Typically, the respiratory tract immune system is strictly controlled to prevent inflammation in response to innocuous antigens or commensal bacteria. When harmful pathogens colonize the respiratory tract, the local immune system becomes activated to eliminate the threat. It was found that typical mice can effectively clear up to 10⁵ pneumococci within 4–12 h (Sun and Metzger, 2008). However, with influenza infection onset, several processes occur that might impact the antibacterial innate immune response, rendering both the upper airways and lungs susceptible to subsequent bacterial infiltration, leading to increased bacterial load and mortality (Hillyer et al., 2004; Ishikawa et al., 2016). These processes include inhibition by type I interferons (IFNs) and depletion of alveolar macrophages.

After influenza infection is established, most viral subtypes replicate in the mucosal epithelial cells of the upper respiratory tract, providing more receptors for bacteria (Hatta et al., 2007; Nakamura et al., 2011). However, some viral subtypes can target both upper and lower respiratory tract tissues (Shinya et al., 2006; Childs et al., 2009; Maines et al., 2009; Munster et al., 2009). In particular, when viral infection precedes the presence of bacteria, access to otherwise inaccessible receptors in the lower respiratory tract might be available to invading bacteria (McCullers and Bartmess, 2003). Here, we consider three main mechanisms associated with this phenomenon.

First, influenza virus proteins can contribute. Neuraminidase, present on the envelope of the influenza virus, is responsible for sialidase activity, required by the virus for budding. After neuraminidase cleaves sialic acid from the termini of glycochains, cryptic receptors on host cells become exposed, and bacteria such as pneumococci can adhere (Foster and Hook, 1998; McCullers and Tuomanen, 2001). Furthermore, disrupted sialylated mucins can provide decoy receptors for bacteria (Plotkowski et al., 1986, 1993). For example, substantial numbers of epithelial cells can be destroyed by virulent viruses such as the mouse-adapted influenza strain PR8, resulting in exposed sites for bacteria to attach in the tracheobronchial tree (Plotkowski et al., 1986; McCullers and Rehg, 2002). Interestingly, neuraminidase proteins are not restricted to viruses. Some bacteria like Streptococcus pneumoniae also provide neuraminidases to access receptors and inhibit host defenses by cleaving sialic acids from protective mucins, allowing efficient infection of host lungs (Camara et al., 1991).

Second, the host inflammatory response to influenza infections can provide additional receptors. The host inflammatory response can alter not only the regulatory state, but also the surface display of multiple proteins, to facilitate pneumococcal invasion (Cundell and Tuomanen, 1994; Miller et al., 2007).

Third, adherence sites might also be provided during wound recovery in the airways (Plotkowski et al., 1993; de Bentzmann et al., 1996; Martin and Leibovich, 2005). There are some differences in terms of the wound recovery between common infection and co-infection. During co-infection with complex pathogens, reduced damage tolerance will occur; however, as a result, repair efficiency will decrease compared to that with a typical infection. This is also one of the causes of increased mortality with co-infection. Although some progress has been made, many potential mechanisms still need to be elucidated. Apical receptors including asialylated glycans (for example, GalNacβ1-Gal) or α5β1 integrins can be expressed on the surfaces of injured cells or those in an intermediate state of differentiation. Bacteria such as Staphylococcus aureus or Pseudomonas aeruginosa can adhere to receptors (Puchelle et al., 2006). Furthermore, bacteria such as S. pneumoniae, Haemophilus influenzae, or S. aureus can bind to exposed areas of incomplete healing via traditional adhesins. These exposed areas can be covered by basement membrane elements such as fibrin and fibrinogen deposition, laminin, or type I and IV collagen. This phenomenon has been observed in the clinics; patients can be easily infected by bacteria while recovering from primary illness (Louria et al., 1959; Peteranderl et al., 2017). In addition, many bacterial virulence factors can attach to elements of the extracellular matrix or basement membrane (Foster and Hook, 1998; McCullers and Tuomanen, 2001).

The viral cytotoxin PB1-F2, not present on all subtypes of influenza, is a non-structural protein encoded by an alternative reading frame on genomic segment 2 of influenza A. PB1-F2 is encoded by a small, variable open reading frame in PB1 that exists in most influenza viruses. This protein is capable of activating AP-1 transcription factors via ERK1/2 kinase, and was confirmed to be a determinant of virulence. The length of PB1-F2 differs according to subtype; full-length PB1-F2 is 90 aa (amino acids). However, that of the A/Puerto Rico/8/24 strain was found to be 87 aa. Most avian influenza viruses have complete PB1-F2-encoding genes. After 1947, PB1-F2 of the H1N1 subtype was found to be cleaved at position 57, and that of the classical swine H1N1 subtype was determined to be cleaved at 11, 25, and 34 aa, which results in reduced viral pathogenicity. Since 1968, variations in the PB1-encoding gene of the H3N2 subtype gradually stabilized, and PB1-F2 truncation gradually represented a novel evolutionary feature. The 2009 H1N1 strain did not show any significant antigenic and pathogenic variation. Patient infection after obvious gastrointestinal symptoms is not associated with viral virulence, but might be caused by individual-specific properties. When viral infection is established, PB1-F2 can induce apoptosis, mediated by mitochondrial permeabilization (McAuley et al., 2007, 2010; Alymova et al., 2011; Leymarie et al., 2013), ultimately providing nutrients to invading opportunistic bacteria, following cytopathic damage and disruption of surfactant in the lungs. Consequently, inhaled or commensal bacteria can become overgrown, adversely affecting host survival (Loosli et al., 1975). In 2016, observations (Sun et al., 2016) showed that effective antibiotic treatment of clinical post-influenza bacteria depends on nicotinamide adenine dinucleotide phosphate oxidase 2 (Nox2), and that a balance exists between Nox2-dependent antibacterial immunity and inflammation. However, influenza infection might disrupt this balance and increase susceptibility to bacterial infection.

Both influenza and bacteria contribute to the immunopathogenicity of co-infection. For example, expression of PB1-F2 has been associated with excessive inflammatory responses, which can lead to increased cellular infiltration of the lungs and airways, together with a cytokine storm (Conenello et al., 2007; McAuley et al., 2007, 2010). Interestingly, it was proposed that nascent hemagglutinin can be cleaved, from its primary state to a fusion-active complex, by bacterial proteases from S. aureus (Tashiro et al., 1987), which might increase influenza viral titers and spread.

Likewise, bacterial cytotoxins such as pneumolysin and Panton-Valentine leukocidin (PVL), can also contribute to immunopathogenicity (Rogolsky, 1979; Tuomanen et al., 1995; Loffler et al., 2013; Wolf et al., 2014). Bacterial components that lead to exacerbated cell death, associated with pore formation or enhanced inflammatory signaling, might synergize with influenza virulence factors (Boulnois et al., 1991). In addition, multiple innate immune mechanisms that involve pathogen recognition receptors also generate inflammatory responses to influenza and/or bacteria (Koppe et al., 2012; Ramos and Fernandez-Sesma, 2012). These inflammatory pathways work together, leading to synergistic activation of the immune system and increased mortality (Joyce et al., 2009; Bucasas et al., 2013; Kimaro et al., 2013).

Moreover, certain pathogenic virulence factors are only evident with the presence of a co-pathogen, as is the case for a mouse model of influenza and H. influenzae co-infection (Wong et al., 2013). However, it has not been possible to identify synergistic pathogenicity genes that facilitate co-infections using traditional virulence screens during single-agent infections (Bellinghausen et al., 2016).

The concept of tolerance used here is different from “immunological tolerance,” which is described as a state of unresponsiveness to self-antigens. Tolerance, in the current context, can reduce the negative impact of pathogen outgrowth and immunopathology, but can be compromised by the influenza virus, resulting in mortality even in the context of effective resistance.

For example, recent observations (Jamieson et al., 2013) demonstrated that influenza can compromise tolerance to tissue damage and result in mortality when mice were co-infected with Legionella pneumophila. It was shown that lethal synergy can be independent of impaired resistance to either influenza or L. pneumophila. Notably, this is different from previously discussed co-infections with influenza and opportunistic bacterial pathogens. Moreover, during the 1918 influenza outbreak, it was observed that human mortality was not directly related to infection rates (Shanks and Brundage, 2012). This study also demonstrated that lethal synergy between the influenza and L. pneumophila is unlikely to be due to failed immune resistance to either agent (Mendel et al., 1998).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.