Host Alternation of Chikungunya Virus Increases Fitness while Restricting Population Diversity and Adaptability to Novel Selective Pressures ^{▿ † ‡}

Cells and viruses.

Baby hamster kidney (BHK), human epithelial (HeLa), and African green monkey kidney (Vero) cells were grown in Dulbecco's minimal essential medium (DMEM) with 10% newborn calf serum (NCS; Gibco) and 1% penicillin and streptomycin (P/S; Sigma) at 37°C with 5% CO₂. Aedes albopictus (C6/36) mosquito cells were maintained in L-15 medium with 10% NCS, 1% P/S, 1% nonessential amino acids, and 1% tryptose phosphate broth at 28°C without CO₂.

Chikungunya virus strain 06-049 (GenBank accession no. AM258994 [39]) was used for passage studies and as the template for an infectious clone. Serum from a febrile patient with arthralgia and myalgia in Reunion Island in 2006 was inoculated into C6/36 cells. CHIKV recovered from the supernatant was passaged once in Aedes pseudoscutellaris cells, and viral RNA extracted from this passage was used as the template for generation of the infectious clone. The starting stock for all passage series was generated from a small-population-size (10⁴ PFU, 0.1 PFU/cell) infection of Vero cells to minimize any preadaptation to insect cells in previous passages. The consensus sequence of this stock matched the reference sequence.

Generation of CHIKV infectious clone.

A plasmid encoding the full cDNA of CHIKV 06-049 was generated by sequential insertion of restriction enzyme-digested cDNA amplicons flanking the genomic RNA into a pBR322 vector (Invitrogen). Viral RNA from the Aedes pseudoscutellaris passage was extracted from supernatant 3 days postinoculation (p.i.) using a QIAamp viral RNA minikit (Qiagen). Reverse transcription-PCR (RT-PCR) amplicons flanking the CHIKV genome were produced from RNA extracts using CHIKV-specific primers and high-fidelity enzymes in the Titan one-tube RT-PCR kit (Roche). RT-PCR amplicons were cloned into intermediate plasmids (TOPO; Invitrogen), sequenced for verification, and then digested and ligated (T4 DNA ligase; Roche) into the vector. Ligated products were transformed into TOP10 competent cells (Invitrogen), and bacterial clones present after 16 h of incubation at 37°C on agar plates containing 100 μg/ml ampicillin were screened for plasmids containing inserts based on restriction digestion banding patterns and by sequencing. The full-length cDNA was resequenced in its entirety after the complete construct was obtained.

The CHIKV clone was used as a reference virus only for fitness assays. It was genetically marked for subsequent competition assays with populations derived from its parent by introducing a synonymous nucleotide (C→T) via site-directed mutagenesis (XL-10 kit; Stratagene) at nucleotide 1980 of the NSP2 gene (NSP2 amino acid 660) to ablate a SacII restriction site (5′-CCGC↓GG3-′→5′-CCGTGG-3′). The location of this site is in a region of the NSP2 gene not known to include regulatory sequence elements under direct RNA sequence selection.

To generate infectious virus from the cloned CHIKV cDNA (henceforth termed “marked clone”), 12 μg of purified plasmid was linearized with NotI, and the purified linear DNA was in vitro transcribed from an SP6 promoter with a mMESSAGE mMACHINE kit (Ambion). RNA was electroporated into BHK cells using an XCell gene pulser electroporator (Bio-Rad) at 1.2 kV. Electroporated BHK cells were then held at 37°C in a 5% CO₂ incubator with 7 ml DMEM. At 48 h postelectroporation, cell culture supernatants containing virus were harvested, aliquoted, and frozen at −80°C. Electroporation harvests were used directly in competition assays.

In vitro replication kinetics of CHIKV clone versus parent.

To compare the replication kinetics of the marked clone with those of its parent CHIKV, growth curve analyses of both viruses were simultaneously conducted in BHK and C6/36 cells (see Fig. S1 in the supplemental material). The marked clone from the electroporation harvest or parent CHIKV from one additional vertebrate cell passage (to control for source cell type, since the marked clone was harvested from BHK cells) was diluted to either 10 or 1 PFU/cell and was inoculated into triplicate wells of BHK or C6/36 cells. After 1 h incubation, inocula were washed from cell monolayers once with 1× phosphate-buffered saline (Gibco), and 800 μl growth medium was added to each well. Fifty microliters of the supernatant was immediately withdrawn as the 0-h sample. A further 50 μl was harvested at 12, 24, and 48 h p.i. Fifty microliters of medium was added to each well at each sampling time to compensate for the reduction in volume with sampling. The virus titer of each sample and inoculum was determined by plaque assay.

Experimental passage.

To simulate natural infection or to artificially bypass the invertebrate or vertebrate cell, CHIKV was alternately passaged between BHK or HeLa and C6/36 cells or serially passaged in BHK, HeLa, or C6/36 cells for a total of seven passages. To initiate each passage, CHIKV was diluted from the earliest available passage (Vero cell, nonclonal, high-titer population) to ≈0.1 PFU/cell and used to infect an ∼80% confluent flask. Virus from each passage was harvested at 48 or 72 h p.i. Output titers from each passage ranged from 4 to 8 log₁₀ PFU/ml (see Fig. S2 in the supplemental material). Consensus genomes from each passage were sequenced at the end of the series. Since different glycosylation patterns of alphaviruses processed in vertebrate versus invertebrate cells can affect pathogenesis (40), we controlled for cell type by comparing each passage 7 (p7) virus to the p1 virus from the same series for all sequence and phenotypic analyses. The first and last passages in the alternating invertebrate ↔ vertebrate cell series occurred in mosquito cells.

RNA extractions and RT-PCRs.

CHIKV RNA from plaque clones or cell culture supernatants was extracted in Trizol (Invitrogen) or with a QIAmp viral RNA minikit. Amplicons were generated using a Titan one-step RT-PCR kit that contains high-fidelity polymerases. For consensus sequencing, primers flanking the full CHIKV genome were used and 3× overlapping chromatograms for each genomic nucleotide were obtained. Complete E1 gene sequences were obtained using PCR primers forward 5′-CTATGCAACTGTCTGAG-3′ and reverse 5′-CCCTTTGAACTACTTCTGT-3′ or partial E1 gene sequences were generated with primers forward 5′-TACGAACACGTAACAGTGATCC-3′ and reverse 5′-CGCTCTTACCGGGTTTGTTG-3′ or forward 5′-GCACCATCTGGCTTTAAGTATTG-3′ and reverse 5′-CGACACGCATAGCACCACGAT-3′.

Virus titrations.

Tenfold dilutions of supernatants containing CHIKV were added to confluent BHK or Vero cell monolayers attached to 12-well plates and incubated for an hour. Titers of the same sample from plaque assays on each cell type differed by ≤0.5 log₁₀ PFU/ml, indicating that infectivity was similar on BHK and Vero cells. After incubation, each cell monolayer was overlaid with a 2-ml solution of 0.8% agar (liquefied 10% agar, ultrapure agarose [Invitrogen] diluted in warm MEM) and allowed to solidify. The plates were incubated at 37°C for 48 or 72 h. Cell monolayers were then fixed with 4% formalin for 30 min, agar plugs were removed, and 0.025% gentian violet (Sigma) in 30% ethanol was overlaid in each well to stain viable cells. Viral titers were recorded as the reciprocal of the highest dilution where plaques are noted. The limit of detection of the assay was 800 PFU/ml (2.9 log₁₀ PFU/ml).

Relative mutation frequency in molecular versus plaque clones.

To determine the proportion of RNA sequences in the population that are capable of plaquing on Vero cells, the mutation frequency in molecular clones, representing all RNAs in the supernatant, was compared to the frequency in RNAs isolated from plaques. Either CHIKV RNA was extracted from cell culture supernatants (molecular clones), or a pipette tip was plunged through agarose overlays to harvest RNA from single plaques (plaque clones). RT-PCRs were conducted on supernatants or plaque clone-derived RNAs as described above. Tips plunged in areas where plaques were not visible were always RT-PCR negative. For molecular clones, purified amplicons from supernatants were cloned into TOPO pCRII vectors (Invitrogen) and transformed into Escherichia coli. Bacterial colonies containing the E1 insert were individually grown, and the E1 amplicon in plasmids was sequenced with the same CHIKV-specific primers used for RT-PCR with conventional techniques. For plaque clones, RT-PCR amplicons from each plaque were sequenced directly.

Genome-to-PFU ratio.

The number of CHIKV genomes in passages of known titer was determined by quantitative real-time RT-PCR. Viral RNA from passaged CHIKV was extracted using Trizol. The TaqMan RNA-to-C_T one-step quantitative RT-PCR kit (Applied Biosystems) was used to quantify the number of genomes in each passage. Each 25-μl reaction mixture contained 2.5 μl of RNA, 100 μM each primer (forward primer, 5′-TCACTCCCTGTTGGACTTGATAGA-3′; reverse primer, 5′-TTGACGAACAGAGTTAGGAACATACC-3′), and 25 pmol of probe 5′-(6-carboxyfluorescein)-AGGTACGCGCTTCAAGTTCGGCG-(black hole quencher)]-3′ as previously described by Lanciotti et al. (24). Reverse transcription was performed at 50°C for 30 min and 95°C for 10 min, followed by 40 cycles at 95°C for 15 s and 60°C for 1 min. A standard curve (y = −3.115x + 42.553, R² = 0.96) was generated using two replicates of in vitro-transcribed purified genomic RNA from a serially 10-fold diluted marked clone. The number of genomes in the marked clone RNA was calculated by dividing the mass (measured by spectrometry [Thermo Scientific Nanodrop]) by the genome molecular mass. Quantification of CHIKV genomes in the passages was calculated by comparing the threshold cycle (C_T) values of the samples to the standard curve. Genome values are expressed as the average of two wells. Genome/PFU ratios are expressed as the mean number of genomes/mean number of PFU for four replicates using values from two PFU concentrations.

Competition fitness assays.

By modifying methodologies from Coffey et al. (6), the relative fitness of passaged CHIKV was compared to that of the marked clone via the direct competition of the two populations under identical conditions. The ablated SacII marker in the clone was used to differentiate marked clone-derived RNA from parent (progenitor of passages) RNA. To confirm the stability of the marker, marked clone stocks were passaged five times in BHK or C6/36 cells and once in colonized Aedes aegypti mosquitoes. The marker did not revert to the parent sequence after any of these passage series. Artificial mixtures of known ratios of the two competitors revealed that the assay could reliably detect the correct ratio when one competitor was up to 10 times more abundant than the other. To control for cell-type-specific effects and to include an internal control, since the replication kinetics of the marked clone versus those in the parent (see Fig. S1 in the supplemental material) were slightly lower in C6/36 cells, marked clone was also competed against parent BHK p1, HeLa p1, and C6/36 p1 cells. Marked clone and passaged competitors were mixed in a 1:1 ratio and then inoculated in triplicate wells in BHK, HeLa, or C6/36 cells at ∼0.01 PFU/cell. Forty-eight hours p.i., RNA was extracted from cell culture supernatants and amplified by RT-PCR using primers forward 5′-GTGCGGCTTCTTCAATATG-3′ and reverse 5′-CACTGTTCTTAAAGGACTC-3′ flanking the marker. The abundance of each competitor was measured as the intensity of restriction enzyme-digested DNA fragments. Two hundred nanograms of purified amplicons was digested with SacII (Roche) for 2 h at 37°C. The amount of marked clone (uncut) or passaged (cut) DNA in ethidium bromide-stained bands in 1% agarose gels was quantified for each sample by densitometry (AlphaInnotech). DNA band intensities were not corrected for size differences or heteroduplexes since the raw intensity of the sum of the parent amplicon digested with SacII was within 95% of the intensity of the undigested parent fragment. Amplicons from inoculum mixtures were always included on each gel adjacent to lanes containing supernatants from competitions. Fitness is represented as the output/input ratio of the passaged/marked clone CHIKV. A fitness value of >1 indicates that the passaged clone is more fit than the marked clone.

Mutation frequency estimations.

The CHIKV envelope E1 gene was selected for sequencing to estimate mutation frequencies in mutant spectra from individual passages. The E1 gene, which encodes a glycoprotein on the surface of the virion, was selected since its variance in natural isolates is greater than that for the nonstructural genes. E1 mutations were significantly more frequent than NSP4 polymerase mutations in sequences from the same population (e.g., for serial BHK p7 cells, 0.9 for E1 versus 0.4 for NSP4). A mean of 71 CHIKV E1 sequences (∼90,000 nucleotides/population) were cloned into TOPO vectors and sequenced. Mutation frequencies were determined by dividing the number of nucleotide polymorphisms in all clones, where polymorphisms at the same genetic locus on multiple clones were counted multiple times, by the number of nucleotides sequenced and then multiplying by the E1 gene length (1,317 nt). Since multiples were rare, including them in the counts did not significantly affect frequency measures. Each frequency estimated from molecular cloning was corrected by subtracting the background mutation frequency, defined as the mutation frequency in TOPO-cloned plasmid DNA sequences. In vitro-transcribed marked clone RNA subjected to the same RT-PCR and TOPO cloning demonstrated a mutation frequency that was not significantly different from that for cloned plasmid (data not shown), indicating that the RT-PCR was not a significant source of mutational error.

Genetic distance from consensus.

The number of mutations by which each RNA sequence differs from the consensus of the population was measured across the same 717-bp portion of the E1 gene. The percentages of sequences that differ by 0, 1, and ≥2 nucleotides from the consensus are represented; clones with more than 2 nucleotide differences compared to the consensus were rare. Mutations encoding a stop amino acid and deletions were also rare; nevertheless, sequences with these mutations were included in the distance measures. Mean pairwise distances between p1 and p7 were calculated using the MEGA (version 4.0) program (45).

Neutralization escape and escape mutations.

To measure whether passaging affects resistance of populations to neutralization by polyclonal antibody, ∼4 log₁₀ Vero cell PFU/ml of p1 or p7 CHIKV was incubated in duplicate for 1 h (for two replicates) at 37°C with a concentration of anti-CHIKV polyclonal antibody that was expected to neutralize ∼80% of p1 plaques. Virus from each p1 or p7 was inoculated on the cell type in which passage occurred; since CHIKV does not plaque on C6/36 cells, viruses from passages on C6/36 cells were inoculated onto BHK cells. CHIKV samples without antibodies were included as controls. After 1 h, virus-antibody or virus-only mixtures were overlaid on confluent monolayers, and the plaques under agar overlays were counted after 3 days. The percentage of particles resisting neutralization was calculated by dividing the number of CHIKV plaques in wells with antibody by the number of plaques in wells without antibody and multiplying by 100.

Sequences of 10 plaque clones/population that escaped neutralization were compared. A 1-kb portion of the E2 gene was amplified from RNA isolated from plaques using primers forward 5′-GAAGCGACAGACGGGACG-3′ and reverse 5′-GCCTCTTGGTATGTGGCCGC-3′. This genome region was selected as it contains described alphavirus neutralization escape epitopes (2, 53). Plaque clone isolations, RNA extractions, and sequencing were performed as described above. The amino acids at each position were determined by analyzing nucleotide peaks in chromatograms; nucleotide mixes were recorded where ≥2 peaks occurred at a given locus in an otherwise clean chromatogram sequence.

Antiviral compound resistance.

p1 and p7 populations were incubated with a cocktail of antiviral compounds, and the mean titers of CHIKV populations that survived antiviral treatment were compared to those of nontreated populations. A cocktail was used to combine the activities of multiple antiviral compounds that may exert their antiviral effects via different mechanisms. Two hours before infection, HeLa cell growth medium was supplemented with 100 μM ribavirin, 50 μg/ml 5-fluorouracil, and 50 μM azacytidine (Sigma), doses that reduce CHIKV titers by ∼3 log₁₀ PFU/ml (see Fig. S3 in the supplemental material). HeLa cells were inoculated in triplicate with p1 or p7 CHIKV at 0.1 PFU/cell. After 1 h adsorption, the antiviral compound cocktail was supplemented in the cell growth medium. The mean titers of CHIKV harvested 72 h p.i. were compared using titrations. Percent resistance for a passage was determined by dividing the mean CHIKV titer in antiviral compound-treated wells by the mean titer in compound-free wells and multiplying by 100.

Statistical analyses.

Mutation frequencies, genome-to-PFU ratios, neutralization, and antiviral compound resistance comparisons across populations were evaluated using χ² tests with Yates' corrections. Genetic distances were compared with Mann-Whitney tests. Mean fitness values of p7 virus versus marked clone to p1 virus versus marked clone in competition assays were compared using Student's t tests. All statistical tests were conducted using GraphPad Prism software. P values of >0.05 were considered nonsignificant (ns).

CHIKV fitness increases after serial and alternating passage.

To determine how serial or alternating passage associates with CHIKV fitness, competition assays comparing p7 and p1 populations were conducted in the cell types in which passaging occurred (Fig. 1). Although marked clone was sometimes slightly less fit than p1, comparing each p7 to its p1 progenitor shows changes in relative fitness due to passaging. Competition assays indicated that fitness increased significantly for all serially passaged populations when the competition was conducted on the cell type in which serial passage occurred. The effect of serial vertebrate passage was reduced fitness on C6/36 cells. The serial C6/36 CHIKV p7 population showed moderate fitness gains on BHK cells and no change in fitness on HeLa cells. Alternately passaged CHIKV from both the C6/36 ↔ BHK and the C6/36 ↔ HeLa series was significantly more fit than the p1 progenitor (p1 C6/36 cell population) on both cell types where alternation occurred. The magnitude of fitness gains (p1 to p7 increase) for alternately passaged CHIKV was nearly as large as (C6/36 ↔ BHK versus serial BHK on BHK cells) or even greater than the fitness gains after serial passage (alternating series versus serial series for all other competitions). These results, novel in the context of CHIKV, confirm previously published findings that alternately passaged arboviruses have increased fitness on both passaged cell types. With the fitness relationships defined for CHIKV, we next sought to evaluate whether genetic diversity correlates with fitness.

Genetic diversity increases the most after serial passage and moderately with alternating passage.

Mutation frequencies were estimated from nucleotide polymorphisms in sequences from the CHIKV E1 gene, a gene encoding a surface glycoprotein that exhibits greater intrahost variability than nonstructural genes. A mean of 71 CHIKV E1 RNAs (∼90,000 nucleotides/population) were analyzed. Common E1 mutations within a population or between passages were rare. When they did occur, common minor variants in the mutant spectrum of the p1 population versus the p7 population from the same passage series were observed at low frequency (in <3% of clones; data not shown). The mutation frequency in the mutant spectrum of each serial p7 population was increased compared to the frequency for its p1 progenitor (although not significantly for serial HeLa passage). Serially BHK passaged CHIKV experienced the largest magnitude of increase (6-fold), with an estimated 0.88 mutations/E1 being found at p7 compared to 0.13 at p1 (χ², P < 0.0001). Mutations in the serial p7 C6/36 population were five times more frequent than those in the p1 population (mutation frequency, 0.32 versus 0.06; χ², P < 0.02). In contrast, the mutation frequencies of alternately passaged CHIKV did not augment to levels observed after serial passage (e.g., 0.06 for C6/36 ↔ BHK p7 versus 0.88 for serial BHK p7), and for the C6/36 ↔ BHK series there was no increase in mutation frequency at p7 compared to that at p1 (0.11 versus 0.06; ns). While the mutation frequency for the C6/36 ↔ HeLa p7 population increased compared to that at p1 (P < 0.02), the absolute frequency did not reach the levels observed after serial HeLa passage. Together with the fitness data from competition assays, our results indicate that augmented fitness does not necessarily correlate with the greatest magnitudes of increases in genetic diversity.

Serially passaged populations contain more nonviable mutants than alternately passaged populations.

Although the fitness of alternately passaged CHIKV significantly increased (Fig. 1) without major increases in genetic diversity (Fig. 2), our prediction that diversity is a requisite for augmented fitness is not supported. This may mean that much of the genetic diversity observed in the mutant spectrum from a serially passaged CHIKV population represents sequences that do not constitute viable particles, while alternating passage between different cell types requires the population to predominantly maintain highly infectious genomes. Since molecular cloning cannot distinguish between viable versus nonviable RNAs because all RNAs in a supernatant are amplified, cloned, and sequenced, we next determined the ratios of mutation frequencies in molecular versus plaque-purified clones, where each population was plaqued on Vero cells, a cell type in which none of the passage series were conducted. Since isolation of virus from plaques requires viral replication, this approach permitted us to understand indirectly what fraction of the RNAs in populations were viable (Fig. 3 A). A mean of 91,189 (range, 41,773 to 141,805) nucleotides were sequenced per population of molecular or plaque clones. For the BHK p1 population, six times more mutations were detected in molecular clone sequences than in plaque-purified virus sequences. The ratio for the serial BHK p7 population was 26:1, indicating that more than four times more of the mutations observed in this population were nonviable compared to its p1 progenitor population. The ratio for the alternating C6/36 ↔ BHK p7 population, on the other hand, was no different from that for the p1 population (6:1), demonstrating that alternating invertebrate ↔ vertebrate cell passage constrains the population to maintain more replication-competent, higher-fitness variants. These data support the concept that releasing CHIKV from its alternating cycle via serial vertebrate cell passage allows CHIKV to more deeply explore sequence space, manifest as more RNAs with nonviable mutations. Alternating passage may cleanse these nonviable mutations from the population; mutations permitted in a vertebrate cell may be removed during invertebrate cell infection, and vice versa. If this is the case, then the genome/PFU ratio, representing the total packaged RNA versus the replication-competent RNA capable of forming plaques, should be greater for a serially passaged population than for an alternately passaged population. To assess this, we performed quantitative real-time RT-PCR to determine directly the number of RNA genomes in passages with measured PFU titers (Fig. 3B). Indeed, the mean ratio for the serial BHK population was 1,184:1, while the ratio for the alternating C6/36 ↔ BHK population was 5-fold less, 223:1. Adaptation to alternating passage may therefore require more replication-competent genomes at the cost of expanded genetic diversity. In contrast, serial passaging results in the accumulation of more mutations, many of which result in nonviable RNAs. Nevertheless, this abundance of nonviable RNAs in the serial BHK population does not excessively occur to the point of a significant loss of infectivity, since fitness is still increased after passage (Fig. 1).

Genetic divergence from consensus increases with serial and alternating passage; magnitude of increase is greatest after serial vertebrate passage.

While mutation frequency (Fig. 2) gives a global estimate of the total number of mutations present within the population, distance measures the divergence of individual RNAs from the consensus sequence. Distance was calculated in two ways: (i) determining the percentage of molecular clones in each population that differ by 0, 1, or ≥2 mutations from the consensus sequence (Fig. 4) and (ii) calculating the mean pairwise distance of each p1 sequence compared to each p7 sequence (Table 1). The mean number of sequences per population for distance calculations was 94 (range, 63 to 124), corresponding to 67,577 nt/population, where the length of each RNA sequence analyzed was 717 bp (approximately 1/2 of the E1 gene). Serial CHIKV passage in C6/36 cells significantly increased the proportion of RNAs that were 1 or ≥2 mutations distant from consensus (P < 0.001), and the serial BHK p7 population was more distant from the consensus than its p1 progenitor (although not significantly). The serial HeLa p7 population was not more distant than its p1 progenitor. Both alternately passaged CHIKV series became more distant from the consensus than their C6/36 p1 progenitor (nearly significant at P < 0.07 for C6/36 ↔ BHK p7, P < 0.01 for C6/36 ↔ HeLa p7).

Mean pairwise distance measures (Table 1) reveal that serial vertebrate passage generated RNA that each differ by nearly 1 mutation (0.9 ± 0.12 mutation for BHK p7, 0.8 ± 0.08 mutation for HeLa p7) from consensus; values for serial C6/36 p7 and both alternately passaged series were more moderate (0.3 to 0.4), indicating that less than 50% of clones in the population contained a mutation from the consensus. Serial C6/36 passage produced CHIKV RNAs that are farther from consensus than p1, but the population is still less distant than serial BHK and HeLa populations are from their respective p1 progenitor populations. These results confirm that serial passage (in vertebrate cells especially) loosens constraints on increases in genetic distance, while alternating passage restricts increases in distance to retain RNAs that are more genetically similar to each other and the consensus.

Serially passaged populations are more adaptable to novel environmental conditions than alternately passaged populations.

To address whether the enhanced ability of serially passaged CHIKV populations to explore sequence space (higher diversity and distance) confers increased adaptability to new environmental conditions compared to that for populations adapted to alternating passage, we tested CHIKV populations under two new conditions by measuring their ability to (i) withstand antiviral compounds and (ii) escape neutralizing antibody.

(i) Antiviral compounds.

We expect that an RNA virus population with high genetic diversity is more likely to contain variants conferring resistance to antiviral compounds and therefore resist compound treatment better than a less diverse population. To address this prediction, CHIKV populations were examined for their relative sensitivity to a cocktail of base analogs: ribavirin, 5′-fluorouracil, and 5′-azacytidine (Fig. 5). These compounds act primarily as RNA mutagens in tissue culture but have been shown to have pleiotropic effects (ribavirin, in particular) (9) for which multiple mechanisms of resistance are possible (31, 34, 35, 41, 59). Resistance of passaged CHIKV populations was measured by comparing the percentage of particles in p1 and p7 populations that survived treatment of HeLa cells with antiviral cocktails at predetermined concentrations that reduced the titer by ≈3 log₁₀ PFU/ml each (see Fig. S3 in the supplemental material). Serial and alternately passaged populations resisted antiviral treatment significantly better than their p1 progenitors (Fig. 5; 1.3% for serial HeLa versus 0.2% for p1 HeLa [P < 0.0002]; 0.02% for C6/36 ↔ HeLa versus 0.004% for p1 C6/36 [P < 0.02]). Furthermore, the serial HeLa population was significantly more resistant to these compounds than the alternating C6/36 ↔ HeLa population (P < 0.0009).

(ii) Neutralization resistance.

To determine whether increased genetic diversity confers enhanced escape from neutralization by antibody, we evaluated the relative capacity of passaged populations to plaque on vertebrate cell types where passage occurred after incubation with a constant concentration of neutralizing antibody (Fig. 6). The percentage of CHIKV particles in serial BHK or serial HeLa p7 populations that resisted neutralization was significantly greater than the percentage of resistant particles in respective p1 progenitor populations (Fig. 6A, 5.2% for serial BHK versus 0.002% for p1 BHK; Fig. 6B, 0.1% for serial HeLa versus 0.03% for p1 HeLa; for both comparisons, P < 0.0001). Although serial C6/36 (1.6%) or alternating C6/36 ↔ BHK (0.3%) also resisted neutralization better than BHK p1, serial BHK p7 exhibited the greatest resistance (serial BHK versus alternating C6/36 ↔ BHK and serial C6/36; for both comparisons, P < 0.0001). Since serial BHK is more genetically diverse than serial C6/36 or C6/36 ↔ BHK (Fig. 2), these results suggest that the presence of more variant RNAs in that population increases the likelihood that some contain neutralization escape mutations, which imparts increased neutralization resistance. To obtain direct genetic evidence to confirm this hypothesis, we sequenced RNAs flanking a portion of the E2 gene that contains described alphavirus neutralization escape loci for 10 plaque-purified clones/population from the BHK series (Table 2). Nearly all RNAs from the progenitor p1 BHK population and the passaged populations (serial BHK, serial C6/36, and alternating C6/36 ↔ BHK) contained mutations at E2-229 and E2-231, suggesting that amino acid changes at these positions are requisite for escape; the E2-231 locus was previously identified in escape of VEEV (2). Sequences from the three passaged populations contained mutations at additional E2 amino acid positions (positions 232, 233, and 234) that were not present in any p1 BHK sequences. Positions 232 and 234 were previously identified to be escape mutations for Ross River virus (53). Compared to the alternating C6/36 ↔ BHK population and the serial C6/36 population, the serial BHK population contained more variant amino acids at position 232 and sequences from both serial BHK and serial C6/36 plaque clones had more variation at position 233 than alternating C6/36 ↔ BHK. At position 234, the serial BHK population expressed a total of 7 possible amino acids, while the serial C6/36 and alternating C6/36 ↔ BHK populations presented a total of only 3 amino acid variations. Further, none of the amino acids at position 234 in the serial BHK population were observed in the other two passaged populations. Although the neutralization epitopes have not been described for CHIKV, the location of high-frequency mutations clustered around known alphavirus neutralization sites in many of the 10 plaque clones from each passage supports their role in antigenic escape. Furthermore, no other mutational clusters were observed elsewhere in our E1 and E2 sequences from passaged populations. These results confirm that increased genetic diversity in the serial BHK population is responsible for enhanced escape from neutralization via a greater number of RNAs encoding variant amino acids.