The respiratory tract microbiome, the pathogen load, and clinical interventions define severity of bacterial pneumonia

Abstract

Bacterial pneumonia is a considerable problem worldwide. Here, we follow the inter-kingdom respiratory tract microbiome (RTM) of a unique cohort of 38 hospitalized patients (n = 97 samples) with pneumonia caused by Legionella pneumophila. The RTM composition is characterized by diversity drops early in hospitalization and ecological species replacement. RTMs with the highest bacterial and fungal loads show low diversity and pathogen enrichment, suggesting high biomass as a biomarker for secondary and/or co-infections. The RTM structure is defined by a "commensal" cluster associated with a healthy RTM and a "pathogen" enriched one, suggesting that the cluster equilibrium drives the microbiome to recovery or dysbiosis. Legionella biomass correlates with disease severity and co-morbidities, while clinical interventions influence the RTM dynamics. Fungi, archaea, and protozoa seem to contribute to progress of pneumonia. Thus, the interplay of the RTM equilibrium, the pathogen load dynamics, and clinical interventions play a critical role in patient recovery.

Keywords: Legionella pneumophila; biomass; microbial ecology; pneumonia; respiratory tract microbiome; severity.

Graphical abstract

Figure 1

Diversity of the RTM of patients with pneumonia due to L. pneumophila (A) Alpha-diversity metrics Shannon, Chao 1, Berger-Parker index (dbp), and biomass represented in violin box plots (n = 87). Bacterial abundance is measured as the number of 16S rRNA copies per milliliter of sample. The number of bacteria ranged from 10⁵ to 10⁹, with a mean of 10⁷ 16S rRNA copies per milliliter of sample. The Shannon index distribution is variable (0.2–6), with most values between 3 and 4 (mean of 3). The richness estimator Chao 1 shows for most of the samples values between ∼100 and 300 predicted ASVs (mean of 221), while the Berger-Parker index ranged mainly from 0.2 to 0.5 with a mean of 0.3. (B) Spearman correlations of bacterial abundance with alpha-diversity metrics Shannon, Chao 1, and dbp (n = 87). The correlation coefficients (rho) and p values are shown. (C) Samples and taxa distribution represented in a non-metric multidimensional scaling (NMDS) plot based on the Bray-Curtis dissimilarity matrix (n = 87). The PERMANOVA-associated p value of the microbial abundance effect in shaping microbial composition is included above the graph. The bacterial abundance variable is plotted as a vector (red arrow), and the most significant taxa that change in correlation with the bacterial abundance are represented with colored arrows. (D) Spearman correlations of fungal abundance with alpha-diversity metrics Shannon, Chao 1, and dbp (n = 29). The correlation coefficients (rho) and p values are shown. (E) Samples and taxa distribution represented in an NMDS plot based on the Bray-Curtis dissimilarity matrix (n = 29). The PERMANOVA-associated p value of the fungal abundance effect in shaping microbial composition is included above the graph. The fungal abundance variable is plotted as a vector (red arrow), and the most significant taxa that change in correlation with the bacterial abundance are represented with colored arrows. n, number of samples.

Figure 2

Putative interactions among dominant bacteria of the RTM of pneumonia patients (A) Heatmap of the dominant bacterial taxa per sample (n = 87). The most significant taxa regarding predominance, relative abundance, and absolute abundance are marked with asterisks. (B) The prokaryotic network of the RTM considers the most abundant taxa (mean over 10³ 16S rRNA copies per milliliter of the sample). The network is based on the ASV abundance tables collapsed at the genus level. Significant positive associations are shown (p < 0.01). Each node represents a taxon, color-coded by phylum. The main clusters of nodes are highlighted. Connections within a cluster are colored gray, and connections between clusters are in red. Dominant, highly connected, and articulation points of taxa are shown. A complete detailed network is shown in Figure S3C. n, number of samples.

Figure 3

RTM diversity and composition per patient (A) Alpha-diversity metrics Shannon, Chao 1, and Berger-Parker index (dbp). Patients with temporal samples were included (15 patients, n = 69). (B) Total bacterial and Legionella load. Abundance represents the number of 16S rRNA copies per milliliter of sample. Samples and taxa distribution are represented in an NMDS plot based on the Bray-Curtis dissimilarity matrix. Taxa distribution is represented with triangles. The PERMANOVA-associated p value of the variable “patient” is included above the graph. (C and D) (C) Temporal variation of the microbiome composition and (D) bacterial load of two patients (O and P). (E) Significant temporal trends of the most abundant bacteria (%). Significance is based on the “Trendyspliner” (test for a significant non-zero trend over time) implemented in the splinectomeR package. n, number of samples.

Figure 4

Statistical association of metadata variables with the alpha and beta diversity, the biomass of the bacterial RTM, and the Legionella load The heatmap shows the p values of the statistical analyses (Wilcoxon signed-rank test, Spearman correlation, PERMANOVA). The definition of the categorical variables and the alpha diversity (Shannon, Chao 1, and Berger-Parker index [dbp]) was based on the Wilcoxon signed-rank test with Bonferroni correction. The association with the alpha diversity was based on the Spearman correlation for continuous variables. The direction of the correlation coefficient is indicated. For the categorical variables “dual therapy of antibiotics,” the positive direction means higher abundance in treated samples; in “gender,” higher abundance in men; in “diabetes,” higher abundance in diabetic individuals; “invasive ventilation,” higher abundance in ventilated individuals. The association of the microbiome composition (beta diversity) with the variables was based on the PERMANOVA test (R adonis2 function). The number of samples and patients is shown at the top. The R² coefficient from the adonis2 result is also represented as a measure of the variance explained by the variable.

Figure 5

Impact of clinical interventions on RTM diversity and composition (A) Effect of mechanic ventilation on alpha-diversity metrics Shannon, Chao 1, and Berger-Parker index (dbp) (n = 53). (B) Total bacterial and Legionella load (n = 53). Abundance: number of 16S rRNA copies per milliliter of sample. The comparison is based on the Wilcoxon signed-rank test with Bonferroni correction. p values are shown. Samples and taxa distribution are represented in an NMDS plot based on the Bray-Curtis dissimilarity matrix. Taxa distribution is represented with triangles. The PERMANOVA-associated p value of the variable “mechanical ventilation” is included above the graph. (C) Effect of dual therapy (macrolide and fluoroquinolone) on alpha-diversity metrics Shannon, Chao 1, and dbp (n = 50). (D) Total bacterial and Legionella load (n = 50). Abundance represents the number of 16S rRNA copies per milliliter of sample. The comparison is based on the Wilcoxon signed-rank test with Bonferroni correction. p values are shown. Samples and taxa distribution are represented in an NMDS plot based on the Bray-Curtis dissimilarity matrix. Taxa distribution is represented with triangles. The PERMANOVA-associated p value of the variable “dual therapy” is included above the graph. n, number of samples.

Figure 6

Predicted inter-kingdom RTM interactions (A) Relative abundance (%) of the overall most prevalent fungal genera of the RTM. The taxonomy is based on the UNITE database from QIIME software (n = 29). (B) The inter-kingdom (bacteria, archaea, and fungi) network considers the most abundant taxa (mean over 10³ 16S rRNA copies per milliliter of the sample and 10 for the ITS) (n = 17). The network is based on the ASV abundance tables collapsed at the last identified taxonomic level. Significant positive associations are shown (p < 0.05). Each node represents a taxon color-coded by phylum. The main clusters of nodes are highlighted. Gray, connections within a cluster; red, connections between clusters; bold, fungal taxa. Only the closest taxa to fungi are shown. A complete, detailed network is shown in Figure S3E. n, number of samples.