Topographic diversity of the respiratory tract mycobiome and alteration in HIV and lung disease

Abstract

Rationale: Microbiome studies typically focus on bacteria, but fungal species are common in many body sites and can have profound effects on the host. Wide gaps exist in the understanding of the fungal microbiome (mycobiome) and its relationship to lung disease.

Objectives: To characterize the mycobiome at different respiratory tract levels in persons with and without HIV infection and in HIV-infected individuals with chronic obstructive pulmonary disease (COPD).

Methods: Oral washes (OW), induced sputa (IS), and bronchoalveolar lavages (BAL) were collected from 56 participants. We performed 18S and internal transcribed spacer sequencing and used the neutral model to identify fungal species that are likely residents of the lung. We used ubiquity-ubiquity plots, random forest, logistic regression, and metastats to compare fungal communities by HIV status and presence of COPD.

Measurements and main results: Mycobiomes of OW, IS, and BAL shared common organisms, but each also had distinct members. Candida was dominant in OW and IS, but BAL had 39 fungal species that were disproportionately more abundant than in the OW. Fungal communities in BAL differed significantly by HIV status and by COPD, with Pneumocystis jirovecii significantly overrepresented in both groups. Other fungal species were also identified as differing in HIV and COPD.

Conclusions: This study systematically examined the respiratory tract mycobiome in a relatively large group. By identifying Pneumocystis and other fungal species as overrepresented in the lung in HIV and in COPD, it is the first to determine alterations in fungal communities associated with lung dysfunction and/or HIV, highlighting the clinical relevance of these findings. Clinical trial registered with www.clinicaltrials.gov (NCT00870857).

Keywords: COPD; HIV; bronchoalveolar lavage; fungi; microbiome.

Figures

Figure 1.

Principal coordinate analysis plot and ubiquity–ubiquity plots (U–U plot) of the mycobiome of oral wash (OW), induced sputum (IS), and bronchoalveolar lavage (BAL) in HIV-uninfected individuals with normal lung function. (A) Principal coordinate analysis plot showing the clustering trend of OW (n = 15), IS (n = 8), and BAL (n = 11) samples (18S data). PCo = principal coordinate. (B–D) U–U plot generated using Corbata (21), comparing the ubiquity of every fungal species present in two biomes of either OW (n = 18), IS (n = 9), or BAL (n = 16) (internal transcribed spacer data). The top five taxa in every biome were labeled, and the P value showing the significance of difference between two biomes was generated in one-sided “abundance-weighted” Kolmogorov-Smirnov (AWKS) in Corbata (21). (B) U–U plot comparing the oral mycobiome (OW) with the lung mycobiome (BAL). (C) U–U plot comparing the oral mycobiome (OW) with the lung mycobiome (IS). (D) U–U plot comparing the lung mycobiome (IS) with the lung mycobiome (BAL). NS = not significant. **P ≤ 0.01; ***P ≤ 0.005.

Figure 2.

Neutral model comparing the lung community (bronchoalveolar lavage [BAL]) with the oral community (oral wash) in the entire cohort and comparing lung communities in HIV infection and in chronic obstructive pulmonary disease (COPD). (A) In the neutral model plot, the solid line represents the frequency predicted in the model and dashed lines are 95% binomial confidence intervals. Although most operational taxonomic units (OTUs) were shared between the oral and lung community (found between the 95% confidence intervals in the neutral model; gray in the plots), 39 fungal species fell outside the upper confidence interval (red) and were considered to have a fitness advantage in the lung (those OTUs that were too close to the axis were omitted for plotting, but a detailed taxonomic list is provided in Table E1). The fungal species that fell below the lower confidence interval (green) did not have a fitness advantage in the lung. (B) Overrepresented species in BAL from HIV-infected and HIV-uninfected individuals. (C) Overrepresented species in BAL from HIV-infected individuals with and without COPD. The intersection of the results from three statistical methods (random forest, logistic regression, and metastats) (adjusted for multiple t tests) was colored in the plot. Only random forest and metastats results are colored for COPD species in C because logistic regression failed to identify any species. Red-labeled species are overrepresented in the disease group compared with the control group, and the blue-labeled species are found to be dominant in the control group. The ranking by importance in the group classification is based on the output from the random forest classifier (internal transcribed spacer data, n = 53 oral wash, 44 BAL).

Figure 3.

Neutral model comparing the lung communities in induced sputum (IS) with the oral community (oral wash) and comparing lung communities in HIV infection and in chronic obstructive pulmonary disease (COPD) using IS. (A) In the neutral model plot, the solid line represents the frequency predicted in the model and dashed lines are 95% binomial confidence intervals. Although most operational taxonomic units (OTUs) were shared between the oral and lung community (found between the 95% confidence intervals in the neutral model; gray in the plots), 225 fungal species fell outside the upper confidence interval (blue) and were considered to have a fitness advantage in the lung (those OTUs that were too close to the axis were omitted for plotting, but a detailed taxonomic list is provided in Table E1). The fungal species that fell below the lower confidence interval (green) did not have a fitness advantage in the lung. (B) Overrepresented species in IS from HIV-infected and HIV-uninfected individuals. (C) Overrepresented species in IS from HIV-infected individuals with and without COPD. The intersection of the results from three statistical methods (random forest, logistic regression, and metastats) (adjusted for multiple t tests) was colored in the plot. Only random forest and metastats results are colored for COPD species in C because logistic regression failed to identify any species. Red-labeled species are overrepresented in the disease group compared with the control group, and the blue-labeled species are found to be dominant in the control group. The ranking by importance in the group classification is based on the output from the random forest classifier (internal transcribed spacer data, n = 53 oral wash, 30 IS).

Figure 4.

Ubiquity–ubiquity plots (U–U plot) comparing the ubiquity of every fungal species present in two disease groups. The top five taxa in every biome were labeled, and the P value showing the significance of difference between two biomes was generated in one-sided “abundance-weighted” Kolmogorov-Smirnov (AWKS) in Corbata (21). (A) U–U plots comparing HIV-infected individuals with HIV-uninfected individuals in oral wash (n = 53), induced sputum (n = 30), and bronchoalveolar lavage (BAL) (n = 44) samples. (B) U–U plots comparing HIV-infected individuals with or without chronic obstructive pulmonary disease (COPD) in oral wash (n = 31), induced sputum (n = 19), and BAL (n = 26) samples (internal transcribed spacer data). NS = not significant. **P ≤ 0.01; ***P ≤ 0.005.