Impact of Pneumococcal and Viral Pneumonia on the Respiratory and Intestinal Tract Microbiomes of Mice

ABSTRACT

With 2.56 million deaths worldwide annually, pneumonia is one of the leading causes of death. The most frequent causative pathogens are Streptococcus pneumoniae and influenza A virus. Lately, the interaction between the pathogens, the host, and its microbiome have gained more attention. The microbiome is known to promote the immune response toward pathogens; however, our knowledge on how infections affect the microbiome is still scarce. Here, the impact of colonization and infection with S. pneumoniae and influenza A virus on the structure and function of the respiratory and gastrointestinal microbiomes of mice was investigated. Using a meta-omics approach, we identified specific differences between the bacterial and viral infection. Pneumococcal colonization had minor effects on the taxonomic composition of the respiratory microbiome, while acute infections caused decreased microbial complexity. In contrast, richness was unaffected following H1N1 infection. Within the gastrointestinal microbiome, we found exclusive changes in structure and function, depending on the pathogen. While pneumococcal colonization had no effects on taxonomic composition of the gastrointestinal microbiome, increased abundance of Akkermansiaceae and Spirochaetaceae as well as decreased amounts of Clostridiaceae were exclusively found during invasive S. pneumoniae infection. The presence of Staphylococcaceae was specific for viral pneumonia. Investigation of the intestinal microbiomés functional composition revealed reduced expression of flagellin and rubrerythrin and increased levels of ATPase during pneumococcal infection, while increased amounts of acetyl coenzyme A (acetyl-CoA) acetyltransferase and enoyl-CoA transferase were unique after H1N1 infection. In conclusion, identification of specific taxonomic and functional profiles of the respiratory and gastrointestinal microbiome allowed the discrimination between bacterial and viral pneumonia.

IMPORTANCE Pneumonia is one of the leading causes of death worldwide. Here, we compared the impact of bacterial- and viral-induced pneumonia on the respiratory and gastrointestinal microbiome. Using a meta-omics approach, we identified specific profiles that allow discrimination between bacterial and viral causative.

INTRODUCTION

In 2019, pneumonia was ranked as the 2nd leading cause of death in low-income countries and the 4th worldwide (1). Pneumonia can be distinguished as community-acquired pneumonia (CAP), meaning the infection has been acquired outside the hospital setting, and hospital-acquired pneumonia (HAP), which is defined as pneumonia occurring after at least 48 h of hospitalization (2). With nearly 35% of cases in Europe and >25% worldwide, Streptococcus pneumoniae is the predominant cause of CAP, followed by influenza A virus (IAV) (3–9). The risk of acquiring pneumonia is increased in the winter months or by host factors, such as age, comorbidities, and tobacco and/or alcohol consumption (8, 10–13).

As part of the commensal flora of the upper respiratory tract (URT), S. pneumoniae can spread to the lower respiratory tract (12, 14). Colonizing and invasive strains have acquired mechanisms to evade the key innate immune responses, which allows them to exploit and proliferate in their ecological niches (14). S. pneumoniae is able to kill closely related members of the microbiome by production of bacteriocins (15, 16) or to outcompete other microbial families of the URT microbiome by the production of hydrogen peroxide (H₂O₂) with the pyruvate oxidase SpxB as a key enzyme (17). Due to these abilities, pneumococci can disrupt the balance of the healthy microbiome, overcome/interrupt the natural colonization barrier, and colonize the URT, a prerequisite to develop CAP (18). This is, for instance, mirrored by an inverse correlation between carriers of pneumococci and staphylococci or a positive correlation of pneumococci with Haemophilus influenzae and Moraxella catarrhalis carriage (19–26). Besides S. pneumoniae, viruses (e.g., IAV) are frequent causes of respiratory disease (27). In contrast to S. pneumoniae, IAVs are obligate pathogens. Under healthy conditions, the URT microbiome has a homeostatic status, and all members are present in optimized proportions (28). When IAVs overcome the natural colonization barrier represented by the microbiome and successfully invade and proliferate in host cells, this balance is disturbed (29–31). The combination of damaged epithelial cells, representing new attachment sites for bacteria (32–35), and alterations in the structure of the microbiome favor outgrowth of opportunistic pathogens, such as Staphylococcus aureus or S. pneumoniae, which ultimately promote secondary bacterial infections (31).

The respiratory and gastrointestinal microbiomes play an important role in disease progression, as it is known that the host’s fitness and immune response can be notably affected by pathogens, medication, or nutrition (36). Furthermore, the microbiome can prevent the initial colonization of pathogenic bacteria (37). The URT microbiome is characterized by high interindividual variability in the community and is influenced by seasonality (20). In the URT microbiome of both adult humans and mice, the phyla (with families shown in parentheses) Firmicutes (Staphylococcaceae, Ruminococcaceae, and Streptococcaceae), Bacteroidetes (Prevotellaceae), Proteobacteria (Pasteurellaceae), and Actinobacteria (Corynebacteriaceae, Propionibacteriaceae, and Bifidobacteriaceae) are predominant (38–40). It is accepted that pathogenesis of disease is often caused by a prior disruption of the nasopharyngeal microbiome, which subsequently causes dysbiosis of the lung microbiome (41). In contrast to the respiratory microbiome, the gastrointestinal tract (GIT) microbiome has been largely explored and is the best-investigated host-associated microbial ecosystem (41). The gastrointestinal tract of the host provides a stable nutrient-rich microenvironment for the microbes, which in turn exert functions like fermentation of dietary fibers and production of metabolites (like short-chain fatty acids [SCFAs]), which are necessary for homeostasis of the immune system (42). At the phylum level, the taxonomic composition of the murine gastrointestinal microbiome resembles the human gastrointestinal microbiome, as both are dominated by Firmicutes and Bacteroidetes, and around 80 microbial gut genera are shared between mice and humans, even though in different abundances (43). GIT microbes affect the host’s immune response locally in the gut compartment (e.g., barrier functions) and distantly at mucosal interfaces, including the lung (41, 44). Recent studies have shown that the healthy GIT microbiome is crucial for the host’s defense against respiratory infections with influenza A virus and S. pneumoniae by modulating functions of effector immune cells: i.e., alveolar macrophages and neutrophils (29, 45–51). In this context, the gut microbiota was shown to distantly influence pulmonary immunity. This observation is commonly referred to as the gut-lung axis (41, 52). The gut-lung axis was proven to be involved in distant immune modulation during respiratory disease. Therefore, the composition and homeostasis of the GIT microbiome are crucial for maintaining a healthy status of the host.

In the last few decades, murine models have contributed immensely to our improved understanding of disease, as they are by far the most frequently used biomedical models for pneumonia (53). Interactions between the pathogen, the host, and its microbiome are increasingly coming into scientific focus. For instance, monocausal influenza or bacterial superinfection has been proven to contribute to dysbiosis of the intestinal microbiome through altered short-chain fatty acid production (54). However, most recent studies investigated progression of pneumonia or host-microbe interplay or focused on changes in the microbial composition, while the impact on the functional composition of the microbial community itself is far less explored (29, 37–39, 46, 54). Direct comparison of the impact of viral pneumonia and bacterial pneumonia on the respiratory and gastrointestinal microbiomes and their effects on its structure and function is still scarce. Thus, the present study used a meta-omics approach to identify common and pathogen-specific disease-related changes in composition and functionality of the respiratory and gastrointestinal murine microbiome during S. pneumoniae and IAV infections.

RESULTS AND DISCUSSION

To assess the impact of pneumococcal colonization or pneumococcal/viral pneumonia on the URT and GIT microbiome composition and functionality, an integrated meta-omics approach combining 16S rRNA gene sequencing and metaproteomics was employed. Two infection trials were performed independently from each other. The pneumococcal pneumonia trial harbored three groups (phosphate-buffered saline [PBS]-treated mice, n = 10; strain 19F-colonized mice [19F_C], n = 10; strain 19F-infected mice with pneumonia [19F_P], n = 12), while the setup of viral pneumonia experiments consisted of two groups (PBS, n = 6; H1N1, n = 12) (see Materials and Methods and Fig. S1 in the supplemental material). Samples for analysis of the URT and GIT microbiome were taken directly after euthanization. Of note, characterization of the respiratory microbiome using metaproteomics was limited and hindered by the low biomass obtained from murine nasal lavage specimens (NALs). In addition, the dominance of high-abundance host proteins and concurrently rare bacterial proteins as well as high salt concentrations hampered a robust meta-proteome analysis of the URT microbiome. Similar challenges have been described in studies of the microbiome from sputum and bronchoalveolar lavage samples in cystic fibrosis or asthma (55, 56). Thus, analysis of the URT microbiome was performed exclusively by 16S rRNA gene sequencing.

Alterations in the URT microbiome during colonization and infection with S. pneumoniae.

Intranasal application with a low infection dose (1 × 10⁷ CFU) of pneumococcal strain 19F (19F_C) resulted in an asymptomatic colonization of mice for at least 16 consecutive days. Classification of this infection dose as low was previously verified by Cuypers and colleagues, who established this mouse model to investigate the innate immune response during bacterial and viral coinfection (57). Initial colonization of mice was characterized by a drop in weight on day 1 post-bacterial inoculation (Fig. 1A). The weight normalized in colonized mice over the next days. Seven days after initial colonization, animals were euthanized and cecum content as well as nasal washes were sampled. Carriage of S. pneumoniae was confirmed by plating NALs on blood agar (Fig. 1B). Analysis of the URT microbiome using 16S rRNA gene sequencing revealed an average of 317 (±39) amplicon sequence variants (ASVs) within the URT of healthy mice. Similarly, 322 (±45) ASVs were detected in the URT microbiome of 19F_C mice (see Table S2 in the supplemental material), suggesting that colonization did not affect the number of ASVs, which would be an indication for dysbiosis. Comparison of taxonomic profiles of healthy and S. pneumoniae-colonized mice revealed that the composition of the URT microbiome was nearly unaffected by pneumococcal colonization (Pearson correlation = 0.99) (Fig. 2). The families Vibrionaceae, Streptococcaceae, Muribaculaceae, Lachnospiraceae, Prevotellaceae, Ruminococcaceae, and Lactobacillaceae were identified as predominant in the URT microbiome of healthy and 19F_C mice, and only Prevotellaceae were significantly increased during pneumococcal colonization. The finding of minor effects of pneumococcal colonization on taxonomic composition was reflected by the principal-component analysis (PCA) plot, showing no separation between the clusters of the healthy and S. pneumoniae-colonized mice (Fig. 3A), and is in accordance with other studies investigating the effects of pneumococcal colonization on the respiratory microbiome (58). Moreover, clinical scoring revealed that the 19F_C group recovered after bacterial colonization and was comparably to the healthy animals. Cuypers and colleagues defined colonization as stable bacterial load in the URT and absence of local and systemic inflammation. Using the same experimental setup, they demonstrated that pneumococcal colonization was locally and systemically asymptomatic at sampling day 7 (57). Asymptomatic carriage of low-abundance pneumococci as part of the commensal flora of the URT microbiome occurs frequently, but development of pneumonia is rare (14). Therefore, our results of a nearly unaffected URT microbiome during bacterial colonization are in line with previous studies.

FIG 1.

Bacterial and viral infections of C57BL/6J mice. (A) One group of female C57BL/6J mice was intranasally colonized with 1 × 10⁷ CFU S. pneumoniae 19F (19F_C), a second group was infected with 1 × 10⁸ CFU S. pneumoniae 19F (19F_P), and a third group was challenged with PBS as a control. Weight was monitored daily over 7 consecutive days. (B) CFU counts in nasal wash specimens (NAL) at the endpoint of the analyses. (C) C57BL/6J mice were intranasally inoculated with 1 × 10⁵ PFU of H1N1 or challenged with PBS, and weight was monitored daily over 5 consecutive days. (D) To exclude potential bacterial contamination, nasal wash specimens were plated on blood agar. Mean values ± standard deviations (SD) are displayed in panels A and C. Each dot represents one mouse, and horizontal lines display mean values (B and D). Two independent experiments with at least five mice per group were performed: total, for panels A and B, PBS, n = 10, 19F_C, n = 10, and 19F_P, n = 12; total for panels C and D, PBS, n = 6, and H1N1, n = 12. The level of significance between the groups was determined using Kruskal-Wallis test with Dunn’s multiple-comparison posttest.

FIG 2.

(A) Impact on the taxonomic composition of the URT microbiome of C57BL/6J mice during S. pneumoniae 19F colonization and infection (PBS, n = 10; 19F_C, n = 10; 19F_P, n = 12). (B) Alterations in composition of the respiratory microbiome after IAV infection (PBS, n = 4; H1N1, n = 12). For better illustration, only high-abundance families (relative abundance > 0.01) with significant changes in relative abundance are shown (ANOVA; P = 0.05).

FIG 3.

Effects on the taxonomic composition of the URT microbiome of S. pneumoniae 19F colonization (19F_C) or infection (19F_P) (A) and influence of viral pneumonia (B), illustrated in a PCA plot. Shown are results for PBS-treated controls (gray crosses and gray squares), S. pneumoniae-colonized (blue triangles) and -infected (orange circles) mice, and H1N1-infected mice (light orange diamonds).

In contrast to colonization, intranasal infection with a high dose (1 × 10⁸ CFU) of the same strain induced severe pneumonia within 2 days (19F_P) (Fig. 1A), characterized by additional weight loss on day 2 postapplication. Declaration of the infection dose as high was also made by Cuypers and colleagues (57). Animals had to be euthanized due to critical clinical scoring. Analyses of bacterial burden in nasal wash specimens (NALs) showed that all mice harbored pneumococci in the nasopharynx. Higher bacterial counts were detected in NALs of the pneumococcal pneumonia group compared to the group of colonized mice (Fig. 1B). Contrary to the 19F_C group, 16S rRNA gene sequencing revealed that pneumococcal pneumonia reduced the identified ASVs to 27 (±6), while the healthy group contained up to 317 (±39) ASVs (Table S2). A similar observation was made for microbial richness according to Shannon’s H index. Levels of microbial richness were comparable between the PBS control group and the 19F_C mice, while significantly decreased microbial richness was observed in the 19F_P group (Fig. S2A). As expected, Streptococcaceae were significantly elevated in the 19F_P cohort, displacing most of the other bacterial families. This observation is in good accordance with the decreased ASVs, microbial richness, and reduced Pearson correlation of 0.18 between the healthy and 19F_P group. Taxonomic profiling showed that Enterobacteriaceae was the only family (above 1% relative abundance) that increased in relative abundance in the URT microbiome during pneumococcal pneumonia, while all other predominant families were decreased (Fig. 2). Since a healthy and diverse URT microbiome has been shown to suppress virulence of pathogens (59), decreased α diversity is considered an important step reflecting dysbiosis along the respiratory microbiome (60–62). Therefore, the observed dysbiosis in the URT microbiome can be linked to the overgrowth of S. pneumoniae in the ecological niche, which is a key to pneumococcal disease (63). To further characterize the impact of pneumococcal pneumonia on the URT microbiome, principal-component analysis (PCA) was performed. While the clusters of the PBS control and 19F_C group were nearly identical, the 19F_P group clusters at a distance from the other groups (Fig. 3A). This finding supports the hypothesis of the dysbiosis within the URT microbiome during pneumococcal pneumonia. Of note, scattering within the individual groups was observed. This can be explained by the fact that the respiratory microbiome itself is known to be highly variable (20). In accordance, scattering was reduced in the 19F_P group, which might be explained by the reduced α diversity.

Effects of H1N1 infection on the URT microbiome.

Similar to pneumococcal infection, viral application of 1 × 10⁵ PFU of non-mouse-adapted H1N1 resulted in a continues disease progression (57). In accordance with the previously published results, H1N1 infection of mice caused continues weight loss, reaching critical clinical scoring 5 days after infection (Fig. 1C). Animals were euthanized, and samples from the respiratory and gastrointestinal tract were taken. To exclude potential bacterial contamination of the nasopharynx, NALs from these animals were plated on blood agar. All animals were negative for pneumococci (Fig. 1D). With an average of 165 (±65) ASVs, IAV-infected mice had slightly increased counts of ASVs compared to the PBS control group (120 ± 85 ASVs) (Table S2). Nevertheless, this increase was not significant. Unlike pneumococcal pneumonia, the viral infection did not reduce microbial richness, reflected by Shannon’s H index (Fig. S2B), and caused only minor changes in the taxonomic profile of the URT (Pearson correlation = 0.97). Analysis of the composition of the URT microbiome identified Pseudomonadaceae, Enterobacteriaceae, Lachnospiraceae, Lactobacillaceae, Staphylococcaceae, Propionibacteraceae, and Muribaculaceae as the predominant families. Only Propionibacteriaceae showed significantly increased abundance in response to H1N1 infection. The PCA plot showed cluster formation, enabling us to distinguish healthy and H1N1-infected animals (Fig. 3B). Nevertheless, the distance between the groups was not as vast as that observed during pneumococcal pneumonia. In line with our findings, Handa and colleagues described that an infection with influenza caused only mild effects on the respiratory microbiome of humans, while significant changes in the GIT microbiome were detected (64). Thus, we also investigated the impact of pneumonia on the GIT microbiome by an integrated meta-omics approach.

Pneumonia alters taxonomic composition of the gastrointestinal tract microbiome.

The GIT microbiome was characterized by an integrated meta-omics analysis of cecal content. Results of the taxonomic profiling by independent analysis of the data sets from 16S rRNA gene sequencing and the metaproteomic approach revealed great similarity according to the Pearson correlation of 0.92 (Table S3 and Fig. S3). The taxonomic profile of the gastrointestinal microbiome is based on an average of ~4,650 (±1,280) protein groups (PGs), identified by metaproteomic analysis, that were assigned to 130 families (based on NCBI taxonomy). Taxonomic data of 16S rRNA gene sequencing are available in the supplemental material (Table S4). Most PGs were identified from samples of the 19F_P mice, while the numbers of detected PGs in other samples were significantly smaller (Table S5). The comparably small number of PGs in the PBS-treated mice of the H1N1 infection can be explained by the lower number of animals in this group (n = 4 versus n = 12). Shannon’s H index revealed that diversity of the microbial community was rather constant, but it was slightly elevated in the GIT microbiome of the 19F_P mice (Table S2 and Fig. S2B). We detected a decreased Firmicutes/Bacteriodetes ratio during pneumococcal pneumonia, while this ratio increased in animals with H1N1-induced pneumonia.

Influence of bacterial and viral infection on the GIT microbiome.

The gut microbiome has a protective role during pneumococcal pneumonia, evidenced by increased bacterial dissemination, inflammation, organ failure, and mortality in microbiota-depleted mice (48). Besides its role in local defense against infection, the gastrointestinal microbiome can also modulate immune responses systemically (50, 52, 65–68). In compliance, the GIT microbiome was shown to regulate/mediate immune defense against respiratory tract IAV infection (29, 69) and organ failure in S. pneumoniae-induced sepsis (48).

Similar to the effects on the respiratory microbiome, colonization of the URT with S. pneumoniae 19F had only minor nonsignificant effects on taxonomic structure of the GIT microbiome (Fig. 4). Using the same setup, Cuypers and colleagues demonstrated that colonization of the URT with S. pneumoniae did not have significant effects on the general local inflammatory response 7 days after intranasal inoculation (57). In line with these findings, the observation of no significant changes in taxonomic composition of the GIT microbiome could be linked to the observed minor alterations in the respiratory microbiome. To further characterize the influence of URT colonization by pneumococci on the microbiome, PCA was performed. Healthy and colonized groups clustered together (Fig. 5A). This supports the results of the taxonomic profiling and indicates that pneumococcal colonization of the URT had no effects on the GIT microbiome.

FIG 4.

Taxonomic profile of the murine gastrointestinal microbiome highlighting similar and countervailing trends, as well as specific changes due to S. pneumoniae 19F or H1N1 pneumonia, based on metaproteomic data. For better illustration, only higher-abundance families with significant changes are shown (ANOVA; P < 0.05). Color code: gray bars, PBS control groups; blue bars, 19F-colonized group; orange bars, 19F-infected group; and light orange, H1N1-infected group. For statistics, see Tables S6A and B.

FIG 5.

PCA plot illustrating the effects of pneumococcal colonization and mild infection (A), as well as the influence of mild H1N1 infection (B) on the taxonomic composition of the gastrointestinal microbiome. PBS-treated control mice (gray crosses and gray squares), S. pneumoniae-colonized (blue triangles) and -infected (orange circles) mice, and H1N1-infected mice (light orange diamonds).

In contrast to the pneumococcal colonization, infection with S. pneumoniae led to several significant changes (analysis of variance [ANOVA]; P = 0.05) in the composition of the intestinal microbiome (Table S5). For instance, Desulfovibrionaceae, Rikenellaceae, Odoribacteraceae, Prevotellaceae, Tannerellaceae, Akkermansiaceae, and Spirochaetaceae significantly increased in relative abundance in response to pneumococcal pneumonia. On the other hand, Eubacteriaceae and Clostridiaceae decreased significantly in relative abundance. Of note, significantly increased levels of Spirochaetaceae and Akkermansiaceae, as well as decreased amounts of Clostridiaceae, were exclusively detected in the 19F_P group (Fig. 4C). Reduced food intake is common during severe pneumonia and can be an explanation for rapid weight loss in the 19F_P group (Fig. 1A and C). The reduced nutrient/fiber intake can affect the GIT microbiome (54). For example, competition between gut commensals at the expense of SCFA producers and to the benefit of bacteria using host mucins as energy source have been described (70). In line with this, we found significantly increased abundance of the mucin-degrading family Akkermansiaceae and elevated numbers of Bacteroidaceae in the GIT microbiome of mice with pneumococcal pneumonia. As the ends of mucin polysaccharides are often equipped with sulfate groups, degradation of mucins leads to the release of these sulfate groups (71). These free sulfate groups can be used as an energy source by sulfate-reducing bacteria. The most dominant family of sulfate reducers in the murine colon are Desulfovibrionaceae (72, 73), and indeed, our data indicated that the abundance of Desulfvibrionaceae increased during pneumococcal pneumonia. We speculate that increased abundance of Desulfovibrionaceae might be linked to a higher level of sulfate caused by mucin degradation. The PCA plot enabled distinction between the PBS, 19F_C, and 19F_P groups. Compared to PCA of the respiratory microbiome, wide scattering within the groups was observed. α diversity and scattering decreased in the URT microbiome during pneumonia, while Shannon’s H index revealed stable microbial richness in the GIT microbiome under all conditions. Thus, scattering within the cohorts could be explained by the dynamic nature of the individual microbiomes.

Similar to the 19F_P mice, the GIT microbiome of H1N1-infected animals exhibited increased levels of Desulfvibrionaceae and Odoribacteraceae. Moreover, Prevotellaceae and Tannerellaceae were detected in significantly decreased abundance, while Eubacteriaceae significantly increased. Comparison of the compositions of the GIT microbiome during bacterial and viral infection revealed countervailing trends (Fig. 4B): e.g., alterations in abundance of the families Prevotellaceae, Tanerellaceae, and Eubacteriaceae. Most interestingly, our analysis identified Staphylococcaceae as the exclusive significantly increased family in the taxonomic profile of the H1N1-infected mice (Fig. 4C). PCA showed separated clusters of the control and H1N1-infected mice, which was similar to the bacterial counterpart (Fig. 5B).

As the composition of the microbiome or the presence of some bacterial species was linked to the host’s health status (74), the exclusively altered bacteria could possibly be used as indicator families for the discrimination between bacterial or viral pneumonia. Langevin and colleagues suggested that microbiome profiling could be used in clinical management, as mild and severe pneumonia caused specific signatures in the microbiome (75). Based on the taxonomic profiles observed in this study, we consider a discrimination between viral and bacterial pneumonia, which could lead to a more target-oriented treatment.

Pneumococcal and H1N1 infections of the URT shape the functional composition of the GIT microbiome.

As changes in taxonomy favor changes in functional composition, alterations in composition of the GIT microbiome can be expected during pneumonia. Recent studies demonstrated that carbohydrate, energy, and xenobiotic metabolism as well as signal transduction pathways (e.g., SCFA metabolism) of the murine GIT microbiome are affected by pneumococcal as well as viral pneumonia (48, 54, 76). Using the metaproteomic annotation tool PROPHANE 2.0, the average number of 4,650 (±1,280) PGs was assigned to a total number of 1,116 protein descriptions (according to the eggNOG database). Analysis of the functional composition of the GIT microbiome revealed common and specific changes in response to bacterial or viral infection.

Pneumococcal colonization and infection shape the functional composition of the GIT microbiome.

Our meta-omics analysis revealed that pneumococcal colonization did not affect the GIT microbiome in its taxonomic structure. Nevertheless, alterations in the functional composition were observed. Changes detected in the functional composition of the 19F_C group were similar to the ones we identified in the 19F_P cohort, albeit they were less pronounced in colonized mice (Fig. 6). We assume that distinguishing between colonized and infected mice based on changes in functional profile, as well as identifying specific markers, would be very challenging. Therefore, changes in the functional composition of the microbiome during pneumococcal colonization and infection are discussed together. We observed decreased amounts of protein groups assigned to the functional categories “energy production and conservation.” In particular, expression of proteins from the ATPase complex decreased (Fig. 6). ATPase consists of several subunits and functions as a motor that pumps ions out of the cell or produces ATP (77). It is also involved in the detoxification of metal-thiolate complexes, the most common form of metals in the intracellular space (78). Decreased expression of ATPase might thus lead to accumulation of toxic metal complexes within the members of the GIT microbiome. This relationship between decreased ATPase activity and concomitant increased amounts of secretion proteins was shown as a response of a bacterial community to the exposure of cadmium in a cultivation experiment using two-dimensional (2D) SDS-PAGE (79). Thus, we identified an increased expression of transport and secretion channels assigned to the functional categories “inorganic ion transport and metabolism” and “cell wall/membrane/envelope biogenesis” (e.g., OmpA, TolA, and TonB) correlating with reduced amounts of ATPases, during pneumococcal colonization and infection (Fig. 6). Increased presence of these PGs might contribute to the maintenance of cell homeostasis. Moreover, the expression of electron transfer flavoprotein and rubrerythrin was reduced in the GIT microbiome of mice suffering from pneumococcal pneumonia. Since electron transfer flavoproteins (ETFs) play a central role in transferring electrons from dehydrogenases to the membrane-bound respiratory-chain (80), reduction of ETFs is in accordance with the decreased expression of the ATPase. Our data showed that rubrerythrin, a protein protecting against oxygen-dependent killing by pathogenic anaerobic organisms (81), was mainly produced by members of the phylum Firmicutes. Congruent with the lower levels of rubrerythrin in 19F_P mice, we observed a specific reduction of Firmicutes in the taxonomic profile of the GIT microbiome during bacterial infection. Congruently, we also observed reduced expression of PGs assigned to the functional category “carbohydrate transport and metabolism” (Fig. 6). The PGs most affected were the enzymes aldolase and glyceraldehyde-3-phosphate dehydrogenase, which were specifically reduced during bacterial pneumonia. Both PGs are involved in ATP production during glycolysis. Thus, this observation is in line with the congruent reduction of PGs assigned to the functional category “energy production and conservation” (i.e., the ATPase complex). Moreover, the abundance of PGs assigned to “cell motility”—i.e., flagellin, the flagellar motor switch protein, and the flagellar proteins FlgE and FliS—was strongly (>2-fold change) reduced specifically in the 19F_P group (Fig. 6). Flagellins are dominant antigens found in members of the Firmicutes—for example, in Crohn´s disease (82). Their expression was proved to correlate with the expression of Toll-like receptor 5 (TLR5), which is known to regulate the bacterial load and species composition of Firmicutes and Bacteroidetes (83, 84). Congruent with this finding, the families Clostridiaceae, Eubacteriaceae, Ruminococcaceae, and Lachnospiraceae, were detected in lower abundance exclusively during S. pneumoniae infection (Fig. 4). A decreased number of Firmicutes during pneumococcal pneumonia might thus be caused by an immune response toward flagellins.

FIG 6.

Voronoi treemaps illustrating changes in the functional composition of the murine gastrointestinal microbiome during pneumococcal colonization and infection, as well as H1N1 infection, based on the eggNOG database. The Voronoi treemap on the top left provides an overview of the corresponding functional category of the eggNOG database, to which the individual PGs have been assigned. On the top right, the influence on the gastrointestinal microbiome by pneumococcal colonization of the URT is shown. The Voronoi treemap on the bottom right monitors effects of severe pneumonia on the functional composition of the GIT microbiome, induced by S. pneumoniae infection. The influence of the H1N1 infection is illustrated in the Voronoi treemap on the bottom left. The size of the fields represents the average abundance of the identified PGs under all conditions. For comparison, the log₂ fold change (FC) (challenged versus healthy) was used. The two-sided gradient orange fields represent an increase in abundance (log₂ FC value of 1.5), the dark blue fields a decrease in abundance (log₂ FC value of −1.5), and the gray fields no change in abundance. Higher saturated color indicates more pronounced alterations in the functional composition of the gastrointestinal microbiome.

PCA demonstrated that the 19F_C and 19F_P cohorts formed clusters, which separated from the PBS control. However, the cluster of the colonized mice was located closer to the PBS control than the cluster of the infected animals (Fig. 7). As changes in taxonomical composition favor changes in the functional structure, separation of the individual clusters was expected. The three clusters showed a distribution from the top right to the bottom left with a great distance between the control and 19F_P cohort. In an intermediate distance from both, the 19F_C cluster was detected. This observation is in line with the observation that colonization caused similar effects on the functional assignments to the 19F_P group, but they were expressed to a lesser extent (Fig. 6).

FIG 7.

PCA plot illustrating the effects of bacterial and viral pneumonia on the functional assignment of the gastrointestinal microbiome. Shown are results for PBS-treated control mice (gray crosses and gray squares), S. pneumoniae-colonized (blue triangles) and -infected (orange circles) mice, and H1N1-infected mice (light orange diamonds).

H1N1 infection alters functional capacity of the GIT microbiome.

In contrast to the bacterial infection, larger amounts of PGs assigned to “energy production and conversion” were identified during H1N1 infection. Especially, the ATPase complex was significantly (P < 0.05) increased in relative abundance. Further examples of PGs that were expressed in countervailing levels in this category were rubrerythrin, electron transfer flavoprotein, rubredoxin, phosphoenolpyruvate carboxykinase, and phosphate butyryltransferase, which were highly abundant during viral infection but decreased during bacterial pneumonia (Fig. 6). Vice versa to the pneumococcal colonization and infection, IAV pneumonia leads to reduced expression of TonB, TolA, and OmpA, PGs belonging to the categories “inorganic ion transport and metabolism” and “cell wall/membrane/envelope biogenesis.” Furthermore, PGs assigned to the categories “carbohydrate metabolism and transport” (i.e., transketolase) and “cell motility” (i.e., FlgE) showed increased expression specifically in H1N1-infected mice (Fig. 6).

In the category “lipid transport and metabolism,” PGs involved in short-chain fatty acid production (i.e., acetyl coenzyme A [acetyl-CoA] acetyltransferase, enoyl-CoA hydratase, and 2-hydroxy-3-oxopropionate reductase) were expressed in larger amounts during viral infection. Moreover, increased expression of lactaldehyde reductase, aldehyde dehydrogenase, phosphate butyryltransferase, acetate kinase, and methylmalonyl-CoA epimerase was observed during pneumococcal and H1N1 pneumonia. Nevertheless, this observation was more pronounced after viral infection (Fig. 6). It is known, that SCFAs produced by members of the GIT microbiome remotely influence the immune response in the lungs and can thereby modulate disease outcome of respiratory infections (46, 85–88). In parallel, increased expression of proteins assigned to amino acid metabolism was detected. Sencio and colleagues confirmed a reduced amount of SCFA in the GIT of H1N1-infected mice, attributed to decreased food intake during viral infection (54). We thus hypothesize that SCFA-producing members of the microbiome seek to synthesize SCFAs from energetically less beneficial sources, such as amino acids, to compensate the reduced production of SCFAs from dietary fibers. Interestingly, observed changes were more distinct during H1N1 pneumonia than in pneumococcal pneumonia.

Using PCA, a distinction between H1N1-infected mice and the corresponding PBS group was possible. Remarkably, the clusters spread over a larger area in the PCA plot (Fig. 7). This finding is similar to the observation of the S. pneumoniae-colonized mice, highlighting minor alterations in the taxonomic composition can cause significant changes in the functional composition of the GIT microbiome.

Conclusion.

In this study, we compared the amounts of influence of the most common causes of pneumonia, S. pneumoniae and IAV, on the URT and GIT microbiome of mice. Our data revealed that pneumococcal pneumonia resulted in significant changes in the taxonomic and functional profiles of the upper respiratory and gastrointestinal microbiome compared to pneumococcal colonization and IAV infection. Analysis of the taxonomic composition showed that diversity of the URT microbiome decreased during pneumococcal pneumonia but remained unaffected during S. pneumoniae colonization and H1N1 infection. Moreover, using 16S rRNA gene sequencing and metaproteomics for taxonomic profiling, specific taxonomic profiles of the URT and GIT microbiome, depending on the respective pathogen, have been identified. These pathogen-specific changes in the taxonomic profiles of the URT and GIT microbiome allow a discrimination between bacterial and viral pneumonia. In addition to the taxonomic profiling, this study provides new information on the functional response of the GIT microbiome to bacterial and viral pneumonia: i.e., expression changes of proteins involved in energy production, carbohydrate utilization, ion transport, cell motility, and lipid metabolism. More specifically, a reduced expression of ATPase and a concomitant increased expression of transport and secretion channels seemed to be exclusively caused by bacterial pneumonia, whereas an increased expression of enzymes for SCFA production was more frequently found during viral pneumonia. Further studies, addressing the composition of the gastrointestinal and respiratory microbiome as well as the host’s immune response, will provide a more comprehensive picture of the host-microbiome interaction during pneumonia.

MATERIALS AND METHODS

Ethics statement.

All animal experiments were carried out in accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals (89), the guidelines of the ethics committee at the University of Greifswald and the German regulations of the Society for Laboratory Animal Science (GVSOLAS), and the European Health Law of the Federation of Laboratory Animal Science Associations (FELASA). All experiments were approved by the Landesamt für Landwirtschaft, Lebensmittelsicherheit und Fischerei Mecklenburg-Vorpommern (LALLFV M-V, Rostock, Germany) and the LALLFV M-V ethical board (permit no. 7221.3-1.1-032/17). All efforts were made to minimize suffering, ensure the highest ethical standard and adhere to the 3R principle (reduction, refinement, and replacement).

Bacterial and viral strains.

Streptococcus pneumoniae 19F (EF3030), a nasopharynx isolate from a child with frequent episodes of otitis media (90, 91), was grown on blood agar plates (Oxoid) and cultivated to mid-log phase (optical density at 600 nm [OD₆₀₀] of 0.35 to 0.40) in Todd-Hewitt broth supplemented with 0.5% (wt/vol) yeast extract (Carl Roth) at 37°C and 5% CO₂. Influenza virus A/Bavaria/74/2009 (H1N1) was propagated as described by Eisfeld and colleagues (92).

Mouse infections, monitoring, and sampling.

To characterize the impact of bacterial and viral pneumonia on the respiratory and gastrointestinal microbiome, we used a murine model that had been previously established to investigate the innate immune response during bacterial and viral coinfections (57). In brief, 7-week-old specific-pathogen-free (SPF) female C57BL/6J mice were purchased from Janvier Labs. Mice were maintained within the biosafety level 2 (BSL2) animal facility for 1 week prior to the infection experiments. All diets were provided ad libitum. Mouse maintenance complete feed (S8189-S095 for the first 7 weeks and V1535-000 from week 8) (SSNIFF Spezialdiäten GmbH) was used in the animal facility. Randomly assigned groups of 5 to 6 mice (8 weeks old) were intranasally colonized or infected under ketamine/xylazine anesthesia with S. pneumoniae 19F or IAV. For colonization with S. pneumoniae 19F, 20 μL PBS containing 1 × 10⁷ CFU was administered. For pneumonia, 1 × 10⁸ CFU in 20 μL PBS were applied. For viral infections, 42 μL PBS containing 100,000 PFU was used. Control mice were mock treated with an equivalent volume of PBS (Fig. S1). Animals were observed daily for weight and clinical monitoring. At different time points, mice were euthanized and cecum content and nasal wash specimens (NALs) in 1 mL PBS were harvested. Individual cecum content samples were evenly aliquoted to an approximate weight of between 50 and 100 μg, depending in the available weight of the initial sample, and used for 16S rRNA gene sequencing or metaproteomics. Samples were stored until further analyses at −80°C. To quantify bacterial load, NALs were serially diluted and plated on blood agar plates (Oxoid). The provided samples are listed in Table S1 in the supplemental material.

16S rRNA gene sequencing and bioinformatic processing.

Nucleic acids of individual cecum content samples were extracted using a bead beating (MP Biomedicals lysing matrix E) phenol-chloroform protocol, followed by nucleic acid precipitation with 3 M Na-acetate and isopropanol. After being washed with 70% (vol/vol) ethanol, DNA pellets were resuspended in diethyl pyrocarbonate (DEPC)-treated Milli-Q water for downstream applications. Nucleic acids of nasal washes were extracted via the Qiagen PowerSoil kit (Qiagen, Hilden, Germany) according to the manufacturer´s protocol. For both sample types, bead beating was performed using the FastPrep-24 5G instrument (MP Biomedicals, Santa Ana, CA, USA) for 45 s at an intensity of 5.5 m/s. Quantification of the DNA content was performed using Qubit double-stranded DNA (dsDNA) broad-range assay kit (Invitrogen) in combination with the Qubit fluorometer 3. Subsequently, DNA extracts were diluted to a final concentration of 5 ng/μL.

For sequence preparation, amplicon and index PCR were performed using the V4 primer pair 515F (5′-GTG-YCA-GCM-GCC-GCG-GTA-A-3′)/806R (5′-GGA-CTA-CNV-GGG-TWT-CTA-AT-3′) with subsequent PCR cleanup between and after both amplifications with AMPure XP beads (93, 94). Quantification of the libraries was done via Invitrogen Qubit dsDNA broad-range assay kit, normalized to a final concentration of 5 pM, denatured with NaOH, and sequenced via Illumina MiSeq with approximate outputs from 30,000 to 130,000 reads per cecum sample and 7,000 to 55,000 reads per sample obtained from nasal washes.

Initial filtering steps as well as truncation and merging of the sequence reads were done using the dada2 pipeline-package v.1.11.1 (R v.3.6.1) in R (95) according to the descriptions in the article by Gierse et al. (96). Amplicon sequence variants (ASVs) were assigned via the SILVA 132 database after removing chimeras, and ASVs identified as chloroplasts and mitochondria were removed. The R packages vegan, ggplot, phyloseq, plyr, and reshape2 were used for advanced bioinformatic processing (e.g., α and β diversities, Bray-Curtis dissimilarities, and nonmetric multidimensional scaling).

Protein extraction, mass spectrometry, and data analysis.

To extract the total protein amount, a TRIzol-based protocol was applied to the cecal samples following the manufacturer’s instructions. Briefly, 100 mg cecal content was resuspended in 1 mL TRIzol and incubated for 5 min. Afterwards, 0.2 mL chloroform was added and centrifugation was performed (15 min at 12,000 × g and 4°C) for separation in the phenol-chloroform, inter, and aqueous phases. Finally, the phenol-chloroform phase was used for protein extraction. To quantify the amount of protein, the Pierce bicinchoninic acid (BCA) protein assay was performed. Afterwards, a 4 to 20% Criterion TGX precast gel (BioRAD, Hercules, CA, USA) was loaded with 30 μg of protein of each sample, run for 1 h at 150 V, and subsequently stained with colloidal Coomassie brilliant blue G-250. After washing of the gel, each lane was cut into 10 pieces, which were further processed to smaller blocks, destained, and tryptic digested. To desalt the peptide-containing solution, ZipTip purification (C₁₈ column) (Merck Millipore, Billerica, MA, USA) was performed according to the manufacturer’s protocol. The desalted peptide mixture was vacuum centrifuged until dry. All samples were analyzed by an Easy nLC 1000 coupled online to an Orbitrap Velos mass spectrometer (Thermo Scientific). An 8-μL peptide mixture was loaded into in-house self-packed columns (inside diameter [i.d.], 100 μm; outside diameter [o.d.], 360 μm; length, 200 mm), which were packed with 3.0-μm Dr. Maisch Reprosil C₁₈ reversed-phase material (ReproSil-Pur 120 C₁₈-AQ), which was first loaded and subsequently eluted with 18 μL of buffer A (0.1% [vol/vol] acetic acid) at a maximum pressure of 22,000,000 Pa. Peptide elution was performed in a nonlinear 100-min gradient of buffer B (0.1% [vol/vol] acetic acid in 99% [vol/vol] acetonitrile) at a constant flow rate of 300 nL/min. Eluted peptides were measured in the Orbitrap with a resolution of R = 30,000 with lock-mass correction activated. Following each MS full scan, up to 20 dependent scans were performed in the linear ion trap based on the precursors’ intensity after collision-induced dissociation (CID) fragmentation.

Mass spectrometry data analysis with Mascot Deamon 2.6.2 (Matrix Science, Ltd., London, United Kingdom), Scaffold 4.10.0, X!Tandem (version X!Tandem Alanine 2017.2.1.4), and annotation pipeline Prophane (97) were performed as described by Gierse et al. (96). For statistical analysis, one-way ANOVA with Benjamini-Hochberg correction (P = 0.05) and Scheffeé post hoc tests were performed. For data visualization, the software tools RStudio (95) and Perseus (98) were used.

Data availability.

All 16S rRNA gene sequences were submitted to European Nucleotide Archive (ENA) under project no. ERP135841 and accession no. PRJEB51232. The mass spectrometry proteomics data have been deposited in the ProteomeXchange Consortium via the PRIDE (99) partner repository with the data set identifiers PXD036279 (PBS), PXD036129 (19F_C), PXD036115 (19F_P) and, PXD036014 (H1N1).