West Nile virus (WNV) genome RNAs with up to three adjacent mutations that disrupt long distance 5′-3′ cyclization sequence basepairs are viable

Abstract

Mosquito-borne flavivirus genomes contain conserved 5′ and 3′ cyclization sequences (CYC) that facilitate long distance RNA-RNA interactions. In previous studies, flavivirus replicon RNA replication was completely inhibited by single or multiple mismatching CYC nt substitutions. In the present study, full-length WNV genomes with one, two or three mismatching CYC substitutions showed reduced replication efficiencies but were viable and generated revertants with increased replication efficiency. Several different three adjacent mismatching CYC substitution mutant RNAs were rescued by a second site mutation that created an additional base pair (nts 147-10913) on the internal genomic side of the 5′-3′ CYC. The finding that full-length genomes with up to three mismatching CYC mutations are viable and can be rescued by a single nt spontaneous mutation indicates that more than three adjacent CYC basepair substitutions would be required to increase the safety of vaccine genomes by creating mismatches in inter-genomic recombinants.

Introduction

West Nile virus (WNV) is a member of the family Flaviviridae, genus Flavivirus. WNV and other members of this genus cause significant human morbidity and mortality across broad geographical regions of the world. The WNV genome is an ∼11 kb single-stranded, positive-sense RNA with a 5′ untranslated region (UTR) of 96 nucleotides (nts) and a 3′UTR that varies from 337 to 649 nts. The genome RNA has a 5′ type 1 cap (Cleaves and Dubin, 1979), terminates at the 3′ end with 5′-CU_OH-3′ (Wengler and Wengler, 1981), and encodes a single open reading frame. The viral proteins: capsid (C), pre-membrane (prM), envelope (E), nonstructural protein (NS) 1, NS2A, NS2B, NS3, NS4A, NS4B, and NS5 are co- and post-translationally processed from the viral polyprotein by viral and cellular proteases (Lindenbach et al., 2007).

Short conserved sequences (CS)1, CS2 and repeated CS2 (RCS2) are present in the 3′ UTRs of all mosquito-borne flavivirus genomes (Hahn et al., 1987; Lindenbach et al., 2007). Interaction between a conserved 8 nt sequence within the 3′ CS1 and an exact complement in the capsid coding region near the 5′ end of flavivirus genomes was predicted to cyclize the genome (Hahn et al., 1987). Tick-borne flavivirus genomes cyclize through the interaction of a different set of 5′ and 3′ conserved sequences (Kofler et al., 2006). Additional terminal genomic sequences on either side of the polyprotein AUG initiation codon (nts 97-99) were subsequently shown to also be involved in long distance 5′-3′ RNA interactions. The 5′ upstream AUG region (UAR) interacts with a complementary 3′ UAR sequence downstream of the 3′ CYC (Alvarez et al., 2005b; Zhang et al., 2008). A functional role for the 5′-3′ UAR interaction was reported by Alvarez et al. (2008). One (dengue 2) or two (WNV) 5′ regions downstream of the AUG codon (DAR) (Dong et al., 2008; Friebe and Harris, 2010) also have been shown to interact with the complementary 3′ DAR sequence(s) located between the 3′ UAR and 3′ CYC sequences (see Fig. 1A). Interaction between 5′ and 3′ flavivirus RNA fragments was demonstrated by atomic force microscopy (Alvarez et al., 2005b), structure probing (Dong et al., 2008; Polacek et al., 2009), and electrophoretic mobility shift assays (Alvarez et al., 2005b; Zhang et al., 2008). RNA fragments with mutations that disrupted CYC basepairing were not able to form the 5′-3′ CYC or UAR interactions, while mismatching UAR mutations disrupted the UAR interaction but had no effect on the CYC interaction (Polacek et al., 2009). It has been proposed that the flavivirus genomic 5′-3′ RNA-RNA interaction initiates between the 5′CS and 3′CS, then the interaction of the 5′-3′ DAR sequences extends the initial long distance interaction, and this assists the interaction of the UAR elements which unwind the bottom of the terminal 3′SL (Friebe et al., 2010).

Figure 1. Effect of single mismatching nt substitutions in either the 5′ or 3′ WNV CYC sequence on virus replication.

(A) Schematic representation of the WNV genomic terminal secondary structures and 5′-3′ RNA-RNA interactions. (B) The 8 conserved 5′ (nts 137-144) and 3′ (nts 10916-10923) CYC nts are shown in bold. Flanking basepairs extend the 5′-3′ interaction in this region to 11 basepairs. Capsid amino acids (aa V-N-M-L) encoded by the 5′ CYC nts are indicated above the nt sequence. (C) The substituted mismatching CYC nts in each mutant sequence are indicated in bold and underlined. The size of the plagues produced by each mutant on an overlaid transfection well is indicated. Plaque diameters are: PP- pinpoint (∼< 1 mm), S- small (∼1.5 mm), M- medium (∼2.5 mm), and L- large (∼3.5 mm). (D) Kinetics of virus production by BHK cells transfected with a mutant RNA. Culture fluids harvested at the indicated times after transfection with 1 μg of viral RNA were titrated for infectivity by plaque assay. Each value shown is the average of duplicate titrations from two independent experiments. Error bars indicate ± standard deviations SD (n = 4). (E) Relative quantification of intracellular WNV genomic RNA by real-time qRT-PCR. Genomic RNA levels detected at 48 and 72 h after transfection are expressed as log₁₀ fold change in relative quantification units compared to the level of input viral RNA present 6 h after transfection. The data for each RNA sample was normalized to cellular GAPDH mRNA in the same sample. Error bars represent ± SD (n = 3). (F) Percentages of plaques with different phenotypes at various times after transfection. BHK cells (80% confluent) in six well plates were transfected with 1 μg of mutant viral RNA. Culture fluids harvested at the indicated times after transfection were titrated by plaque assay. The numbers of plaques of each size were counted and expressed as a percent of the total number of plaques.

Multiple previous studies with flavivirus replicons demonstrated that the 5′-3′ CYC interaction facilitates viral RNA replication but is not required for genome translation (Alvarez et al., 2005a; Corver et al., 2003; Khromykh et al., 2001; Kofler et al., 2006; Lo et al., 2003). Replicons with CYC nt mutations that were predicted to disrupt one or more CYC basepairs were unable to replicate (Khromykh et al., 2001; Lo et al., 2003). The only exception was mutant m2-5′ described by Suzuki et al., (2008) in which the U₁₃₇-A₁₀₉₂₃ basepair was mismatched by a U137→A substitution. In addition, mutant replicons with multiple substituted alternative CYC basepairs either replicated poorly (Alvarez et al., 2005a; Khromykh et al., 2001; Lo et al., 2003) or were replication incompetent (Suzuki et al., 2008). However, the effect of “flipping” a single CYC basepair on replicons varied with their position (Suzuki et al., 2008). “Flipping” the basepair at or next to the genome terminal side of the 5′- 3′ CYC interaction was lethal, while “flipping” single basepairs in the central region had little or no effect. “Flipping” basepairs on the internal genome side reduced replication efficiency to various degrees. Based on the high degree of conservation of the CYC sequences among mosquito-borne flaviviruses and the demonstrated functional requirement for the 5′- 3′ CYC interaction for viral RNA replication, Suzuki et al. (2008) proposed that alternative CYC basepairs at two nonadjacent and functionally neutral central positions be engineered into vaccine genomes to reduce the viability of inter-genomic recombinants between wild and vaccine genomes by creating mismatched CYC sequences. Although Suzuki et al. (2008) observed reversion of one or both of the mismatched CYC nts when they passaged a packaged replicon with the two nonadjacent mismatches, no previous studies have systematically analyzed the functional consequences or reversion frequencies of single or multiple mismatched basepairs at different positions in the CYC sequences in a full length infectious genome in the context of a complete virus life cycle.

In the present study, the viability of WNV infectious clone RNAs with single or multiple mismatching substitutions at different positions in the 5′ or 3′ CYC nts was assessed. Mutant genomes with one, two or three mismatches were viable but these mutations negatively affected virus replication to varying degrees. However, five mismatching mutations were lethal. The data suggested that terminal and flanking basepairs on both sides of the CYC region play important roles in the initiation and/or establishment of the 5′-3′ long distance RNA-RNA interaction. While all single and double mismatching substitutions reverted during passage of progeny virus, a number of different combinations of three adjacent CYC mismatching substitutions were rescued by a second site mutation that created an additional 5′-3′ basepair on the internal genomic side of the CYC region. The finding that up to three CYC mismatching mutations are not lethal for full-length genomes indicates that designing vaccine genomes with only two CYC substituted basepairs as previously suggested by (Suzuki et al., 2008) would not be a valid strategy for improving the safety of live vaccine candidates. However, data from the present study indicate that vaccine genomes with five adjacent substituted CYC basepairs would yield significantly attenuated spontaneous recombinants.

Materials and Methods

Cells

Baby hamster kidney-21/WI2 cells (hereafter referred to as BHK cells) (Vaheri et al., 1965) were maintained in Eagle's minimum essential media (MEM) supplemented with 5% heat-inactivated fetal bovine serum (FBS) and 10 μg/ml gentamycin (Invitrogen) at 37° C in 4.5% CO₂.

Site-directed mutagenesis

The construction of the WNV W956 infectious clone in a pBR322 vector was previously described (Yamshchikov et al., 2001). Two shuttle vectors were made to facilitate mutation of 5′ and 3′ CYC nts. The 5′ terminal genomic sequence shuttle vector, pWNV-Trunc I, was constructed by first digesting WNV W956 cDNA with restriction enzymes MluI and XbaI to remove the 3′ and middle portions of the viral cDNA. The plasmid DNA with the remaining WNV 918 5′ nts was then ligated to a foreign sequence (an 800 nt MluI/XbaI DNA fragment from the mouse Oas1b gene). The 3′ terminal sequence shuttle vector, pWNV-Trunc II, was generated by first digesting the WNV W956 cDNA with SphI and XbaI to obtain the 3243 nt 3′ terminal fragment of the viral cDNA. This fragment was then subcloned into the pGEM^®-3Zf(+) vector (Promega). Mutations were introduced into the CYC sequences in the shuttle vectors using the Quickchange II site-directed mutagenesis kit (Stratagene) according to the manufacturer's protocol. To generate mutant infectious clones, substituted pWNV-Trunc I cDNAs were digested with MluI and Xba1 or substituted pWNV-Trunc II cDNAs were digested with SphI and XbaI and the respective gel-purified mutant WNV cDNA fragment was ligated into an appropriately digested parental infectious clone cDNA. To create infectious clones with mutations in both the 5′ and 3′ CYCs, the 3′ and 5′ regions were sequentially replaced with mutated fragments. The sequences in the appropriate regions of each mutant shuttle vector as well as of each mutant infectious clone were checked by DNA sequencing. Two or three cDNA clones for each CYC mutant were separately transcribed into viral RNA and used for replicate cell transfections.

Transfection of in vitro transcribed viral RNA and assay of progeny virus replication

Parental or mutated infectious clone DNA was linearized by digestion with XbaI and then purified using a Qiaquick PCR Purification kit (Qiagen). Capped viral RNA was transcribed in vitro using an SP6 mMessage mMachine High Yield Capped RNA Transcription kit (Ambion) according to the manufacturer's protocol. BHK cell monolayers (∼80% confluent) in six-well plates were transfected with viral RNA as described previously (Elghonemy et al., 2005). Briefly, one well of a six-well plate was transfected with 0.1 μg and two wells were each transfected with 1 μg of genomic RNA mixed with Opti-MEM (Invitrogen) and DMRIE-C (Invitrogen). After a 2 h incubation at 37° C, the transfection media was removed and the monolayers were transfected with viral RNA. One well transfected with 0.1 μg and one well transfected with 1 μg of genomic RNA were overlaid with a 1:1 (vol/vol) mixture of 1% Seakem ME agarose (BioWhittaker Molecular Applications) and 2X MEM containing 5% FCS. At 72 h after transfection, the agarose plugs were removed and plaques were visualized by staining the cells with 0.05% crystal violet in 10% ethanol. Two mls of MEM containing 5% FCS were added to the remaining well (transfected with 1 μg of genomic RNA). Virus infectivity titers in media harvested at 72 h after transfection from non-overlaid wells were determined by plaque assay on BHK cells as previously described (Elghonemy et al., 2005).

Serial passage of progeny virus

Media was harvested from non-overlaid wells at 72 h after transfection with 1 μg of genomic RNA and 0.1 ml was used to infect fresh BHK monolayers in six-well plates. At each successive virus passage, 0.1 ml of media was transferred to a fresh monolayer in a six-well plate 72 hr after infection.

Sequence analysis of viral RNA

Viral RNA was extracted from picked plaques with TRI reagent LS (Molecular Research Center, Inc.) according to the manufacturer's protocol. A cDNA copy of the desired region of the viral RNA was then amplified by RT-PCR and cloned into pTOPO-TA 2.1 DNA (Invitogen). The DNA from 10 clones was sequenced for each viral RNA sample.

Analysis of the kinetics of virus production after transfection or infection

Duplicate wells of BHK cells in a six-well plate were transfected with 1 μg of viral RNA for 2 h and then the transfection media was replaced with fresh MEM containing 5% FCS. At 30, 48, 56, 60 and 96 h after transfection, 0.3 ml aliquots of media were harvested and viral infectivity titers were determined by plaque assay.

The growth kinetics of progeny virus were assessed as described previously (Elghonemy et al., 2005). Briefly, duplicate confluent BHK monolayers in T25 flasks were infected at a MOI of 0.1. The inoculum was removed after a 1 h adsorption, the monolayers were washed 3 times and fresh MEM medium was added. Aliquots (0.3 ml) of culture fluid were harvested at 1, 8, 24, 32, 48, 56 and 72 h after infection and virus infectivity was assessed by plaque assay.

Analysis of intracellular viral RNA levels by quantitative real-time RT-PCR

BHK monolayers in six-well plates were washed once with Opti-MEM and then transfected with 200 ng of viral RNA. At various times after transfection, some wells were washed three times with 5% FCS MEM and total intracellular RNA was extracted using TRI reagent. The relative amount of intracellular viral genomic RNA was determined by real-time RT-PCR using NS1 region primers and a TaqMan One-Step RT-PCR kit (Applied Biosystems) and an Applied Biosystems 7500 real-time PCR system as previously described (Davis et al., 2007). The mRNA of the housekeeping gene, glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (Applied Biosystems) was used as the endogenous control and was assayed in each sample using Taqman rodent GAPDH control reagent primers and probe (Applied biosystem). The data were analyzed using the relative quantification software from Applied Biosystems. Intracellular genomic RNA levels at 48 and 72 h after transfection were expressed as the fold change compared to the level of viral RNA present at 6 h after transfection.

RNA binding assay

Interaction between 5′ and 3′ viral RNA fragments was assessed using an electromobility shift assay described by Alvarez et al., (2005). The radiolabeled 3′ parental RNA probe was in vitro transcribed from a PCR product corresponding to the WNV 3′ terminal 110 nts in the presence of [α-³²P] GTP (3000 ci/mmol) using a Maxiscript T7 kit according to the manufacturer's protocol (Ambion). After 4 h of transcription, the DNA template was digested by DNase 1 and the RNA product was purified on a 5% polyacrylamide-urea gel. After ethanol precipitation, the RNA was suspended in RNase-free water. The radioactivity of the product RNA was measured in a Beckman LS6500 scintillation counter and the specific activity of the probe (∼1.3 × 10⁷ cpm/μg) was calculated as previously described (Blackwell and Brinton, 1997). Unlabeled parental and mutant WNV 5′ terminal 200 nt RNAs consisting of the 96 nt 5′ UTR plus the first 104 nts of the capsid gene were in vitro transcribed from PCR templates as described above (with the omission of [α-³²P] GTP), then purified on NucAway columns (Ambion) and quantified spectrophotometrically. The integrity of the RNA products was verified by electrophoresis on an agarose gel. To analyze RNA duplex formation, the ³²P-labeled 3′ RNA probe (30,000 cpm) was mixed with increasing concentrations of 5′ RNA in a total volume of 20 μl of buffer containing 5 mM HEPES, pH 7.9, 100 mM KCl, 5 mM MgCl₂, 3.5% glycerol, and 500 ng tRNA. The reaction mixture was heat denatured at 85°C for 5 min and then allowed to cool slowly at room temperature. The RNA-RNA complex was resolved by 5% non-denaturing polyacrylamide gel electrophoresis in 1X TBE buffer. The RNA bands were visualized on a Fuji BAS 1800 analyzer (Fuji Photo Film Co., Japan) and the percent of the RNA probe shifted was quantified using Fuzi Multigauge softwere (V3.1).

RNA structure prediction

Optimal RNA secondary structures were predicted by Mfold (version 3.2) (Zuker, 2003).

Results

Analysis of the effect of single mismatching nt substitutions in either the 5′ or 3′ WNV CYC sequence on virus replication

In previous replicon and infectious clone studies that demonstrated the importance of the 5′-3′ CYC interaction for flavivirus RNA replication, either all or most of the 8 conserved 5′ and/or 3′-CYC nts were simultaneously mutated to disrupt long distance CYC region basepairs (Alvarez et al., 2005a; Khromykh et al., 2001; Kofler et al., 2006; Lo et al., 2003). In the current study, the functional importance of individual CYC base-pairs in the context of a complete virus lifecycle was first analyzed by introducing single nt mismatching substitutions in either the WNV 5′ CYC (5′-UCAAUAUG-3′) or 3′ CYC (5′-CAUAUUGA-3′) sequences in the WNV W956 infectious clone (Fig. 1B). The AUG in the 5′ CYC is not the polyprotein initiation codon but encodes the fifteenth amino acid of the capsid region. Each of the single nt mutations made in the 5′ CYC resulted in the substitution of one amino acid. BHK cells which are Type I interferon non-responsive were used for viral RNA transfection and mutant virus passage to avoid the possibility of confounding effects due to differential activation of the interferon response by mutants with different replication efficiencies.

Since basepairing between the 5′-3′ nts flanking both sides of the 8 nt conserved CYC sequences was predicted to extend the WNV long distance RNA-RNA interaction in this region to 11 basepairs (Fig. 1B), the effects of mismatching substitutions in some flanking 5′-3′ basepairs were also analyzed. A basepair is not indicated between nts U₁₃₅ and A₁₀₉₂₅ (Fig. 1B) since these nts were shown not to pair in in vitro RNase probing experiments (Dong et al., 2008); U₁₃₅ pairs with G₁₁₆ at the bottom of an adjacent conserved 5′ stem loop structure named the capsid hairpin (cHP) that is present in both the “free” and 5′-3′ paired 5′ sequences (Clyde and Harris, 2006). Single mismatching nt substitutions were introduced at one of six positions in the 3′CYC to generate mutants 3′1 (A₁₀₉₂₃→U), 3′2 (G₁₀₉₂₂→C), 3′3 (U₁₀₉₂₀→A), 3′4 (A₁₀₉₁₉→U), 3′5 (C₁₀₉₁₆ →G), and 3′7 (A₁₀₉₁₉ →C) and in one 3′ CYC flanking nt to generate mutant 3′6 (G₁₀₉₁₅→C) (Fig. 1C). BHK cell monolayers in six-well plates were transfected with an in vitro transcribed mutant viral RNA as described in Materials and Methods. The plaque phenotypes of mutant progeny viruses were assessed on agarose-overlaid transfection wells at 72 h after transfection. Culture fluids harvested at 72 h from non-overlaid wells were titrated by plaque assay and 0.1 ml was used to infect a fresh six-well monolayer to initiate the first passage. At each subsequent passage, 0.1 ml of 72 h culture fluid was used to infect a fresh monolayer. Viral RNA extracted from picked plaques or from media harvested from transfection wells was amplified by RT-PCR, cloned and sequenced.

For mutants 3′1 and 3′2, faint pinpoint plaques (<1 mm) were detected on the overlaid transfection wells. After a single passage of the progeny virus, only large plaques (∼3.5 nm) were detected that contained the parental sequence. Both medium (∼2.5 mm) and large sized plaques were observed on the transfection wells for mutants 3′3 and 3′4 and also for two additional single mismatching nt mutants U₁₀₉₂₁→A and U₁₀₉₁₈→A in (data not shown). Viral RNA from medium sized plaques produced by each of these mutants contained only the mutant sequence while viral RNA from large plaques contained the parental sequence, indicating that reversion had occurred on the transfection well. Only small plaques containing the mutant viral RNA were detected for mutants 3′5 and 3′6 and only medium sized plaques containing the mutant sequences were detected for mutant 3′7 on the overlaid transfection wells. Large plaques that contained the parental RNA sequence were detected by passage one for each of these mutants. Mutants 3′4 and 3′7 with different substitutions at position 10919 both produced medium sized plaques on the transfection wells. However, the time required for reversion of mutant 3′4 (U→A) was faster than that for mutant 3′7 (C→A). Substitution of the 3′ terminal 3′ CYC nts A₁₀₉₂₃ (mutant 3′1) and G₁₀₉₂₂ (mutant 3′2) reduced plaque size to the greatest degree, while substitution of C₁₀₉₁₆ (mutant 3′5) and G₁₀₉₁₅ (mutant 3′6) flanking the 5′ side of the 3′ CYC had the next greatest negative effect.

Mismatching nt substitutions were also made in the 5′CYC nts to generate mutants 5′1 (U₁₃₇→A), 5′2 (U₁₄₁→A), 5′3 (A₁₄₀→C) and 5′4 (A₁₄₂→C) (Fig. 1C). Each of these mutants produced medium sized plaques on the transfection wells and reverted to the parental sequence during first passage. Although the amino acid changed in each of these 5′CYC mutants differed [mutant 5′1 (V→ D), 5′2 (N→K), 5′3 (N→T) and 5′4 (M→L)], the plaque sizes and reversion times for these mutants were the same suggesting that the various substituted capsid amino acids did not differentially effect virus replication. A previous study also found that a mutant WNV infectious clone with two alternative amino acids (VY*I*L) in the same region of the capsid protein replicated as efficiently as the parental virus in both mammalian (Vero and BHK) and mosquito cells (Suzuki et al., 2008).

To assess the effect of the CYC substitutions on virus replication, the kinetics of virus production by non-overlaid BHK monolayers transfected with 1 μg of a mutant viral RNA was analyzed. Although the peak virus titers produced by mutants 3′3, 3′5 and 5′1 by 72 h after transfection were similar, mutant 5′1 produced less virus at earlier times after infection (Fig. 1D). The peak virus yield produced by mutant 3′1 was significantly lower than those of the other mutants and no virus was detected until 56 h after transfection.

The levels of intracellular genomic RNA were also assessed by real time qRT-PCR. Viral RNA levels measured at 48 and 72 h were expressed as the fold change compared to the amount of input viral RNA measured at 6 h after transfection. Parental genomic RNA levels were 1.5 and 2.6 times higher than the 6 h RNA level at 48 and 72 h after transfection, respectively (Fig 1E). For mutants 3′1, 3′3, 3′5 and 5′1, the intracellular viral RNA levels at 48 h was lower than the input RNA at 6 h indicating that they replicated less efficiently than the parental RNA. However, by 72 h after transfection, the level of intracellular viral RNA for each mutant was higher than that of input RNA. The observed relative viral RNA replication efficiencies of these mutants was 3′3>3′5>5′1>3′1. These data were consistent with the growth kinetics of these mutants (Fig. 1D). Although the same basepair was mismatched in mutants 5′1 and 3′1, mutation of the 3′ CYC nt (A₁₀₉₂₃→U) had a much greater negative effect on plaque size as well as virus yield and intracellular viral RNA levels than did mutation of the 5′ CYC partner (U₁₃₇→ A).

Because mutants 3′3 and 3′4 were observed to produce some revertant plaques on the overlaid transfection wells, the reversion kinetics of selected mutants were compared by analyzing the plaque phenotypes of progeny virus populations harvested from non-overlaid wells at 30, 48, 56, 72 and 96 h after transfection of 1μg of mutant viral RNA (Fig. 1F). By 30 h, similar numbers of large and medium plaques were detected for mutant 3′3, no plaques were detected for mutant 3′1, only small plaques were observed for mutant 3′5 and only medium plaques were detected for mutant 5′1. Large plaques were first detected for mutant 3′5 at 48 h and for mutants 3′1 and 5′1 at 72 h. A faster reversion rate was consistently observed on non-overlaid wells compared to overlaid transfection wells. The data indicate that the relative reversion efficiencies of these mutants were 3′3>3′5>5′1>3′1 suggesting that a higher mutant RNA replication efficiency correlated with faster reversion. Two or three different infectious clones were separately tested for each single substitution mutant and the same results were generated by the replicates.

Analysis of the effect of two mismatching CYC nt substitutions on WNV replication

Two adjacent nts were next substituted in the 3′ CYC to produce mutants 3′9 (AU_{10917 – 10918} →UA), 3′10 (AU_{10919 – 10920} →UA), and 3′11 (UG_{10921 – 10922} →AC) (Fig. 2) that were each predicted by Mfold to disrupt two adjacent 5′-3′CYC base-pairs. For each of these mutant RNAs, pinpoint plaques were detected on the transfection plates but medium sized plaques appeared after one or two passages. Viral RNA in the medium plaques had partially reverted (position 10918 in mutant 3′9 and position 10922 in mutant 3′11). Two different partial revertants, 3′10P1 (AA_{10919 – 10920}) and 3′10P1′ (UU_{10919 – 10920}), were detected after the first passage of mutant 3′10. The 3′10P1 revertant was more than twice as abundant as 3′10P1′. Large plaques were first detected during the third or fourth passages for each of these mutants and sequencing confirmed that the viral RNA in the large plaques had the parental sequence. Virus production by mutant 3′10 on a non-overlaid transfection well was low and delayed (Fig. 3C) and intracellular viral RNA levels in the transfection wells were low at both 48 and 72 h after transfection (Fig. 3C).

Figure 2. Effect of two mismatching CYC nt substitutions on virus replication.

The substituted nts in each mutant sequence are underlined and bolded. Capsid amino acid changes in the 5′ CYC mutants and the passage (P) at which reversion occurred are indicated. Plaque diameters produced by mutants and partial revertants are indicated (see Fig.1 legend for designations).

Figure 3. Effect of three or five adjacent mismatching CYC nt substitutions on WNV replication.

(A) The three or five adjacent substituted nts in each mutant sequence are underlined and bolded. Capsid amino acid changes in the 5′ CYC mutants, mutant and partial revertant plaque sizes (see Fig. 1 legend for designations) and the passage (P) at which partial reversion occurred are indicated. In the partial revertant sequences, retained substituted nts are underlined and bolded, reverted nts are underlined and second site mutations are underlined, bolded and marked with an asterisk. ND-not detected. (B) Plaques produced by the engineered 5′8 mutant at 72 h after mutant RNA transfection. (C) The kinetics of virus production from BHK cells transfected with mutant RNAs that have two mismatching CYC substitutions were analyzed as described in the legend of Fig. 1. Error bars represent ± SD (n = 4). (D) Relative quantification of intracellular WNV genomic RNA levels by real-time qRT-PCR was done as described in Materials and Methods. The data from each RNA sample was normalized to cellular GAPDH mRNA in the same sample. Error bars represent the ± SD (n = 3). (E) The kinetics of virus production by BHK cells transfected with mutant RNAs with three mismatching CYC substitutions. Error bars indicate ± SD (n = 4).

Mutants with two substitutions in the 5′ CYC, 5′5 (A₁₄₀→U and A₁₄₂→C), 5′6 (A₁₄₀→C and A₁₄₃→G) and 5′7 (UC_137-138→CG) were also made (Fig 2). Mfold analysis of the 5′5 RNA predicted that the U141-A10919 basepair between the two mismatched basepairs would not form so that this mutant had three not two mismatched CYC basepairs. Each of these 5′ mutants produced pinpoint plagues. Medium sized plaques that appeared during passage one or two contained viral RNA that had a reversion of the one of the substituted nts: 5′5 (U₁₄₀→A), 5′6 (C₁₄₀→A) and 5′7 (C₁₃₇→U). Large plaques containing parental RNA were detected after one or two additional passages. For each of these mutant RNAs, two amino acids in the capsid protein coding region were changed (Fig. 2). However, even though mutants 5′5, 5′6 and 5′7 had different amino acid substitutions, they each produced pinpoint plaques and sequentially reverted with similar kinetics. Analysis of growth kinetics (Fig. 3C) and intracellular viral RNA levels (Fig. 3D) indicated that mutant 5′6 replicated slightly more efficiently than mutant 3′10. Two or three different infectious clones were separately tested for each of the two mismatching substitution mutants and the same results were obtained with the replicates.

Analysis of the effect of three adjacent mismatching CYC nt substitutions on WNV replication

Even though all of the one or two mismatching CYC substitutions analyzed reduced the efficiency of virus replication to varying degrees, all of these mutants were viable and generated revertants. The viability of three adjacent mismatching CYC nt substitutions was next analyzed with mutants 3′13 (UAU_10918-10920→AUA), 3′14 (AUU_10919-1092→UAA), and 3′15 (AUA_10917-10919→UAU)(Fig. 3A). One additional mutant 3′16 (GCA_10915-10917→CGU) was substituted at the last two positions at the 5′ end of the 3′ CYC and at the adjacent flanking nt. Mutant 3′13 produced pinpoint plaques. Medium sized plaques were observed after passage two and contained viral RNA with two different partial revertant sequences, AUA_{(10918-10920)}→ AAU in 3′13P2 and AUA_{(10918-10920)}→ UUU in 3′13P2′ in a ratio of 2:1 (Fig. 3A). By passage four, only large plaques were detected that had either the parental sequence (which could have resulted from reversion of the remaining substituted nt in either 3′13P2 or 3′13P2′) or a sequence with the 3′13P2 “A” substitution at position 10920 retained and a second site mutation A₁₄₀→U in the 5′CYC that paired with the substituted A₁₀₉₂₀. The revertant with the “flipped” 140-10920 basepair replicated as efficiently as parental virus. A previous WNV replicon study also reported that “flipping” the CYC basepair at this position had no effect on viral RNA replication efficiency (Suzuki et al., 2008).

Mutants 3′14 and 3′15 produced pinpoint plaques. For both of these mutants, viral RNA from picked medium plaques that appeared at passage three retained the 3 substituted 3′ CYC nts but had acquired a second site mutation that created an additional 5′-3′ basepair at the same position (nts 147-10913) in a CYC flanking region. The second site mutation in the 3′14P3 RNA was in the 5′ flanking region of the 3′ CYC (C₁₀₉₁₃→U*) (the same end of the genome as the substituted nts) and created an A-U basepair. In the 3′15P3 RNA, the second site mutation occurred in the 3′ flanking region of the 5′ CYC (A₁₄₇→G*) (the other end of the genome) and created a G-C basepair. No additional spontaneous mutations or increases in plaque size occurred during seven additional serial passages of these two second site revertants. Mutant 3′16 produced faint pinpoint plaques and medium sized plaques were detected at passage four. Viral RNA from picked medium sized plaques had the substituted 3′ nt U₁₀₉₁₇ deleted and a second site mutation A₁₄₇→G at the 5′ end of the genome that created an new additional CYC region 5′-3′ basepair (nts 147-10913). The A₁₄₇→G substitution was silent. No additional spontaneous mutations or increases in plaque size occurred during seven additional serial passages of this revertant. Mutants 3′13 and 3′16 produced low yields of virus and showed delayed replication (Fig. 3E) and the levels of mutant 3′13 intracellular viral RNA at both 48 and 72 h after transfection were significantly lower than input RNA level at 6 h (Fig. 3D).

Two mutants with three adjacent mismatching substitutions in the 5′ CYC, mutant 5′8 (AAU_139-141→ UUA) and mutant 5′9 (AUA_140-142→UAU), were also analyzed (Fig. 3A). Both of these mutants produced pinpoint plaques. Medium sized plaques were detected at passage three for mutant 5′8 and passage four for mutant 5′9. Viral RNA from picked medium plaques retained the 3 substituted nts and had the same 5′ second site mutation (A₁₄₇→G*) that created the additional 147-10913 basepair as detected in the mutant 3′15 and 3′16 revertant RNAs. In the 5′8 and 5′9 pseudorevertant RNAs, the second site mutation occurred in the same CYC sequence that contained the substitutions while in the 3′15 and 3′16 RNAs, the second site mutation occurred in the opposite CYC sequence. Mutant 5′8 produced low yields of virus and showed delayed replication but replicated slightly less efficiently than mutant 3′16 (Fig. 3E). The levels of mutant 5′8 intracellular viral RNA at both 48 and 72 h after transfection were also low (Fig. 3D). No additional spontaneous mutations or increases in plaque size occurred through seven serial passages of the 5′8 and 5′9 pseudorevertants. Even though the mutant 5′8 and 5′9 RNAs contained substitutions that changed one (5′8) or two (5′9) capsid amino acids (Fig. 3A), their plaque sizes, reversion rates were the similar to those of the 3′ CYC three nt mutants. These results suggest that the amino acid changes in these mutants were not the cause of their reduced virus production. Two or three different infectious clones were separately tested for each mutant with three mismatching substitutions and the same results were observed for the replicates.

To confirm that the A₁₄₇→G* mutation was all that was needed to enhance the replication of mutants with three adjacent mismatching CYC nt substitutions, an engineered mutant (5′8 Eng) was constructed by introducing the A₁₄₇→G* mutation into the 5′8 mutant sequence. In contrast to the original mutant 5′8 RNA that produced faint pinpoint plaques through passage three, the 5′8 Eng RNA produced medium plaques on the transfection wells (Fig. 3B). Analysis of the growth kinetics of 5′8 Eng, showed that it replicated much more efficiently than mutant 5′8 but less efficiently than parental RNA (Fig. 3E). The intracellular viral RNA levels for Eng 3′8 at both 48 and 72 h after transfection were also much higher than those produced by mutant 5′8 RNA but lower than those observed for parental RNA (Fig. 3D). The 5′8 Eng infectious clone was constructed twice and the same results were obtained with the duplicate clones. The results indicate that mutant genomes with three adjacent CYC mismatches are viable and that the additional flanking basepair created in the majority of the revertants generated by a spontaneous second site mutation significantly enhanced their replication efficiency.

Analysis of the effect of five adjacent mismatching CYC nt substitutions on WNV replication

Since all of the mutant viral RNAs with one, two or three mismatching CYC substitutions were viable and acquired revertant or second site mutations that enhanced their replication efficiency, five adjacent mismatching CYC nt substitutions were next made in the middle of either the 5′ or 3′ CYC sequence to generate mutant 3′17 (AUAUU_10917-10921→UAUAA) and mutant 5′13 (AAUAU_139-143→UUAUA) (Fig 3A). Neither of these mutant viral RNAs produced plaques on the transfection wells or during six serial passages, indicating that five adjacent CYC mismatches were lethal for the virus. Two or three different infectious clones were tested separately for each five substitution mutant and the same results were observed with the replicates. The substitutions in the 5′13 RNA changed an internal capsid coding region methionine codon to an amber stop-codon.

Analysis of the effect of mismatches generated by substitution of a different nt on WNV replication

The strategy used to generate CYC mismatches up to this point in the study was to substitute the same base in one CYC as was present in the other. The effect on viability and reversion efficiency of substituting alternative mismatching nts was next analyzed. Three or five of the A and U nts in the middle of either the 5′ or 3′ CYC were substituted with G or C to create mutants, 3′18 (AUU_10919-10921→CCC), 5′10 (AAU_139-141→GGG), and 5′12 (AAUAU_139-143→GGGGG) (Fig. 3A). Mutant 3′18, which had three adjacent CYC mismatches, produced faint pinpoint plaques at passage four that contained only the mutant sequence. Medium size plaques appeared at passage six and the viral RNA in these plaques had a C₁₀₉₂₀→U reversion that restored the parental A-U basepair and a second site mutation in the 5′ CYC (U₁₄₁→G) that created an alternative G-C basepair. This second site mutation changed one capsid amino acid. No other sequence changes were observed during seven additional passages and the remaining single mismatch was retained. An Mfold analysis predicted that the 139 and 140 nts in both the mutant 5′10 and 5′12 RNAs would form G-U basepairs with 3′ CYC nts. However, the substituted 142 nt in mutant 5′12 was not predicted to pair with a 3′ CYC nt. Therefore, mutant 5′10 had one not three 5′ or 3′ CYC mismatches and mutant 5′12 had three not five mismatches in addition to two new G-U basepairs. Mutant 5′10 produced small plaques. Medium sized plaques detected at passage two contained viral RNA that retained the two substituted G nts that formed G-U pairs. However, the unpaired mutant G at position 141 had reverted restoring the parental basepair. No other sequence changes were observed during additional passages. Mutant 5′12 produced pinpoint plaques through passage six. Only medium sized plaques were detected at passage seven. The viral RNA from medium plaques contained a G₁₄₁→U reversion that recreated the parental U-A basepair and a second site mutation in the 3′ CYC (A₁₀₉₁₉→C) that created a G-C pair. These two changes “repaired” all of the CYC mismatches and allowed the 142-10918 base pair to form. No additional spontaneous mutations or increases in plaque size were detected during four additional serial passages of this revertant. Two or three different infectious clones were separately tested for each of these mutants and the same results were obtained with the replicates. Although the mutant 5′10 revertant retained the one amino acid substitution and the 5′12 revertant retained the two amino acid substitutions of the original mutants, both produced medium sized plaques. Neither the 5′10 nor the 5′12 revertant had any mismatches between their 5′ and 3′ CYC sequences. The medium sized plaques produced by these revertants suggest that CYC G-U basepairs are not optimal as A-U pairs for this long distance interaction.

Analysis of the effect of substituting CYC basepairs on WNV replication

The five center positions of the 5′-3′ CYC long distance interaction are occupied by A-U basepairs in a particular arrangement (Fig. 1B). To test the positional importance of the CYC basepairs at these positions, 3 or 5 of these basepairs were “flipped” to generate mutant 5′8+3′14, mutant 5′9+3′13, and mutant 5′13+3′17 or three of the A-U basepairs were substituted with G-C pairs mutant to create mutant 5′10+3′18 (Fig. 4A). The 5′8+3′14 mutant produced large plaques on the transfection wells and no spontaneous reversions/mutations or changes in plaque size were detected through seven subsequent passages (Fig. 4A). Mutant 5′9+3′13 produced medium sized plaques on the transfection wells and no spontaneous reversions/mutations or changes in plaque size were detected through seven subsequent passages (Fig. 4A). Analysis of the kinetics of virus production after RNA transfection showed that mutant 5′8+3′14 grew as efficiently as the parental virus while mutant 5′9+3′13 grew slightly less efficiently (Fig 4B). Assessment of the relative levels of intracellular viral RNA at 48 and 72 h after transfection by quantitative real time qRT-PCR indicated that mutant 5′8+3′14 produced parental levels of intracellular genomic RNA, while mutant 5′9+3′13 produced slightly lower levels (Fig. 4C).

Figure 4. Effect of substitution of either three or five CYC basepairs on replication of WNV RNA.

(A) For each mutant, the three or five adjacent substituted nts in both CYCs are underlined and bolded. Capsid amino acid changes and the passage (P) at which reversion occurred are indicated. Plaque diameters produced by mutants and partial revertants are indicated (see Fig.1 legend for designations). The second site mutation is shown as a white letter in a black box. (B) The kinetics of virus production from BHK cells transfected with mutant RNAs were analyzed as described in the legend of Fig. 1. Error bars represent ± SD (n = 4). (C) Relative quantification of intracellular WNV genomic RNA levels by real-time qRT-PCR was done as described in Materials and Methods. Each RNA sample was normalized to cellular GAPDH mRNA. Error bars represent the ± SD (n = 3).

Mutant 5′13+3′17 produced opaque pinpoint plaques on the transfection wells and after passage one (Fig 4A). Only the mutant RNA sequence was detected in culture fluid harvested from the transfection plate and after passage one. Mutant 5′13+3′17 showed delayed virus replication and produced lower yields of virus (Fig. 4B) and lower levels of intracellular viral RNA at both 48 and 72 h after transfection (Fig. 4C) than the other basepair substitution mutants. As noted previously, the 5′13 substitutions replaced the second capsid region methionine codon with an amber stop codon. The production of plaques by this mutant indicated that infectious virions were produced with an N-terminally deleted C protein. Medium size plaques were first detected at passage two. Viral RNA from picked medium sized plaques had an A₁₄₃→G change that replaced the amber stop codon with a tryptophan codon rather than the parental methionine codon and created a G-U basepair instead of the parental A-U basepair. No additional spontaneous mutations or changes in plaque size were detected through seven subsequent passages of this pseudorevertant.

Mutant 5′10+3′18 produced medium sized plaques on the transfection wells and during subsequent passages (Fig. 4A). No spontaneous reversions/mutations or changes in plaque size were detected through seven subsequent passages of this mutant. Mutant 5′10+3′18 replicated less efficiently than mutants 5′8+3′14 and 5′9 +3′13 but more efficiently than mutant 5′13+3′17 (Fig. 4B). Mutant 5′10+3′18 intracellular viral RNA levels were higher than those of mutant 5′13+3′17 but lower than those of mutants 5′8+3′14 and 5′9 +3′13 at both 48 and 72 h after transfection (Fig. 4C).

Effect of mismatching CYC nt mutations on the in vitro 5′-3′ RNA-RNA interaction

The optimal secondary structure of an RNA fragment consisting of the parental 200 5′ terminal and 110 3′ terminal WNV genomic nts connected by a 10 nt poly A linker was predicted by Mfold (Fig. 5A). The predicted ΔG for this structure was -125.84 kCal/mol. Introduction of one mismatching mutation into the 5′ CYC sequence (mutant 5′2) reduced the thermodynamic stability of the secondary structure (ΔG of -121.74 kCal/mol) (Fig 5A). The mutant 5′5, mutant 5′8 and mutant 5′12 RNA structures were each predicted by Mfold to have a three nt mismatch and similar decreased thermodynamic stabilities (ΔG of ∼ -118 kCal/mol). The presence of the two G-U basepairs in mutant 5′12 RNA did not negatively or positively affect the thermodynamic stability of the predicted structure. The five mismatching substitutions in the lethal 5′13 mutant were predicted to significantly alter the RNA structure in the CYC region and to change some of the 5′-3′ nt pairing partners. The structure predicted for this mutant RNA had the least favorable thermodynamic stability (ΔG of -116.06 kCal/mol) (Fig. 5A). The introduction of the second site mutation A₁₄₇→G into the 5′8 mutant sequence increased the thermodynamic stability of the RNA secondary structure (ΔG was changed from -118.44 to - 120.90 kCal/mol).

Figure 5. Analysis of in vitro 5′ or 3′ RNA-RNA interactions.

(A) Mfold predicted secondary structures for interactions between 110 nt 3′ and 200 nt 5′ RNAs joined by a 10 nt A linker. Folds for each parental terminal sequence alone as well as for the parental 3′ RNA linked to either that a parental or mutant 5′ sequence are shown. The CYC sequences are indicated by lines. Free energy ΔG values for each structure are in kCal/mol. (B) CYC region sequences of the parental 3′ probe and the parental and mutant 5′ RNAs. (C) Analysis of parental 5′-3′ RNA-RNA interactions by gel mobility shift assay. The parental ³²P-labeled 110 nt 3′ probe (30,000 cpm) was incubated with increasing amounts of the parental 200 nt 5′ RNA. The 5′ RNA concentration in nM is indicated at the top of the gel. The RNA-RNA complex and free probe are indicated by arrows. The percentage of probe bound with each concentration of the 5′ RNA is indicated at the bottom of the gel. (D) Analysis of the interactions of mutant 5′ RNAs with the parental 3′ RNA probe.

To directly investigate the effect of mismatching CYC substitutions on 5′-3′ RNA basepairing, the interactions of various mutant 5′ RNAs with a parental 3′ RNA probe were analyzed by gel mobility shift assay. Unlabeled 200 nt 5′ terminal RNA fragments with one, two, three or five mismatching CYC nt substitutions (Fig. 5B) and parental ³²P-labelled 3′ probe were synthesized in vitro. A parental or mutant 5′ RNA was incubated with the parental ³²P-labelled 3′ terminal 110 nt probe in the presence of tRNA and the formation of an RNA-RNA complex was assessed after separation on nondenaturing gels. Only the parental 3′ probe was used in these experiments to allow a more accurate comparison of the effects of different 5′ mutations on the RNA-RNA interaction. In initial experiments, the concentration of parental 5′ RNA required to shift more than 50% of the 3′ probe was determined to be 100 nM (Fig. 5C). Subsequent experiments were done with 100, 200, and 400 nM of 5′ RNA. Mutant 5′2 RNA shifted only about half as much 3′ probe as the parental 5′ RNA (Fig. 5D). A lower amount of the 3′ probe was shifted by mutant 5′5 RNA and only minimal amounts were shifted by the 5′8, 5′13 and 5′12 RNAs. The slightly more efficient binding of the 5′5 RNA could be due to a possible weak interaction between the U₁₄₁-A₁₀₉₁₉ nts. However, although a weak interaction might also be possible between the G₁₄₂-U₁₀₉₁₈ nts in mutant 5′12, the RNA-RNA interaction with this mutant was similar to that of the three mismatch mutant 5′8. The percent of probe shifted by the Eng 5′8 RNA was about 10 times higher than the amount shifted by the mutant 5′8 RNA indicating that the existence of the 147-10913 basepair significantly enhanced the efficiency of the 5′-3′ RNA-RNA interaction but not to parental levels. For each of the mutants tested, the relative efficiency of the 5′-3′ RNA-RNA interaction correlated with the relative replication efficiency observed in the previous experiments.

Discussion

Genome circularization facilitated by long distance RNA-RNA interactions between 5′ and 3′ terminal sequences has been reported for different families of positive strand RNA viruses (Filomatori et al., 2006; Guo et al., 2001; Hahn et al., 1987; Klovins and van Duin, 1999). The genomes of mosquito-borne flaviviruses contain 8 nt conserved 5′ and 3′ CYC sequences flanked by one or two non-conserved nts that can also pair and extend the 5′-3′ CYC region RNA-RNA interaction. Although additional terminal genomic regions, such as the UAR and DAR sequences, are functionally involved in flavivirus genome 5′-3′ RNA-RNA interactions, the CYC region interaction has been shown to be critical for the initiation/establishment of the 5′-3′ RNA-RNA interaction (Alvarez et al. (2008); Friebe et al., 2010; Polacek et al., 2009). The conserved 8 nt 5′-3′ CYC region consists of five A-U basepairs in the middle region flanked by one G-C basepair on the internal genome side and by a G-C and a A-U base pair on the genome terminal side (Fig. 1B). Although the 5′-3′ RNA-RNA interactions of several plant RNA viruses have been shown to regulate both genome translation as well as replication (Barry and Miller, 2002; Fabian and White, 2004; Miller and White, 2006), multiple previous functional studies on flavivirus 5′-3′ CYC sequences established that the 5′-3′ CYC interaction does not affect genome translation but is required for viral RNA replication (Alvarez et al., 2005a; Corver et al., 2003; Filomatori et al., 2006; Khromykh et al., 2001; Kofler et al., 2006; Lo et al., 2003). However, since the majority of the previous studies on flavivirus CYCs were done with replicons, the flavivirus 5′-3′ RNA-RNA CYC interaction had not been well characterized in the context of a complete virus replication cycle.

In the present study, a flavivirus infectious clone was used to analyze the functional consequences of nt substitutions that mismatched 5′-3′ CYC basepairs. Mutant genomes with one, two or three CYC basepair mismatches were able to replicate, although usually at a reduced level compared to the parental RNA, and to generate spontaneous rescuing revertants. In contrast, single (with one exception) and multiple CYC basepair mismatching mutations were lethal in replicons (Khromykh et al., 2001; Lo et al., 2003; Suzuki et al., 2008). One exception was the WNV mutant replicon m2-5′ that had only a slightly reduced replication efficiency compared to the parental replicon (Suzuki et al., 2008). The m2-5′ mutant replicon contained the same substitution as our mutant 5′1. Mutant 5′1 virus showed delayed and less efficient replication compared to the parental virus and produced medium sized plaques. However, other replicons with either a single CYC mismatching substitution or a double CYC mismatching substitution tested in the same previous study, including the m2-3′ replicon with the same substitution as our mutant 3′1, were unable to replicate. These data indicate that replicons do not accurately predict the effect of CYC mutations on full length genomes.

Mismatching single base pairs at the genome terminal end of the CYC region (mutants 3′1 and 3′2) had the greatest negative effect on virus and RNA replication suggesting that basepairing in this region was functionally important. However, comparison of the effects of individually substituting the partners of the same basepair in mutants 5′1 and 3′1 indicated that the 3′ CYC nt (A₁₀₉₂₃→U) mutation had a much greater negative effect on RNA and virus replication than the 5′CYC (U₁₃₇→A) substitution and suggested that an A at position 10923 as well as the 5′-3′ basepair at this position were functionally important. Data obtained by Suzuki et al. (2008) with m2-3′ (the same substitution as our mutant 3′1) also indicated that an A at position 10923 was functionally important. It is not clear why A₁₀₉₂₃ is so critical for function. This nt is predicted to be unpaired in the 3′UTR structure and so is not involved in an alternate pairing interaction. It is possible that the long distance interaction is nucleated by this nt. Data obtained in a previous study of long distance RNA-RNA interactions in a plant virus genome suggested that the formation of an A-U pair could initiate a long distance RNA-RNA interaction (Guo et al., 2001). Consistent with the 5′-3′ CYC interaction beginning from this side, the first nt to revert to generate the partial revertants 3′9P1, 3′11P1, 5′5P2, 5′6P1, and 5′7P2 was the nt on the 3′ end of the 3′ CYC or the 5′ side of the 5′ CYC.

With the exception of the single A₁₀₉₂₃→U and G₁₀₉₂₂→C mutations, two mismatching nt substitutions at all postions tested had a greater negative effect on virus replication than did a single nt mismatching substitution and three mismatching nt substitutions had an even greater negative effect. However, all of these mutants were viable and generated revertants that replicated more efficiently. Mutant 3′13 was the only one of the three A or U nt substitution mutants tested for which reversion of any of the substituted nts occurred. For mutants 3′14, 3′15, 3′16, 5′8 and 5′9, rescue occurred by a second site mutation that created an additional basepair at the same position (nts 147-10913) on the genomic side of the 5′-3′ CYC. For mutants, 3′15, 3′16, 5′8 and 5′9, the second site mutation was A₁₄₇→G, while for mutant 3′14, it was C₁₀₉₁₃→U indicating that either a G-C or an A-U pair at this position was functionally sufficient. The observation that the formation of a 147-10913 basepair significantly decreased the free energy of the 5′-3′ RNA structure and increased the efficiency of the in vitro 5′-3′ RNA-RNA interaction (Eng 5′8) even with three CYC basepair mismatches still present suggested that extension of the CYC interaction by one basepair was sufficient to restabilize the CYC interaction on the internal genome side. This new basepair significantly increased plaque size, viral RNA replication efficiency and virus yield, but not to parental levels. The in vitro RNA-RNA interaction data suggest that the inability to attain parental replication efficiecy was due to the effect of the three retained CYC mismatches on the efficiency of the long distance RNA-RNA interaction.

The presence of two conserved G-C basepairs on the genomic side of the CYC region as well as the finding that mismatching either of these basepairs (mutants 3′5 and 3′6) had a greater negative effect on virus and RNA replication than did mismatching basepairs in the central region of the CYC together with the frequent detection of a new rescuing basepair on this side of the CYC suggested that a minimum number of basepairs as well as some stronger basepairs are needed in this region for the establishment/stabilization of the long distance RNA-RNA interaction. All of the second site mutants that retained three adjacent CYC mismatches had at least 4 basepairs remaining on the internal genomic side of the mismatched CYC basepairs, with the exception of the second site 3′16 mutant which was the only three mismatch mutant that generated the 147-10913 basepair that did not maintain the three mismatches (Fig. 3A). Although the G₁₄₄-C₁₀₉₁₆ and C₁₄₅-G₁₀₉₁₇ basepairs were mismatched in the 3′16 mutant, pairing of these two substituted nts was restored by deletion of one of the substituted nts. The 3′16 revertant results support the requirement for multiple basepairs in this region of the CYC 5′-3′ interaction. While additional regions such as the DAR and UAR that extend the 5′-3′ interaction from the CYC toward the terminal regions of the genome have been reported, no known 5′-3′ interaction regions have yet been identified that extend the long distance interaction on the genome internal side of the CYC. The nts adjacent to the 5′-3′ CYC interaction on the internal genome side in both the 5′ and 3′ sequences are predicted to be unpaired.

Interestingly, mutants with G or C mismatching nt substitutions in the central CYC region were not rescued by the creation of a 147-10913 basepair. Instead, mutants 3′18 and 5′12 were rescued by the combination of a nt reversion (that recreated the parental basepair) and a second site mutation that created an alternative central G-C basepair. Second site rescuing mutations were detected in the same CYC sequence that contained the substitutions for mutants 3′14, 5′8, and 5′9 and in the other CYC sequence in mutants 3′15, 3′16, 3′18 and 5′12. The detection of second site rescuing mutations in either CYC with similar frequency suggests that the main driving force for mutant selection is increasing the efficiency of the 5′-3′ RNA-RNA interaction.

While mutant RNAs with three adjacent CYC mismatches were viable, introduction of five adjacent mismatching substitutions was lethal. The five adjacent mismatches in mutant 3′17 were predicted to create a large bulge (data not shown) while those in mutant 5′13 were predicted to alter the structure and some of the CYC region pairing partners (Fig. 5A). The free energies of the predicted 5′-3′ RNA structures of both of these mutant RNAs were higher (mutant 3′17, ΔG of -115.76 kCal/mol and mutant 5′13, ΔG of -116.06 kCal/mol) than those of any of the other mutants tested and no in vitro RNA-RNA interaction was detected with the mutant 5′13 RNA.

The observation, that mutant viruses with different 5′ CYC single mismatching substitutions that changed one capsid amino acid produced the same size or larger sized plaques than a number of the mutant viruses with single mismatching substitutions in the 3′ CYC, suggested that these amino acid changes did not negatively affect virus spread. In most cases, the replication efficiencies of the 5′ CYC mutants correlated with the number of CYC basepair mismatches present rather than a particular amino acid substitution. This was the case for the replacement of V by D in mutant 5′1, replacement of N by either K or T in mutants 5′2 and 5′3, and the replacement of M by L or R in mutant 5′4 and the partial revertants of mutants 5′5 and 5′6. The hypothesis that the medium plaque size produced by each of these viruses was due to the presence of the CYC mismatch rather than the amino acid change was also supported by the observation that the mutant 5′7 partial revertant had the parental amino acid sequence but retained one CYC basepair mismatch and produced medium sized plaques. Two previous studies also reported that substitutions or deletions of amino acids in this region of the capsid protein were well tolerated by flaviviruses (Corver et al., 2003; Suzuki et al., 2008). Interestingly, the nt reversion and second site mutation that occurred in the mutant 3′18 revertant recreated two adjacent basepairs but changed a parental amino acid. This partial revertant retained a single CYC mismatch and amino acid substitution during additional serial passage and produced medium sized plaques.

However, not all capsid amino acid changes may be neutral. Mutant 5′10+3′18 and the revertants of mutants 5′10 and 5′12 produced medium sized plaques and had a slightly reduced replication efficiency compared to the parental RNA even though all of the 5′-3′ CYC nts were paired and the free energies of the predicted 5′-3′ RNA structures were similar (5′10 revertant, -125.04 kCal/mol and 5′12 revertant, -126.96 kCal/mol) or lower (mutant 5′10+3′18,-131.84 kCal/mol) to that of the parental RNA structure. The only common feature of each of these viruses was the N to G amino acid substitution.

The free energies of the 5′-3′ RNA structures of mutants 5′8+3′14 and 5′9+3′13 were very similar to that of the parental RNA (each ∼-125 kCal/mol) but mutant 5′8+3′14 replicated more efficiently than mutant 5′9+3′13. Mutant 5′9+3′13 had two amino acid changes, N was replaced by I and M was replaced by L, while mutant 5′8+3′14 had one, an N to L change. Reduction in plaque size consistently correlated with a reduction in intracellular viral RNA levels, less efficient virus replication, a higher free energy for the predicted 5′-3′ RNA structure and a less efficient in vitro 5′-3′ RNA-RNA interaction suggesting that mismatched or substituted CYC basepairs that negatively affected the 5′-3′ long distance RNA-RNA interaction and not the efficiency of virion spread were the main reason for the observed mutant phenotypes. Reversion to recreate mismatched basepairs was frequently detected but spontaneous mutation to restore a parental amino acid in the capsid protein was only observed with mutant 5′13+3′17 to change an amber stop codon. Neither mutant 5′13+3′17 nor its pseudorevertant had any CYC basepair mismatches and the structures of these two RNAs had similar free energies (mutant 5′13+3′17 was -126.04 kCal/mol; pseudorevertant with a W codon was -125.74 kCal/mol).

The second AUG in mutant 5′13 was changed to an amber stop codon. Both the 5′13 and 3′17 mutations alone caused the mismatching of the same five adjacent 5′-3′ CYC nts (Fig. 3A) and prevented RNA replication whether or not translation of the polyprotein started at the first AUG or an AUG after the third one. For the compensatory mutant 5′13+3′17, only the mutant sequence was detected in virions in the culture fluid harvested from the transfection plate and these virions produced opaque pinpoint plaques and peak titers of ∼10⁵ PFU/ml. For these results to have been obtained, the polyprotein would have to have had to have been initiated at a downstream AUG such as the third in frame AUG in the mutant RNA. It is not known whether the presence of the short ORF beginning at the normal initiation codon may have facilitated subsequent initiation at the subsequent in frame AUG. The capsid proteins produced by the mutant would have had at least a 20 amino acid deletion at the N-terminus. The capsid proteins of virions produced by the mutant 5′13+3′17 pseudorevertant would not be truncated since the stop codon in the CYC sequence was replaced by that of amino acid W. The pseudorevertant produced clear medium sized plagues and the substituted amino acid in the pseudorevertant genome was maintained through 7 passages.

In summary, the data from the present study indicate that both complete basepairing between the 5′-3′ CYCs as well as particular nts at a few positions are required for optimal virus replication efficiency. Based on data obtained with mutant replicons, a previous study identified only two central region CYC positions at which a basepair could be “flipped” without loss of replication efficiency (Suzuki et al. (2008). The neutral effect of these two basepair changes was confirmed with a mutant WNV produced from an infectious clone. Data from the present study showed that three or more CYC basepairs could be “flipped” or substituted by an alternative basepair, including a G-C pair, with either no or only a small negative effect on replication efficiency (Fig. 4). Single and double CYC mismatches in mutant RNAs rapidly reverted and a single nt second site mutation that created a new base pair was sufficient to rescue genomes with three adjacent CYC mismatches. These data indicate that the 5′-3′ CYC interaction in the context of a full length genome is functionally more “flexible” than the complete conservation of the 8 nt CYC sequences among flavivirus genomes and the demonstration that the CYC interaction is required for the initiation of the 5′-3′ interaction would suggest. The severe negative effect of most single CYC basepair mismatches on the replication of replicon RNAs was not seen with full length viral RNAs and rapid reversion of single and double CYC mismatches regenerated parental virus replication efficiency. These results indicate that conclusions drawn from replicon studies may not be directly applicable to virus infections. A previous study proposed the utilization of CYC basepair substitutions in live virus vaccines as a means of improving their safety. Although there is no evidence of recombination between field strains of flaviviruses (de Silva and Messer, 2004; Monath et al., 2005; Murphy et al., 2004) and a lab study showed that recombination rarely occurs under “forced” conditions (Taucher et al., 2010), recombination between live vaccine and wild genomes of other positive strand RNA viruses, such as poliovirus and bovine diarrhea virus, has been documented (Arita et al., 2005; Becher et al., 2001; Guillot et al., 2000) and some concern has been expressed about the possibility of live flavivirus vaccine recombinants (Seligman and Gould, 2004). Because of the conservation of CYC sequences in WNV strains, engineering alternative CYC basepairs into live vaccine RNAs would create mismatched 5′-3′ CYC basepairs at these positions in recombinants generated between vaccine and wild genomes or between the two RNAs of single cycle two-component vaccines (Shustov et al., 2007). The data obtained in the present study showing that up to three mismatching CYC mutations are not lethal for full-length genomes indicate that designing vaccine genomes with only two CYC substituted basepairs as previously suggested by Suzuki et al. (2008) would not improve the safety of live vaccine candidates. However, vaccine genomes with five adjacent substituted CYC basepairs would yield significantly attenuated spontaneous recombinants.