The role of respiratory viruses in the transmission of Streptococcus pneumoniae is poorly understood. Key questions, such as which serotypes are most fit for transmission and disease and whether influenza virus alters these parameters in a serotype-specific manner, have not been adequately studied. In a novel model of transmission in ferrets, we demonstrated that pneumococcal transmission and disease were enhanced if donors had previously been infected with influenza virus. Bacterial titers in nasal wash, the incidence of mucosal and invasive disease, and the percentage of contacts that were infected all increased. In contact ferrets, viral infection increased their susceptibility to S. pneumoniae acquisition both in terms of the percentage infected and the distance over which they could acquire infection. These influenza-mediated effects on colonization, transmission, and disease were dependent on the pneumococcal strain. Overall, these data argue that the relationship between respiratory viral infections, acquisition of pneumococci, and development of disease in humans needs further study to be better understood.
Streptococcus pneumoniae, the pneumococcus, is the most common etiologic agent associated with otitis media, community-acquired pneumonia, and invasive diseases, such as sepsis and meningitis . Because pneumococci can express 1 of <90 different capsules, and because the local and regional distribution of pneumococcal serotypes based on these capsules can differ considerably according to geography, period studied, or age, prevention of pneumococcal disease by vaccination is complex. The regional nature of serotype distribution and concerns over serotype replacement are driving the development of new, higher-valency conjugate vaccines to address the shortcomings of presently used vaccines .
The interplay between serotype distributions, disease incidence, and vaccine implementation is driving an intense interest in understanding the patterns of pneumococcal epidemiology. In this context, an important unanswered question is which serotypes are most likely to cause disease. Typical study designs that address this question may be confounded by differences in the duration of carriage between strains and by our lack of knowledge of the timing of invasion relative to the onset of carriage. Although it is thought that prior influenza virus infection enhances the incidence and severity of bacterial disease (as reviewed in  and ), the role of respiratory viruses, such as influenza virus, has not been assessed in recent studies of pneumococcal epidemiology. We have hypothesized that prior or concomitant infection with influenza viruses may favor particular serotypes or clonal types, or that it may alter the invasive disease potential in a strain-specific manner .
A second major unanswered question in pneumococcal epidemiology is whether there are pneumococcal serotype- and/ or strain-specific differences in transmission and whether these are modified by viral infections. Is the prevalence of particular serotypes determined more by the ability to persist in the nasopharynx or by an advantage in transmission? Does influenza alter this dynamic, enhancing transmission and favoring certain serotypes and/or strains? Limited data from small longitudinal studies conducted in the early 20th century suggest that respiratory viruses enhance transmission of S. pneumoniae [5, 6]. Definitively answering these questions as they apply to humans would require large longitudinal studies with frequent sampling of the cohorts. Because such studies would be expensive and logistically difficult, we sought to develop an animal model that was reflective of both human disease and transmission to address relevant hypotheses. In this report, we developed a novel model of transmission of S. pneumoniae in ferrets. We then evaluated the ability of different serotypes of S. pneumoniae to colonize, cause secondary bacterial infections in mice and ferrets, and transmit between ferrets.
Infectious agents. The pneumococcal isolates were chosen to belong to different serotypes and/or clonal types and have similar invasive disease potential in mice, as described elsewhere , with the exception of TIGR4, which is a highly virulent type 4 clinical strain that is commonly used in murine models . BHN78 (type 14, ST124), BHN54 (7F, ST191), and BHN60 (9V, ST838) are invasive isolates, and BHN97 (19F, ST425) is a carrier isolate. TIGR4 (type 4) is a human clinical strain that is commonly used in mouse models of pneumococcal disease . BHN54 and BHN97 were engineered to express luciferase, as described elsewhere . The St Jude strain of mouse-adapted influenza virus A/Puerto Rico/8/34 (H1N1; PR8), generated by reverse genetics , and influenza virus A/Sydney/5/97 (H3N2) were grown in Madin-Darby canine kidney cells.
Animal models. Eight-week-old BALB/cbJ mice (Jackson Laboratories) were used in a dual-infection model, as described elsewhere . Young adult (age, 7–8 weeks) outbred ferrets that were serologically negative for influenza were bred in the Animal Resources Center at St Jude Children's Research Hospital and used in a dual-infection model, as described elsewhere . Bacterial and viral titers were determined from lung homogenates or nasal washes, as described elsewhere [7, 12]. All experiments were conducted in biosafety level 2 facilities, in a manner in accordance with the guidelines of the committee on care and use of laboratory animals.
Bioluminescent imaging. Anesthetized mice or ferrets were imaged for 1 min (mice) or 2 min (ferrets) by use of an IVIS CCD camera (Xenogen). Total photon emission from selected and defined areas within the images of each animal was quantified using the Living Image software package (version 2.20; Xenogen), as described elsewhere [7, 9, 12].
Transmission model. Ferrets were infected with 1×105 TCID50 influenza virus or 1×107 cfu of S. pneumoniae given intranasally in a volume of 400 µL of sterile phosphate-buffered saline while anesthetized with isoflurane. Pairs of donor ferrets were housed alone in open caging in closed cubicles either for 3 days (the pneumo only and contact flu groups) or for 6 days if donor ferrets sequentially received influenza virus then pneumococcus (the donor flu and both flu groups) before introduction of contact ferrets. Pairs of contact ferrets that either were naive to any infectious agents (the pneumo only and donor flu groups) or had been infected with influenza virus 3 days previously (the contact flu and both flu groups) were then placed in the same cage as the donor ferrets, in a separate cage in the same cubicle 3 m apart, or in a separate cage in a facing cubicle 10 m apart (with intervening doors left open) overnight (for 14–16 h). At this point, the donor ferrets were removed to a separate cubicle, all doors were closed, and no further contact was allowed between the groups of ferrets. There is no pressure differential between cubicles to create airflow from one group of ferrets to another, but air is pulled across the face of both cubicles at a high rate (8–10 exchanges per hour). The environment is controlled at 24°C and 30% humidity with a light:dark cycle of 12 h:12 h. Care was taken to ensure that inadvertent transmission by animal care attendants or through common source exposures was not possible.
Microarray analysis. Comparative genomic hybridizations were performed using a reference design, as described elsewhere [13, 14], comparing the presence or absence of specific genes linked to invasion as well as 41 accessory regions (ARs) (defined elsewhere ) between the isolates with 2 reference strains (TIGR4 and R6). Analyses were performed with GenePix Pro (verson 6.0; MDS Analytical Technologies) and the R Project for Statistical Computing (version 2.4.0), by use of a Bayesian linear model, as described elsewhere . Genes were considered to be absent if they had a P value of <.01 with an M value (ie, a log2-fold change) ⩾−1, present if they had an M value ⩽−0.8, or not predictive if neither of these criteria applied.
Statistical analyses. Comparison of survival between groups of mice was done using the log rank Χ2 test on the Kaplan-Meier survival data. Comparison of bacterial and viral titers was done using analysis of variance. P<.05 was considered to denote statistical significance for these comparisons. SigmaStat software for Windows (version 3.11; SysStat Software) was used for all statistical analyses.
Pneumococcal strain-dependent differences in secondary bacterial infections in mice. The first hypothesis to be tested was that influenza virus infection differentially affects the expression of disease from different pneumococcal strains. Several clinical S. pneumoniae isolates of different clonal types of known invasive disease potential in humans  and mice  were administered to influenza-infected mice by use of a dose that was nonlethal considering the bacteria alone. Considerable differences in mortality were observed that were pneumococcal strain dependent (Figure 1A). The ability to kill mice correlated with bacterial lung load (Figure 1B), and only the highly virulent TIGR4 caused bacteremia.
To assess nasal colonization and site-specific disease expression by use of bioluminescent imaging, we chose 2 pneumococcal strains that caused an intermediate percentage of mortality in association with influenza, BHN54 (ST191−7F) and BHN97 (ST425−19F), and engineered them to express luciferase. BHN97 colonized 100% of the mice assessed (Figure 1C) and persisted for a median of 33 days in animals infected only with S. pneumoniae, compared with 56 days in animals preinfected with influenza. By contrast, BHN54 could only be found transiently in animals infected with influenza virus (median duration of persistence of the pneumococcal strain, 1 day). Both strains could cause otitis media, although this was more common with BHN54 after influenza. Neither strain caused pneumonia with any frequency in the absence of influenza virus, but 100% of mice in both groups that were preinfected with influenza virus developed pneumonia (Figure 1C). These data indicate that (1) there are differences in the support that influenza virus is able to provide to different strains of pneumococcus, and (2) in mice, the predominant effect of preinfection with influenza virus is to increase duration of carriage and enhance bacterial pneumonia in a strain-dependent manner.
Pneumococcal strain-dependent enhancement of secondary bacterial disease in ferrets after influenza. We previously showed that influenza virus can increase pneumococcal nasal titers and increase the incidence of secondary bacterial sinusitis and otitis media in young adult ferrets . To determine whether these effects were pneumococcal strain specific, influenza- infected and influenza-naive ferrets were challenged with pneumococcal strains BHN97 or BHN54 and were assessed daily by nasal wash and bioluminescent imaging. For the first 3 days after infection, BHN97 was recovered from nasal wash at titers significantly higher than those of BHN54 (Figure 2E), a disparity that was no longer evident after 72 h. In the absence of influenza virus, 1 (20%) of 5 ferrets in each group developed secondary bacterial infections that were detectable by bioluminescent imaging. Four of 5 ferrets infected with BHN97 after influenza virus had secondary infections, including otitis media (Figure 2A); sinusitis (Figure 2B); disseminated disease, including meningitis (Figure 2C and 2D); and meningitis with bacteremia (3 of the 4 ferrets had positive blood culture results). By comparison, influenza virus did not enhance the incidence of infections with BHN54, because only 1 (20%) of 5 of ferrets had a secondary bacterial infection (sinusitis with negative blood culture results). All ferrets infected with influenza virus were lethargic and were noted to sneeze. Ferrets infected with pneumococcus alone did not sneeze, and they did not show overt clinical signs unless they developed meningitis, at which time they became obtunded. On the basis of these data for mice and ferrets, we conclude that prior influenza virus infection enhances the incidence and severity of pneumococcal disease in a strain-dependent manner.
Influenza-mediated enhancement of pneumococcal transmission. The second major question we wished to address was whether influenza virus altered transmission of S. pneumoniae. Transmission was studied using pairs of infected and uninfected ferrets, as described in Methods. All directly inoculated donor ferrets were infected with S. pneumoniae, and pneumococcus was able to transmit from infected ferrets to contact ferrets in the same cage or up to a distance of 1 m (Table 1). To our knowledge, this is the first report of a small animal model of natural transmission of S. pneumoniae. In the donors, infection with influenza virus increased the incidence of pneumococcal acquisition in contact ferrets from 25% to 75% in same-cage contacts and from 50% to 83% in ferrets 1 m away, but it did not affect the absence of transmission to ferrets 3.5 m away. In contact ferrets, infection with influenza virus had a more robust effect, because 100% of contacts, including those 3.5 meters away, were infected. As summarized in Figure 3A, primary infection occurred in every instance when pneumococci were introduced directly into the nose of anesthetized ferrets. Transmission to close contacts, either in the same cage or in a close, adjacent cage, occurred 50%–70% of the time if the contacts had not been previously infected with influenza virus. Influenza rendered contact ferrets extremely susceptible to acquisition of pneumococcus, because 100% of ferrets, even those 3.5 m away, were infected.
Because of space constraints, the experimental design used in this study involved ferrets housed together in pairs for all experiments. Thus, it was theoretically possible for one contact ferret of a pair to acquire pneumococcus and then transmit it to the other ferret in the pair, which could alter the interpretation of these outcomes. Of the 21 ferrets that were infected at a distance, 17 had positive results at the initial assessment within 24 h of exposure, whereas 4 had positive results on the second day (Table 2), with a mean time to positive results of 1.2 days after exposure. Ferrets exposed by direct contact had similar or slightly longer times to positive results, with 9 of 16 having positive results on the first day, 5 having positive results on the second day, and 2 having positive results on the third day, with a mean time to positivity of 1.6 days after exposure. When pairs of ferrets were examined, discordant results where one ferret shed and the other did not occurred only once. From these data, we conclude that, in some cases, secondary transmission within pairs of contact ferrets likely occurred in the same-cage contact ferrets. Although we do not believe this alters the conclusions reached on the effect of influenza on transmission, it does mean that the true efficiency of transmission in the model is likely overestimated by looking at the raw percentages, and the length of exposure to infected animals (16 h of exposure to donor ferrets in the present study) required to transmit cannot be estimated.
An examination of bacterial titers from primary (direct inoculation) compared with secondary (via transmission) infections showed no differences in nasal washes between virusinfected ferrets (Figure 4A). However, secondary acquisition of pneumococcus in virus-naive animals resulted in lower titers and more-rapid clearance of bacteria, compared with direct inoculation of a high titer of bacteria. Viral titers were different 24 h after primary inoculation, compared with secondary acquisition, but they could not be distinguished beginning 2 days after infection or exposure (Figure 4B). Secondary bacterial infections were detected by bioluminescent imaging more commonly in virus-infected ferrets than in virus-naive animals after primary inoculation, but the incidence of secondary disease was lower after natural acquisition than with direct inoculation (Figure 3B and Table 1). Donor ferrets that were naive to influenza virus but were exposed to virus-infected contacts all developed influenza virus infections. Viral titers in these ferrets, which contracted pneumococcal infection before influenza, did not differ from those in animals primarily infected with influenza (data not shown), and none of these ferrets developed otitis media, sinusitis, pneumonia, or invasive disease.
Differences in patterns of accessory regions and gene composition between BHN97 and BHN54. Because BHN97 and BHN54 differed substantially in their ability to colonize both mice and ferrets and transmit between ferrets, we sought to determine genetic differences between the strains by use of a whole-genome microarray approach . Examination of the accessory regions present in the genomes of these 2 strains (Figure 5A) revealed several instances where an accessory region was present in BHN97 and absent in BHN54, or vice versa. Accessory regions 4, 10, 40, 41 (present in BHN97), and 29 (present in BHN54) have been associated with clonal complexes and serotypes that have a high potential for invasive disease . Of note, accessory region 6, which is found in nearly all isolates from serotypes with the highest invasive disease potential, is present in both strains, whereas accessory region 11, which encodes invasive pili that enhance adherence and colonization of the 19F isolate of ST162 , was absent in both. The differences in accessory region loci should provide target genes for understanding the colonization and transmission phenotypes of these strains.
Analysis of the isolates for the presence or absence of genes associated with invasion and virulence in signature tagged mutagenesis (STM) studies [13, 17–19] delineated a number of genes that could be associated with the enhanced virulence and invasive capacity of TIGR4 (Figure 1), compared with the other isolates studied (Figure 5B). Of interest, BHN97, which was more virulent in association with influenza virus before infection than was BHN54, has considerably fewer of these genes than the other isolates. Only 2 genes from this analysis, SP_0396 (mtlF, a putative mannitol-specific enzyme involved in carbohydrate transport) and SP_1045 (a hypothetical protein), are clearly present in BHN97 and absent in BHN54 and thus represent candidates to explain the enhanced relative virulence of this strain in mice infected with influenza virus. However, because the published STM studies used invasiveness in mice that are not infected with influenza to identify these genes, it is likely that there are other potential genes that are important in the context of prior influenza infection but do not fall out in screens in the absence of influenza. Thus, further work to identify virulence genes in S. pneumoniae that are context specific (ie, important in the postinfluenzal host) is required.
The influence of respiratory viruses on pneumococcal transmission has been studied in humans in a cursory fashion only. In a longitudinal study of families, Gwaltney et al  demonstrated that in 56% of cases where a defined episode of transmission could be documented, the donor had symptoms of an upper respiratory tract infection. They suggested that increased pneumococcal titers in the nasopharynx, as mediated by viral coinfection; modification by viruses of the site of pneumococcal colonization from the nasopharynx to the anterior nares; or increased production and dissemination of respiratory virus secretions due to intercurrent viral illness was responsible for this phenomenon. A study of prevalence in adults showed that carriage could be detected in adults twice as often when they had an upper respiratory tract infection, implying that acquisition of pneumococci was favored during viral coinfections . In this report using the ferret model, both transmission from the donor and acquisition by contacts were enhanced by prior influenza virus infection. The effect on the contacts seemed stronger, because only when the contacts were virally infected was transmission possible over 3.5 m, and enhanced nasal titers as a sole mechanism seems unlikely because influenza virus enhanced nasal titers of BHN54 to levels similar to that of BHN97, yet transmission of BHN54 could not be demonstrated. Similar effects by other respiratory viruses should be assessed.
We did not assess the theory espoused by Gwaltney et al  that influenza facilitates a transition from the posterior nasopharynx to the anterior nares; simultaneous culturing of both sites in this model would be revealing. We also did not assess the mode of transmission. It is likely that aerosol transmission was required for acquisition by contact ferrets placed 3.5 m away, because large droplets are unlikely to be capable of crossing this space, particularly in a room with such a high rate of air exchange. Sneezing by ferrets infected with influenza virus may have contributed either to formation of aerosols or dissemination across that distance. However, contacts infected with influenza virus were capable of acquiring pneumococcus from donors 3.5 m away that were not infected with influenza and were not sneezing. Changes in respiratory secretions and increased dissemination could be one of the mechanisms by which influenza virus infection enhances transmission, however, because influenza caused significant sneezing and increased the quantity and thickness of secretions. This is more likely to have contributed to direct contact within the same cage and to transmission across short spaces where large droplet transmission appears to be possible.
In this model, secondary bacterial disease manifests as increased mucosal disease (otitis media and sinusitis), as well as increased pneumonia and invasive disease. Of interest, this finding was pneumococcal strain specific. We have confirmed through whole-genome microarray analysis that several accessory regions differ between the strains, with 4 of the 5 regions previously implicated in virulence [13, 16] present in BHN97, which was superior in colonization and transmission, compared with BHN54. These data should provide targets to explore the mechanisms underlying these differences in pathogenesis. Analysis of specific genes identified in STM studies [17–19] also revealed differences in the strains studied that could be related to relative disparities in virulence. However, these analyses are limited because there are likely virulence factors that are specifically important in the context of postinfluenza pneumonia that cannot be identified through screens in the absence of influenza.
A final, interesting question is that suggested by studies from Gray et al [21, 22]: does pneumococcal disease typically occur shortly after acquisition of a new strain, before development of type-specific immunity, or can disease in colonized individuals occur at any time, independent of duration of carriage? Henderson et al  argued that timing of acquisition did not matter for subsequent development of otitis media, but they excluded all cases for which an intercurrent viral infection was suspected. Syrjanen et al  examined this issue in more detail and found that acute otitis media was most common shortly after acquisition of a new serotype, particularly if a viral infection was present at the time of presentation. If respiratory viruses, such as influenza virus, enhance acquisition of new strains and also increase the chances of disease, then understanding the interaction between viruses and S. pneumoniae, particularly understanding serotype-specific differences in the ability to benefit by viral coinfection, may be critical information for prevention of disease. If the competing concept is true, that invasion and disease can occur at any time after acquisition, perhaps aided by a superimposed viral infection, then length of carriage is a more important concept. In these studies, it is interesting to note that otitis media does occur in mice colonized with pneumococcus that are subsequently challenged with influenza virus , but pneumonia or invasive disease are not seen in either mice or ferrets in this scenario. These data favor the hypothesis of Gray and colleagues and argue that point prevalence studies of carriage, although easier to perform, are not as useful as longitudinal studies that can determine timing of acquisition.