Species-level, metagenomic and proteomic analysis of microbe-immune interactions in severe asthma

Abstract

Background: The airway microbiome in severe asthma has not been characterised at species-level by metagenomic sequencing, nor have the relationships between specific species and mucosal immune responses in 'type-2 low', neutrophilic asthma been defined. We performed an integrated species-level metagenomic data with inflammatory mediators to characterise prevalence of dominant potentially pathogenic organisms and host immune responses.

Methods: Sputum and nasal lavage samples were analysed using long-read metagenomic sequencing with Nanopore and qPCR in two cross-sectional adult severe asthma cohorts, Wessex (n = 66) and Oxford (n = 30). We integrated species-level data with clinical parameters and 39 selected airway proteins measured by immunoassay and O-link.

Results: The sputum microbiome in health and mild asthma displayed comparable microbial diversity. By contrast, 23% (19/81) of severe asthma microbiomes were dominated by a single respiratory pathogen, namely H. influenzae (n = 10), M. catarrhalis (n = 4), S. pneumoniae (n = 4) and P. aeruginosa (n = 1). Neutrophilic asthma was associated with H. influenzae, M. catarrhalis, S. pneumoniae and T. whipplei with elevated type-1 cytokines and proteases; eosinophilic asthma with higher M. catarrhalis, but lower H. influenzae, and S. pneumoniae abundance. H. influenzae load correlated with Eosinophil Cationic Protein, elastase and IL-10. R. mucilaginosa associated positively with IL-6 and negatively with FGF. Bayesian network analysis also revealed close and distinct relationships of H. influenzae and M. catarrhalis with type-1 airway inflammation. The microbiomes and cytokine milieu were distinct between upper and lower airways.

Conclusions: This species-level integrated analysis reveals central, but distinct associations between potentially pathogenic bacteria and airways inflammation in severe asthma.

Keywords: Asthma; bacteria; cytokine; microbiome; sputum.

Figure 1. Relative abundance of species in sputum

Heatmap of relative abundance of species from induced sputum sequenced samples using ONT; Z-scores, denoted by shade, represent the relative abundance of each taxon within each individual patient (columns). Cohorts include Wessex (WES; Healthy [H]: n=5, Mild [M]: n=10 and Severe asthma [S]: n=51) and Oxford (OX; Severe asthma[S]: n=30). Clustering by species (rows) and individuals (columns) using Euclidian distance performed independently on samples and most differentially abundant bacterial species. Sputum inflammatory phenotypes are eosinophilic (E), neutrophilic (N), mixed granulocytic (M) and paucigranulocytic (P). Positive pathogen specific PCR results (>1x106copies/ml) are indicated in black: S. aureus (Saur), P. aeruginosa (PsA), S. pneumoniae (Spneu), M. catarrhalis (Mcat) and H. influenzae (Hinf). Total eubacteria (16S) PCR load is denoted by shade (x106 copies/ml).

Figure 2. Integrated analyses of microbiome and inflammatory mediators in sputum

Integrated cytokine analysis (A) Heatmap of scaled and centralised sputum cytokines in health, mild and severe asthma. Z-score centralised cytokine measurements denoted by shade. Independent clustering performed by subject (columns) and cytokine (rows). Clustering distances shown are representative of correlation distances. In severe asthma, presence of a dominant pathogenic organism on metagenomic sequencing is indicated as HI (H. inf), MC (M. cat) and SP (S. pneu); the pathogen-dominant microbiome in severe asthma (data only shown for participants with severe asthma) is associated with (B) sputum neutrophilia, (C) elevated type-1 cytokines and (D) anti-inflammatory mediators in sputum supernatant. Median/IQR shown −/+ Pathogen dominance ([C] Mann-Whitney test, [D]/[E] unpaired t-test, adjusted for multiple comparisons with BH procedure [FDR 0.05]. *P≤0.05, **P≤0.01).

Figure 3. Heatmap displaying correlations between integrated sputum microbiome and cytokine data

The 50 most variable bacterial species across all subjects were selected. Spearman correlation coefficients between species and cytokines denoted by shade. Independent clustering performed by cytokine (columns) and species (rows). Statistically significant correlations have been annotated as ^$P=0.05, *P< 0.05 and **P<0.01. Adjustment for multiple comparisons was made using BH- procedure (FDR 0.05).

Figure 4. Integrated Bayesian Network Analysis

Bayesian network showing strongest interactions between most differentially abundant bacterial species, inflammatory mediators, and clinical features in severe asthma (WES, n=51). Nodes are coloured by class: bacterial species (blue), infection status (red), inflammatory mediators (green) and clinical features (orange). Nodes without strong interactions have been excluded. Line thickness represents strength of interaction (Euclidean distance) as shown in the scale. Interactions with a positive linear correlation are shown with a green edge.

Figure 5. Upper and lower airway microbiome and cytokines

(A) Heatmap of relative abundances of bacterial species on metagenomic sequencing of paired nasal lavage (NL) and sputum (SP) samples from severe asthmatics (n=17). Z-score denoted by shade. Independent clustering by samples (columns) and 23 most variable species (rows) using Euclidian distance. (B) Correlation matrix of the most significantly correlated (P<<0.001) proteins in paired nasal lavage (NAS) and sputum (SPU) cytokines (O-link). Significantly correlated pairs displayed (p<<0.001, adjusted for multiple comparisons with Benjamini–Hochberg procedure [FDR 0.05]).