Report

Role of Tissue Protection in Lethal Respiratory Viral-Bacterial Coinfection

See allHide authors and affiliations

Science  07 Jun 2013:
Vol. 340, Issue 6137, pp. 1230-1234
DOI: 10.1126/science.1233632

Tolerance Needed

It's a common enough occurrence: You're sick as a dog with a cold, but the person in the office next door just has a few sniffles. What accounts for this difference? Most commonly, these sorts of differences are chalked up to differences in resistance—perhaps you have higher viral loads than your office mate. But such differences can also involve differences in the ability to tolerate the same amount of virus. Deciphering the contribution of resistance versus tolerance, however, is difficult. Jamieson et al. (p. 1230, published online 25 April) studied a mouse model of viral and bacterial co-infection where tolerance and resistance could be separated. Mice infected with influenza virus were more likely to succumb to a secondary infection with Legionella pneumophila as a result of impaired tolerance to tissue damage, rather than because of a difference in bacterial burden.

Abstract

Secondary bacterial pneumonia leads to increased morbidity and mortality from influenza virus infections. What causes this increased susceptibility, however, is not well defined. Host defense from infection relies not only on immune resistance mechanisms but also on the ability to tolerate a given level of pathogen burden. Failure of either resistance or tolerance can contribute to disease severity, making it hard to distinguish their relative contribution. We employ a coinfection mouse model of influenza virus and Legionella pneumophila in which we can separate resistance and tolerance. We demonstrate that influenza virus can promote susceptibility to lethal bacterial coinfection, even when bacterial infection is controlled by the immune system. We propose that this failure of host defense is due to impaired ability to tolerate tissue damage.

Resistance and tolerance are two distinct strategies of host defense from infections: the former is based on pathogen detection and elimination, whereas the latter relies on host adaptation to a given level of pathogen burden (14). This distinction is important because infectious disease morbidity and mortality can be due to failed resistance or failed tolerance, which may, in turn, dictate different therapeutic options. Thus, a lethal outcome of microbial infection is usually ascribed to either high pathogen virulence or low host resistance (for example, caused by immunosuppression or immunodeficiency). Pathogen virulence can be due to direct damage to the host by toxins and virulence factors (intrinsic virulence) or, more commonly, due to excessive inflammatory response with collateral tissue damage (extrinsic virulence). However, insufficient tissue protection and repair could also be an important contributor to infectious disease phenotypes (4).

The upper respiratory tract is exposed to numerous pathogens simultaneously, and viral-bacterial coinfection in the lung is a common clinical manifestation [reviewed in (58)]. Complications from secondary bacterial infection are a leading cause of morbidity and mortality associated with influenza virus infection (58). Influenza virus can suppress the immune response to a bacterial infection, which can lead to increased bacterial load and decreased survival. This has been shown in both clinical studies and mouse models for multiple bacterial pathogens, including Streptococcus pneumoniae, Haemophilus influenzae, S. pyogenes, and Staphylococcus aureus [reviewed in (58)]. Bacterial overgrowth in these models complicates the analysis of other possible causes of morbidity and mortality. Therefore, we sought an alternative model of coinfection lacking this complication.

Legionella pneumophila has recently been recognized as a clinically relevant complication of influenza virus infection (9). When mice were infected with a sublethal dose of influenza virus and then coinfected with a sublethal dose of L. pneumophila, 100% of coinfected mice died within 1 week of coinfection, whereas all mice survived single infections (Fig. 1A). An established viral infection was necessary for lethality, because mice infected with influenza at the same time or 3 days after infection with L. pneumophila survived coinfection (Fig. 1A). A resolved influenza virus infection no longer affected the ability to survive coinfection with L. pneumophila, because mice infected with bacteria 10 or 14 days after viral infection also survived (Fig. 1A). The mice coinfected with L. pneumophila 3 days after influenza virus infection also had other signs of morbidity, including decreased body weight and temperature (Fig. 1, B and C).

Fig. 1 Decreased tolerance with unchanged resistance of mice coinfected with influenza virus and L. pneumophila.

(A) Survival of mice infected intranasally with 300 plaque-forming units (PFUs) of influenza virus and 0, 3, 6, 10, or 14 days (d) later with 1 × 106 L. pneumophila. To better mirror a human infection, mice were infected with a strain of L. pneumophila lacking the flaA gene (JR32∆flaA), which encodes flagellin. (B) Weight of mice infected with influenza virus or L. pneumophila alone or coinfected 3 days after influenza virus. (C) Body temperature of mice coinfected 3 days after influenza virus compared with the singly infected controls. (D) Lung bacterial load in mice coinfected 3 days after influenza virus or mice infected with L. pneumophila alone. Day 0 indicates colony-forming units (CFUs) in the lung 1 hour after infection. (E) Lung viral load in mice coinfected 3 days after influenza virus or mice infected in influenza virus alone. (F) Survival of mice coinfected 3 days after influenza virus infection with LP01ΔdotA or the thymidine auxotroph LP02 strains of L. pneumophila. (G) Survival of mice coinfected 3 days after infection with inactivated (IA) influenza virus. Data are combined from at least three independent experiments with at least five mice in each group (*P ≤ 0.05; **P ≤ 0.001). Data were analyzed with the logrank test, generalized linear model, or analysis of variance (ANOVA). Error bars represent SD from the mean.

Importantly, despite the dramatic difference in host survival, there was no significant difference in the viral or bacterial pathogen burden after single infections or coinfections (Fig. 1, D and E). Moreover, there was no detectable systemic dissemination of influenza or L. pneumophila after infections (fig. S1). These results indicate that lethal synergy between influenza virus and L. pneumophila was not due to impaired resistance to either of these pathogens. This is in contrast to coinfections with influenza virus and opportunistic bacterial pathogens where bacterial overgrowth and systemic dissemination are commonly observed (58, 1013).

To further address the role of pathogen virulence, we used an attenuated mutant strain of L. pneumophila, which lacks the Dot/Icm type IV secretion system and is therefore unable to secrete virulence factors (14). Administration of dotA mutant or thymidine auxotroph (LP02) L. pneumophila, which are severely attenuated in vivo, still resulted in 100% mortality of coinfected mice (Fig. 1F and fig. S2) (14). These results indicate that bacterial virulence or growth is not essential for lethal synergy of influenza–L. pneumophila coinfection. Furthermore, mortality is unlikely to be due to failed immune resistance. However, administration of formalin-inactivated influenza virus did not synergize with the subsequent L. pneumophila coinfection (Fig. 1G), indicating that a productive virus infection is necessary to make the host sensitive to secondary bacterial infection. Moreover, treatment of mice with neuraminidase inhibitors (NAIs) increased survival and decreased weight loss and hypothermia after coinfection (fig. S3, A to C), presumably because NAIs suppressed viral load (fig. S3D) (15).

We next examined whether mortality of influenza–L. pneumophila coinfection was due to excessive inflammatory response. Influenza virus activates three innate immune signaling pathways: the (i) Toll-like receptor–MyD88, (ii) RIG-I–interferon-α/β, and (iii) Nlrp3–caspase-1–interleukin-1 pathways (TLR, Toll-like receptor; IFN, interferon; IL, interleukin) (16, 17). L. pneumophila is recognized by the innate immune system via several mechanisms, including the Naip5/Birc1e-dependent pathway, which requires an intact Dot/Icm secretion system, and TLRs (1822). Gene expression analysis of the lungs after single infection and coinfection indicated that some of the inflammatory genes, including tumor necrosis factor–α (TNF-α), nitric oxide synthase 2 (Nos2), and several chemokines, were expressed at higher levels in coinfected compared with single-infected mice (fig. S4). TNF-α and IL-1β protein levels were also elevated in the broncho-aveolar lavage fluid (BALF) at day 3 after coinfection (Fig. 2A). Moreover, there was a significant increase in neutrophil infiltration in the lungs of coinfected mice compared with singly infected controls (Fig. 2, B and C). TNF-α, IL-1β, Nos2, and neutrophils are all known to play important roles in immunopathology, including in the context of influenza virus infection (2325). However, we found that genetic deletions of TNF-α, caspase-1, MyD88, TLR2/4, and Nos2 or antibody-mediated depletion of neutrophils or natural killer cells did not rescue coinfected mice from mortality (Fig. 2, D to J). Similarly, Rag2−/− mice, which lack an adaptive immune system, also succumbed to lethal coinfection, indicating that lymphocyte-mediated immunopathology is not essential for the lethal outcome of coinfection (fig. S5A). Virus-induced IFN-α/β can interfere with antibacterial responses (26). However, IFN-α/β receptor (IFNAR)–deficient mice (Ifnar1−/−) were still susceptible to coinfection (fig. S5B). Rag2 and IFNAR knockout mice were also susceptible to influenza infection alone, Nos2−/− mice were susceptible to infection with L. pneumophila alone, and Myd88−/− mice were susceptible to both single infections. However, in all cases, the mortality from coinfections was kinetically distinguishable from that from single infections and was similar to the mortality kinetics of wild-type (WT) mice (Figs. 1A and 2, D to J, and fig. S5, A and B). Finally, systemic treatment of the mice with synthetic glucocorticoid dexamethasone or antioxidant N-acetyl cysteine did not rescue them from mortality of coinfection (fig. S5, C and D). Collectively, the elimination of all major immune and inflammatory pathways triggered by either the viral or bacterial infection did not rescue the lethal synergy. These results suggest that the lethal outcome of coinfection in our model was not solely due to excessive inflammatory response or immunopathology.

Fig. 2 Decreasing inflammatory pathways does not increase survival of coinfected mice.

(A) TNF-α and IL-1β levels in the BALF of mice infected with JR32∆flaA 3 days after influenza virus infection. (B) Immune cell infiltrate in BALF of mice coinfected 3 days after influenza virus infection. (C) Types of infiltrating cells in the BALF. Survival of (D) Tnf−/−, (E) Casp1−/−, (F) Tlr2−/−Tlr4−/−, (G) Myd88−/−, (H) Nos2−/−, (I) Gr-1–depleted, and (J) NK1.1–depleted mice coinfected 3 days after infection with influenza virus compared with singly infected controls (Ctrl.). Data are combined from at least three independent experiments with at least five mice in each group (*P ≤ 0.05; **P ≤ 0.001; ***P ≤ 0.0001). Data were analyzed with the logrank test, t test, or ANOVA. Error bars represent SD from the mean.

Because neither bacterial growth or virulence nor host immune responses were individually required to cause lethality in coinfection, we next combined host immunodeficiency and bacterial attenuation. We used a severely attenuated L. pneumophila strain LP02∆dotA∆flaA, which lacks flaA and dotA and is also a thymidine auxotroph (27). This strain lacks flagellin and is unable to replicate and secrete effectors, thus lacking major immunostimulatory factors, except for cell-wall components detectable by TLR2 and TLR4. Therefore, we used this strain to coinfect TLR2/TLR4 double-deficient mice. Tlr2−/−/Tlr4−/− mice coinfected with influenza virus and LP02∆dotA∆flaA had a small increase in survival compared with WT mice infected with the same strain of bacteria (Fig. 3A); however, most mice still succumbed to coinfection. Tlr2−/−/Tlr4−/− mice coinfected with LP02∆dotA∆flaA had decreased immune cell infiltrate into the BALF at day 3 after infection, when compared with C57BL/6 mice infected with either LP02∆dotA∆flaA or Jr32∆flaA (Figs. 2, B and C, and 3, B and C). Thus, severely attenuated, nonreplicating L. pneumophila still caused mortality in coinfected Tlr2−/−/Tlr4−/− mice, despite almost a complete lack of immunostimulatory signals.

Fig. 3 Increased tissue damage in coinfected lungs.

(A) Survival of Tlr2−/−/Tlr4−/− and C57BL/6 mice coinfected with the LP02∆dotA∆flaA strain of L. pneumophila. (B) Types of infiltrating cells in BALF of Tlr2−/−/Tlr4−/− and C57BL/6 mice coinfected with strain LP02∆dotA∆flaA 3 days after influenza virus infection. (C) Amount of infiltrating immune cells in BALF of Tlr2−/−/Tlr4−/− and C57BL/6 mice coinfected with strain LP02∆dotA∆flaA 3 days after influenza virus infection. (D) Red blood cells (RBCs) and (E) albumin in the BALF day 4 after bacterial infection in C57BL/6 mice infected with influenza virus and 3 days later with the JR32∆flaA strain. (F) Representative images of airways and (G) histological damage scoring of lung sections 4 days after bacterial infection from C57BL/6 mice infected with influenza virus and 3 days later with the JR32∆flaA strain. Data are combined from at least three independent experiments with at least five mice in each group (*P ≤ 0.05; **P ≤ 0.001). Data were analyzed using the logrank test, t test, or ANOVA. Error bars represent SD from the mean.

The lethal outcome of influenza–L. pneumophila coinfection, despite normal control of pathogen growth, suggests that the mortality could be due to a failed tolerance to tissue damage caused by coinfection. Coinfected mice had high levels of red blood cells and albumin in the BALF (Fig. 3, D and E), indicating a damage to the lung epithelial-capillary barrier (28). The lung epithelial damage was further confirmed by histological analysis (Fig. 3, F and G). The principal difference among the singly infected and coinfected mice was in the degree of airway epithelial necrosis, with the coinfected lungs having a significant increase in epithelial cell damage. Extensive damage to the airway epithelia, with secondary alveolar collapse (Fig. 3, D to G), is presumably responsible for the mortality of the coinfection.

Consistent with the histological evidence of lung tissue damage, a gene expression analysis revealed that a cohort of genes involved in tissue protection and repair was specifically down-regulated in coinfected compared with singly infected or uninfected mice. This cohort included genes that are essential for tissue and cellular repair and development in the lung, such as Mdk, Adamts2, Timp4, Slpi, Mmp2, Mmp9, Vegfc, Itgb7, and Itga1 (29), as well as genes involved in stress response in lung tissue, such as Gcnt2, Hif3a, Stra13, Hmox1, and Aifm1 (30) (fig. S6).

We next tested whether the defective expression of the tissue-repair program is responsible for mortality of coinfection. Amphiregulin (AREG), an epithelial growth factor family member, was recently found to contribute to tissue homeostasis in the lung during influenza infection (31). Although AREG did not have a significant effect in WT mice, administration of AREG to Tlr2−/−/Tlr4−/− mice coinfected with influenza virus and the LP02∆dotA∆flaA strain of L. pneumophila significantly increased survival while decreasing weight loss and hypothermia (Fig. 4, A to C). AREG treatment resulted in decreased lung damage, as indicated by histopathological analysis, decreased albumin level in the BALF, and decreased pulmonary infiltrate (Fig. 4, D to G). Importantly, AREG treatment significantly decreased mortality of coinfection (Fig. 4A) but did not affect the viral and bacterial burdens (Fig. 4, H and I). The reason AREG administration did not rescue WT mice from coinfection is likely because, in this case, the disease is too severe and may require a more optimal regimen of AREG administration or, perhaps, additional methods of promoting tissue protection and repair.

Fig. 4 Targeting tolerance mechanisms increases survival of coinfected mice.

(A) Survival, (B) weight loss, and (C) temperature of Tlr2−/−/Tlr4−/− and C57BL/6 mice coinfected with the LP02∆dotA∆flaA L. pneumophila strain and treated with AREG daily. (D) Albumin levels in BALF from coinfected mice treated with AREG. (E) Histological damage scores, (F) representative samples of lung airway histology, and (G) percent area of inflammation per lung of coinfected lungs from Tlr2−/−/Tlr4−/− and C57BL/6 mice coinfected with the LP02∆dotA∆flaA L. pneumophila strain and treated with AREG. (H) PFUs and (I) CFUs in the lungs of coinfected Tlr2−/−/Tlr4−/− after treatment with AREG. Data are combined from at least three independent experiments with at least five mice in each group (*P ≤ 0.05; **P ≤ 0.001). Data were analyzed with the logrank test, t test, or ANOVA. Error bars represent SD from the mean.

Collectively, these results demonstrate that (i) lethal synergy of influenza virus and bacterial coinfection can result from loss of tolerance to infection-induced tissue damage, (ii) morbidity and mortality of coinfection can be independent of pathogen burden or excessive inflammatory response, and (iii) promoting tissue repair can, in principle, rescue coinfected animals from morbidity and mortality, even without affecting pathogen burden. Finally, our influenza–L. pneumophila coinfection model demonstrates the distinction between resistance and tolerance as separate host defense strategies that can both contribute to morbidity and mortality of infectious disease.

Supplementary Materials

www.sciencemag.org/cgi/content/full/science.1233632/DC1

Supplementary Text

Figs. S1 to S6

References (32, 33)

References and Notes

  1. Acknowledgments: We thank S. Holley and C. Annicelli for technical assistance; T. Ichinohe, M. Linehan, and A. Iwasaki for viral strains and advice; T. Ren, M. Fontana, R. Vance, K. Archer, S. Shin, and C. Roy for L. pneumophila strains and advice; M. Gillum for assistance with experiments; and M. Mueller and C. Lassnig for mouse infection infrastructure. The data presented in the manuscript are tabulated in the main paper and in the supplementary materials. This work was supported by the Howard Hughes Medical Institute (R.M.), NIH grants R01 046688 and AI R01 055502 (R.M.), the Ellison Foundation (R.M.), the New England Regional Center of Excellence (R.M.), and FWF (Austrian Science Fund) grant P25235-B13 (A.M.J.). A.M.J. was a Berger Foundation fellow of the Damon Runyon Cancer Research Foundation. The authors have no conflicts of interest.
View Abstract

Navigate This Article