Genome plasticity of triple-reassortant H1N1 influenza A virus during infection of vaccinated pigs

Abstract

To gain insight into the evolution of influenza A viruses (IAVs) during infection of vaccinated pigs, we experimentally infected a 3-week-old naive pig with a triple-reassortant H1N1 IAV and placed the seeder pig in direct contact with a group of age-matched vaccinated pigs (n = 10). We indexed the genetic diversity and evolution of the virus at an intra-host level by deep sequencing the entire genome directly from nasal swabs collected at two separate samplings during infection. We obtained 13 IAV metagenomes from 13 samples, which included the virus inoculum and two samples from each of the six pigs that tested positive for IAV during the study. The infection produced a population of heterogeneous alleles (sequence variants) that was dynamic over time. Overall, 794 polymorphisms were identified amongst all samples, which yielded 327 alleles, 214 of which were unique sequences. A total of 43 distinct haemagglutinin proteins were translated, two of which were observed in multiple pigs, whereas the neuraminidase (NA) was conserved and only one dominant NA was found throughout the study. The genetic diversity of IAVs changed dynamically within and between pigs. However, most of the substitutions observed in the internal gene segments were synonymous. Our results demonstrated remarkable IAV diversity, and the complex, rapid and dynamic evolution of IAV during infection of vaccinated pigs that can only be appreciated with repeated sampling of individual animals and deep sequence analysis.

Introduction

Influenza A viruses (IAVs) are distributed globally and can infect a wide range of host species, including humans (Russell et al., 2008), birds (Chen & Holmes, 2010), pigs (Vincent et al., 2008), horses (Murcia et al., 2010), dogs (Hoelzer et al., 2010), cats (Ali et al., 2011) and seals (Blanc et al., 2009). Wild waterfowl are considered the natural IAV reservoir (Taubenberger & Kash, 2010) and a genetically distinct lineage of viruses has also been identified in bats (Tong et al., 2013). A swine-origin H1N1 IAV was responsible for the first pandemic of the twenty-first century (Rambaut & Holmes, 2009) and was associated with >200 000 human deaths (Simonsen et al., 2013). In recent decades, the genetic diversity of swine IAVs in North America has increased significantly due to the emergence of triple-reassortant H3N2 viruses in the late 1990s (Kitikoon et al., 2013), the numerous introductions of human-origin viruses in pigs including the 2009 pandemic virus (Nelson et al., 2012) and the large-scale movement of pigs between different US regions (Nelson et al., 2011). It is estimated that >90 % of swine herds in the mid-western US are infected with IAVs (Corzo et al., 2013) and that pigs can be exposed to different IAVs during their lifetime (Diaz et al., 2015; Ducatez et al., 2011).

IAVs belongs to the family Orthomyxoviridae and have a segmented genome composed of eight negative-sense ssRNA segments that encode at least 12 proteins: polymerase basic 2 (PB2), polymerase basic 1 (PB1), polymerase acid (PA), haemagglutinin (HA), nucleoprotein (NP), neuraminidase (NA), matrix (M) and non-structural protein (NS). RNA viruses have a high mutation rate that increases their genetic diversity over time (Belshaw et al., 2007; Domingo et al., 2012; Lemey et al., 2006) and the segmented nature of the IAV genome allows the virus to exchange (reassort) gene segments with other IAVs, thus contributing to the overall genetic diversity of IAVs.

The main antigenic proteins of the virus, HA and NA, determine IAV subtype. In pigs, H1N1, H1N2 and H3N2 are the most prevalent IAV subtypes (Torremorell et al., 2012). In North American swine there are six antigenically and phylogenetically distinct H1 groups (α, β, γ1, γ2, δ1 and δ2) (Anderson et al., 2015; Lorusso et al., 2011) and four H3 groups (I, II, III and IV) (Kitikoon et al., 2013). Multiple IAV subtypes can co-circulate in swine herds and persist at the population level for prolonged periods of time (Corzo et al., 2013; Diaz et al., 2015). Additionally, multiple alleles (sequence variants of the same virus) can coexist during IAV infection of pigs (Murcia et al., 2012), and the same virus can evolve differently in the upper and lower respiratory tract of pigs (Wei et al., 2014). Furthermore, nucleotide substitutions within the HA antigenic sites can occur shortly after infection of pigs with no significant differences noticed between pigs with or without immunity to the virus (Diaz et al., 2013; Murcia et al., 2012).

In the USA, most pigs may be exposed to one or more IAVs during their lifetime. Hence, the majority of pigs may have immunity to one or several IAVs strains when they are challenged with currently circulating IAVs. However, there is a lack of knowledge about how the virus evolves in swine populations that are seropositive to different IAVs. Therefore, the objective of this study was to explore the genetic diversity of the complete genome of a triple-reassortant H1N1 IAV population during experimental infection of vaccinated pigs. A vaccine with multiple IAV strains was used to mimic field conditions where pigs are usually exposed to different IAVs. We identified several polymorphisms using next-generation sequencing (NGS) technologies directly from nasal swabs and reconstructed 13 complete genomes of the within-host viral populations (metagenomes) using polymorphism overlapping sequence fragment analysis, demonstrating the complex, rapid and dynamic evolution of IAV during infection of vaccinated pigs.

Results

Animal IAV infection status before and after contact with the seeder pig

To study the genomic plasticity of a triple-reassortant H1N1 IAV during infection of vaccinated pigs, we obtained 11 pigs free of IAV, vaccinated 10 and infected one to serve as a seeder. All pigs (n = 11) were IAV-negative by real-time reverse transcription-PCR (RRT-PCR) and seronegative by NP ELISA prior to vaccination. Two weeks after the booster vaccination, and before contact with the seeder pig, nine pigs tested ELISA-positive to IAV [sample-to-negative ratio (S/N) < 0.6] and one was a suspect (S/N = 0.803). Three pigs were negative [HA inhibition (HI) titre < 1 : 20] and seven had HI titres ≤ 1 : 40 to the challenge virus. All vaccinated pigs had HI titres ≥ 1 : 20 to the vaccine viruses (Table 1).

Table 1. IAV serology results by ELISA and HI tests for pigs prior to the start of the study (before vaccination), after vaccination and after infection.

For simplicity, animals that tested IAV RRT-PCR-positive after exposure to the seeder pig (n=5) are renamed A1–A5 and animals that tested negative are renamed A6–A10. ELISA results are expressed as S/N and considered positive, suspect or negative when S/N<0.6, 0.6<S/N<0.9 or S/N>0.9, respectively. HI titres are expressed as the reciprocal dilutions and considered positive at >1 : 20. The reciprocal mean titre was compared between pigs that tested RRT-PCR-positive and -negative after exposure to the seeder pig and considered statistically significant if the P value for the Kruskal–Wallis test was <0.05.

Animal	PCR	Before vaccination ELISA	Two weeks after second vaccination (prior contact with the seeder pig)						Two weeks after infection
			ELISA	HI test*				ELISA	HI test*
				00239	031	110600	MO069		00239	031	110600	MO069
Seeder	Positive	0.967	0.869	Neg†	Neg	Neg	Neg	0.32	80	160	0	40
A1	Positive	0.963	0.503	Neg	20	80	160	0.194	320	320	80	160
A2	Positive	1.700	0.803	20	20	160	320	0.129	640	640	80	320
A3	Positive	0.979	0.331	Neg	40	160	320	0.135	160	160	80	320
A4	Positive	0.925	0.332	Neg	20	80	320	0.202	160	320	80	320
A5	Positive	1.009	0.362	20	40	320	640	0.164	640	640	320	640
Mean		1.115	0.466	10	28	160	352	0.165	384	416	128	352
A6	Negative	0.958	0.546	40	40	320	640	0.434	20	80	320	640
A7	Negative	0.968	0.187	20	40	320	640	0.219	40	80	320	640
A8	Negative	1.018	0.545	20	20	160	320	0.247	20	40	80	320
A9	Negative	0.968	0.430	40	20	640	640	0.275	40	80	160	640
A10	Negative	0.981	0.193	20	40	320	640	0.162	20	80	80	320
Mean		0.979	0.380	28	32	352	576	0.2674	28	72	192	512
P value (Kruskal–Wallis test)		0.91	0.75	0.03	0.55	0.048	0.06	0.04	0.03	0.007	0.28	0.16

^*IAV isolates used in the HI test: 00239 (challenge virus), A/Swine/IA/00239/2004(H1N1); 031 (vaccine virus), A/Swine/NorthCarolina/031/2005(H1N1); 110600 (vaccine virus), A/Swine/Iowa/110600/2000(H1N1); MO069 (vaccine virus), A/Swine/Missouri/069/2005(H3N2). †: Neg=Negative.

The seeder pig remained negative before challenge and tested IAV-positive by RRT-PCR 48 h after challenge. After introduction of the seeder pig into the isolation unit with the remaining pigs (n = 10), five animals tested IAV-positive and five tested negative by RRT-PCR during the study period (Table 2). Prior to the introduction of the seeder pig, the mean S/N ELISA titre was not statistically different (P = 0.75) between pigs that tested positive and negative by RRT-PCR[Q1]. However, 14 days after the introduction of the seeder pig the S/N ELISA titre was lower (P = 0.04) in pigs that tested RRT-PCR-positive compared with pigs that tested RRT-PCR-negative (Table 1). Additionally, we also found statistically significant differences (P < 0.05) in the HI titres before and after contact to the seeder pig, between pigs that tested positive and pigs that tested negative by RRT-PCR (Table 1).

Table 2. IAV RRT-PCR results and samples selected for deep genome sequencing.

Nasal swabs were collected and tested for 14 days post-contact. No pig tested positive after 8 days post-contact. Results from five pigs that did not test positive at any point during this study are not shown. The days when pigs tested positive are indicated by ‘+’ and the RRT-PCR C_t value for each sample is shown. A total of 12 pig samples were selected for complete genome sequencing: seeder pig, day 2 (SD2, dark blue) and day 4 (SD4, light blue); animal A1, day 5 (A1D5, dark green) and day 7 (A1D7, light green), animal A2, day 4 (A2D4, dark purple) and day 8 (A2D8, light purple); animal A3, day 5 (A3D5, dark brown) and day 6 (A3D6, light brown); animal A4, day 3 (A4D3, dark grey), and day 6 (A4D6, light grey); animal A5, day 4 (A5D4, dark pink) and day 6 (A5D6, light pink). Sequences obtained from each sample are colour coded in Fig. 1 according to the colours indicated in this table.

Extensive allelic variation was identified by sequence analysis during infection

The complete genome of IAV was amplified and sequenced from 13 samples, which included the inoculum virus before challenge, two samples from the seeder pig at days 2 and 4 (SD2 and SD4), and two samples from each of the five infected pigs after contact (A1D5, A1D7, A2D4, A2D8, A3D5, A3D6, A4D3, A4D6, A5D4 and A5D6, where ‘A’ refers to the animal number and ‘D’ refers to the day of study). The most frequent allele in the inoculum virus for segments 1, 4, 5, 7 and 8 was also the most frequent allele in the pig samples analysed. In contrast, the most frequent allele for segments 2, 3 and 6 was different in pig samples compared with the inoculum virus. Many nucleotide polymorphisms (n = 794) were found in all samples throughout the course of the study and were distributed in all gene segments. However, there was great variability in the number of polymorphisms between samples and gene segments (Table S1, available in the online Supplementary Material). The overlapping sequence fragments analysis estimated a total of 327 alleles, of which 214 were unique sequences (Table 3).

Table 3. Number of alleles distributed by sample and gene segment.

The last column of the table indicates the ratio between synonymous mutations (dS) and non-synonymous mutations (dN) for each gene segment. In total, 214 out of 327 alleles found were unique sequences. na, No dN mutations were found in segment 6, hence no dS/dN ratio was estimated.

Segment	Inoculum	SD2	SD4	A1D5	A1D7	A2D4	A2D8	A3D5	A3D6	A4D3	A4D6	A5D4	A5D6	Unique sequences	dS/dN
1 (PB2)	4	1	2	2	0	1	2	1	1	4	2	1	1	11	15.7
2 (PB1)	2	2	1	2	1	2	2	2	2	1	1	1	1	5	3.7
3 (PA)	8	2	2	4	0	2	1	2	0	64	2	2	2	77	12.5
4 (HA)	2	8	2	2	32	2	2	1	1	1	2	2	1	44	0.8
5 (NP)	16	2	2	2	8	2	2	2	8	4	2	2	2	31	10.9
6 (NA)	2	1	1	1	2	1	1	1	1	1	2	1	1	3	na
7 (M)	2	2	1	1	4	1	1	1	1	1	1	1	1	7	41.0
8 (NS)	2	1	1	1	32	1	1	1	1	4	1	1	1	36	6.1
Total	38	19	12	15	79	12	12	11	15	80	13	11	10	214

Three of the 41 original polymorphisms present in the inoculum virus were not identified in any of the pig samples analysed and not all alleles in the inoculum were identified in the pig samples sequenced. Moreover, only four emergent alleles (defined as alleles not present in the inoculum virus) were found in multiple pigs (two in segment 2, and one each in segments 3 and 4). Finally, for all but the HA segment, the crude ratio of synonymous to non-synonymous substitutions (dS/dN) was >1 (Table 3).

Although there was small or no variation in the number of alleles detected between the two samples sequenced of most pigs that tested positive, the allele frequency changed significantly within animals 1 and 4 (Table 3). We noticed that prior to contact with the seeder pig, these two pigs (animals 1 and 4) were negative by HI to the challenge virus and had the lowest HI titre (1 : 80) to A/Swine/Iowa/110600/2000(H1N1), which was the vaccine virus closest to the challenge virus at the nucleotide level (Table 1).

HA and NA antigenic proteins diverge independently during infection

Whilst 44 different alleles were found in HA, only three alleles were found in NA during this study (Fig. 1). The 44 HA alleles identified (Fig. 1a) yielded 43 different predicted peptides (Table 4). The starting virus inoculum contained two HA alleles that only differed in 1 nt within the HA2 region. In the seeder pig on day 2 (SD2), we found eight HA alleles (Fig. 1a) with variations in the amino acid sequence within both the HA1 and HA2 regions (Table 4). However, the emergent alleles of SD2 were not found in any other pig samples analysed. All the other HA emergent alleles were unique to an animal except for one that was detected in two pigs (A1D5 and A2D8); however, the latter emergent variant (A1D5/A2D8, highlighted in Fig. 1a and Table 4) contained a HA1 region identical to the inoculum alleles. Additionally, we found 32 HA alleles in A1D7; at the nucleotide level, half of these 32 alleles were closer to an allele identified in A4D6, whilst the other half were closer to an inoculum allele (Fig. 1a). At the protein level, these 32 variants contained polymorphisms in all three regions of the HA, signal peptide, HA1, and HA2. Overall, amino acid substitutions within HA1 were only found in alleles obtained from SD2 and A1D7 (Table 4); four of these substitutions happened within antigen sites previously described for HA subtype H1 (Table 4), and their location and nature are illustrated in Fig. 2.

Fig. 1.

Median-joining networks of HA and NA alleles found during experimental IAV infection of vaccinated pigs. Each circle represents a sequence variant (allele) and each colour represents the sample where that sequence was found. Red numbers indicate the number of nucleotide differences between sequences (not all numbers are included for brevity). Within each network, the branch length is proportional to the number of differences between alleles. An asterisk indicates an emergent HA allele (not present in the inoculum virus) that was found in more than one pig. The colour code of this figure coincides with the colour code of Table 2.

Table 4. clustal_x alignment of the complete hypothetical HA proteins found by sample.

Only polymorphic sites amongst alleles identified in this study are shown. Superscripts A, B and D indicate the antigenic site where changes in HA1 were observed. The first two rows indicate amino acids found in the vaccine viruses at the polymorphic sites identified in the samples sequenced. The reference amino acid for each polymorphic position is shown in inoculum allele 1. Non-highlighted proteins are unique variants and proteins highlighted with the same shading are 100 % identical amongst them. A total of 58 functional HA sequences were identified amongst all samples. These sequences represented 44 unique alleles and translated 43 different hypothetical HAs. SP, signal peptide; IAV031, A/Swine/NorthCarolina/031/2005(H1N1); IAV110600, A/Swine/Iowa/110600/2000(H1N1). na, Not applicable [the complete sequence for A/Swine/NorthCarolina/031/2005(H1N1) is not available].

^*Identical proteins within a sample translated from two different alleles.

Fig. 2.

Three-dimensional models illustrating the HA1 region of the HA and the polymorphic amino acids found during the study. Polymorphic amino acids [histidine (H), serine (S), asparagine (N), lysine (K), glutamate (E), leucine (L), arginine (R) and proline (P)] are indicated and coloured according to their physical properties: [Q1] polar (red), charged (green) and hydrophobic (yellow). (a) Reference amino acids at polymorphic sites in HA1 for A/Swine/IA/00239/2004(H1N1) (challenge virus). (b) Amino acid residues predicted from the sample sequenced from the seeder pig at day 2 (SD2). (c) Amino acid residues predicted from the sample sequenced from animal 1 at day 7 (A1D7).

In contrast, at the NA level only three alleles were found during this study and all of them translated the same NA protein. Two of these alleles were present in the starting inoculum virus. The majority inoculum allele was not detected in most samples sequenced (except in A1D7) and the minority allele became fixed in most of the pig samples except for A4D6, in which a third emergent NA allele was found (Fig. 1b).

Discussion

To better understand the evolution of IAVs during infection of vaccinated pigs we used deep genome sequencing to compare the viral genetic diversity at two separate sampling points during infection. We demonstrated that the genetic makeup of the virus changed in all gene segments as the virus replicated within the group of animals, yielding a complex collection of viral genomes with similar and distinct variants. The infection produced a population of heterogeneous alleles by gene segment (usually two or more) that was dynamic over time. Therefore, our results indicate that the genetic heterogeneity of IAVs during infection of partially immune pigs is significant and it might have been underestimated. Under this scenario, controlling the transmission of IAVs in pigs is challenging because under natural conditions a large proportion of pigs have maternal or active immunity to different IAV strains (Corzo et al., 2013; Torremorell et al., 2012), IAV are endemic in swine populations (Vincent et al., 2008), multiple genetic lineages of the virus can co-circulate in pigs (Diaz et al., 2015; Ducatez et al., 2011), and pigs are moved and mixed in large batches of animals during their production stage (Knauer & Hostetler, 2013; Oh & Whitley, 2011).

In our study, the genetic diversity of IAVs changed dynamically throughout the course of infection. Two samples that corresponded to two different pigs (A1D7 and A4D3) had a higher number of alleles compared with the rest of the samples. As we did not sequence all samples from all pigs and our sample size was limited, we cannot be certain that high numbers of alleles were not present in all pigs at some point during infection. However, our results proved that the diversity of IAVs could change within a vaccinated pig throughout the course of infection. Interestingly, these two pigs (animals 1 and 4) were negative by HI to the challenge virus (titre < 1 : 20) before exposure to the seeder pig and had the lowest HI against A/Swine/Iowa/110600/2000(H1N1), which is the closest vaccine virus compared with the challenge virus. These findings suggest that in pigs the level of antibodies against IAVs might influence the overall diversity of the virus during infection. However, this observation needs to be corroborated in future studies, particularly in the context of heterologous vaccination to infection. Most vaccines are heterologous to circulating viruses and only provide partial protection to infection; therefore, the variability in the immune response to IAV vaccination may influence virus evolution. Individual host factors such as response to social stress (de Groot et al., 2001), individual host genetics and animal behaviour may also affect the immune response to viral infections. In addition, it remains unclear to what extent different alleles are selected for or whether rapid changes in the viral population are primarily stochastic.

In the samples sequenced in this study, the HA segment was more likely to undergo non-synonymous mutations compared with the remaining segments, including NA. Only one HA emergent allele was found in more than one pig and this allele was identical in the HA1 region to the inoculum virus. In agreement with our results, other studies in pigs have shown that nucleotide substitutions can occur in the HA segment very early after infection in pigs with immunity to IAV and that allele fixation can occur amongst infected animals (Diaz et al., 2013; Murcia et al., 2012). However, the time required for these substitutions to become fixed at a population level is still unknown. Our results are also consistent with a previous study of the HA1 region of the HA indicating that IAVs in pigs are not being transmitted as a single genotype, but rather as a population of viruses that may be closely related to each other (Murcia et al., 2012). However, our results also showed that nucleotide substitutions can occur in the signal peptide and the HA2 region of the HA, which were not evaluated by Murcia et al. (2012). In addition, other factors in our study, such as the group housing conditions, which facilitated greater interaction between pigs compared with previous studies where pairs of individuals were used to measure intra-host diversity of IAV (Hensley et al., 2009; Murcia et al., 2010, 2012), may have had an effect on the increased overall genetic diversity.

In contrast, we did not find non-synonymous mutations in NA, and there was therefore no evidence of coevolution of HA and NA or epistatic interactions. Although the dominant NA allele in the inoculum was not observed amongst the majority of samples sequenced, including the seeder, the allele was observed in sample A1D7, indicating that it had likely persisted at low levels during transmission. This genotype ‘recovery’ has been described during replication for other RNA viruses, such as polioviruses (Domingo et al., 1985). It is possible that unique alleles found in pigs were present in the inoculum at a low prevalence and that we were not able to identify them in the inoculum itself. Additionally, unique alleles could have been present in the pig samples that were not sequenced.

Multiple studies have evaluated the intra-host diversity of RNA viruses overtime (Debbink et al., 2014; Salemi, 2013; Tu et al., 2013). In IAVs, this research has focused on HA (Hoelzer et al., 2010; Murcia et al., 2010, 2013). To the best of our knowledge, our study is the first to evaluate the intra-host diversity of the complete genome of IAV during infection of pigs using NGS, and our results are comparable with a recent study in children (Bourret et al., 2015). We uncovered an additional layer of complexity in the evolution of IAVs during infection of immune pigs by demonstrating that all IAV gene segments replicate as a population of alleles that may or may not be transmitted. Although our methods were not able to capture reassortment events within hosts, the intra-host diversity observed here certainly provides opportunity for novel reassortant viruses to emerge. In other species, the evaluation of intra-host reassortment of IAV has shown that two viruses that are closely related to each other reassort at different rates depending on their co-infecting dose (Tao et al., 2014). Genome reassortment should be further investigated as it is a potential source of genetic diversity to swine IAV. It is not clear to what extent the diversity observed in our study is deleterious and not likely to be transmitted onward over longer time periods in pigs. Further understanding of the intra-host dynamics of co-infection and reassortment remains an important outstanding question in IAV evolution.

The dynamic nature of polymorphisms found in our study highlights that IAV genetic diversity ought to be studied directly in the original biological sample (i.e. nasal swab). The study of genetic diversity and evolution of IAV populations through cell culture might be misleading. The cell culture of IAV leads to loss of diversity as cell culture selects for the fastest growing virus in a new environment. As an example, the frequently used Madin–Darby canine kidney (MDCK) and Vero cell lines have differential preferences for IAV variants, which has led to selective rescue of specific alleles during serial passages (Roedig et al., 2011). As IAV diversity and population dynamics are complex and shaped by many factors, including viral fitness, mutation rate, host factors and stochastic events that may produce bottlenecks, a better estimate of IAV diversity at the population level can be obtained directly from the original sample. However, amplifying the complete genome from IAV isolates, where the viral concentration is exponentially higher than the original sample, might be easier for certain studies.

It is important to mention different external sources of potential variation and bias, including sequencing (Ross et al., 2013), depth of coverage (Bidzhieva et al., 2014), PCR (Archer et al., 2012; Cummings et al., 2010) and sampling or intra-assay bias or error. The platform we used, i.e. 454, is mature and errors due to sequencing are considered non-issues as we avoid reads with a Phred score < 20, polymorphisms in homonucleotide runs, not represented in both strands and not represented in unique independent sequence runs (Archer et al., 2012; Chen-Harris et al., 2013). The high confidence differences HCDiff file that we used takes these three precautions into account. The 454 has proven to accurately detect human immunodeficiency virus mutants at a prevalence as low as 0.1 % (Shao et al., 2013). Nevertheless, the variability on NGS reads mapped and depth of coverage throughout the complete genome of IAVs remains an issue to better estimate the genetic diversity of the viral populations (Bidzhieva et al., 2014; Bourret et al., 2013). Additionally, PCR (especially when the polymerase is stalled) generates in vitro recombinants that inflate and distort the estimates of the number and structure of the true alleles (Cummings et al., 2010; S. Enomoto, unpublished data). To avoid PCR errors, we used a high-fidelity PCR system that uses a blend of DNA polymerases including one isolated from Pyrococcus furiosus (Pfu) which has 3′ → 5′ exonuclease (proofreading) activity, with a 10-fold improved error rate compared with Taq DNA polymerase (André et al., 1997; Lundberg et al., 1991). High-fidelity polymerases have been shown to enhance the PCR and sequencing conditions (Hedman et al., 2009; Wu et al., 2010), improving the accuracy to estimate microbial diversity.

In conclusion, the swine IAV population in an experimental setting was complex. Although we recognize our sampling bias to estimate the complete genetic composition of the viral population during transmission, our findings demonstrate that the diversity of IAV can change dynamically during infection of vaccinated pigs. New sequencing technologies and bioinformatics algorithms might provide more precise estimates in future studies. In this study, the polymorphisms were abundant, dynamic and not limited to HA and NA. Some variants were maintained whilst others were not identified amongst the samples sequenced. Direct sampling and deep sequencing allowed us to investigate the dynamic plasticity of the IAV population during IAV infection in a small group of pigs. We envision that the plasticity of the IAV genome under field conditions is no less complex as different IAV subtypes can coexist and susceptible animals are continuously introduced into infected populations. Our study emphasizes the need to study IAV evolution directly from the infected host using new-generation sequencing approaches, which will help design better strategies to control influenza in animals and people. More studies are needed in order to evaluate whether the changes observed in this study are due to vaccination or whether they are also found in non-immune pigs.

Methods

Study design

Eleven 3-week-old specific-pathogen-free piglets were selected from a serologically IAV-negative swine herd and moved to the University of Minnesota animal research units. The IAV-negative status was confirmed by testing individual nasal swabs with RRT-PCR targeting the M gene (Slomka et al., 2010; Spackman & Suarez, 2008) and serum samples by ELISA (Influenza Ab Test kit; IDEXX Laboratories) for antibodies against NP (Ciacci-Zanella et al., 2010).

Viral RNA was eluted using 50 μl of each sample into 50 μl elution buffer using an Ambion MagMax virus RNA isolation kit (Life Technologies). An AgPath-ID One-Step RT-PCR reagent kit (Life Technologies) was used to detect IAV. PCR mix containing 5 μl RNA, 12.5 μl 2 ×  buffer, 1.0 μl 25 ×  enzyme mix, 1.67 μl detection enhancer, 5 pmol each primer and 1.5 pmol probe was run on a LightCycler 480 system (Hoffmann-La Roche) at 45 °C for 10 min, followed by 95 °C for 10 min, and 45 cycles at 94 °C for 1 s and 60 °C for 30 s. Fluorescence was recorded at 60 °C and a sample was considered positive if the C_t was < 40. This PCR protocol can detect IAVs in samples containing ≥ 200 copies of the target amplicon, and has 100 and 95 % diagnostic sensitivity and specificity, respectively (Slomka et al., 2010).

Ten pigs were vaccinated 1 day after arrival and 2 weeks later with 2 ml of a licensed inactivated trivalent IAV vaccine (FluSure; Zoetis Animal Health), containing the δ and γ clusters of H1N1 [A/Swine/NorthCarolina/031/2005(H1N1) and A/Swine/Iowa/110600/2000(H1N1), respectively] and one H3N2 [A/Swine/Missouri/069/2005(H3N2)]. Two weeks after the second vaccination, nasal swabs and blood samples were collected from all pigs and tested for IAV by RRT-PCR (Spackman & Suarez, 2008) and ELISA, respectively. Additionally, blood samples were tested by HI tests against the challenge and vaccine viruses before and after infection as described previously (Direksin et al., 2002). The mean ELISA and HI titres were compared between vaccinated pigs that tested RRT-PCR-positive or -negative during this study, and considered statistically significant if the P value for the non-parametric one-way ANOVA Kruskal–Wallis test was < 0.05.

One unvaccinated pig was inoculated with IAV in a separate room to serve as a seeder pig to infect the other pigs. An aliquot of 2 ml 1 × 10⁶ TCID₅₀ ml^− 1 A/Swine/IA/00239/2004(H1N1) IAV (GenBank accession number EU139832.1) grown in MDCK cells (Meguro et al., 1979) was used to challenge the seeder pig intranasally and intratracheally. The A/Swine/IA/00239/2004(H1N1) clusters within the β H1 swine IAVs (Lorusso et al., 2011). This virus was selected because it has been fully characterized, genetically and antigenically (Anderson et al., 2015), and it has been used in several pathogenesis (Vincent et al., 2007) and transmission studies (Allerson et al., 2013; Diaz et al., 2013; Romagosa et al., 2011). The challenge virus was 91.5 and 73.7 % identical at the nucleotide level to the H1 γ and δ vaccine virus strains, respectively. The infection was confirmed 48 h later by RRT-PCR and the seeder pig was placed in contact with the rest of the pigs. Nasal swabs were collected from all pigs daily for 14 days into 1.8 ml viral transport medium (minimum essential medium plus 2 % BSA and 1 % penicillin/streptomycin) and an aliquot of the transport medium was used for RRT-PCR testing. All pigs were euthanized on day 14 and all procedures for this study were approved by the University of Minnesota Institutional Animal Care and Use Committee (protocol number 0908A71965).

Sample selection, genome amplification and sequence identification

To explore the within host variability of IAV during infection, two IAV-positive samples from each pig were conveniently selected for complete genome amplification and sequencing using NGS technologies (Table 2). Samples with the lowest C_t value and best genome amplification were targeted for sequencing. The IAV genome was amplified using a modified protocol of Zhou et al. (2009). Briefly, the viral RNA was purified from the swabs using a QIAamp Viral RNA Mini kit (Qiagen). IAV cDNA was created from viral RNA using primer MBtuni12(M) (ACGCGTGATCAGCRAAAGCAGG) and Superscript III First Strand Synthesis SuperMix (Invitrogen) cDNA was amplified in a PCR (five cycles of 94 °C 15 s, 45 °C 30 s, 68 °C 180 s and 31 cycles of 94 °C 15 s, 57 °C 30 s, 68 °C 180 s) consisting of PicoMax High Fidelity DNA Polymerase (Agilent), MBtuni12(M) and MBtuni13 (ACGCGTGATCAGTAGAAACAAGG). PCR products were verified by gel electrophoresis and purified using a QIAquick Spin kit (Qiagen). Purified cDNAs from the virus inoculum and 12 pig samples (Table 2) were submitted to the Genomics Center at the University of Minnesota for library preparation and 454 sequencing (454 GS-FLX; Roche Diagnostics) as described in detail by Ramakrishnan et al. (2009).

The 454 inoculum reads were assembled with Newbler 2.6 (Roche Diagnostics) using a reference template obtained from GenBank (Table S2) and the inoculum consensus sequence was used as the reference genome (Table S3) to assemble the 454 reads from each pig sample. The polymorphisms present in each sample were extracted from the 454 HCDiff.txt files created during each assembly in Newbler 2.6. These file include only highly confident differences which are defined as variants identified in at least three unique reads, and present in forward and reverse reads.

Allele identification and overlapping reading test

Alleles (sequence variants) were defined as complete functional gene segments identified by aligning overlapping sequence fragments. The Newbler output, 454 HCDiff.txt, is a file of sequence alignments surrounding all the high confidence polymorphic loci. A Ruby (Goto et al., 2010) script was written to test the linkage of two adjacent loci by enumerating the occurrence of the four sequence combinations: consensus–consensus, consensus–variant, variant–consensus and variant–variant. If >80 % of the sequences occurred only as two sequence combinations, the two loci were considered linked. Presence or absence of polymorphisms at each locus was encoded as 1 or 0, respectively. The alleles were deduced by linking together the adjacent intervals between the two polymorphic loci and its functionality verified using the National Center for Biotechnology Information FLu ANnotation tool (flan; http://www.ncbi.nlm.nih.gov/genomes/FLU/Database/annotation.cgi) (Bao et al., 2007). Additionally, if the distance that separated two polymorphisms was longer than the length of the reads obtained, then those two polymorphisms were considered not linked. For example, if two adjacent polymorphic loci were linked and recovered as 00 and 11, the segment contained two alleles rather than four alleles. The raw 454 reads, the allele sequences obtained, and the Ruby scripts for overlapping sequence fragments analysis and allele extraction are available upon request.

Sequence analysis

To illustrate the phylogenetic relationship between sequences, alleles were aligned to the reference genome using DNA-Alignment and median-joining networks were estimated using Network (Bandelt et al., 1999). Each network was annotated with Network Publisher (Fluxus Technology) and Adobe Illustrator CC (Adobe Systems). Additionally, for the first ORF we estimated the mean number of synonymous (dS) and non-synonymous (dN) mutations and their ratio (dS/dN) amongst sequences (Korber, 2000; Nei & Gojobori, 1986) using the Synonymous and Non-synonymous Analysis Program (snap; www.hiv.lanl.gov).

HA and NA protein analysis

For HA and NA, hypothetical proteins were inferred from nucleotide sequence, aligned using clustal_x (Larkin et al., 2007) and compared. The amino acid differences amongst HA sequences were mapped to the known H1 antigenic sites (Caton et al., 1982; Deem & Pan, 2009), modelled using the tools available at http://swissmodel.expasy.org/ (Arnold et al., 2006) and illustrated using PyMOL version 1.5.0.4 (https://www.pymol.org/). The HA1 IAV template used for our protein model was A/Swine/Iowa/15/30(H1N1) (Protein Data Bank ID: 1RUY). This template was used because this virus is from swine origin, the HA has been crystallized and it is available for public use.

Supplementary Data

Supplementary Data