Materials and Methods

Infectious agentsThe Mount Sinai strain of mouse-adapted influenza virus A/Puerto Rico/8/34(H1N1), hereafter referred to as “PR8,” was grown in MDCK cells from stock from the influenza virus repository at St. Jude Children’s Research Hospital (Memphis). The dose of PR8 lethal for 50% of infected mice (MLD₅₀) was the equivalent of 3000 TCID₅₀ and 140 MID₅₀ (the dose infectious for 50% of mice). S. pneumoniae D39, a type 2 encapsulated strain, was grown in Todd-Hewitt broth (Difco Laboratories). The MLD₅₀ for pneumococcus was equivalent to 5×10⁵ cfu, as quantitated on tryptic soy agar (Difco Laboratories) supplemented with 3% (vol/vol) sheep erythrocytes

MiceFemale BALB/c ByJ mice (8–10 weeks old; Jackson Laboratory) were maintained in a biosafety level 2 facility in the Animal Resource Center at St. Jude Children’s Research Hospital. All experimental procedures were done while mice were under general anesthesia with inhaled isoflurane 2.5% (Baxter Healthcare)

Infectious modelInfectious agents were diluted in sterile PBS and administered intranasally in a volume of 100 μL (50 μL/nostril) to anesthetized mice held in an upright position. Mice were weighed and monitored at least daily for illness and mortality. Mice found to be moribund were euthanized and considered to have died on that day. For experiments using PAFr inhibitor, CV-6209 (Biomol Research Laboratories) was diluted in sterile PBS and administered via tail-vein injection in a volume of 100 μL every 8 h

Blood culturesApproximately 110 μL of blood was obtained from mice via retro-orbital puncture with polished, sterile, glass Pasteur pipettes 24 h after infection with pneumococcus and was transferred into Isolator 1.5 microbial tubes (Wampole Laboratories). Quantitation of colony counts by the Isolator 1.5 system was done by 10-fold dilutions on tryptic soy agar plates supplemented with 3% sheep erythrocytes (vol/vol). The assay can quantitate colony counts between 10 and 10¹⁰ cfu/mL of blood

Lung titersMice were euthanized by cervical dislocation. Lungs were removed under sterile conditions, washed 3 times in sterile PBS, and placed into 500 μL of sterile PBS. Lung homogenates were used directly for bacterial cultures or were spun at 10 g for 5 min, and the supernatants were used for determination of virus titers. Quantitation of pneumococcal colony counts was done by 10-fold dilutions on tryptic soy agar plates supplemented with 3% sheep erythrocytes (vol/vol). Identification of pneumococcal colonies was done by visual inspection and recognition of standard colony morphology and characteristic hemolysis on blood agar. Virus titers were determined by 10-fold serial dilutions on MDCK cell monolayers to obtain the TCID₅₀

Pathologic analysisLungs were removed immediately after euthanasia and fixed in 10% neutral buffered formalin. After 24 h of fixation, the lungs were embedded in paraffin, sliced into 5-μm sections, stained with hematoxylin-eosin, and examined microscopically for histopathologic alterations

The lungs from each mouse were examined by a blinded observer (J.E.R.) to determine the percentage of lung parenchyma with inflammation and whether the airways (bronchioles and alveolar ducts) contained inflammatory cells, epithelial necrosis, or epithelial hyperplasia. Each microscopic field (magnification, ×250) of the lung parenchyma evaluated was assigned a score of 1+, 2+, or 3+, on the basis of degree of involvement. Lung parenchyma involvement of <25% received a score of 1+, lung parenchyma involvement of 25%–50% received a score of 2+, and lung parenchyma involvement of >50% received a score of 3+. Each airway in the lung was scored from 0 to 3+, with the morphologic changes of inflammation, necrosis, and hyperplasia having equal weight of 1+ each. An overall pathologic score of pulmonary alteration for each mouse was determined by dividing the sum score of the parenchymal fields and of the airways by the maximum combined score possible for the 2 parameters

Statistical analysisComparison of survival between groups was done with the Mantel-Cox χ² test on the Kaplan-Meier survival data. Comparison of quantitative bacterial counts in blood and lungs between groups was done with the Wilcoxon&amp;rank sum test. P<.05 was considered to be significant for these comparisons. Evaluation of differences in lung pathologic analysis scores was done by pairwise comparisons with the exact Wilcoxon&amp;rank sum test and Student’s t test. Because of the small sample size (3 animals/group), significance in the pathologic comparisons could be claimed only when P<.10. Because results for both tests were similar, only those for the Student’s t test are reported

Results

Lethal synergismTo determine whether synergistic mortality could be seen in a mouse model of dual infection with influenza virus and pneumococcus, mice were given influenza virus and pneumococcus either alone, simultaneously, or in sequence 1 week apart. Groups of 20 mice were infected on day −7 and then again on day 0 with either 0.3 MLD₅₀ of influenza virus, 0.2 MLD₅₀ of pneumococcus, both together, or sterile PBS (mock infection) in various combinations. These doses were chosen so that mice would be ill but mortality would be predicted to be infrequent in the absence of synergism. Mice were monitored for 21 days after the second challenge (day 0); all mice survived for at least the initial 7 days between infections. Survival after challenge with the second agent is plotted in figure 1. Lung virus titers peaked at days 2–3 after inoculation with PR8 at ∼2×10⁷ TCID₅₀ and decreased by only ∼1 log on day 7 at the time of pneumococcal challenge (data not shown)

Mice infected with either influenza virus or pneumococcus alone at day 0 after mock infection exhibited mortality rates of 35% and 15%, respectively. Infection with both agents simultaneously (in a total volume of 100 μL) at day 0 after mock infection gave an additive effect, with 60% of the animals dying 2–11 days after the dual infection. Mice infected with pneumococcus at day −7 and challenged with influenza virus at day 0 were protected from influenza virus infection with no mortality, no clinical symptoms, and no weight loss (weight loss and morbidity data not shown), which is identical to the day 0 results seen with control mice challenged only with PBS at both time points (P<.001, for difference in survival, vs. the group treated with virus alone). In contrast, all mice infected with influenza virus at day −7 and pneumococcus at day 0 uniformly succumbed to infection in <24 h (P<.001, for difference in survival, vs. all other groups), demonstrating that synergistic mortality occurs in the model

Timing of infectionAfter the demonstration that synergistic mortality was seen when influenza virus infection preceded pneumococcal challenge by 1 week, the number of days between administration of the 2 infectious agents was altered to determine the optimal timing necessary to achieve this effect. Groups of 6 mice were infected with 0.3 MLD₅₀ of influenza virus at day 0 and challenged with 0.2 of MLD₅₀ pneumococcus at days −7, 0 (simultaneously), 3, 5, 7, 9, 14, or 21 and were monitored for 21 days after the second infection. The percentage of mice surviving and the mean duration of survival (i.e., the number of days the mice survived after the second infection, considering only mice who died) are shown in figure 2

As in the first experiment, an additive mortality was seen with simultaneous infection, no mortality was observed when pneumococcal infection preceded influenza virus infection, and 100% lethality occurred when pneumococcal challenge followed influenza virus infection by 7 days. One hundred percent mortality was also observed in the day 3 and day 5 groups, although death was most rapid in the day 7 group, occurring in <24 h in all mice, compared with a mean duration of survival of 3.3 days for the group challenged on day 3 after influenza virus infection and 2.5 days for the group challenged on day 5. The synergistic effect was lost by day 21. These data suggest that there is a temporal sequence of events that must occur to prime for synergistic lethality and indicate that, in this model, pneumococcal challenge 7 days after influenza virus infection results in the most rapid and complete mortality

Dose response for influenzaOne potential contributor to the synergistic mortality seen with pneumococcal challenge after influenza virus infection was the debilitation of the mice at the doses of PR8 used in the previous experiments (0.3 MLD₅₀), which may have been most severe at 7 days after influenza virus infection. We therefore infected groups of 10 mice with sequentially 2-fold-lower doses of influenza virus, from 140 MID₅₀ (equivalent to 1.0 MLD₅₀) to 0.5 MID₅₀, then challenged with 0.2 MLD₅₀ of pneumococcus (1×10⁵ cfu) at day 7 after infection. The dose at which 50% of mice exhibit morbidity (i.e., clinical signs of infection such as weight loss, ruffled fur, huddling, hunched posture, shivering, and tachypnea) for this PR8 stock is 10 MID₅₀. High mortality was observed at doses of influenza virus as low as 1 MID₅₀ (table 1), with only 30% survival, compared with 90% in control mice, indicating that synergism occurs at doses below that needed to engender clinical symptoms or weight loss. Therefore, other synergistic mechanisms besides debilitation from influenza must be operative in the model. The mean duration of survival (among mice that died) was directly related to the influenza virus dose (table 1)

Dose response for pneumococcusAfter the demonstration that low doses of influenza virus can be used in the model with maintenance of the synergistic effect, an experiment was undertaken to determine the minimum dose of pneumococci needed for synergistic killing. Groups of 10 mice were infected with 0.13 MLD₅₀ of influenza virus (equivalent to 18 MID₅₀, the lowest dose at which 100% mortality was observed in the prior experiment) and challenged at day 7 with sequentially 4-fold-lower doses of pneumococcus ranging from 1×10⁵ cfu (equivalent to 0.2 MLD₅₀) to 100 cfu. Data in table 2 indicate that synergistic mortality can be observed with low doses of both influenza virus and pneumococcus, although 100% mortality was seen only at the highest dose tested. Of interest, although only 50% mortality was observed in the group receiving 100 cfu of pneumococcus, the mean duration of survival (among mice that died) was similar to that for the highest group. Thus, it is seems apparent that the rapidity of demise is related to the timing of administration of the 2 agents and the dose of influenza virus but not the dose of pneumococcus, whereas all 3 parameters factor into the percent mortality

Pathologic abnormalities in the synergism modelThe pulmonary alterations in the mouse lungs involved the pulmonary airways and/or the pulmonary parenchyma, but the degree of tissue involvement depended on the treatment received. Airway involvement was primarily limited to mice infected with virus and was characterized by inflammatory cells in the lumen and/or mucosa, as well as epithelial necrosis and epithelial hyperplasia. Parenchymal involvement was focal, and the percentage of infected tissue was variable. There were no significant differences, compared with controls, in the lungs of mice infected with pneumococcus alone (figures 3A and 3B ; control mice not shown). However, virus-infected mice at 48 h after challenge at day 0 (i.e., 9 days after infection with influenza virus) had multiple parenchymal foci with alveolar inflammation, alveolar epithelial cell hypertrophy and hyperplasia, and occasional alveolar necrosis and fibrin deposition. Inflammatory cell infiltrates with lymphocytes, neutrophils, and macrophages were seen, particularly around blood vessels. The pulmonary alterations of mice infected with virus then bacteria were significantly more extensive (P<.005) than those of mice given either virus or bacteria alone (figure 3). Extensive and severe consolidation of affected lobes and obliteration of alveolar architecture were striking findings. Inflammatory infiltrates, epithelial cell hypertrophy and hyperplasia, and fibrin deposition were common (figure 3E and 3F )

Effect of PAFr inhibition on survivalIt has been demonstrated that up-regulation of PAFr during an inflammatory response in the lung facilitates bacterial invasion so as to cause bacteremia [19]. We hypothesized that, if pneumococcus is using PAFr for adherence and invasion in the lungs after up-regulation during an influenza infection, inhibition of PAFr would ameliorate secondary pneumococcal pneumonia and prevent bacteremia in the mouse model of lethal synergism. We first assessed mortality in the presence or absence of CV-6209, a competitive antagonist of PAF at the PAFr [36]. The dose used was sufficient to completely prevent hypotension, shock, and death after intravenous (iv) challenge with 1 μg of PAF (Biomol Research Laboratories) in mice (data not shown). Groups of 25 mice were infected with 100 TCID₅₀ of influenza virus PR8 and then challenged 7 days later with 100 cfu of pneumococcus D39. These doses were chosen on the basis of results of earlier experiments to give an intermediate mortality rate, so that positive or negative effects of the drug could be determined. Five mice were mock infected with PBS and then challenged 7 days later with 100 cfu of pneumococcus D39 as a control group. Mice were administered either sterile PBS or CV-6209 1 mg/kg iv every 8 h for a total of 8 doses, beginning 30 min before pneumococcal challenge. There was no significant effect on survival from inhibition of PAFr by CV-6209 (figure 4). However, mice treated with CV-6209 demonstrated delayed mortality (no mice died in the first 48 h; P<.001) and showed few clinical symptoms in the first 48 h, compared with mice receiving PBS who were clinically quite ill (J.A.M., personal observation). Treatment with CV-6209 for 7 days instead of 3 did not alter outcome (data not shown)

Virus and bacterial titers after PAFr inhibitionTo determine what effect the PAFr inhibitor had on the model beyond delayed mortality, we repeated the experiment with CV-6209 treatment, as detailed in the previous section, and assayed mice for lung virus titers. There were no significant differences in lung virus titers 48 h after pneumococcal challenge in mice treated with CV-6209 (n=6; mean titer, 5.25 TCID₅₀), compared with pneumococcus-infected mice treated with PBS (n=6; mean titer, 5.0 TCID₅₀) or mice that were challenged with PBS instead of pneumococcus (n=3; mean titer, 5.25 TCID₅₀)

We next quantitated blood cultures in groups of mice (n=12 per group) at 24 h after pneumococcal challenge. Blood cultures were positive in 9 (75%) of 12 mice treated with CV-6209, compared with 5 (42%) of 12 pneumococcal controls (figure 5). This difference was not statistically significant (P=.0532, Wilcoxon&amp;rank sum test). Because of the apparent (although not statistically significant) increase in prevalence of positive blood cultures in mice receiving CV-6209, the opposite of what we hypothesized, we repeated the experiments, administering PBS as a mock infection instead of influenza virus and challenging with pneumococcus as before (figure 5). Among pneumococcus-infected mice treated with CV-6209, 2 (18%) of 11 had positive blood cultures, compared with 0 (0%) of 12 controls. Although not statistically significant, this was a surprising result, because this low dose of pneumococcus D39, when administered intranasally, was typically cleared from the mouse in <24 h and did not result in detectable lung titers or cause bacteremia without pretreatment with influenza

Quantitative bacterial counts were obtained from the lungs of mice at 24, 48, and 72 h after pneumococcal challenge. Four groups of pneumococcal-infected mice received either (1) influenza pretreatment and then CV-6209 or PBS or (2) PBS pretreatment (mock infection) and then CV-6209 or PBS. As seen in figure 6, mice pretreated with influenza who received CV-6209 had higher bacterial lung titers each day (P<.05, 24-h vs. 72-h titers within the group), whereas mice pretreated with influenza but receiving PBS had a decrease in their titers by 72 h (P<.05, 24-h vs. 72-h titers within the group). Because of the failure to clear bacteria in the group receiving CV-6209, as occurs in control mice, the difference in bacterial lung titers at 72 h was significantly different between the groups (P<.05). A similar result was seen in the absence of influenza pretreatment, although the effect was not as pronounced. Bacteria could be cultured from the lungs of mice treated with CV-6209 during the first 48 h but not at all from control mice. Thus, CV-6209 seemed to be predisposing to bacteremia, but the evaluation of the effect of influenza pretreatment was not conclusive

Pathologic abnormalities in mice treated with CV-6209In mice infected with pneumococcus alone, there was no significant difference (P>.10) in overall pulmonary pathologic abnormalities, comparing mice receiving inhibitor with control mice (figure 7A and 7B ). There was also no significant difference (P>.10) in the overall pulmonary reaction of the mice treated with CV-6209 after influenza and pneumococcal infection, compared with those receiving PBS (figure 7C and 7D ). However, neutrophil migration in the lungs appeared to be delayed, because there were fewer neutrophils in the lungs of mice receiving the inhibitor at 24 h. At 48 h, this trend reversed, and neutrophils were present in greater numbers in the lungs of mice receiving CV-6209. There was a tendency for mice receiving CV-6209 to have more inflammatory cells surrounding the pulmonary blood vessels than did the mice receiving PBS (figure 7B )

Discussion

The long-term goal of studying synergism between influenza virus and bacteria is to determine the mechanism(s) involved so that targets can be identified for intervention in human disease. An important first step toward achieving this goal was the development and characterization of a mouse model that could reproduce the excess mortality seen in epidemiologic studies. After the recognition that bacterial superinfection frequently accompanied fatal influenza during the 1918 Spanish influenza epidemic [4, 8, 10], attempts were made to examine the interaction between influenza virus and bacteria in animal models. Early experiments explored the interaction between influenza virus and H. influenzae in pigs [37] and then in mice [38]. Studies in the late 1940s then extended these observations to pneumococcus, demonstrating that sublethal infection of mice with a pneumovirulent influenza virus could prime for lethal pneumonia after inhalation of aerosolized pneumococci [39]. In recent years, study has centered on the chinchilla model [40] and bacterial-viral synergism in otitis media. In contrast, the role of viruses in predisposing for bacterial pneumonia has received little attention, despite being a leading contributor to mortality worldwide

Although the impetus for thoroughly characterizing this mouse model of lethal synergism was to provide a framework for further studies exploring mechanisms, some observations can be made from these initial data. Mice treated with doses of influenza and pneumococcus near the MLD₅₀ of each agent were highly bacteremic [35] and died rapidly from overwhelming sepsis (figure 1). However, lower doses of the infectious agents result in a more prolonged disease course and mortality from secondary bacterial pneumonia (tables 1 and 2 and figure 3). Thus, the model will be useful for studies of both sepsis and pneumonia, and the desired mortality and length of illness for the purpose of scientific study can be controlled. The observation that lethality can be seen at doses of both influenza virus and pneumococcus below the levels at which clinical signs and weight loss are seen when administered separately supports the concept of a synergistic, rather than an additive, process

Two observations can be made from the studies of the timing of administration of the infectious agents. First, a prior pneumococcal infection appeared to protect mice from morbidity and mortality from a pneumovirulent influenza virus, despite no reduction in virus titers in these mice (weight loss and virus titer data not shown). The reason for this is not known, but it may be that nonspecific elements of the pulmonary immune response are stimulated by the pneumococcal infection and blunt the cytokine responses that lead to anorexia, weight loss, and death in mice during influenza infection. Second, the delay between the establishment of influenza infection and challenge with pneumococcus necessary to produce synergism (figure 2) implies that host factors are involved in the interaction. Influenza virus must be directly or indirectly altering the host in some way that makes it more susceptible to pneumococcal infection, an alteration that returns to normal within 2–3 weeks of infection. The pathologic finding that the pneumonic consolidation is spread extensively throughout the lung after influenza infection (figure 3) argues that this alteration occurs at the level of the pneumococcal receptor(s) or host defense, allowing for spread of pneumococcus beyond a focal, lobar process. This echoes pathologic descriptions of “influenzo-pneumococcal septicaemia” during the 1918 pandemic: “True lobar pneumonia is not what is found post mortem in the majority of cases. Whole lobes may be consolidated…” [4 p. 7], and “The most common condition was one in which almost entire lobes were consolidated.…The term ‘diffuse bronchopneumonia’ probably best describes the condition” [8 p. 4]. In addition, bacteremia is increased after influenza virus preinfection (figure 5 and [35]), which may be due to the same mechanisms underlying the increased severity of pneumonia or may be an entirely different pathway

We previously characterized the mouse model of lethal synergism between influenza virus and pneumococcus [35], so we sought to use this same model to explore a potential mechanism for more extensive pneumonia and increased bacteremia following influenza virus infection. Up-regulation of PAFr would be expected during the host cytokine response to influenza. Since PAFr is permissive for pneumococcal attachment and invasion, PAFr could play a central role in synergy. On the basis of in vitro data showing that competitive antagonism of the PAFr blocked pneumococcal adherence to cytokine-stimulated A549 cells [19], we reasoned that a similar process in vitro would prevent pneumococcal adherence and invasion in the lungs of mice infected with influenza virus. We hypothesized that competitive inhibition of PAFr with CV-6209 would decrease mortality, bacteremia, and lung titers of pneumococcus. Although no overall effect on mortality was observed, a delay in mortality was seen (improved survival at 48 h). However, inhibition of PAFr with CV-6209 appeared to increase bacteremia and lung titers. The moderate phenotype implies that the system is more complex than we had anticipated. We think it most likely that there are more receptors involved in the synergistic interaction and that our focus on PAFr alone in a complex system does not permit an in-depth assessment of the involvement of this potential mechanism. However, other possibilities exist as well. Pneumococcus does not need to activate the signaling functions of PAFr and might, therefore, bind at a site on the molecule not occupied by the inhibitor which competitively blocks the active site to which PAF binds. In addition, it has been suggested that there is some heterogeneity of PAF receptors in different cell types [36], so CV-6209 may inhibit one receptor that mediates chemokine functions of PAF, but not inhibit another that can be used by pneumococcus. In addition, the role of the various choline-binding proteins, modulation of their expression through phase variation, and a variety of other pneumococcal adherence or virulence factors is yet to be delineated (reviewed in [41])

Although we did not demonstrate a reduction in mortality using the PAFr antagonist, other useful information has come out of this study. The failure of mice treated with CV-6209 to clear bacteria from the lungs and blood leads to some important conclusions. If we are indeed blocking PAFr in the lung, then pneumococcus can cause pneumonia and bacteremia by pathways independent of PAFr. This downplays, to some extent, the primacy of this pathway in the pathogenesis of lung disease [42, 43] and adds a level of complexity, if strategies for interrupting pneumococcal disease by targeting lung receptors [42, 44] are to succeed. More research into the role of PAFr in pathogenesis and the role of other potential receptors, such as C3 in the lung [45] or polymeric immunoglobulin receptor in the nasopharynx [46], is needed. Further study of the effects of PAFr inhibition in this model, particularly an examination of cytokine levels, is warranted as well. Why clearance of bacteria was prevented by CV-6209 is unclear. Inflammatory cells of all types were present in the lungs of necropsied mice at 48 h in numbers equal to or greater than in controls; thus, recruitment into the lungs was unaffected. This implies that the function of these cells was affected. Since PAFr is found on many cell types including macrophages and neutrophils, CV-6209 may have had a direct effect on the inflammatory cells responsible for clearance of bacteria. Finally, the delayed mortality in the presence of CV-6209 remains unexplained. The specific mechanism of death in pneumococcal pneumonia is poorly understood [47], but it seems to be inextricably bound up in the host response to pneumococcus. The delayed mortality in this model may reflect an interruption or delay of that host response allowing time for uncontrolled growth of S. pneumoniae to overwhelm the host

Findings from this study relevant to synergism between influenza and pneumococcus are summarized in the following list:

1. Influenza virus infection primes for development of lethal pneumococcal pneumonia and bacteremia in a dose dependent fashion;

2. Seven days between infectious agents produces the greatest effect;

3. More extensive and severe pathologic findings are seen than in classic lobar pneumonia;

4. Preceding pneumococcal infection improves survival from influenza; and

5. Pneumococcus can cause pneumonia and bacteremia by pathways independent of the platelet activating factor receptor