A longitudinal characterization of the Non-Cystic Fibrosis Bronchiectasis airway microbiome

Abstract

A diverse microbiota exists within the airways of individuals with non-cystic fibrosis bronchiectasis (nCFB). How the lung microbiome evolves over time, and whether changes within the microbiome correlate with future disease progression is not yet known. We assessed the microbial community structure of 133 serial sputa and subsequent disease course of 29 nCFB patients collected over a span of 4-16 years using 16S rRNA paired-end sequencing. Interestingly, no significant shifts in the microbial community of individuals were observed during extended follow-up suggesting the microbiome remains relatively stable over prolonged periods. Samples that were Pseudomonas aeruginosa culture positive displayed markedly different microbial community structures compared to those that were positive for Haemophilus influenzae. Importantly, patients with sputum of lower microbial community diversity were more likely to experience subsequent lung function decline as defined by annual change in ≥-1 FEV₁% predicted. Shannon diversity values <1 were more prevalent in patients with FEV₁ decline (P = 0.002). However, the relative abundance of particular core microbiota constituents did not associate with risk of decline. Here we present data confirming that the microbiome of nCFB individuals is generally stable, and that microbiome-based measurements may have a prognostic role as biomarkers for nCFB.

Figure 1

Taxonomic summaries of OTU present in >1% relative abundance of all samples (n = 133) obtained from a cohort of 29 nCFB patients as a function of the years from first sample. Samples for each patient are grouped together by a unique patient identifier – as indicated by the grey heading. Patient’s numbers from the Decliner group are indicated in magenta. Analysis of the taxonomic summary reveals a unique microbiome associated with each patient and varying degrees of inter and intra-patient diversity. Samples collected under antibiotic pressure are indicated by red text indidcating year of collection.

Figure 2

Clustering of samples by patient during stable periods (excluding exacerbation events) was observed (n = 121). Samples obtained from each patient are designated by the patient number adjacent to each point. Samples were categorized by color based on the time since baseline sample. Patients classified as Decliners are indicated by a circle, whereas patients with a Stable course are represented by triangles. Significant clustering was observed using PERMANOVA analysis (p = 0.03).

Figure 3

The microbial community was observed to remain relatively constant over the study period as measured by Observed OTUs (A), Shannon-Wiener (B) and Simpson diversity (C), and Bray-Curtis dissimilarity (D). Samples were categorized based on the time interval relative to first sample as: 0 (n = 29), 0–2 years (n = 36), 2–4 (n = 22), and 4 + (n = 46) years since initial sample collection. The median and interquartile ranges (IQR) are represented by the middle, top, and bottom lines of each boxplot. No statistically significant comparisons were observed.

Figure 4

Sputum samples collected from patients (n = 44 samples, 11 patients) who were categorized as Decliners (≥1 FEV₁% per year) generally displayed lower levels of alpha diversity as compared to Stables (n = 89, 18 patients). No significant differences were detected using the Observed OTU (A), Shannon (B), and Simpson (C) diversity indices. The median and interquartile ranges of each group are represented by the middle, top, and bottom lines of each boxplot. Statistical analysis was performed using the Wilcoxon-signed rank test.

Figure 5

Observed OTU (A), Shannon (B), and Simpson (C) diversity, and Bray-Curtis dissimilarity (D) illustrate clustering of P. aeruginosa and H. influenzae positive samples. Differences in the sputum microbial diversity of samples collected in relation to the growth of Pseudomonas aeruginosa (PA, n = 60) and Haemophilus influenzae (HI, n = 17) and in samples with neither (Neither, n = 56). The ellipses for each group represent the 90% confidence interval. The median and interquartile ranges (IQR) are represented by the middle, top, and bottom lines of each boxplot. Statistically significant comparisons were shown using the p-value.

Figure 6

Acute fluoroquinolones use was associated with reduced community richness as represented by the Observed OTU. Samples collected during acute fluoroquinolone use (FQ, n = 55) were compared to samples collected in their absence (No-FQ, n = 78). Significant differences were seen in Observed OTUs (A), but not for Shannon (B) or Simpson diversity (C). Significant community-wide differences were found using PERMANOVA when permutations were constrained by patient and visualized using PCoA (p = 0.01).