Introduction Reassortment is the process by which viruses carrying segmented genomes exchange gene segments. The reshuffling of genetic material achieved through reassortment supports rapid production of variant viruses that can be markedly different, genotypically and phenotypically, from the parental strains. The more gradual process of genetic drift, resulting from errors in genome replication, and the process of reassortment come together to generate vast genomic diversity among influenza A viruses. It is this diversity that, in turn, permits the rapid evolution of influenza viruses and the generation of novel pandemic and epidemic strains. The contribution of reassortment to the emergence of pandemic influenza viruses is well established: the 1957 and 1968 pandemics arose following reassortment events between avian and human influenza viruses that allowed novel HA subtypes to gain widespread circulation in the human population [1], [2], [3]. Reassortment furthermore played a prominent role in the creation of the H5N1 viruses that continue to circulate in poultry of Southeast Asia [4], and in the H1N1 swine influenza viruses that emerged in humans in April 2009 [5], [6]. Thus, epidemiological studies indicate that reassortment is an important means of viral diversification and often facilitates inter-species transmission. In addition to its role in pandemic influenza, phylogenetic studies have revealed the importance of reassortment between co-circulating viruses of the same subtype in generating a diverse pool of seasonal influenza viruses [7], [8], [9], [10], [11], [12], [13], [14]. This diversity in turn allows for the selection of variants that escape pre-existing immunity in the population and thereby cause widespread epidemics: evidence suggests that the unusually severe epidemics of 2003, 1951 and 1947 were each caused by strains generated through intra-subtype reassortment among co-circulating clades [8], [10]. Previous efforts to study influenza virus reassortment in the lab have been of three main types. First, beginning with the work of Lubeck et al. in 1979, several research groups have examined the phenomenon of segment mismatch, in which the gene segments of two differing strains are found to assort in a non-random fashion due to functional incompatibilities between the viral proteins or RNA segments [15], [16], [17], [18]. It is clear from this literature that strain differences between parental viruses limit the fitness of many reassortant progeny and thereby restrict the number of different genotypes that arise, or are detected, following co-infection. Thus, segment mismatch is a potent determinant of reassortment efficiency. Second, a number of risk assessment type studies have addressed the potential for variants with increased virulence or transmissibility to arise through reassortment between two strains of epidemiologic importance [19], [20], [21], [22], [23], [24], [25], [26], [27]. Third, since reassortment between circulating strains and the egg-adapted A/Puerto Rico/8/34 virus has been used since 1969 to generate vaccine seed strains that grow well in embryonated chicken's eggs, significant research effort has been put into optimizing this procedure [28], [29]. Research to date is lacking on the conditions of co-infection that are most favorable for reassortment, and it is therefore unclear under what circumstances we can expect to see novel influenza A viruses arising in nature. In part, this knowledge gap has arisen because, when one studies reassortment between two dissimilar strains, the effects of other parameters are confounded by those of segment mismatch. Herein, we report a novel method for the study of reassortment in the absence of segment mismatch and the application of this method to determine the baseline frequency of reassortment under unbiased conditions, and the impacts of infection dose and timing on this baseline. Results A system for the study of unbiased reassortment In order to obtain data that are not confounded by segment mismatch, we have designed an approach that employs a pair of phenotypically identical but genotypically distinct influenza viruses. Reverse genetics was used to introduce silent mutations into each gene segment of A/Panama/2007/99 (H3N2) [Pan/99] virus such that the segments of the resultant variant, or rPan/99var, virus can be distinguished from those of the rPan/99wt virus using molecular techniques (described below). The mutations introduced were selected carefully such that the rPan/99var viruses were not attenuated in growth relative to the rPan/99wt strain (Figure 1). As a result, all 256 different progeny that might arise following co-infection with rPan/99wt and rPan/99var viruses are expected to be of equal fitness. Because there are no selective pressures acting differentially on the various progeny strains, co-infection with rPan/99wt and rPan/99var viruses constitutes an unbiased system in which to study reassortment. 10.1371/journal.ppat.1003421.g001 Figure 1 rPan/99 wt and var viruses show similar growth phenotypes in MDCK cells and guinea pigs. A) MDCK cells were infected at an MOI of 0.001 PFU/cell with the indicated viruses. For rPan/99wt-HIS virus, n = 6 dishes; for rPan/99var2-HA virus, n = 3 dishes. B) Groups of three guinea pigs were inoculated intranasally with 1000 PFU of the indicated virus. Virus titers in nasal washings are plotted vs. day post-infection. Average values +/− standard deviations are shown. As shown in Figure 2, the silent mutations differentiating rPan/99wt and rPan/99var viruses allow the full genotypes of progeny arising from mixed infections to be determined using high resolution melt (HRM) analysis [30]. This method exploits the fact that sequence differences between two double stranded DNA (dsDNA) molecules confer differences in melting properties. These differences in melting properties can in turn be detected as changes in fluorescence when dsDNAs labelled with a saturating fluorescent dye are heated (e.g. from 65°C to 95°C), since the dye will cease to fluoresce as the DNA melts into single strands. As described in more detail in the methods section, we have applied HRM analysis to clonal virus isolates derived from co-infection in order to identify each gene segment as either wt or var in origin. 10.1371/journal.ppat.1003421.g002 Figure 2 Identification of wild-type and variant virus gene segments by high resolution melt analysis. Examples of the Difference RFU (relative fluorescence units) curves generated by the Precision Melt Analysis software are shown for each vRNA segment. Curves colored red clustered with the rPan/99wt control and curves colored green clustered with the rPan/99var control. Reassortment occurs with high efficiency under unbiased conditions We have defined the baseline frequency of reassortment as the percentage of progeny viruses with reassortant genotypes that arises after a single cycle of replication, given high levels of co-infection at the cellular level and an absence of segment mismatch or other selection pressures that would promote parental genotypes over reassortant ones. To determine this baseline value, rPan/99wt-HIS and rPan/99var-HA viruses were used to co-infect MDCK cells at a multiplicity of infection of 10 PFU/cell of each virus. Epitope tagged versions of the wt and var strains were used so that the number of cells infected with each virus and the number of cells co-infected could be determined by flow cytometry. Co-infection rates of 99.4% were achieved in each of two independent samples. As shown in Figure 3, the resultant frequencies of reassortment were 87.6% and 89.2% (average = 88.4%). While this result indicates that reassortant viruses arise with high frequency under the unbiased conditions described, the theoretically optimal efficiency of 254/256, or 99.2%, was not achieved: parental progeny viruses were over-represented relative to the expected frequency of 0.8% (p 35 in 5/6 guinea pigs). Progeny virus isolates obtained from the nasal wash samples were then genotyped and scored as wt, var or reassortant. The results indicate that a delay of up to 12 hours between primary and secondary infections with rPan/99 virus does not reduce reassortment frequency (Table 2). In fact, higher levels of reassortment were seen in guinea pigs with a 12 h interval between infections than in those infected with both viruses simultaneously (p = 0.02, Student's t-test; Figure 6). In contrast, secondary infection 18 h after primary infection resulted in a low frequency of reassortant progeny (6.7% on average), whereas no reassortants were detected when the two infections were staggered by 24 h. Thus, while a brief delay between primary and secondary influenza virus infections may actually increase the potential for reassortment, no reassortment was seen with a delay of 24 h or more. 10.1371/journal.ppat.1003421.g006 Figure 6 Infections separated by less than 18 h led to robust reassortment in vivo. Groups of three guinea pigs were infected with 1000 PFU rPan/99var virus and, either at the same time (0 h group), or after the indicated time interval, infected with 1000 PFU rPan/99wt virus. Plaque isolates derived from nasal washings collected 48 h after wt virus infection were genotyped by HRM analysis. The average +/− standard deviation of the percentage of isolates with reassortant genotypes is shown. 10.1371/journal.ppat.1003421.t002 Table 2 Super-infection up to 12 h after primary infection leads to robust reassortment in vivo. Guinea pig no. Time interval (h) Prevalence of HA RNA in bulk nasal wash fluid (Ct value)1 Genotypes of virus isolates (%)2 HAwt HAvar Reassortant wt var 1 03 25.0 24.6 14 52 33 2 0 26.0 25.6 18 27 55 3 0 25.3 26.4 33 57 10 4 6 28.1 27.1 29 29 43 5 6 26.2 26.8 14 64 23 6 6 24.8 25.2 19 62 19 7 12 25.4 25.6 42 33 26 8 12 27.2 26.7 45 23 32 9 12 26.2 24.6 57 0 43 10 18 29.1 23.3 10 5 86 11 18 25.7 23.0 5 24 71 12 18 29.8 23.2 5 0 95 13 24 31.5 22.7 0 0 100 14 24 33.7 26.0 0 0 100 15 24 32.5 25.8 0 0 100 4wt n/a 27.9 >40 n/a n/a n/a 4var6 n/a >40 25.9 n/a n/a n/a 1 Quantitative PCR was performed using primers specific for wild-type or var virus HA segments. Average of two replicates is shown. Ct values 0.5 cm apart) plaques were picked using 5 ml serological pipettes, ejected into 160 ul of PBS in 1.5 ml tubes, and then stored at −80°C. 2) RNA was extracted from the agar plugs using the Qiagen QiaAmp Viral RNA kit, with the following modifications to the manufacturer's protocol: carrier RNA was not used, agar plugs in PBS were heated at 65°C for 5 min prior to mixing with AVL lysis buffer, and 40 ul water was used for the elution step. 3) Twelve microliters of RNA was reverse transcribed using Maxima reverse transcriptase (Fermentas) according to the manufacturer's instructions. 4) cDNA was used as template in qPCR reactions. Four microliters of 1∶4 diluted cDNA were combined with the appropriate primers (0.4 uM final concentration; see Supplementary Table S1 for primer sequences) and Precision Melt Supermix (BioRad) in wells of a white, thin wall, 384 well plate (BioRad). qPCR and melt analyses were carried out in a CFX384 Real-Time PCR Detection System, as per the instructions provided with the Precision Melt Supermix. Data were analysed using Precision Melt Analysis software (BioRad). Viruses were scored as reassortant if the genome comprised a mixture of wt and var gene segments in any proportion (e.g. both 7∶1 reassortants and 4∶4 reassortants were treated in the same way). Occasionally, one gene segment from a given isolate could not be typed as wt or var with high confidence (this was the case with approximately 2.5% of segments); such isolates were scored as wt or var parental viruses if all other gene segments were wt or var, respectively. If greater than one segment could not be typed, the isolate was excluded from the analysis. Quantification of HA segment in bulk nasal wash fluids The HA segments of rPan/99 wt and rPan/99var6 viruses differ by 6 nucleotides in two clusters: T308C/C311A/C314T and A464T/C467G/T470A. Forward and reverse primers encompassing these mutation clusters were designed: HAwt 295F/HAwt 481R and HAvar 295F/HAvar 481R. These primers specifically amplify the wt or var HA segments, respectively, allowing their quantification by conventional qPCR methods. Thus, RNA extracted directly from nasal lavage fluids was subjected to reverse transcription followed by qPCR using SsoFast Evagreen Supermix (BioRad), according to the manufacturer's instructions. qPCR was performed with a CFX384 Real-Time PCR Detection System and results were analysed using CFX Manager software (BioRad). Statistical analyses A one-sided exact test was applied to the data shown in Figure 3 to determine whether the proportion of isolates with parental genotypes was statistically greater than 2/256, or 0.008 (the expected value if reassortment occurred with full efficiency). Two-sided Fisher's exact tests were used to compare proportions of reassortant vs. parental progeny for data shown in Figures 4 and 5, while chi-squared tests were applied to compare proportions of singly vs. doubly infected cells for flow cytometric data shown in Figures 4, 5 and 7. Finally, unpaired, two-sided Student's t-tests were applied to data shown in Figure 6 and Tables 1 and 2. Supporting Information Table S1 Nucleotide sequences of primers used for HRM analysis. (DOCX) Click here for additional data file.
