|Home | About | Journals | Submit | Contact Us | Français|
Influenza A(H1N1) viruses entered the U.S. swine population following the 1918 pandemic and remained genetically stable for roughly 80 years. In 1998, there was an outbreak of influenza-like illness among swine that was caused by A(H3N2) viruses containing the triple reassortant internal gene (TRIG) cassette. Following the TRIG cassette emergence, numerous reassortant viruses were isolated in nature, suggesting that the TRIG virus had an enhanced ability to reassort compared to the classical swine virus. The present study was designed to quantify the relative reassortment capacities of classical and TRIG swine viruses. Reverse genetic viruses were generated from the classical H1N1 virus A/swine/MN/37866/1999 (MN/99), the TRIG virus A/swine/NC/18161/2002 (NC/02), and a seasonal human H3N2 virus, A/TX/6/1996 (TX/96), to measure in vitro reassortment and growth potentials. After coinfection with NC/02 or MN/99 plus TX/96, H1/H3 double-positive cells were identified. Delayed TX/96 infection was fully excluded by both swine viruses. We then analyzed reassortant H3 viruses. Seventy-seven of 81 (95.1%) TX/96-NC/02 reassortants contained at least one polymerase gene segment from NC/02, whereas only 34 of 61 (55.7%) MN/99-TX/96 reassortants contained at least one polymerase gene segment from MN/99. Additionally, 38 of 81 (46.9%) NC/02-TX/96 reassortants contained all NC/02 polymerase gene segments, while none of the MN/99-TX/96 reassortants contained all MN/99 polymerase genes. There were 21 H3 reassortants between MN/99 and TX/96, compared to only 17 H3 reassortants between NC/02 and TX/96. Overall, the results indicate that there are no distinct differences in the ability of the TRIG to reassort with a human virus compared to the classical swine virus.
IMPORTANCE There appear to be no differences in the abilities of classical swine and TRIG swine viruses to exclude a second virus, suggesting that under the right circumstances both viruses have similar opportunities to reassort. The increased percentage of TRIG polymerase gene segments in reassortant H3 viruses indicates that these viruses may be more compatible with gene segments from other viruses; however, this needs to be investigated further. Nevertheless, the classical swine virus also showed the ability to reassort, suggesting that factors other than reassortment capacity alone are responsible for the different epidemiologies of TRIG and classical swine viruses. The post-TRIG diversity was likely driven by increased intensive farming practices rather than virologic properties. Our results indicate that host ecology can be a significant factor in viral evolution.
Influenza surveillance has demonstrated that swine play an important role in the maintenance and evolution of influenza viruses. Most recently, the 2009 H1N1 pandemic virus was linked to swine, which have been proposed to be “mixing vessels” for influenza viruses because their respiratory tract is lined with both the avian influenza A virus α2,3-linked sialic acid receptor and the mammalian influenza virus α2,6-linked sialic acid receptor. Swine have also been infected experimentally with avian influenza viruses (1,–3).
Around 1918, an H1N1 influenza virus (classical swine H1N1 virus) entered U.S. swine herds and circulated for approximately 80 years, during which little antigenic or genetic variation was detected (4). A 1978 study identified a relatively high prevalence of the classical swine H1N1 virus, with nearly 25% of all swine showing serologic evidence of exposure to it (5). In contrast, the seroprevalence of H3N2 viruses was only around 1.4% in U.S. pigs, and the H3N2 isolates collected from 1976 to 1977 were genetically similar to isolates from humans (5). However, it was evident that H3N2 viruses were capable of sustained pig-to-pig transmission because they were frequently isolated from swine in Europe and Asia (6). The relative ratios of these viral subtypes remained similar even 10 years later: in 1988 and 1989, 51% of U.S. swine herds tested positive for H1N1 viruses, whereas only 1.1% tested seropositive for H3N2 viruses (7,–9).
In the mid- to late 1990s, the viruses circulating among U.S. swine changed dramatically. From 1997 to 1998, the seroprevalence of H1N1 viruses declined to approximately 27.7% of U.S. pigs; however, 8% of pigs were seropositive for human H3N2 viruses (10). In August 1998, there was an outbreak of severe influenza-like illness in swine in North Carolina, followed by similar outbreaks in Minnesota, Iowa, and Texas. These outbreaks were all caused by H3N2 viruses (11). The swine H3N2 viruses subsequently spread to Nebraska, Colorado, Oklahoma, Wisconsin, and Illinois and also showed increased virulence compared to that of the classical H1N1 virus (12). Sequence analysis of the viruses collected during the swine outbreaks revealed that 2 unique influenza virus genotypes were involved, indicating the possibility that several reassortment events had occurred. One genotype was a reassortant virus in which the polymerase basic 1 (PB1), hemagglutinin (HA), and neuraminidase (NA) genes were of human influenza virus origin and the remaining segments were from the classical swine H1N1 virus (11). The second genotype was a triple reassortant virus with polymerase basic 2 (PB2) and polymerase acidic (PA) genes of avian influenza virus origin, PB1, HA, and NA genes of human influenza virus origin, and nucleoprotein (NP), matrix (M), and nonstructural (NS) gene segments originating from classical swine influenza viruses (12). A third genotype, isolated from swine in Ontario and not associated with the U.S. outbreaks, was a wholly human H3N2 virus that jumped from humans to pigs but was unable to be transmitted further among pigs (8). Viruses with the triple reassortant internal gene (TRIG) cassette (comprising the PB2, PB1, PA, NP, M, and NS genes) propagated among U.S. swine herds and reassorted further, thereby acquiring new HA and NA genes (8, 13, 14). In a serosurveillance study from 1998 to 2001, 31.7% of 3,561 swine samples were seropositive for influenza virus, with 77.3% of the seropositive samples being H1 seropositive and 22.7% being H3 seropositive (15). Since then, however, the TRIG H1N1 viruses have completely displaced the classical swine viruses in U.S. pig populations, suggesting that viruses with the TRIG cassette have increased fitness compared to that of previously circulating swine viruses.
After the generation of the TRIG cassette in swine viruses, there was an increase in viral diversity arising from viruses with different HA and NA gene combinations. An H1N2 virus with a classical swine virus H1 gene and the human virus N2 gene was isolated from an Indiana farm in 1999 (16). Similarly, an H3N1 virus containing the TRIG cassette was isolated in Minnesota in 2004 (17). Aside from novel reassortants that have emerged in swine, antigenically distinct subgroups of H1 viruses have developed. These subgroups have been classified as α, β, and γ viruses, all of which are generated by antigenic drift in HA, and the human-like δ H1 viruses, which are generated through reassortment with human H1N1 viruses (18, 19). Another reassortment between avian and swine TRIG influenza viruses led to the detection of H2N3 viruses on 2 U.S. farms (20). Although the future direction of influenza virus evolution in U.S. swine is unclear, there is a stark contrast between the evolution of the viruses that circulated from 1918 to 1998 and the evolution of those that have circulated since 1998. The underlying reasons for this difference remain unclear, though one hypothesis is that a virologic property of the TRIG viruses increases their ability to acquire novel HA and/or NA genes (21, 22), which would be consistent with observations in nature and in the laboratory. The present study was designed to investigate this hypothesis.
To ensure viral clonality, reverse genetic viruses derived from A/swine/MN/37866/1999 (MN/99), A/swine/NC/18161/2002 (NC/02), and A/TX/6/1996 (TX/96) (Table 1) were rescued with the 8-plasmid system by using the pHW2000 plasmid as previously described (23). The MN/99 virus was chosen because it was a contemporary, wholly classical swine influenza virus, whereas NC/02 was a contemporary TRIG-containing swine influenza virus and TX/96 was a wholly human influenza virus that was isolated around the time that the TRIG evolved. Briefly, each gene segment was amplified by reverse transcription-PCR and ligated into the dual-promoter pHW2000 plasmid. All 8 plasmids were simultaneously transfected into a 90%- to 95%-confluent MDCK-293T cell coculture by using TransIT-LT1 transfection reagent (Mirus Bio LLC, Madison, WI). The cell supernatants were then injected into 10-day-old embryonated chicken eggs and incubated at 35°C for 48 h. The allantoic fluid was then harvested and reinjected into embryonated chicken eggs for a second passage. The virus was harvested, aliquoted, and stored at −80°C. Each vial was thawed only once before being used in any experiment. All virus stocks were verified via Sanger sequencing before use.
Monoclonal antibodies that were selectively neutralizing for NC/02 and MN/99 or for TX/96 were identified in the St. Jude Children's Research Hospital repository via hemagglutination inhibition assays and purified from ascites fluid by use of a 1-ml HiTrap protein G affinity column (GE Healthcare, Pittsburgh, PA). The purified anti-H1 antibody was fluorescently labeled with Invitrogen Alexa Fluor 488 (Life Technologies, Grand Island, NY), and the anti-H3 antibody was fluorescently labeled with Invitrogen Alexa Fluor 555 (Life Technologies). The fluorescently labeled antibodies were used to stain infected MDCK cells. The stained cells were analyzed on a BD FACSCanto II flow cytometer (BD Biosciences, Franklin Lakes, NJ) to quantify coinfected cells.
The procedure for isolating and growing primary swine respiratory epithelial cells (pSRECs) was adapted from that of Bateman et al. (24). Pig tracheas were collected from slaughtered animals, and the epithelial cells were detached by digestion with a mixture of pronase (Roche, Indianapolis, IN) and DNase (Sigma-Aldrich, St. Louis, MO). The pSRECs were grown in tissue culture flasks or plates coated with 5% (wt/vol) type VI human placental collagen (Sigma-Aldrich) in bronchial epithelial cell growth medium (BEGM) (Lonza, Walkersville, MD) supplemented with BEGM BulletKits and 20% fetal calf serum.
The rescued viruses (Table 1) were used to infect 80%- to 90%-confluent pSREC monolayers at a multiplicity of infection (MOI) of 0.01. After approximately 1 h of incubation with virus, the cells were gently washed with 1× phosphate-buffered saline (PBS), and infection medium with 0.5% tosylsulfonyl phenylalanyl chloromethyl ketone (TPCK)-trypsin was added. The cell culture plates were incubated at 37°C in 5% CO2. Samples collected at 10, 12, 14, 16, 18, 20, 24, 36, and 48 h postinfection (hpi) were stored at −80°C until titration. All 50% tissue culture infective dose (TCID50) titers were determined on MDCK cell monolayers at 80% to 90% confluence and calculated by using the Reed-Muench method (25).
The reassortment frequency was determined by growing pSRECs to approximately 80% to 90% confluence and then simultaneously infecting them with TX/96 and either NC/02 or MN/99, each at an MOI of 0.01. The infected cells were incubated in BEGM supplemented with 0.5 μg/ml TPCK-trypsin at 37°C in 5% CO2 for 48 h. The viral supernatants were collected and stored at −80°C until the virus could be plaque purified.
The viral supernatants were serially diluted from 1:104 to 1:106 in 8 ml of infection medium, and then 80 μl of an anti-H1 monoclonal antibody raised against NC/02 was added to each supernatant dilution to give a final antibody dilution of 1:100. The supernatants were thereby enriched for viruses possessing the H3 antigen in order to increase the sequencing efficiency of only viruses containing the H3 gene segment. The diluted supernatants were then incubated on ice for 1 h before undergoing viral adsorption on a 95%- to 100%-confluent MDCK cell monolayer. After adsorption, the cells were washed and overlaid with 1× infection medium and 0.9% Bacto agar (BD Biosciences) supplemented with 0.5 μg/ml TPCK-trypsin and then incubated at 37°C in 5% CO2 for 48 h. Identifiable isolated plaques were aspirated with a wide-bore pipette tip, and each agar plug was ejected into a well of a 96-well plate containing an approximately 80%-confluent MDCK cell monolayer and incubated for an additional 72 h.
Sequencing was performed using the Illumina MiSeq platform (Illumina, San Diego, CA) as follows. RNA was extracted with a MagMAX-96 AI/ND viral RNA isolation kit (Applied Biosystems, Grand Island, NY). Purified RNA was reverse transcribed with SuperScript III reverse transcriptase (Life Technologies), and all 8 gene segments were amplified with universal influenza virus primers and Phusion high-fidelity PCR master mix (New England BioLabs, Ipswich, MA). By using a Covaris E220 focused ultrasonicator, the amplified samples were sheared into 400- to 500-bp fragments and processed using a Bioo Scientific NEXTflex DNA sequencing kit (Bioo Scientific, Austin, TX). All sequencing data were analyzed with CLC Bio Genomics Workbench v.7 software (Qiagen, Boston, MA).
Twenty-two 2-week-old pigs (Northwoods Pork, Parker Prairie, MN) were divided into 4 groups of 5 pigs and 1 group of 2 pigs. All pigs were housed in a biosafety level 2 (BSL-2) laboratory space with straw-lined pens in individual cubicles.
Two pigs from each infection group were inoculated intranasally with 1 × 106 to 2 × 106 PFU/ml of TX/96, RG-MN/TX/NC-PB1, RG-MN/TX, or RG-NC/TX by using metered-pump nasal spray bottles (Health Care Logistics, Circleville, OH). The inoculated pigs were housed with 3 virus-negative contact animals to assess viral transmissibility. Each pig was monitored daily for clinical symptoms. On the first day postinfection (dpi) and every second day thereafter, each nostril of every pig was swabbed. The swabs were stored in 500 μl of viral transport medium (VTM) at −80°C until titration on MDCK cells to calculate the TCID50. Animal studies were conducted in accordance with the St. Jude Children's Research Hospital Animal Care and Use Committee guidelines.
TCID50 titers and coinfected cell percentages were compared by 2-way analysis of variance (ANOVA) with the Bonferroni posttest, as calculated by GraphPad Prism for Windows, version 5.03 (GraphPad Software, San Diego, CA). Proportions of polymerase gene segment incorporation were compared using Fisher's exact test. Associations between gene segments were assessed by calculating Spearman's rank correlation coefficient (SCC). The SCC was defined by the following formula: , where Ri is the rank of the first segment, Si is the rank of the second segment, is the mean of the Ri values, and is the mean of the Si values (26). The SCC value was estimated for each pair of binary-coded gene segments (in the MN/99-TX/96 conversion, TX/96 = 0 and MN/99 = 1; in the NC/02-TX/96 conversion, TX/96 = 0 and NC/02 = 1), and then the statistical significance of the coefficients was controlled based on the false discovery rate (27). Statistical analysis was performed using SAS, Windows, version 9.3 (SAS Institute, Cary, NC). A P value of ≤0.05 was considered significant unless otherwise specified.
We designed a virus exclusion experiment to determine if a classical swine H1N1 virus (MN/99) was more capable than a TRIG H1N1 virus (NC/02) of excluding a superinfection by a human H3N2 virus (TX/96). If so, this would explain the lack of reassortment among classical swine viruses that is observed in nature. Flow cytometry data revealed that during simultaneous infection with 2 viruses, each at an MOI of 0.01, nearly 3 times more cells had surface coexpression of H1 and H3 HAs in the NC/02-TX/96 coinfection group (26.1%) than in the MN/99-TX/96 coinfection group (8.8%), and the percentage of H1 virus-infected cells was considerably higher than that of H3 virus-infected cells (Fig. 1A and andB).B). This difference can likely be attributed to the total number of cells counted rather than to a virus characteristic. Subsequent experiments revealed that 8.55% of cells were coinfected with NC/02-TX/96 and 14.2% were coinfected with MN/99-TX/96 (data not shown). Interestingly, when we delayed the TX/96 infection until 24 h after the H1 virus infection, the superinfection was completely excluded by both MN/99 and NC/02 (Fig. 1C and andD).D). Similarly, when we delayed the infections with the H1 viruses until 24 h after the TX/96 infection, TX/96 fully excluded superinfection by the H1 viruses (Fig. 1E and andF).F). To attempt to gain better resolution between influenza virus-positive and -negative cells, the experiment was repeated at an MOI of 1 for each virus. The resulting flow data were nearly identical to those for an MOI of 0.01 (data not shown). Thus, secondary viral infections were clearly excluded 24 h after the initial infection; however, despite a quantitative difference in the number of double-stained cells, both of the H1N1 viruses evidently permitted coinfected cells, suggesting that there was ample opportunity for generating reassortants.
Although no differences in the percentage of coinfected cells could be identified between the two swine viruses at 24 hpi, we thought that there might be differences observed at earlier time points. The TX/96 virus infection was delayed for various times, from 4 hpi to 24 hpi, and the cells were analyzed by flow cytometry as previously described. There was no statistical difference between the percentages of coinfected cells when either MN/99 or NC/02 preceded TX/96 (Table 2), suggesting that both swine viruses share similar abilities to coinfect and exclude a subsequent influenza virus infection.
Although both of the swine H1N1 viruses were able to coinfect cells with TX/96, and therefore theoretically generate reassortants, the replication capacities of such reassortants might have differed substantially. To investigate this possibility, we created specific reassortants of both swine H1N1 viruses (Table 1), and to better mimic the natural host, we assayed their growth on pSRECs. The parental strains, NC/02, MN/99, and TX/96, showed no significant growth differences in pSRECs (Fig. 2). Similarly, we observed no differences between the 6 + 2 reassortant genotypes of NC/02 with the HA and NA gene segments from TX/96 (NC/TX) and MN/99 with the HA and NA gene segments from TX/96 (MN/TX) and their respective parental strains (Fig. 2). The kinetics of additional reassortants with the NC/02 polymerase segments on the MN/99 backbone were tested, and no significant differences were observed (data not shown). These data show that growth deficiencies did not limit the generation of NC/TX or MN/TX 6 + 2 reassortant strains.
As there was no difference in the ability of the swine viruses to grow in pSRECs, we assessed their ability to reassort with the human H3N2 virus. To quantify the frequency of reassortment in cells of porcine origin, we coinfected pSRECs with TX/96 and either MN/99 or NC/02 at an MOI of 0.01 and then genotyped plaque-purified progeny viruses. We identified 36 different genotypes arising from the coinfection with MN/99 and TX/96, with the most common genotype being the wild-type TX/96 virus (Table 3). Conversely, only 19 different reassortant genotypes were identified after reassortment between NC/02 and TX/96, of which 10 were found only once. Among the 93 plaques, we identified the MN/99 NP gene segment in 69 (74.2%) of the viruses, the NS gene segment in 59 (63.4%), and the M gene segment in 50 (53.8%) (Table 3). Of the 61 H3-enriched viruses identified, 50 (82%) contained both the TX/96 PB2 and PB1 gene segments and 57 (93.4%) contained both the TX/96 PB1 and HA gene segments; this is consistent with the association between the PB1 and HA gene segments in naturally occurring reassortant viruses. Additionally, of the H3 reassortants between MN/99 and TX/96, 51 (83.6%) contained both the HA and NA genes from TX/96, whereas only 14 (17.3%) of the reassortants between NC/02 and TX/96 contained both the HA and NA genes from TX/96.
We calculated SCCs to assess whether certain gene segments were more likely than others to reassort together, and we adjusted the P values based on the false discovery rate (Table 4). The strongest gene correlation observed was that between the MN/99 M and NS gene segments (SCC = 0.619; P < 0.001). The MN/99 M segment was also strongly correlated with the PA segment (SCC = 0.447; P < 0.001). Interestingly, the MN/99 NP gene segment was strongly correlated with both the M and NS gene segments of MN/99 (SCCs = 0.439 and 0.538, respectively; P < 0.001).
There was less viral diversity overall among the progeny of coinfection with the NC/02 and TX/96 viruses, with only 21 different genotypes identified (Table 5). The most prevalent genotype, found in 33 (34.4%) of the 96 progeny viruses genotyped, was an H3N1 virus with the HA gene segment from TX/96 and the remaining 7 gene segments from NC/02, similar to the genotypes transiently observed in nature. After H3 HA enrichment, 85 (88.5%) of the 96 genotyped viruses were H3 viruses. In contrast to the MN/99-TX/96 reassortants, only 27 (31.8%) of the 85 H3 viruses contained the TX/96 PB1 gene, suggesting that the human PB1 gene has no advantage over the homologous TRIG virus human-like PB1 gene when paired with the TX/96 HA gene. Interestingly, 77 (95.1%) of the 81 non-wild-type H3 viruses recovered from the coinfection with TX/96 and NC/02 had at least one polymerase gene segment that originated from the NC/02 virus (Table 5), whereas only 34 (55.7%) of the 61 non-wild-type H3 viruses recovered from the coinfection with TX/96 and MN/99 contained at least one polymerase gene segment that originated from MN/99 (Table 3). The difference in proportion for the inclusion of at least one polymerase gene segment was significant based on Fisher's exact test (P < 0.0001). Additionally, 38 (46.9%) of the 81 H3 reassortant viruses recovered from the coinfection with TX/96 and NC/02 derived all of their polymerase gene segments from NC/02, whereas no MN/99-TX/96 reassortants derived all of their polymerase genes from MN/99. The difference in percentage for inclusion of all 3 polymerase segments was also significant (P < 0.0001; Fisher's exact test).
NC/02 and TX/96 reassortment genotypes showed an increased number of correlated gene segments compared to those identified from reassortments between MN/99 and TX/96. The NC/02 NP gene segment showed a strong correlation with 4 NC/02 gene segments, including PA (SCC = 0.652; P < 0.001), NA (SCC = 0.488; P < 0.001), M (SCC = 0.777; P < 0.001), and NS (SCC = 0.656; P < 0.001) (Table 6). The NC/02 PA gene segment also correlated with several gene segments in addition to the NP segment, and it had the strongest correlations with PB1 (SCC = 0.404; P < 0.001), M (SCC = 0.838; P < 0.001), and NS (SCC = 0.402; P < 0.001) (Table 6). Additionally, the NC/02 NA and M gene segments were correlated (SCC = 0.435; P < 0.001), as were the M and NS gene segments (SCC = 0.495; P < 0.001) (Table 6). Interestingly, the 4 strong MN/99 gene correlations were also observed for the NC/02 virus. These data demonstrate that both the classical swine virus and the TRIG virus are capable of reassortment, and although there appears to be no underlying growth influence, it is evident that the polymerase genes of the TRIG cassette reassort with the H3 virus more readily than those of the classical swine virus.
Although the in vitro coinfections produced numerous reassortants, this did not necessarily indicate which viruses would successfully infect and be transmitted among hosts in vivo. Therefore, we inoculated pigs with TX/96, MN/TX/NC-PB1 (because of the strong association between HA and PB1), MN/TX, or NC/TX to test the ability of these viruses to infect and be transmitted among cohoused animals. As expected, the wholly human TX/96 virus infected the inoculated animals but was not detected in any of the contact animals (Fig. 3A). This virus reached an average peak titer of 104.08 TCID50/ml from 1 to 3 dpi in infected animals, and the infection was cleared prior to day 7 postinfection.
The MN/TX/NC-PB1 virus showed infection kinetics similar to those of TX/96. Pigs inoculated with MN/TX/NC-PB1 shed virus from around 1 to 7 dpi, with an average peak titer of 103.987 TCID50/ml on day 5 postinfection (Fig. 3B). This virus was also not transmitted to any of the contact animals.
Pigs inoculated with MN/TX had an average peak titer of 105.282 TCID50/ml at 3 dpi, and the virus was transmitted to all 3 contact animals. The contact animals began shedding virus around 3 dpi, and the shed virus reached a peak titer of 105 TCID50/ml at 5 to 7 dpi (Fig. 3C). Both the inoculated pigs and the contact animals shed virus for approximately 6 to 7 days.
The NC/TX virus also efficiently infected inoculated animals and was transmitted to all of the contact animals. The virus reached an average peak titer of 104.909 TCID50/ml at 1 dpi (Fig. 3D). The virus reached slightly lower peak titers in the contact animals, with an average titer of 104.41 TCID50/ml around 5 dpi. Both the inoculated and contact groups shed virus for approximately 5 to 6 days (Fig. 3D).
Both the MN/TX and NC/TX viruses were transmitted to all of the contact animals, suggesting that the classical swine virus had growth and transmission capabilities similar to those of the TRIG virus after acquisition of the human virus H3 and N2 genes.
Although there was no evidence of genetic diversification of the classical H1N1 virus in the United States before 1998, our data suggest that this lack of diversity was unlikely to have been because of an intrinsic property of the virus. Both of the swine viruses investigated, the TRIG virus NC/02 and the classical virus MN/99, supported superinfection of MDCK cells with the human virus TX/96 during simultaneous inoculations, indicating that there were similar opportunities for reassortment between MN/99 and NC/02. However, the results of the delayed-infection experiments clearly show that if infection with a second virus is delayed until at least 24 h after a primary infection, even at a low MOI, the second virus will be excluded from infecting the cells. This mirrors the results of other studies of reassortment between genetically similar viruses, in which increasing the time between infections reduced the incidence of coinfection and reassortment (22). One possible explanation for the decrease in viral superinfection over time is that infected cells export HA and NA to their cell surfaces, where NA cleaves surface sialic acids, thereby preventing other virus particles from binding to and infecting the same cells (28). Other mechanisms are also possible, however, such as induction of cellular antiviral responses. This aspect of superinfection exclusion needs to be studied further to determine the time course of exclusion as well as to further explore the mechanism of exclusion. Nevertheless, the biologic significance of viral exclusion is difficult to determine in vitro, especially considering the difference in viral dynamics in vivo.
The lack of a difference in superinfection exclusion led us to believe that there might be genetic incompatibilities between TX/96 and MN/99 that prevent reassortment between their genomes in nature. When the human HA and NA genes were rescued in both the MN/99 and NC/02 viruses, there were no significant differences in the viral growth properties in pSRECs. NC/TX grew similarly to the NC/02 wild-type virus, and MN/TX grew with kinetics similar to those of MN/99. Interestingly, TX/96 appeared not to grow well in pigs but grew at rates similar to those of the swine viruses in pSRECs, showing that host properties can alter viral kinetics in vivo. The lack of growth differences between the different reverse genetic viruses clearly demonstrates that the swine viral backbones can maintain similar growth kinetics with the TX/96 HA and NA genes.
Since the data indicated that nothing prevented MN/99 from reassorting with TX/96, pSRECs were simultaneously coinfected with TX/96 and NC/02 or MN/99, and the progeny virions were characterized by next-generation sequencing to identify and quantify the reassortants. MN/99 not only reassorted with TX/96 but also generated a larger variety of reassortants than did the coinfection with NC/02 and TX/96. The most surprising observation was that 67 (82.7%) of the identified H3 reassortant viruses produced by NC/02 and TX/96 were of the H3N1 subtype, whereas MN/99 and TX/96 produced that subtype in only 10 (16.4%) of the H3 reassortant viruses. The H3N1 subtype has previously been identified in pigs, but only at a very low frequency (17), suggesting that it has a favorable viral genetic composition but cannot be transmitted efficiently among pigs. Of the H3 reassortants between MN/99 and TX/96, 34 (55.7%) contained at least one polymerase gene segment from MN/99, whereas 77 (95.1%) of the H3 reassortants between NC/02 and TX/96 contained at least one polymerase gene segment from NC/02. To break this down further, 4 (6.6%) of the MN/99 H3 reassortants contained both the PB2 and PA gene segments of MN/99, whereas 47 (58%) of the NC/02 H3 reassortants contained both the PB2 and PA gene segments of NC/02. Additionally, the MN/99 PB1 gene segment was found in only 4 (6.6%) of the H3 reassortants, whereas the NC/02 PB1 gene segment was found in 58 (71.6%) of the H3 reassortants. Finally, no H3 reassortants contained all 3 polymerase gene segments from MN/99, but 38 (46.9%) of them contained all 3 polymerase gene segments from NC/02. These percentages clearly demonstrate that the wholly NC/02 polymerase complex and individual polymerase gene segments were more often identified in combination with the HA and/or NA gene segments originating from the human virus than were the MN/99 polymerase gene segments. The NC/02 polymerase genes were more frequently integrated with the TX/96 HA and NA genes than were those of MN/99, suggesting that a virologic aspect increases the relative reassortment frequency between NC/02 and TX/96. This in turn suggests that the NC/02 polymerase gene segments are more promiscuous in their ability to replicate with novel HA and NA gene segments but that this does not increase viral growth kinetics.
To further analyze the reassortant viruses, we calculated SCCs to determine if certain gene segments were more likely than others to be present with other specific gene segments. The positive correlation values obtained and the lack of any significant negative values suggest that the gene segments are more likely to associate with segments from the same virus than with segments from a new virus. The positive correlations observed between the MN/99 M, NS, and NP gene segments are probably the result of the protein interactions during the influenza virus life cycle: the M protein interacts with the nuclear export protein as newly formed ribonucleoproteins are exported from the nucleus (29, 30). These correlations are mirrored for the NC/02 virus, suggesting that these segments must be well matched in order for the progeny virions to be viable. Another correlation shared by both coinfections is the interaction between M and PA. No such interactions have actually been reported, but the data suggest that there is an interaction that requires these gene segments to be well adapted to each other. Interestingly, the only polymerase segments to correlate were the NC/02 PB1 and PA genes. This was expected, because the polymerase proteins must interact for effective RNA replication. The NA and M correlation may be the result of interactions of the NA transmembrane domain, as it is embedded on the viral surface. This relationship is also observed in the 2009 H1N1 pandemic virus, wherein the reassortant NA and M genes are both derived from Eurasian swine-like viruses (31).
Finally, to determine if reassortment with the human HA and NA genes from TX/96 would affect transmission of the swine viruses in vivo, we infected pigs with TX/96, MN/TX/NC-PB1, MN/TX, or NC/TX. Although TX/96 infected the inoculated animals, it was not transmitted to any of the contact animals. Similarly, MN/TX/NC-PB1 established an infection in inoculated animals but was not transmitted to any contact animals. In the pSRECs, we observed no difference in the ability of this virus to grow (data not shown), but the PB1 gene reassortant appeared to be slightly attenuated in vivo, with reduced viral shedding and a reduced duration of shedding. These factors may contribute to the lack of transmission, but this needs to be investigated further. However, when the TX/96 HA and NA genes were rescued with either of the 2 swine virus genes, the viruses could be transmitted to all of the contact animals. This provides compelling evidence that the internal genes play a crucial role as host determinant factors when there is no difference in HA receptor specificity (32).
Ultimately, these data suggest that there is no inherent viral characteristic that prevents the classical swine viruses from reassorting. Both the classical swine viruses and the TRIG viruses were capable of reassorting with the human H3N2 virus, producing progeny virions with growth characteristics similar to those of the parental strains. Not only could the reassortant viruses grow in primary swine cell cultures, but they also could infect and be transmitted among pigs, with similar efficiencies. The emergence of triple reassortant swine viruses in nature probably resulted from a combination of host factors, the environment, and their ability to rapidly establish a robust infection which led to their eventual predominance.
In the late 1980s, traditional pig farming began to shift to intensive farming in the United States. A USDA report indicates that between 1992 and 2004, the number of hog farms in the United States fell by 70%, while the total number of hogs remained the same (33). Traditional pig farms tended to be smaller farms where the piglets were bred and raised before being sent to market, with little to no movement between different swine herds. Intensive farms are much larger than traditional farms and are “industrialized,” with specialized functions. Piglets are farrowed and weaned before being moved to a finishing farm, where they are housed in confinement with many other animals. This provides an ideal situation for influenza viruses from different locales to be brought together in a single farm. It is possible that this change in environment drove swine influenza virus reassortment more than any inherent virologic property and that the TRIG viruses were simply the dominant viruses at the time, perhaps as a result of the development of immunity to the classical swine viruses. This report further illustrates and supports the concept that host restriction and transmission are polygenic traits that extend beyond the HA and NA gene segments. The TRIG backbone has a proven ability to reassort with HA and NA gene segments from different lineages and to be transmitted readily among both swine and humans; however, it is probably the viral and host ecology rather than solely virologic properties that contributes to the reassortment.
We thank Ashley Webb, Jennifer DeBeauchamp, Jeri-Carol Crumpton, and the St. Jude Animal Resource Center for their technical assistance, Jacco Boon for his expert advice and discussion, and Keith Laycock for manuscript review and preparation.
This work was supported by the American Lebanese Syrian Associated Charities (ALSAC) and by the National Institute of Allergy and Infectious Diseases Center of Excellence for Influenza Research and Surveillance (contract HHSN272201400006C).