The influenza virus genes that will dictate the severity of the next influenza virus pandemic currently exist in nature. However, most deaths associated with a future pandemic may likely be due to secondary infections caused by Gram-positive respiratory pathogens. Previous reports have demonstrated that certain versions of the influenza virus protein PB1-F2 play a role in the development of secondary bacterial pneumonia associated with Streptococcus pneumoniae, but our current surveillance efforts do not typically include efforts to sequence this viral gene. Surveillance strategies must be expanded to include genes associated with virulence in laboratory models of primary influenza virus infection and influenza virus:bacteria super-infection. Identification of potentially deadly reassortant virus strains before these circulate within humans may allow adequate time to implement plans for treatment and/or prevention.
The World Health Organization estimates that influenza virus epidemics are associated with 250–500,000 deaths and 3–5 million illnesses worldwide on an annual basis. It is well-accepted that influenza virus infection predisposes hosts to secondary bacterial pneumonia, and the Gram-positive pathogens Streptococcus pneumoniae, Staphylococcus aureus and Streptococcus pyogenes have all been associated with excess mortality during influenza epidemics and pandemics. The virulence associated with a circulating influenza virus appears to be a multi-genic trait, with specific reassortment events yielding viruses with the potential to cause devastating pandemics, like the one that occurred in 1918. Currently, the highly pathogenic avian influenza A viruses of the H5N1 subtype represent a significant pandemic threat that could be realized if these viruses acquire the ability to transmit from human-to-human. In addition to identifying the genetic changes associated with adequate transmissibility of these viruses, we must also understand the viral genes that modulate virulence in the context of both primary influenza virus infection and influenza virus:bacteria super-infection. One way to accomplish this goal is to identify viral genetic signatures associated with virulence that occur in the natural setting, ideally in the dominant intermediate mammalian host, the swine.
Influenza virus virulence factors have been largely defined using models of primary influenza virus infection, with relatively few studies evaluating the contribution of these factors toward disease within models of influenza virus:bacteria super-infection. The influenza virus protein PB1-F2, discovered in 2001 (Chen et al., Nat Med 2001), has been studied as a viral virulence factor that contributes to morbidity and mortality within models of influenza virus:S. pneumoniae super-infection. Viruses that circulated during the 1918, 1957 and 1968 influenza pandemics expressed full-length, virulent variants of this 87–90 amino acid viral protein, while future seasonal virus isolates expressed variants of this protein that were less virulent due to the loss of specific amino acids through either mutation or truncation (McAuley et al., PLoS Pathog 2010). Using experiments designed to focus solely on the PB1-F2 protein, prior studies have identified the key amino acids expressed by this protein that predispose the host to lethal super-infections. While studies designed to focus on individual viral virulence factors are critical for evaluating the mechanisms through which a given virulence factor affects overall morbidity and mortality, one must also consider whether variant forms of these genes contribute to virulence when expressed by naturally-occurring virus isolates.
In a recent manuscript, our group tested the hypothesis that the laboratory-defined virulence associated with the PB1-F2 gene could be used to predict the outcome of secondary bacterial infections initiated with naturally-occurring swine influenza virus isolates. To test this hypothesis, natural swine influenza virus variants were incorporated into murine models of influenza virus:bacteria super-infection, with either S. pneumoniae, S. aureus or S. pyogenes acting as secondary bacterial invaders. One unique aspect of this study is its evaluation of the virulence associated with naturally-occurring influenza virus variants, using the PB1-F2 genotype as the sole criterion for virus selection. In addition, this is also the first study where three distinct Gram-positive respiratory pathogens were directly compared in a model of influenza virus:bacteria super-infection. The major findings from this study are summarized here, with emphasis on the association between laboratory-defined virulence and naturally-occurring virus reassortants that circulate within pigs.
To date, the virulence associated with PB1-F2, defined in mice using both models of primary influenza virus infection and models of influenza virus:bacteria super-infection, has been mapped to five specific amino acid residues (62L, 66S, 75R, 79R and 82L). Naturally occurring swine influenza viruses that were selected for inclusion in our studies were grouped based on the number of virulence-associated amino acids expressed. In addition, one group of viruses expressed truncated forms of PB1-F2 that were either 11 amino acids or 57 amino acids in length. Based on the length of these truncated proteins, none of the virulence-associated amino acids indicated above were present in the PB1-F2 proteins expressed by these viruses.
In general, our findings demonstrate that outcomes after super-infection in mice are directly associated with an increasing number of virulent PB1-F2 amino acids expressed, regardless of the bacterial species introduced as the secondary invader. Of note, the virus that demonstrated the most lethal phenotype within our model was the only virus that expressed an asparagine at position 66 instead of a serine. This N66S mutation has demonstrated increased virulence in models of primary influenza virus infection with the 1918 H1N1 virus, and may have contributed to the high mortality observed during this pandemic (Conenello et al., J Virol 2011). Furthermore, one surprising observation was that viruses expressing truncated PB1-F2 proteins, with predicted avirulent phenotypes, demonstrated a particular preference for S. pyogenes as a secondary invader, compared with S. pneumoniae and S. aureus. This demonstrates that variation in the bacterial species can also contribute to the severity of the super-infection, providing evidence that all Gram-positive respiratory pathogens cannot be considered equivalent in their ability to induce lethal outcomes in murine super-infection models. Thus, survival after super-infection can differ greatly based on host, viral, and bacterial contributions to these polymicrobial infections.
The obvious caveat to our findings is that additional virulence factors expressed by these naturally-occurring swine influenza virus variants have the potential to contribute to the phenotype observed. However, our ability to use a laboratory-characterized virulence factor to predict the outcomes of a super-infection supports current efforts to expand the surveillance of naturally-occurring influenza virus variants to include the sequencing of virulence-associated genes, like PB1-F2. Because whole-genome sequencing of viruses can be done rapidly and inexpensively now, I propose that this replace current strain typing methods. Alternatively, either microarray or PCR-based methods designed to detect specific amino acid residues that have been identified as molecular signatures of virulence could also be employed.
Since the genes expressed by prior pandemic influenza viruses are known to have evolved within intermediate mammalian species during the pre-pandemic phase (Vijaykrishna et al., Science 2010), pigs represent an excellent target for such surveillance efforts. In fact, our model specifically incorporated primary swine influenza virus isolates that already express virulent PB1-F2 variants, which demonstrates that reassortment events could occur at any time, with potentially deadly outcomes predicted. Experiments designed to evaluate the contribution of additional viral genes toward the virulence observed with these specific viruses have been initiated, with full consideration for contributions of more than one gene toward severe super-infections. At the present time, interspecies transmission of H3N2 variant (H3N2v) influenza viruses from swine to humans has increased in the United States (CDC, MMWR 2012). An assessment of the PB1-F2 sequences present in five of these recent H3N2v human isolates indicates that the PB1-F2 gene expressed is full-length (90 amino acids), it contains just two amino acids associated with virulence (62L and 82L), and there is an N present at position 66. Thus, focusing solely on the PB1-F2 genotype, I would predict that these H3N2v viruses would be less likely to predispose toward severe secondary bacterial infections with Gram-positive respiratory pathogens, at least in a murine model.
In summary, our group recently evaluated naturally-occurring swine influenza viruses within murine models of influenza virus:bacteria super-infection. The swine influenza viruses utilized were selected based on the predicted virulence associated with the PB1-F2 gene expressed. This study was not designed to identify the specific genotypes associated with virulence, rather our intent was to determine whether increased surveillance efforts could provide insight into outcomes from influenza virus:bacteria super-infection. Although some exceptions to the general rule were identified, we were able to broadly predict severity of secondary bacterial infection based solely on the PB1-F2 expressed by these influenza viruses. Since secondary bacterial infections contribute significantly to the morbidity and mortality observed after primary influenza virus illness, our ability to use surveillance to predict the incidence of these complicated infections will allow us to improve future approaches toward treatment and prevention.