Quantitative analysis of the human airway microbial ecology reveals a pervasive signature for cystic fibrosis

Abstract

Cystic fibrosis (CF) is an autosomal recessive disease caused by mutations in the gene encoding the CF transmembrane conductance regulator. Disruption of electrolyte homeostasis at mucosal surfaces leads to severe lung, pancreatic, intestinal, hepatic, and reproductive abnormalities. Loss of lung function as a result of chronic lung disease is the primary cause of death from CF. Using high-throughput sequencing to survey microbes in the sputum of 16 CF patients and 9 control individuals, we identified diverse microbial communities in the healthy samples, contravening conventional wisdom that healthy airways are not significantly colonized. Comparing these communities with those from the CF patients revealed significant differences in microbial ecology, including differential representation of uncultivated phylotypes. Despite patient-specific differences, our analysis revealed a focal microbial profile characteristic of CF. The profile differentiated case and control groups even when classically recognized CF pathogens were excluded. As a control, lung explant tissues were also processed from a group of patients with pulmonary disease. The findings in lung tissue corroborated the presence of taxa identified in the sputum samples. Comparing the sequencing results with clinical data indicated that diminished microbial diversity is associated with severity of pulmonary inflammation within our adult CF cohort.

Fig. 1

Phylum-level analysis of 16S rRNA gene sequences from CF patients and healthy control individuals. (A) Representation of microbial phyla in subject sputum. (B) Diversity of microbiota in the sputum of CF and control patients. (C) Ratio of Firmicutes to Bacteroidetes in the sputum of CF and control patients. (D) Representation of uncultivated organisms in the sputum of CF and control patients according to our DMI (see Materials and Methods for definitions).

Fig. 2

PCA of sequence classification occurrence. (A) Phylum-level and (B) family-level classification data are presented on the same scale in the two representations in each part. Data variance in each populated category (here, 12 phyla or 41 families) is normalized to minimize the influence of large variations in the abundance of one group across samples on the overall analysis. To test the PCA for robustness and dependence on sequencing depth (e.g., counting statistics for rare species), we repetitively rarefied the data, sampling 1000 sequences 100 times (with replacement) from each sample, and performed PCA using all 2500 (25 samples by 100 subsamples) subsamples. Points correspond to individual bootstrap replicate samples, with study subject clusters marked where distinct. The 16 CF samples (red points) cluster separately from controls (blue points). Solid gray lines show the projected original coordinates (corresponding to microbial taxonomic groups). (C) Mean interindividual PCA distance for the two study populations. Mean distances in the family PCA space among the 900 control subsamples (n = 360,099) and 1600 CF subsamples (n = 1,200,099) are plotted in consideration of an increasing number of PCA components (PC1, PC1, and PC2, and PC1, PC2, and PC3). Controls are separated to a much greater degree than are CF samples (t test for unequal variances, P < 10⁻⁶). (D) Ratio of Veillonellaceae to Micrococcaceae can be used to segregate subject population by clinical status. Box plots represent data quartiles with outliers indicated.

Fig. 3

Phylum-level correlations with clinical parameters. (A and B) Scatter plots reveal correlation of the Shannon diversity index (A) and the Simpson diversity index (B) with the inflammatory severity index.