Comparison of Mutation Patterns in Full-Genome A/H3N2 Influenza Sequences Obtained Directly from Clinical Samples and the Same Samples after a Single MDCK Passage

Human influenza viruses can be isolated efficiently from clinical samples using Madin-Darby canine kidney (MDCK) cells. However, this process is known to induce mutations in the virus as it adapts to this non-human cell-line. We performed a systematic study to record the pattern of MDCK-induced mutations observed across the whole influenza A/H3N2 genome. Seventy-seven clinical samples collected from 2009-2011 were included in the study. Two full influenza genomes were obtained for each sample: one from virus obtained directly from the clinical sample and one from the matching isolate cultured in MDCK cells. Comparison of the full-genome sequences obtained from each of these sources showed that 42% of the 77 isolates had acquired at least one MDCK-induced mutation. The presence or absence of these mutations was independent of viral load or sample origin (in-patients versus out-patients). Notably, all the five hemagglutinin missense mutations were observed at the hemaggutinin 1 domain only, particularly within or proximal to the receptor binding sites and antigenic site of the virus. Furthermore, 23% of the 77 isolates had undergone a MDCK-induced missense mutation, D151G/N, in the neuraminidase segment. This mutation has been found to be associated with reduced drug sensitivity towards the neuraminidase inhibitors and increased viral receptor binding efficiency to host cells. In contrast, none of the neuraminidase sequences obtained directly from the clinical samples contained the D151G/N mutation, suggesting that this mutation may be an indicator of MDCK culture-induced changes. These D151 mutations can confound the interpretation of the hemagglutination inhibition assay and neuraminidase inhibitor resistance results when these are based on MDCK isolates. Such isolates are currently in routine use in the WHO influenza vaccine and drug-resistance surveillance programs. Potential data interpretation miscalls can therefore be avoided by careful exclusion of such D151 mutants after further sequence analysis.