Among subjects undergoing lung lobectomy for suspected or confirmed lung cancer at the Minneapolis VA Medical Center (MVAMC) and who reported no use of systemic antibiotics or systemic corticosteroids in the prior 1 month, 44 consented to study participation (Table 1). Consistent with the VA patient population, most subjects were male. Thirty-six of 44 (81.8%) met spirometric criteria for COPD, and over half (20, 55.6%) of those with COPD had mild obstruction. Few of those with COPD (2, 5.6%) were using inhaled corticosteroids (ICS). Subjects with COPD (vs. without COPD) reported a greater number of pack-years of tobacco exposure (49.5 vs. 25) and were more likely to report current alcohol use (66.7% vs. 25%). Most subjects were found to have lung adenocarcinoma (26, 59.1%), although 2 subjects had more than one lung malignancy and 3 subjects did not have a malignancy. Additionally, in 2 subjects a pathologic diagnosis was not available.

All subject samples and negative controls underwent biomass quantification with droplet digital PCR (ddPCR) prior to microbiome sequencing. Both negative control sample types had significantly less biomass than all subject sample types (generalized estimating equations [GEE] model with single-step adjustment, p < 0.001; Table S1, Fig. 1). Lung and bronchus samples had lower biomass than nasopharyngeal, oral wash, and sputum samples (GEE, all p < 0.001). Oral wash and sputum samples had higher biomass than lung, bronchus, and nasopharyngeal samples (GEE, all p < 0.001). Nasopharyngeal sample biomass was significantly different from the other sample types (GEE, p < 0.001).

Upper and lower airway biomass. Horizontal lines represent the median value in each group, while the top and bottom of the boxes represent the 75th and 25th percentile values, respectively. Both negative control sample types had significantly lower biomass than all patient sample types (GEE, all p < 0.001). The lowest biomass samples (lung, bronchus) had lower biomass than all other subject sample types (GEE, all p < 0.001). The highest biomass samples (oral wash, sputum) had higher biomass than all other subject samples (GEE, all p < 0.001). Lung and bronchus samples were not significantly different from each other, and oral washes and sputum samples were not significantly different from each other

α-diversity statistics (Simpson diversity, Shannon diversity, and Chao1 diversity) were calculated for all samples and illustrated by sample type. Simpson diversity findings were generally consistent across the other α-diversity metrics and are presented here. Oral wash and sputum samples are significantly more diverse than bronchus and lung samples (GEE, p < 0.0001; Fig. 2). Bronchus, lung, and nasopharyngeal sample Simpson diversity are not significantly different from each other. Shannon diversity findings are similar to Simpson diversity findings, however Chao1 diversity findings demonstrate that nasopharyngeal samples are similar in richness to oral wash and sputum samples, which all have greater richness than bronchus and lung samples (Figure S3). Bronchus Simpson diversity associates with within-subject oral wash and sputum Simpson diversity, but not within-subject lung or nasopharyngeal diversity (GEE, Table S2).

Oral wash and sputum samples have greater Simpson diversity than lung, bronchus, and nasopharyngeal samples. Simpson diversity was determined for each sample and illustrated by sample site. Horizontal bars represent the median value for each sample while the top and bottom of the boxes represent the 75th and 25th percentile values, respectively. Oral wash and sputum samples are significantly more diverse than lung, bronchus, and nasopharyngeal samples (GEE, p < 0.0001)

We also investigated whether 6 relevant clinical characteristics were associated with α-diversity at any of the 5 anatomic sites. Older age was associated with greater Simpson diversity in the oral and nasopharyngeal sites (GEE, Coefficient Estimate [CE] 0.0022, 95% Confidence Interval [CI, 0.00024, 0.0042], p = 0.022; and LR, CE 0.016, 95% CI [0.0032, 0.029], p = 0.019, respectively; Fig. 3a and b). Current tobacco use was associated with greater lung and bronchus Simpson diversity (GEE, CE 0.22, 95% CI [0.17, 0.28], p < 0.0001, Fig. 3c), but not diversity at other sites.

Clinical characteristics are associated with upper and lower airway Simpson diversity. A.. Older age is associated with greater oral wash Simpson diversity. Oral wash 1 and oral wash 2 are illustrated by shape, and the regression lines represent the association between older age and greater Simpson diversity (GEE, CE 0.0022, 95% CI [0.00024, 0.0042], p = 0.022). B. Older age is associated with greater nasopharyngeal Simpson diversity. The regression line represents the association between older age and greater Simpson diversity (LR, CE 0.016, 95% CI [0.0032, 0.029], p = 0.019). C. Current tobacco use is associated with greater lung and bronchus Simpson diversity (GEE, CE 0.22, 95% CI [0.17, 0.28], p < 0.0001). D. Last professional dental cleaning more than 6 months prior to surgery (vs. within the last 6 months) was associated with lower lung and bronchus Simpson diversity (GEE, CE 0.21, 95% CI [0.12, 0.29], p < 0.0001). Edentulous subjects were excluded from this analysis. Data were dichotomized at the 6 month timepoint for the GEE analysis, however three timepoints (within the last 6 months, 6–12 months ago, and more than 1 year ago) are presented here for illustrative purposes. E. Lung adenocarcinoma (vs. other pathologic finding) was associated with lower lung and bronchus Simpson diversity (GEE, CE -0.11, 95% CI [-0.21, -0.015], p = 0.024)

Dental care habits were also associated with Simpson diversity in lower airway samples. Ten of the 44 subjects were edentulous and therefore not included in this analysis (Figure S4). Among the 34 remaining subjects, 15 reported that their last professional dental cleaning was within the prior 6 months, while 19 reported that their last professional dental cleaning was more than 6 months prior. Self-reported last profession dental cleaning more than 6 months prior was associated with lower lung and bronchus Simpson diversity (GEE, CE 0.21, 95% CI [0.12, 0.29], p < 0.0001, Fig. 3d). A pathologic diagnosis of lung adenocarcinoma was made in 26 (59.1%) of subjects, while the remaining subjects were diagnosed with other malignancies (13, 29.5%) or non-malignant findings (5, 11.4%). Presence of a lung adenocarcinoma (vs. other pathologic findings) was associated with lower bronchus and lung Simpson diversity (GEE, CE -0.11, 95% CI [-0.21, -0.015], p = 0.024, Fig. 3e). Unless mentioned above, tests of association between site-specific Simpson diversity and these 6 clinical characteristics (including FEV1pp, ICS use, and pack-years of tobacco use) were not significant.

β-diversity was assessed using Bray-Curtis dissimilarity and illustrated with principal coordinates analysis (PCoA; Fig. 4). Coordinate 1, representing 16.6% of the variance within the dataset, separates nasopharyngeal, lung and bronchus samples from oral wash and sputum samples. Coordinate 2, representing 6% of the variance, separates nasopharyngeal samples from the other sample types. β-diversity was also illustrated using weighted UniFrac, with similar separation by anatomic site (Figure S5).

Principal coordinate analysis (PCoA) illustrates sample clustering by sampling site. All subject samples were assessed by Bray-Curtis dissimilarity and illustrated by PCoA of Coordinate 1 (16.6% of variance) and Coordinate 2 (6% of variance). Lung and bronchus samples are similar to each other, and cluster separately from the nasopharyngeal samples and the oral wash and sputum samples. Oral wash and sputum samples cluster with each other, separate from the other samples

After separating samples based on site, correlations between clinical characteristics and β-diversity were assessed with PERMANOVA analyses utilizing Bray-Curtis dissimilarity. Last professional dental cleaning, dichotomized as within the last 6 months vs. more than 6 months prior, was associated with clustering among lung samples (p = 0.027, R2 = 0.016; Fig. 5). Several other clinical characteristics resulted in p-values between 0.05 and 0.10, which did not reach statistical significance (Figure S6). FEV1pp, ICS use, and current tobacco use were not associated with microbiome composition in our dataset. PERMANOVA analyses utilizing the weighted UniFrac matrix yielded very similar results (Table S3).

Principal coordinate analysis (PCoA) illustrates lung sample clustering by self-reported last professional dental cleaning. All lung samples were assessed by Bray-Curtis dissimilarity and illustrated by PCoA of Coordinate 1 (8.1% of variance) and Coordinate 2 (7.2% of variance). Samples were dichotomized by self-reported last dental cleaning (within the last 6 months vs. more than 6 months ago) for analysis, although samples are labeled here with additional detail. Self-reported last professional dental cleaning was associated with lung microbiome composition (p = 0.027, R2 = 0.016)

We investigated correlations between relevant clinical factors and taxonomic composition. Increasing age was associated with greater abundance of 3 oral taxa (Table 2). Current tobacco use was associated with changes in abundance of multiple oral and lung taxa, including greater abundance of pulmonary pathogens Mycoplasmoides and Haemophilus in lower airway samples (Table 3). Self-reported recent professional dental cleaning (within the last 6 months) was associated with greater bronchial Actinomyces and lung Streptococcus abundance, compared with less recent professional dental cleaning (more than 6 months ago; Table 4). A diagnosis of lung adenocarcinoma (vs. no diagnosis of lung adenocarcinoma) was associated with lower Lawsonella abundance in lung samples (Table 5). The use of ICS was associated with changes in taxa abundance in the oral, lung, nasopharyngeal, and sputum microbiome. Notably, these included greater abundance of Haemophilus among oral samples, greater Staphylococcus among lung samples, and lower Fusobacterium, Gemella, and Acinetobacter among lung samples (Table 6). Finally, greater FEV1pp was associated with lower Escherichia/Shigella abundance in the lung microbiome (Table 7).