In this study, case C, who was treated with oseltamivir, most likely infected case D, who was receiving prophylactic doses of oseltamivir. Case C was sampled prior to receiving oseltamivir, while case D was sampled five days after oseltamivir prophylaxis commenced. It has been suggested that resistant viruses carrying the NA H275Y substitution suffer no fitness losses compared with those with NA 275H wild-type in the absence of drug intervention , and since samples were taken from cases A-C prior to oseltamivir-treatment, it is not possible to conclude whether resistance was acquired during their treatment and resistant virus transmitted to case D, or whether resistance emerged during prophylaxis in case D.
However, several pieces of evidence support case C infecting case D (although it cannot be ruled out that the infection originated from case A or B). Firstly, cases A-C tested positive for A(H1N1)pdm09 influenza on June 2nd and they were subsequently isolated. At this time case D remained negative for A(H1N1)pdm09 influenza. Case C and D were in a close, personal relationship and case D had only been in contact with case C (rather than cases A or B) when case D developed symptoms. Secondly, looking to the consensus sequences from patients C and D lends strong support to their epidemiological linkage. For all 8 segments it holds true that the consensus sequences of the virus populations are identical, with the exception of the H275Y conferring mutation in the NA gene from case D. Additionally, performing a BLASTn search against the GenBank nucleotide database (online version as of 14/3 2012, http://www.ncbi.nlm.nih.gov/blast) revealed that the HA, NP and NS genes were uniquely identical, meaning that they matched no other sequences in GenBank. The PB1 sequence matched only a single entry completely, and PB2 matched only 3 other entries (none of which share a patient origin with the PB1 hit), leaving only the PA and M genes matching multiple, globally distributed, entries in the database. For NA the sequences derived from patient C also matched multiple hits whereas those encoding resistance in patient D had a single perfect match.
To date, few studies have reported on the intra-host evolutionary dynamics of influenza viruses. One major reason for this has been the limited capacity to detect low frequency virus variants using conventional methods. However, new deep sequencing technologies offer an efficient and economic way to obtain a snapshot of the entire virus population. During this study we demonstrated the diversity between intra-host variants in cases of oseltamivir–sensitive and oseltamavir–resistant A(H1N1)pdm09 infections. While the master sequences of these cases remained identical, but for the mutation encoding the NA H275Y substitution, the subpopulations of the virus quasispecies varied significantly. Interestingly, case D contained a mixed population at nucleotide 823 on the NA gene, resulting in 3-14% (depending on the sequencing platform used) of the viruses remaining sensitive to oseltamivir. This result is similar to other reports of mixed populations occurring in H275Y variant A(H1N1)pdm09 viruses , and together they suggest either that a small proportion of the sensitive viruses survive the treatment, or that the mutation conferring oseltamivir sensitivity rapidly re-establishes itself in the virus population after the bottleneck. These data also suggest that the NA 275Y containing virus may be less fit than the sensitive NA 275H virus, a hypothesis that is supported by experiments in a ferret model , although contradictory to the 275Y variant outgrowing the 275H wild-type in MDCK cells. However, other authors did not find the resistant 275Y virus to be attenuated in other animal models . In order for the oseltamivir-resistant virus to be stable and more fit than the non-resistant counterpart, other changes in the NA or HA genes may have to be present as has been shown for the former seasonal A(H1N1) virus during 2008 .
In addition to harboring a subpopulation of oseltamivir-sensitive viruses, case D was noteworthy for the occurrence of seven additional intra-host variants. Only four of the seven mutations have been reported in previous influenza studies, three of which have been associated with A(H1N1)pdm09 virus (including that conferring H275Y substitution). Firstly, Baz et al.  reported that a NA S153N substitution occurred concomitantly with the H275Y substitution in an oseltamivir-resistant virus. Secondly, the mutation at residue 600 of the PA gene appears in numerous consensus sequences submitted to the NCBI database. Thirdly, the mutation conferring HA I282V substitution in A(H5N1) influenza virus, was reported by Chakrabarti et al. , but it was not seen in combination with NA H275Y substitution so is unlikely to be linked to oseltamivir-resistance. The occurrence of these intra-host variants in case D may be related to maintenance of virus fitness. However, these mutations occur in non-conserved regions of the influenza A virus genome [18–21], suggesting that they may have arisen as deleterious mutations during a population expansion, with no direct correlation to the NA H275Y conferring mutation.
Given one nucleotide change per 10,000 nucleotides during replication and that most infections are caused by 10–1,000 virions which likely possess varying numbers of nucleotide differences in their genomes, one can expect that each influenza A virion is a possible intra-host variant. However, we identify relatively few variant sites, probably because currently available sequence analysis software do not allow robust intra-host variant analyses and manual curation is necessary. Hence, we believe that with improved bioinformatic tools we would detect more subpopulation variation in our sample set. However, while our levels of intra-host variation seem low, they are consistent with other second-generation sequencing studies. Ghedin et al.  used a 10% cut-off for including a variant. This conservatively high cut-off ensured that variants were not determined by substitution errors (0.03% per nucleotide on the Roche GS FLX). Using this cut-off, the authors detected only 10 sites containing variation throughout the genome. Similarly Ramakrishnan et al.  detected between 0 and 10 sites with variation, depending on the sample.
The aim of our study was to characterize the intra-host variation occurring in the virus populations in a transmission chain. There are, however, some limitations to the methodology and processes that can influence the accurate quantification of subpopulation variants: i) cell culturing processes; ii) RT-PCR amplification; iii) emulsion PCR (emPCR) and bridge PCR (in GS FLX and GAIIx sequencing, respectively); and iv) the sequencing process.
Firstly, second-generation sequences for cases C and D were determined from viral culture isolates rather than from clinical specimens directly. It is possible that some of the substitutions observed may have occurred during passage of virus in cells, although reports of genetic variation occurring during MDCK cell passage are predominantly linked to changes in the NA and HA genes . MDCK cell cultures have been reported to select for new variants, particularly for mutations leading to enhanced growth of virus in these cells . However, given that both cases C and D were cultured in much the same manner, with only one passage more required for case D, they allow direct comparison, with the H275Y conferring mutation being the primary difference between the samples. Additionally, reports of oseltamivir-resistant viruses outgrowing sensitive viruses in MDCK and rhesus monkey kidney cells might suggest that the 3-14% (GAIIx and GS FLX results, respectively) of oseltamivir-sensitive virus that we detected may be an under-estimate of persisting oseltamivir-sensitive viruses in the case D patient [9, 16]. Thus, a direct analysis of clinical specimens for the polymorphisms at the sites listed in Table 3 would be required to confirm these observations. However, due to the low virus concentration in the case C and D clinical samples, we were unable to convert the uncultured samples into libraries. Future generations of sequencing platforms, such as the PacBio RS, requiring lower starting amounts of template and the ability to sequence single influenza segments may resolve these issues.
Secondly, PCR-induced errors were minimized in the preparation of the One-Step RT-PCRs and library amplification steps by using Platinum and Phusion (respectively), both high fidelity polymerases. Further, for the PCR replication step (5 separate reactions for the Illumina GAIIx library template), a small number of amplification cycles (18 cycles total) were used in the preparation of the amplified Illumina GAIIx library.
However, despite the aforementioned precautions, evidence of amplification biases were still evident in the variant sites detected in the NA gene of case D. The H275Y conferring mutation was present in 86% and 97% of sequences using the GS FLX and GAIIx platforms, respectively. This difference is likely to represent biases introduced during RT-PCR or emPCR/bridge PCR, rather than during cell culturing given that these libraries were constructed based on the same cell culture products. Repeating/duplicating the sequencing process could resolve this discrepancy. It should be noted that the GAIIx library was built using 5 replicate RT-PCR reactions; hence this could represent a more reliable variant frequency measurement. The differences could also represent biases introduced due to the differences in GS FLX and GAIIx library amplification and sequencing methods.
Moreover, we detected the NA S153N conferring mutation in 7% of the sequences obtained on the GAIIx platform but not those from the GS FLX platform. This could be explained by: i) a nearly 3-fold decrease in coverage of this site using the GS FLX platform (51-fold coverage compared to 138-fold coverage on the GAIIx platform); ii) the S153N conferring mutation arose as a sequencing error on the GAIIx sequencer; or iii) preferential PCR or sequencing bias in the GS FLX library for the dominant 153S encoding sequences. Given that the S153N encoding variant was present in 10/138 sequences, post-filtering, this anomaly is most likely explained by hypotheses i) or iii).
Given that oseltamivir treatment would cause a virus bottleneck, one might expect to see a reduction in the genetic diversity of the virus population . However, this is not what was observed in case D, as was seen in another A(H1N1)pdm09 study . Specifically, Ghedin et al.  compared virus sequences in specimens taken from an individual prior to and after the emergence of oseltamivir-resistance and detected no variation in the oseltamivir-sensitive sample, compared to 10 nucleotide positions showing variation in the resistant sample. There were no common positions of nucleotide variation, excluding the H275Y conferring mutation, between the resistant virus specimens analyzed in the two studies.
How might one explain these observations? Previously it was shown that RNA viruses can undergo rapid evolution following bottleneck events, leading to the accumulation of neutral and deleterious mutations and, rarely, the emergence of biologically fit variants [23, 24]. Therefore, in the present study, it would appear that the introduction of oseltamivir either as treatment to case C or as prophylaxis to case D created selective pressure that caused the drug-resistant viruses to rapidly accumulate in case D. However, as we had only one collection date for each case in this study, we could not determine whether or not the levels of virus variants remained constant throughout the infections. For future studies of intra-host variation it will be interesting to investigate samples taken over time-series, both within single infections to monitor the variation within an individual, and within longer transmission chains. The latter will enable characterization of whether the unexpected diversity post-bottleneck is maintained, or lost.
In conclusion, we used deep sequencing to characterize the diversity of influenza viruses within a transmission chain that involved the development of oseltamivir resistance. As in previous studies, we detected the presence of low levels of oseltamivir-sensitive virus in the case harboring oseltamivir-resistant viruses, suggesting either survival through the treatment bottleneck, or rapid re-evolution of the sensitivity-conferring mutation post bottleneck. In addition, we paradoxically detected higher levels of intra-host variation in the case harboring oseltamivir-resistant virus, compared to the case carrying oseltamivir-sensitive virus. This contributes evidence to support the hypothesis that post-bottleneck, virus population expansions result in the generation of greater levels of genetic diversity than are otherwise found in the population.