Direct sampling of cystic fibrosis lungs indicates that DNA-based analyses of upper-airway specimens can misrepresent lung microbiota

Abstract

Recent work using culture-independent methods suggests that the lungs of cystic fibrosis (CF) patients harbor a vast array of bacteria not conventionally implicated in CF lung disease. However, sampling lung secretions in living subjects requires that expectorated specimens or collection devices pass through the oropharynx. Thus, contamination could confound results. Here, we compared culture-independent analyses of throat and sputum specimens to samples directly obtained from the lungs at the time of transplantation. We found that CF lungs with advanced disease contained relatively homogenous populations of typical CF pathogens. In contrast, upper-airway specimens from the same subjects contained higher levels of microbial diversity and organisms not typically considered CF pathogens. Furthermore, sputum exhibited day-to-day variation in the abundance of nontypical organisms, even in the absence of clinical changes. These findings suggest that oropharyngeal contamination could limit the accuracy of DNA-based measurements on upper-airway specimens. This work highlights the importance of sampling procedures for microbiome studies and suggests that methods that account for contamination are needed when DNA-based methods are used on clinical specimens.

Fig. 1.

Relative abundance of bacterial taxa in explanted CF lungs. The colored segments of each bar represent the proportion of reads mapping to different bacterial taxa, when data from all lobar samples from each subject’s right and left lung are summed. The legend identifies taxa present at ≥1% abundance. Lower abundance taxa are identified in Dataset S1, Table S1 and Fig. S5.

Fig. 2.

Comparison of taxa identified in upper-airway and lung samples. The colored segments of each bar represent the proportion of reads mapping to different bacterial taxa. Lung data are the sum of all lobar samples from each subject, reproduced from Fig. 1 for comparison. ^§Samples that produced less than 5,000 reads (see Dataset S1, Tables S3 and S4). (A) Throat specimens (T) contained more bacterial diversity and nontypical organisms than lung samples (L) collected from the same subjects several hours later. (B) Paired sputum (S) and lung (L) samples identified the same species as dominant. However, sputum samples from some patients contained diverse collections of nontypical organisms not found in the lungs. *Sample obtained 1 mo before transplant; **sample obtained 3 d before transplant.

Fig. 3.

Comparisons of microbial diversity in throat, sputum, and lung samples. (A) Plots of the Shannon diversity index of throat, sputum, and lobar lung samples. Sputum and throat data points indicate samples from different subjects, and lung data points indicate different lobar samples (6–12 per lung pair). The horizontal line indicates mean values. *P < 0.05. (B) Proportion of organisms considered nontypical in CF (see Results) in throat, sputum, and lung samples. Sputum and throat data points indicate samples from different subjects, and lung data points indicate lobar samples (6–12 per lung pair). The horizontal line indicates mean values. *P < 0.05 by Mann–Whitney U test.

Fig. 4.

Relative abundance of bacterial taxa in explant lung lobes. The colored segments of each bar represent the proportion of reads mapping to different bacterial taxa. Most lungs resembled that from subject 1 (A) and showed minimal regional differences. Some lobar regions from the lungs of subjects 3 (B) and 9 (C) showed differences in the abundance of typical pathogens. See Fig. S5 for data from other subjects. Ling., lingula; Ling.², lingula second sample; LLL, left lower lobe; LLL² , left lower lobe second sample; L main, left main-stem bronchus; LUL, left upper lobe; RLL, right lower lobe; RLL², right lower lobe second sample; RLL³, right lower lobe third sample; R main, right main-stem bronchus; RML, right middle lobe; RML², right middle lobe second sample; RUL, right upper lobe; Sum, aggregate of all lung samples from each subject. ^§Samples that produced less than 5,000 reads (see Dataset S1, Tables S8, S10, and S16).

Fig. 5.

Relative abundance of bacterial taxa in serial sputum samples from clinically stable subjects. The colored segments of each bar represent the proportion of reads mapping to different bacterial taxa. Specimens from subject 11 (A) and 12 (B) showed variation in the abundance of nontypical organisms; however, Pseudomonas dominated in each. Specimens from subject 13 (C) showed day to day variation that changed the identity of the dominant taxa. ^§Samples that produced less than 5,000 reads (see Dataset S1, Table S18).