Evolution of the influenza A virus genome during development of oseltamivir resistance in vitro

Abstract

Influenza A virus (IAV) is a major cause of morbidity and mortality throughout the world. Current antiviral therapies include oseltamivir, a neuraminidase inhibitor that prevents the release of nascent viral particles from infected cells. However, the IAV genome can evolve rapidly, and oseltamivir resistance mutations have been detected in numerous clinical samples. Using an in vitro evolution platform and whole-genome population sequencing, we investigated the population genomics of IAV during the development of oseltamivir resistance. Strain A/Brisbane/59/2007 (H1N1) was grown in Madin-Darby canine kidney cells with or without escalating concentrations of oseltamivir over serial passages. Following drug treatment, the H274Y resistance mutation fixed reproducibly within the population. The presence of the H274Y mutation in the viral population, at either a low or a high frequency, led to measurable changes in the neuraminidase inhibition assay. Surprisingly, fixation of the resistance mutation was not accompanied by alterations of viral population diversity or differentiation, and oseltamivir did not alter the selective environment. While the neighboring K248E mutation was also a target of positive selection prior to H274Y fixation, H274Y was the primary beneficial mutation in the population. In addition, once evolved, the H274Y mutation persisted after the withdrawal of the drug, even when not fixed in viral populations. We conclude that only selection of H274Y is required for oseltamivir resistance and that H274Y is not deleterious in the absence of the drug. These collective results could offer an explanation for the recent reproducible rise in oseltamivir resistance in seasonal H1N1 IAV strains in humans.

FIG 1

Experimental trajectories. (A) A/Brisbane/59/2007 (H1N1) was serially passaged in MDCK cells 13 times in the presence of escalating micromolar concentrations of oseltamivir (open boxes) or no drug (shaded boxes). (B) In a second experiment, A/Brisbane/59/2007 (H1N1) was subjected to similar growth conditions. Additionally, oseltamivir was withdrawn after passage 10 (P10) and P13, and higher concentrations of oseltamivir were used from P14 to P18. (C) Viral quantities were compared for no-drug versus oseltamivir-treated samples from experiment 1. Horizontal bars indicate the average values for samples. No significant differences are observed between input (P, 0.75 by the Mann-Whitney U test) or output (P, 0.28 by the Mann-Whitney U test) copies. (D) Viral amplification was compared for no-drug samples and oseltamivir-treated samples; again, no significant difference is observed (P, 0.99 by the Mann-Whitney test).

FIG 2

(A) Analysis of error frequency distribution. RNA error controls were processed and sequenced in a manner identical to that for viral population samples. These data were used to apply a filtering threshold to the experimental sequence data. The distribution of mutations in the error control data was highly skewed to very low frequencies. By filtering data to remove all variation of ≤0.17%, ≥95% of errors could be removed (arrow), allowing for SNPs to be called in the population with high confidence. (B) Allele frequency spectrum of A/Brisbane/59/2007. Only a very small number of sites in the 13.5-kb IAV genome exhibited polymorphisms or changes during serial passages in the absence of oseltamivir. P4 (no drug) is designated by red columns and P13 (no drug) by black columns. (C) Serial passaging with increasing doses of oseltamivir did not increase the sequence diversity of influenza virus. Bars outlined in red, P4 (no drug); bars outlined in black, P13 (with oseltamivir). Data for experiment 1 are shown; data for experiment 2 are similar.

FIG 3

IAV population diversity and divergence in the presence and absence of oseltamivir. The nucleotide diversity of IAV populations was measured for each passage of in vitro growth. Population divergence was measured by comparing two populations at the same passage using the F_ST statistic, a metric of population differentiation. (A) The presence of oseltamivir did not cause a significant change in the amount of variation in the total population. (B) Oseltamivir did not cause changes in variation in segment 6 (neuraminidase [NA]). (C) Nucleotide diversity was influenced by the amount of input virus and by fold amplification. Trends are shown by the dashed lines. R² (Pearson correlation), 0.49 for input virus versus nucleotide diversity and 0.46 for fold amplification versus nucleotide diversity. Data from experiment 1 are shown. (D) Two populations grown with and without oseltamivir were no more divergent than two populations grown without drug from the two experiments.

FIG 4

SNP frequency changes during in vitro growth with or without oseltamivir. The frequencies of all SNPs across the IAV genome were tracked for 13 (A and B) or 18 (C and D) passages. The presence of oseltamivir caused an increase in the number of SNPs that fixed during passaging. The effect of oseltamivir was especially pronounced in experiment 1, where a sudden increase in the frequency of a cluster of SNPs can be observed beginning at P6. However, the majority of SNP frequency changes were not affected by the presence of oseltamivir.

FIG 5

Frequency of the H274Y mutation versus the passage number in A/Brisbane/59/2007 grown in MDCK cells in the presence or absence of oseltamivir. At oseltamivir concentrations of >1 μM, the known resistance mutation H274Y emerged and fixed in the population in both experiment 1 (filled triangles) and experiment 2 (filled circles). Prior to drug treatment, the frequency of H274Y in the population was below 0.1%. The frequency of H274Y was below 0.1% in the absence of oseltamivir in both experiments (gray lines). Oseltamivir was added at P4 and beyond; the concentration (along the right y axis) is indicated by the heavy black line.

FIG 6

Genomewide distribution of selection coefficients. Selection coefficients for all SNPs across the genome were calculated using time-sampled changes in frequency. The distributions were centered on an s of zero, suggesting that most SNPs are neutral and are subject to genetic drift. Further, growth in oseltamivir had little effect on the distribution, suggesting that the genomewide selective environment is not altered by the presence of the drug.

FIG 7

Selection coefficients for neuraminidase in A/Brisbane/59/2007. Selection coefficients were calculated for the periods corresponding to the development of drug resistance (P4 onward) with no drug (A) and with increasing concentrations of oseltamivir (B). Data were compiled from experiments 1 and 2. The selection coefficients were mapped to the crystal structure of neuraminidase in complex with oseltamivir (PDB 3CL0). The H274Y resistance mutation is the most beneficial mutation in the population. The K248E mutation is adjacent to H274Y and is a target of positive selection with oseltamivir in one of two experiments. The spectrum of colors ranges from the lowest-ranked (violet) to the highest-ranked (red) selection coefficients.

FIG 8

Impact of oseltamivir withdrawal on H274Y frequency in A/Brisbane/59/2007 and correlation with IC₅₀s for neuraminidase. Shown is the frequency of the H274Y mutation for select samples in the second experimental trajectory with A/Brisbane/59/2007. After P13, the removal of the drug has no effect on H274Y frequency (open circles). The removal of the drug at P10, before H274Y reached fixation, did not lead to a significant change in H274Y frequency (open triangles). Mean oseltamivir IC₅₀s for influenza virus samples beyond P13 and P10 versus those for control virus (wild type) are given. Values were determined using a chemiluminescent neuraminidase activity inhibition assay. For samples in which H274Y remains fixed in the population (open circles), the IC₅₀ is >100-fold that for the wild-type control virus (means, 49.8 ± 6.8 nM versus 0.10 ± 0.01 nM; P, <0.001 by Student's t test). For samples in which the frequency of H274Y is ∼40% in the population (open triangles), the IC₅₀ is slightly but significantly greater than that for the wild-type control virus (means, 0.17 ± 0.02 nM versus 0.10 ± 0.01 nM; P, <0.001 by Student's t test).