Replicative RNA recombination, which is more efficient than nonreplicative recombination under tissue culture conditions, is thought to be the predominant mechanism of ongoing recombination among enteroviruses and rhinoviruses in nature. Replicative RNA recombination occurs when nascent RNA products exchange one viral RNA template for another during RNA replication (25). RNA structures, among other features of viral RNA, influence the frequency and location of recombination (26). The frequency of poliovirus RNA recombination can be measured using two noninfectious viral RNA templates (donor and recipient) and murine cells, which lack the cellular receptor for poliovirus (21). Recombination between noninfectious donor and recipient RNAs produces infectious virus, which is subsequently quantified by plaque assay on human cells (21). 3Dpol and/or nascent RNA products have the inherent ability to exchange one template for another in the absence of other viral proteins (27); however, specific features of the polymerase have not yet been implicated in RNA recombination.
Our investigation of RNA recombination is underpinned by the atomic structure of 3Dpol elongation complexes (28, 29) and a series of genetically stable alanine substitution mutations in 3Dpol (30). We hypothesized that phylogenetically conserved features of 3Dpol may exist to mediate RNA recombination. Furthermore, we reasoned that alanine substitution mutations that disrupt 3Dpol-RNA interactions within elongation complexes might increase or decrease the magnitude of RNA recombination. Consequently, we characterized a series of alanine substitution mutations in 3Dpol, focusing on conserved residues in the thumb domain that interact directly with viral RNA template and product strands. We found that one residue, 3Dpol Leu420, was critically important for efficient RNA recombination. We discuss how structural features of 3Dpol mediate distinct aspects of viral RNA replication and consider how 3Dpol Leu420 might facilitate replicative RNA recombination.
Poliovirus type 1 (Mahoney) and cDNA clones thereof were used for this study (30,–32). A full-length cDNA clone was used to produce infectious poliovirus RNA (Fig. 1, Poliovirus). A 9-base deletion in 3Dpol (6965GGU GAU GAU6973) was engineered using QuikChange XL-II (Stratagene, Inc.) to make a noninfectious, RNA replication-incompetent derivative of poliovirus (Fig. 1, 3Dpol ΔGDD). An in-frame deletion of VP2 and VP3 capsid gene sequences (nucleotides [nt] 1175 to 2956), referred to as RNA2 in previous studies (33), was used to make an RNA replication-competent subgenomic replicon (Fig. 1, ΔCapsid). An RNA replication-incompetent derivative of ΔCapsid was engineered by deleting internal ribosome entry site (IRES) nucleotides 285 to 679 (Fig. 1, ΔCapsid-ΔIRES). 3Dpol substitution mutations (G64S, K126A, R128A, K133A, K276A, N409A, D412A, S416A, L419A, and L420A) were engineered into full-length infectious poliovirus and ΔCapsid cDNA clones as previously described (30).
Viral RNAs were produced by T7 transcription of MluI-linearized cDNA clones (Ampliscribe T7 high-yield transcription kit; Cellscript Inc.). Cells were transfected as previously described (30, 31). HeLa, L929, and 3T3 cells were plated in 35-mm 6-well dishes ~24 h before transfection, with ~106 cells per well. Two micrograms of viral RNA was transfected into each well in triplicate (i.e., three independent samples for every experimental condition). 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 (30). The mean titers from triplicates were plotted, with standard deviations (error bars). Statistical significance in RNA recombination experiments was obtained using an unpaired two-tailed Student's t test with GraphPad (La Jolla, CA) Prism software.
HeLa cells were plated in 35-mm 6-well dishes ~24 h before infection, with ~106 cells per well. The cells were infected with poliovirus at a multiplicity of infection (MOI) of 0.1 PFU per cell. Following 1 h of virus adsorption, the inocula were removed and the cells were incubated for 24 h at 37°C in 2 ml of medium with or without 600 μM ribavirin (Rb) (Sigma-Aldrich). The cells were then subjected to three rounds of freeze-thaw, the supernatant was cleared of cellular debris, and the titers of viruses were determined by plaque assay. The mean titer from triplicates was plotted, with standard deviations.
TOPO-TA cloning and cDNA sequencing of virion RNA were described previously (30). Briefly, poliovirus was purified from 8 ml of culture medium by centrifugation through 30% (wt/vol) sucrose cushions, virion RNA was isolated by phenol-chloroform-isoamyl alcohol extraction, and cDNA was synthesized using Superscript III (InVitrogen) and a primer complementary to nucleotides 7415 to 7440 of the poliovirus 3′ nontranslated region (NTR) (5′ 7440CTCCGAATTAAAGAAAAATTTACCCC7415 3′). cDNA corresponding to the 3CD region of poliovirus RNA was amplified by PCR with high-fidelity Phusion DNA polymerase (New England BioLabs). The PCR products were analyzed by agarose gel electrophoresis and TOPO-TA cloned (InVitrogen) according to the manufacturer's instructions. Plasmids containing poliovirus cDNA were extracted using a QIAPrep Spin Miniprep kit (Qiagen) and sequenced in the University of Colorado Cancer Center DNA Sequencing Core Laboratory. Poliovirus sequences (1,500 bases; poliovirus nt 5836 to 7336) were obtained from 30 TOPO-TA clones to yield 45,000 cumulative bases of sequence for each experimental condition/data point. P values were obtained based on two-tailed chi-square tests with Yates correction of a 2-by-2 contingency table using GraphPad QuickCalcs.
Wild-type (WT) and mutant polioviruses were passed 20 times in HeLa cells via limiting-dilution serial passage, with three independent lineages per virus. Limiting-dilution serial passage of poliovirus was achieved by examining the cytopathic effect (CPE) in 96-well plates containing HeLa cells. Limiting dilution was evident when CPE was observed in less than 15% of infected wells. After 20 serial bottlenecks via limiting-dilution serial passage, virus populations were expanded, purified, and converted into cDNA using Superscript III (InVitrogen) and three primers [oligo(dT) and primers complementary to the PV RNA genome at nucleotides 2802 to 2828 and 5995 to 6023]. A full-length genome was sequenced from three independent lineages per virus. The number of mutations per genome was plotted, with standard deviations.
Polioviruses containing 3Dpol mutations were mixed 10:1 with wild-type poliovirus and used to infect 107 HeLa cells at an MOI of 10 PFU per cell. The infections were allowed to proceed until complete CPE was observed. Virus was harvested, and the titer was determined by plaque assay. Passages 1 through 5 were performed at an MOI of 10 PFU per cell. Virion RNA was isolated from each passage, converted into cDNA, and PCR amplified as described above. PCR products were cloned by TOPO-TA cloning as described above. Ten clones from each passage were sequenced or screened by restriction digestion to determine the proportion of each virus (3Dpol mutant versus wild type).
In vitro biochemical assays were used to measure the initiation rate, Km, Vmax, CTP/dCTP discrimination index, and elongation complex stability of wild-type and mutant polymerases. The methods were previously described by Kortus et al. (31).
Based on other studies (21), we designed infectious and noninfectious viral RNAs to distinguish between replicative and nonreplicative RNA recombination (Fig. 1A). Murine cells, which naturally lack the poliovirus receptor, were used to avoid multiple rounds of virus amplification beyond that within transfected cells. Poliovirus RNA produced more than 106 PFU per ml when transfected into L929 cells (Fig. 1B). Deleting three amino acids from the 3Dpol active site rendered poliovirus RNA noninfectious and RNA replication incompetent (Fig. 1, 3Dpol ΔGDD). A deletion within the capsid genes rendered poliovirus RNA noninfectious (Fig. 1, Δcapsid) without inhibiting RNA replication (33). A deletion in the IRES was used to disable RNA replication in the Δcapsid construct (Fig. 1, Δcapsid-ΔIRES).
RNA recombination was evident when infectious virus arose from two noninfectious RNAs within cotransfected L929 cells (Fig. 1). Replicative RNA recombination was detected when 3Dpol ΔGDD and Δcapsid RNAs were cotransfected, yielding ~103 PFU per ml (Fig. 1B). Nonreplicative RNA recombination was detected when RNA replication-incompetent RNAs were cotransfected (Fig. 1B, 3Dpol ΔGDD + Δcapsid-ΔIRES). Replicative RNA recombination was ~100-fold more efficient than nonreplicative RNA recombination under the conditions of these experiments (Fig. 1) (103 PFU per ml for replicative recombination versus 10 PFU per ml for nonreplicative recombination). Consistent with established conventions (21) and the RNA replication competence/incompetence of noninfectious RNAs, Δcapsid RNA was considered the donor during replicative RNA recombination, whereas 3Dpol ΔGDD RNA was considered the recipient (Fig. 1A).
Previously, we characterized a series of alanine substitution mutations in the finger and thumb domains of 3Dpol (Fig. 2A) (30). Alanine substitution mutations at the residues shown in red are stably maintained in poliovirus, with nominal effects on the specific infectivity of RNA, plaque size, or one-step growth phenotypes in HeLa cells (30). Alanine substitution mutations in the finger domains target arginine and lysine residues that interact with the phosphodiester backbone of viral RNA templates and products (Fig. 2A, K126, R128, K133, and K276), whereas mutations in the thumb domain are present in an alpha-helix that lines the minor groove of the double-stranded RNA (dsRNA) product (Fig. 2A, N409A, D412A, S416A, L419A, and L420A). The sequence and structural orientation of these 3Dpol residues are extremely well conserved among enteroviruses, rhinoviruses, and other picornaviruses (28, 30, 34). The active-site GDD residues in the palm domain are highlighted (Fig. 2A, magenta).
To study the impacts of 3Dpol mutations on replicative RNA recombination, we engineered mutations into the 3Dpol gene of Δcapsid donor and cotransfected cells with Δcapsid donor and ΔGDD recipient RNAs (Fig. 2B). A G64S mutation in 3Dpol was included in the study due to its previously described roles in ribavirin resistance (35,–38) and RNA recombination (21). We examined the impacts of 3Dpol mutations on virus replication (Fig. 2C) and RNA recombination (Fig. 2D) in murine L929 cells. Wild-type poliovirus RNA produced more than 106 PFU per ml when transfected into L929 cells (Fig. 2C, WT). Consistent with previous studies in HeLa cells (21, 30), G64S, K126A, R128A, K133A, K276A, N409A, D412A, S416A, L419A, and L420A 3Dpol mutations had nominal effects on the magnitude of poliovirus replication in L929 cells, with titers consistently above 106 PFU per ml (Fig. 2C). These results indicate that the 3Dpol mutations do not inhibit poliovirus replication in L929 cells.
RNA recombination was measured by the production of infectious virus in L929 cells cotransfected with Δcapsid donor and 3Dpol ΔGDD recipient RNAs (Fig. 2B and andD).D). A Δcapsid donor with wild-type 3Dpol produced ~104 PFU per ml of infectious virus when cotransfected with a 3Dpol ΔGDD recipient (Fig. 2D, WT). The L420A 3Dpol mutation reduced RNA recombination by 58-fold in L929 cells compared to wild-type 3Dpol (Fig. 2D and Table 1). In contrast, substitution mutations at other locations in 3Dpol failed to inhibit RNA recombination to this degree (Table 1, G64S, K126A, R128A, K133A, K276A, N409A, D412A, S416A, and L419A). In fact, several alanine substitution mutations in 3Dpol resulted in slight increases in RNA recombination (Fig. 2D and Tables 1 and and2).2). Increased magnitudes of RNA recombination with statistical significance were evident for K276A, D412A, and S416A 3Dpol mutations (Tables 1 and and2).2). Increased magnitudes of RNA recombination near statistical significance were evident for N409 and L419 3Dpol mutations (Tables 1 and and2).2). It is important to emphasize that the increased magnitudes of RNA recombination due to K276A, D412A, and S416A 3Dpol mutations were quite small in L929 cells (1.5- to 3.5-fold increases) in comparison to the dramatic 58-fold decrease in RNA recombination due to the L420A 3Dpol mutation. Thus, the most compelling finding in L929 cells was the dramatic decrease in RNA recombination associated with the L420A 3Dpol mutation.
Next, we sought to further validate the poliovirus RNA recombination effects in another murine cell line. The evidence of RNA recombination in L929 cells (Fig. 1 and and22 and Tables 1 and and2)2) was recapitulated in 3T3 cells (Fig. 3 and Table 1). 3T3 cells supported the replication of poliovirus (Fig. 3A, Poliovirus), whereas lethal mutations in viral RNA prevented virus replication (Fig. 3A, 3Dpol ΔGDD, ΔCapsid, and ΔCapsid-ΔIRES). Cotransfection of two noninfectious RNAs led to the production of infectious virus: greater than 103 PFU per ml for replicative RNA recombination and ~10 PFU per ml for nonreplicative RNA recombination (Fig. 3A). As shown in L929 cells, 3Dpol mutations impacted the frequency of RNA recombination in 3T3 cells (Fig. 3B and Tables 1 and and3).3). R128A and L420A 3Dpol mutations consistently decreased the magnitude of RNA recombination, whereas K276A, N409A, D412A, S416A, and L419A mutations consistently increased the magnitude of RNA recombination compared to wild-type 3Dpol (Fig. 2D and and3B3B and Table 1).
Together, the data from L929 and 3T3 cells show that alanine substitution mutations in 3Dpol affect the frequency of RNA recombination in murine cells (Table 1). The L420A mutation in 3Dpol had the greatest effect, a 58-fold decrease in L929 cells and a 16-fold decrease in 3T3 cells. Differences in the magnitudes of effects of 3Dpol mutations between L929 and 3T3 cells are likely due to experimental variation rather than meaningful biological differences in the two cell lines. Minor changes in the wild-type (benchmark) titer, in particular, can skew the fold change numbers from one experiment to the next. Regardless, the data reveal consistent effects, with some 3Dpol mutations increasing the frequency of RNA recombination (K276A, N409A, D412A, S416A, and L419A) while others decrease the magnitude of RNA recombination (R128A and L420A).
In order to more comprehensively understand 3Dpol structure-function and the phenotypic effects of the L420A mutation in particular, we examined ribavirin sensitivity (Fig. 4) and the fidelity of RNA replication (Table 4). These experiments were performed using poliovirus-infected HeLa cells, consistent with other studies (37, 38). As previously reported (30), the 3Dpol mutations under investigation here do not inhibit virus replication in HeLa cells (Fig. 4A). Some of the 3Dpol mutations under investigation here do, however, affect ribavirin sensitivity (Fig. 4B). Previous investigations established that wild-type poliovirus is sensitive to inhibition by ribavirin whereas a G64S 3Dpol mutation renders poliovirus resistant to ribavirin (37, 38), consistent with the WT and G64S poliovirus controls used here (Fig. 4). D412A and S416A 3Dpol mutations neither increase nor decrease sensitivity to ribavirin compared to WT poliovirus (Fig. 4B). In contrast, alanine substitution mutations at other locations in 3Dpol increased ribavirin sensitivity, with the largest increases in ribavirin sensitivity due to R128A and L420A 3Dpol mutations (Fig. 4B).
Because the antiviral effect of ribavirin correlates directly with its mutagenic activity (35, 36), we examined the fidelity of RNA replication in the absence and presence of ribavirin (±Rb) (Table 4). Consistent with other studies (35), ribavirin treatment resulted in increased numbers of mutations in viral RNAs (Table 4; P value < 0.0001 comparing ±Rb conditions for each virus). Wild-type poliovirus RNA had 1.2 mutations per genome in the absence of ribavirin and 17.7 mutations per genome in the presence of ribavirin. The G64S 3Dpol mutation increased the fidelity of viral RNA replication, with 0.3 mutations per genome in the absence of ribavirin and 8.3 mutations per genome in the presence of ribavirin. Poliovirus containing the S416A 3Dpol mutation had 0 mutations per genome in the absence of ribavirin and 15.7 mutations per genome in the presence of ribavirin (Table 4). Poliovirus containing the L420A 3Dpol mutation had 1.3 mutations per genome in the absence of ribavirin and 21.0 mutations per genome in the presence of ribavirin (Table 4).
As expected, the frequency of ribavirin-induced mutations in viral RNA (Table 4) correlated with the magnitude of ribavirin sensitivity (Fig. 4B). The G64S mutation increased the fidelity of viral RNA replication in both the absence and presence of ribavirin (Table 4) and correlated with ~100-fold-higher titers of virus in the presence of ribavirin (Fig. 4B). Poliovirus with the S416A mutation acquired 15.7 mutations per genome in the presence of ribavirin (Table 4) and was inhibited by ribavirin to the same degree as wild-type virus (Fig. 4B). The L420A 3Dpol mutation increased poliovirus sensitivity to ribavirin (Fig. 4B) coincident with an increased accumulation of ribavirin-induced mutations in viral RNA (Table 4) (21.0 mutations per genome for L420A; P = 0.21 versus the wild type; P = 0.036 versus S416A). Importantly, the L420A mutation did not substantially affect the fidelity of viral RNA replication in the absence of ribavirin compared to wild-type virus (Table 4) (1.2 mutations per genome for the wild type and 1.3 mutations per genome for L420A). These data indicate that ribavirin-induced mutations accumulate to slightly higher magnitudes when RNA recombination is disabled.
Two approaches were used to examine how 3Dpol mutations affect the fidelity of RNA synthesis and RNA replication: biochemical assays of purified polymerase (Table 5) and a biological assay involving serial bottleneck passages in conjunction with cDNA sequencing (Fig. 5). Biochemical assays address the fidelity of RNA synthesis in vitro, whereas serial bottleneck passage and cDNA sequencing address the fidelity of RNA replication in cells. We chose to pursue a more detailed examination of L420A and adjacent thumb alpha-helix mutations in this series of experiments due to their structural proximity in the polymerase (alanine substitutions in a discrete thumb alpha-helix) and their functional impact on RNA recombination (S416A increased the frequency of RNA recombination, whereas L420A decreased the frequency of RNA recombination).
As shown in Table 5, we measured the initiation rate, Km, Vmax, CTP/dCTP discrimination index, and elongation complex stability of the wild-type and mutant polymerases (31). As previously reported by others (35), a G64S mutation in 3Dpol substantially increased the fidelity of RNA synthesis by lowering the rate of elongation (Table 5, G64S versus WT). This increased fidelity resulted in a higher CTP-versus-dCTP discrimination factor (307 versus 117 for the wild type) in our in vitro fidelity assay, which reflects catalytic sensitivity to a mispositioned ribose 2′-hydroxyl group, as would be found with a mismatched nucleotide triphosphate in the active site (39). Elongation rates and the fidelity of RNA synthesis were largely unaffected by the alanine substitution mutations in the thumb alpha-helix, although S416A exhibited slightly higher fidelity than other thumb alpha-helix mutants (Table 5, N409A, D412A, S416A, L419A, and L420A versus WT). In contrast, all of the alanine substitution mutations in the thumb alpha-helix had negative effects on the stability of 3Dpol elongation complexes, with the largest decreases due to the N409A, S416A, and L420A mutations. These data show that alanine substitution mutations in the thumb alpha-helix tend to decrease the stability of 3Dpol elongation complexes; however, none of these altered biochemical phenotypes correlate precisely with increases or decreases in the frequency of RNA recombination. Furthermore, the ribavirin sensitivity of L420A poliovirus does not correspond to changes in the fidelity of RNA synthesis. The fidelities of L419A and L420A polymerases are indistinguishable in vitro (Table 5), yet L420A poliovirus is substantially more susceptible to ribavirin in vivo (Fig. 4B).
Limiting-dilution serial passage is another way to examine the fidelity of viral RNA replication (Fig. 5). Wild-type and mutant polioviruses were passed 20 times in HeLa cells via limiting-dilution serial passage, with three independent lineages per virus. Limiting dilution produces a genetic bottleneck that fixes neutral and nonlethal mutations into viral populations via a founder effect. After 20 serial bottlenecks, we sequenced poliovirus cDNA from each lineage (nucleotides 1 to 7440) to detect the frequency of mutations in the genome. Wild-type poliovirus accumulated 8 to 10 mutations per genome after 20 serial bottlenecks. Virus with a high-fidelity G64S mutation accumulated fewer mutations under these conditions, 3 or 4 mutations per genome (Fig. 5A and andB).B). Virus with S416A or L420A mutations, mutations that increase and decrease frequencies of RNA recombination, respectively, accumulated mutations at magnitudes similar to those of wild-type virus (Fig. 5A and andB).B). Thus, under these conditions, the frequency of recombination does not appear to affect the overall fidelity of RNA replication in the absence of a mutagenic nucleotide. Furthermore, these data indicate that S416A and L420A polymerases exhibit nearly wild-type fidelity of RNA replication in vivo.
As we elaborate on in Discussion below, these data suggest that L420A poliovirus is more susceptible to ribavirin due to decreases in the frequency of replicative RNA recombination rather than changes in the fidelity of RNA synthesis (Table 5 and Fig. 4B and and5).5). This possibility is theoretically appealing, as recombination exists, in part, to purge deleterious mutations from viral RNA (40).
Here, we examined the phenotypic effects of 3Dpol mutations on viral fitness (Fig. 6). The fitness of one poliovirus relative to another is easily measured by serial passage in coinfected HeLa cells (35, 41). 3Dpol R128A, K276A, D412A, and L420A mutations rendered poliovirus less fit than wild-type virus. Wild-type poliovirus overwhelmed these mutants after five serial passages despite being underrepresented by 10:1 in the initial infection. K126A and N409A 3Dpol mutations were less debilitating, with only transient decreases within the population compared to the wild type. K133A, S416A, and L419A 3Dpol mutants were dominant in the virus population through all five passages and thus relatively fit compared to the wild type and other 3Dpol mutants. It is odd to see one virus rise and then wane in abundance during passage in a competition assay, as seen for the WT virus in the K126A and N409A competition data (Fig. 6). In an independent competition assay, using low-MOI coinfections, we did not see WT virus rise and then wane in abundance relative to K126A and N409A viruses (data not shown). We chose to include the data from the high-MOI experiment, as it is generally representative of the outcomes under both low- and high-MOI conditions. R128A and L420A 3Dpol mutations consistently reduced the competitive fitness of virus under both low- and high-MOI serial-passage conditions. Thus, the L420A mutation decreased poliovirus fitness in coinfected HeLa cells.
3Dpol is intimately associated with viral RNA replication, RNA recombination, and the biology of picornavirus species groups. In this study, we identified phylogenetically conserved residues in 3Dpol that influence the efficiency of viral RNA recombination. Some 3Dpol mutations increased the frequency of RNA recombination (K276A, N409A, D412A, S416A, and L419A), while others decreased the magnitude of RNA recombination (R128A and L420A). The L420A 3Dpol mutation was notable because it dramatically inhibited RNA recombination without affecting other vital aspects of viral RNA synthesis and virus replication. The conservation of 3Dpol Leu420 in enteroviruses, rhinoviruses, and other picornaviruses (30), in conjunction with our experimental data (Fig. 2 and and3),3), suggests that this feature of 3Dpol is fundamentally important for the mechanisms of replicative RNA recombination, as well as the formation and maintenance of picornavirus species groups.
Some of the alanine substitutions in our study (S416A, L419A, and L420A) involved changes in the sequence of an RNA element within the 3D gene (42), but the RNA sequences within the 3D gene are unlikely to affect the frequency of replicative RNA recombination. Two of these alanine substitutions (S416A and L419A) exhibited increased RNA recombination, whereas the third (L420A) exhibited dramatically reduced RNA recombination. The role of 3Dpol in replicative RNA recombination is generally accepted (21, 25), while the functions of RNA elements in the 3D gene remain ambiguous (42, 43). Consequently, we attribute phenotypic changes in the frequency of RNA recombination to the alanine substitutions in 3Dpol rather than to changes in RNA sequences encoding 3Dpol.
The structural features of 3Dpol elongation complexes (Fig. 2A) provide a conceptual basis to understand how nascent RNA products can abandon one template for another during RNA recombination, and we reasoned that alanine substitution mutations that disrupt 3Dpol-RNA interactions within elongation complexes might increase or decrease the magnitude of RNA recombination (28,–31, 34, 44). 3Dpol Leu420 is located at the end of a thumb domain alpha-helix that packs into the minor groove of dsRNA adjacent to the active site of 3Dpol (Fig. 2A). An L420A mutation inhibited RNA recombination, whereas other alanine substitutions located further up the RNA-interacting face of the alpha-helix increased RNA recombination, albeit only slightly (i.e., N409A, D412A, S416A, and L419A). Alanine substitution mutations at R128 and K133 that disrupt charge-charge interactions with the phosphodiester backbone of the product strand from the opposite side of the emerging RNA duplex (Fig. 2A) had less impact on the magnitude of RNA recombination.
The structural orientations of Leu419 and Leu420 within 3Dpol elongation complexes provide important insights regarding RNA replication and RNA recombination (Fig. 7). These two mutations have comparable effects on reiterative transcription and polyadenylation of viral RNA, increasing the size of poly(A) tails in viral RNA populations by ~20% (30), but divergent effects on RNA recombination (Fig. 2D and and3B3B and Table 1). The L419A mutation increased the frequency of RNA recombination slightly, whereas the L420A mutation dramatically decreased RNA recombination. Leucines 419 and 420 form similar but distinct hydrophobic contacts with the ribose C-4′ carbons on both RNA strands (Fig. 7); Leu419 contacts the template strand at the −6 position, while Leu420 contacts the product strand at the −3 position, where the −1 position corresponds to the priming base pair located in the active site. Biochemical assays of 3Dpol initiation rates, elongation rates, and elongation complex stabilities show that the L419A mutation retains essentially wild-type behavior. In contrast, the L420A mutation shows slow initiation in an assay that is rate limited by RNA binding, and it forms unstable elongation complexes that fall apart 3-fold faster than the wild type (Table 5). Both these observations suggest L420A weakens the interactions with the RNA product strand, implying that they are important for recombination efficiency. For further comparison, the nearby D412A and S416A mutations similarly lower initiation rates and form unstable complexes, but these mutations do not show any changes in recombination frequency. Structurally, these polar amino acids are inserted into the middle of the minor groove, where they mediate the positioning of bound water molecules at the protein-RNA interface (Fig. 7). The reduced initiation rates observed for these mutants are likely due to a reduction in RNA binding via indirect hydration effects that do not favor one RNA strand over the other.
Recombination is a process whereby a new RNA genome is created by more than one template strand. Replicative RNA recombination involves template switching during RNA synthesis (Fig. 2B); however, the mechanisms by which template switching occurs are largely undefined. A number of distinct mechanisms can be postulated when considering the composition and structure of 3Dpol elongation complexes (Fig. 2A): (i) the nascent product RNA dissociates from the 3Dpol elongation complex, randomly hybridizes via complementarity to a new template, and recruits a new 3Dpol, and RNA synthesis resumes; (ii) the RNA template dissociates, leaving a 3Dpol-product RNA complex that finds a new template; (iii) the product dissociates, leaving a 3Dpol-template complex that associates with a new partial product strand. As explained below, we favor the first possibility mentioned above.
The L420A mutation has been observed to decrease recombination frequency 16- to 60-fold, and it would thus be predicted to either increase elongation complex stability, resulting in fewer abortive replication products, or decrease the resumption of RNA synthesis on the new template, reducing the ability to rescue abortive products via extension on a new template. The in vitro biochemical data showing that L420A destabilizes 3Dpol elongation complexes (Table 5) is not consistent with an increased-processivity model. Instead, the reduced initiation rate of the L420A mutation is consistent with a model where Leu420 is important for the formation of new 3Dpol elongation complexes after an abortive RNA product abandons one template for another. Several structural features of 3Dpol elongation complexes make this model the most plausible. First, the extensive contacts made between 3Dpol and 10 to 12 nucleotides of the template strand make it unlikely that a 3Dpol-product complex will abandon one template for another, whereas the smaller set of contacts between 3Dpol and the nascent RNA product strand favor the likelihood that the nascent RNA can abandon one elongation complex for another. Second, the 3Dpol Leu420 contact with the product RNA strand is only 3 nucleotides away from the active site, and the interaction could be important for properly positioning the RNA product with the new template into the active site. Third, 3Dpol initiation normally involves a viral VPg protein primer and proteolytic activation of 3Dpol by cleavage of the 3CD protease (3CDpro) precursor. In contrast, recombination uses an RNA primer for what is likely a rare and rate-limiting event whose efficiency is critically dependent on the alignment and structural details of the 3Dpol-RNA complex. With these insights in mind, future experiments should be designed to elucidate the molecular details of how the nascent RNA product abandons one template for another during RNA recombination. A better understanding of translocation (45,–47), the process whereby the RNA template and product move away from the active site following catalysis, should also enhance our ability to predict the structural rearrangements involved in RNA recombination. 3Dpol oligomers (48,–50) and the architecture of RNA replication complexes (51) might be important in the mechanisms of RNA recombination, as well.
Ribavirin provokes error catastrophe during virus replication by increasing the accumulation of G→A and C→U transition mutations in viral RNA (36, 37). Among the more striking effects of the L420A 3Dpol mutation are its negative effect on the frequency of RNA recombination (Fig. 2D and and3B3B and Table 1) and its profound increase in sensitivity to ribavirin (Fig. 4). We think these phenotypic effects are functionally related because RNA recombination can purge accumulating errors from viral RNA genomes (40). Thus, viral RNA polymerases have the capacity to introduce errors into viral RNA genomes during RNA replication, along with an inherent capacity to remove errors from genomes via RNA recombination. An L420A 3Dpol mutation increased poliovirus sensitivity to ribavirin (Fig. 4B) coincident with increased accumulation of G→A and C→U transition mutations in viral RNA (Table 4). Importantly, in the absence of ribavirin, the L420A mutation did not substantially affect the fidelity of viral RNA synthesis in vitro (Table 5) or the fidelity of RNA replication in cells (Fig. 5). These data suggest that RNA recombination counteracts the antiviral activity of ribavirin and that ribavirin-induced mutations in viral RNA can accumulate to higher magnitudes when RNA recombination is disabled. Our data comparing the frequency of ribavirin-induced mutations in viral RNA are consistent with this possibility, as the L420A virus accumulated more ribavirin-induced mutations per genome than WT, G64S, and S416A viruses (Table 4). In contrast, we found that a G64S mutation in 3Dpol mediates resistance to ribavirin without affecting the frequency of RNA recombination. Rather, the G64S mutation reduces the rate of RNA elongation, thereby increasing the fidelity of RNA synthesis in both the absence and presence of ribavirin (Table 5 and Fig. 4 and and5)5) (35). Coincident with its effects on RNA recombination and ribavirin sensitivity, the L420A mutation reduced the competitive fitness of poliovirus in HeLa cells (Fig. 6). Together, these observations suggest that viral polymerases have the inherent capacity to balance the fidelity of RNA replication with the frequency of RNA recombination to maintain the overall fitness of the virus population.
Circle sequencing (cirseq) deep-sequencing methods can be used in the future to more thoroughly characterize the manner in which 3Dpol mutations and RNA recombination impact the fidelity of viral RNA replication (52, 53). Cirseq methods would be especially helpful when 45,000 bases of TOPO-TA cDNA reveal limited numbers of mutations per genome (Table 4). For instance, TOPO-TA cDNA cloning and sequencing revealed no mutations in S416A poliovirus virion RNA in the absence of ribavirin (Table 4), yet mutations accumulated in S416A genomes during serial bottleneck transmission (Fig. 5A), and CTP/dCTP discrimination assays indicated that S416A polymerase exhibits (in)fidelity similar to that of wild-type polymerase (Table 5). Thus, the absence of mutations in S416A poliovirus virion RNA when grown in the absence of ribavirin, as determined by TOPO-TA cDNA cloning and sequencing (Table 4), is perplexing. The modestly elevated frequency of recombination associated with the S416A mutation (Fig. 2 and and33 and Table 1) may increase the fidelity of RNA replication in the absence of ribavirin (Table 4). Cirseq deep-sequencing methods would be useful in further evaluation of the S416A mutation. In previous work (30), we found that the S416A mutation altered the size distribution of poly(A) tails in virion RNA. Considering the absolute conservation of S416 in all enteroviruses and rhinoviruses (30), it is notable that S416A poliovirus is relatively fit in HeLa cells (Fig. 6).
In light of our data and the interpretations above, it is noteworthy that the Andino laboratory recently described a 3Dpol mutation (D79H) that inhibits RNA recombination without any effect on ribavirin sensitivity (40). Furthermore, the Evans laboratory reported that ribavirin can increase or decrease the frequency of replicative RNA recombination, depending on the concentration of ribavirin (21). In addition, they reported that a G64S mutation in 3Dpol reduced the frequency of RNA recombination as much as 20-fold (21), contrary to our results for the G64S mutation (Fig. 2 and and33 and Table 1). Finally, based on theoretical considerations, we expected that 3Dpol mutations that increase (S416A) or decrease (L420A) the RNA recombination frequency would influence the accumulation of mutations during serial bottleneck transmission. However, we found that they had no significant effect on the accumulation of mutations in viral RNA during serial bottleneck transmission compared to wild-type virus (Fig. 5). With these observations in mind, it is clear that additional experimentation will be required to fully understand how the frequency of RNA recombination influences the fidelity of RNA replication, ribavirin sensitivity, and virus evolution.
RNA recombination is important in the formation of picornavirus species groups and the ongoing evolution of viruses within species groups (4, 5). In this study, we identified a conserved feature of the picornavirus polymerase-RNA molecular interface that is required for efficient RNA recombination.
This work was supported by Public Health Service grants from the National Institutes of Health (AI059130 to O.B.P. and AI042189 to D.J.B.).
We thank Courtney Springer for the contribution of data.