Commensal Oral Microbiota, Disease Severity, and Mortality in Fibrotic Lung Disease

Abstract

Rationale

Oral microbiota associate with diseases of the mouth and serve as a source of lung microbiota. However, the role of oral microbiota in lung disease is unknown.

Objectives

To determine associations between oral microbiota and disease severity and death in idiopathic pulmonary fibrosis (IPF).

Methods

We analyzed 16S rRNA gene and shotgun metagenomic sequencing data of buccal swabs from 511 patients with IPF in the multicenter CleanUP-IPF (Study of Clinical Efficacy of Antimicrobial Therapy Strategy Using Pragmatic Design in IPF) trial. Buccal swabs were collected from usual care and antimicrobial cohorts. Microbiome data were correlated with measures of disease severity using principal component analysis and linear regression models. Associations between the buccal microbiome and mortality were determined using Cox additive models, Kaplan-Meier analysis, and Cox proportional hazards models.

Measurements and Main Results

Greater buccal microbial diversity associated with lower FVC at baseline (mean difference, −3.60; 95% confidence interval [CI], −5.92 to −1.29% predicted FVC per 1-unit increment). The buccal proportion of Streptococcus correlated positively with FVC (mean difference, 0.80; 95% CI, 0.16 to 1.43% predicted per 10% increase) (n = 490). Greater microbial diversity was associated with an increased risk of death (hazard ratio, 1.73; 95% CI, 1.03–2.90), whereas a greater proportion of Streptococcus was associated with a reduced risk of death (HR, 0.85; 95% CI, 0.73 to 0.99). The Streptococcus genus was mainly composed of Streptococcus mitis species.

Conclusions

Increasing buccal microbial diversity predicts disease severity and death in IPF. The oral commensal S. mitis spp associates with preserved lung function and improved survival.

At a Glance Commentary

Scientific Knowledge on the Subject

The role of oral microbiota in the pathogenesis of idiopathic pulmonary fibrosis is unknown.

What This Study Adds to the Field

Lower diversity of the buccal microbiome and higher relative abundance of the genus Streptococcus associates with preserved FVC and a reduced relative rate of death in patients with idiopathic pulmonary fibrosis.

Microbiota have been causally implicated in myriad disease pathogeneses and are known to regulate immunity (1, 2). In lung disease, lung-associated microbiota are consistently correlated with progression and mortality (3, 4). Fibrotic lung diseases carry high morbidity, and the most common form, idiopathic pulmonary fibrosis (IPF), is progressive and fatal (5). In IPF, lung microbial community composition correlates with alveolar inflammation, host defense, and key radiographic features of disease and is associated with mortality (6–11).

Microbiota of the oropharynx are generally accepted as a primary source of lung microbiota through dispersion and/or aspiration along the respiratory tract (12–16). Microbial diversity, or the richness and evenness of species, varies across oral habitats (17). The lowest relative microbial diversity in the oropharynx occurs at the buccal mucosa, where the genus Streptococcus is dominant (17–19). Higher intestinal microbial diversity is generally accepted to associate with physiological and immunological homeostasis (20). However, lower microbial diversity can be protective and a key feature of health in environments such as the female genital tract (21, 22). Although the impact of oral microbiota in lung disease is not clearly established, changes in the microbial ecology of the oropharynx can predict respiratory disease in early life (23). Despite a substantial role in human health (24), the oral microbiome in lung disease is incompletely understood. The oral microbiome in IPF has never been described, and associations between key features of the oral microbiome and fibrotic lung disease are unknown.

The CleanUP-IPF (Study of Clinical Efficacy of Antimicrobial Therapy Strategy Using Pragmatic Design in Idiopathic Pulmonary Fibrosis) clinical trial afforded an unprecedented opportunity to better understand the oral microbiome, IPF disease severity, and outcomes through the acquisition of buccal swabs (25). The CleanUP-IPF trial tested the impact of long-term antimicrobials on time to first nonelective respiratory hospitalization or death in IPF. Findings in previous smaller studies suggested improved outcomes with cotrimoxazole or azithromycin (26, 27). However, the results of CleanUP-IPF and a further study Effect of Co-trimoxazole (Trimethoprim-Sulfamethoxazole) vs Placebo on Death, Lung Transplant, or Hospital Admission in Patients With Moderate and Severe Idiopathic Pulmonary Fibrosis (EME-TIPAC) were not supportive of long-term antimicrobial use in all patients (25, 28). In both CleanUP-IPF and EME-TIPAC, study participants were not randomized based on a pretrial assessment of mucosal microbiota. An a priori approach to characterizing the microbiome may have enriched the study, given the plausible impact of antimicrobials on lung bacterial burden, diversity, and associated inflammatory signaling (6, 9, 11). Here, we hypothesized that lower microbial diversity and key buccal bacteria would independently associate with preserved lung function and improved survival in patients with IPF. To test our hypothesis, we performed a prospective observational study, using data from buccal swabs, microbial 16S rRNA amplicon, and metagenomic sequencing and correlated our findings with disease severity and death (25).

Online Methods

Study Design

CleanUP-IPF was an NHLBI-funded clinical trial (NCT02759120) (25). This was a multicenter randomized control trial with three cohorts assigned to receive 1) usual standard of care; 2) long-term cotrimoxazole; or 3) long-term doxycycline treatment. Buccal swabs were collected from patients who were enrolled and randomized. Institutional review board approval and written consent by participants were obtained at each of the participating sites. We determined associations between buccal microbiota and pulmonary physiology at baseline and then studied these associations with clinical outcomes (i.e., death).

16S rRNA Gene Amplicon Sequencing and Data Analysis

The University of California San Diego Microbiome Core performed sample extractions and library preparation using protocols and primers published on the Earth Microbiome Project website (https://earthmicrobiome.org). Data generated in this study have been deposited on Qiita under study ID 13516. See the online supplement for further details.

Shotgun Metagenomics Sample Preparation and Analysis

Extracted DNA from representative samples were sequenced on the Illumina NovaSeq (Advanced Genomics Core, University of Michigan) in a 151-bp paired-end format with 100 million reads per sample target. See the online supplement for further details.

Statistical Analysis

In brief, we visualized the community composition of clinical specimens using principal component analysis (PCA) and interrogated which taxa drove clustering via biplot analysis, as described previously (6, 8, 10). We determined statistical significance in community composition comparisons using the adonis function of the vegan package, which performs permutational multivariate ANOVA (PERMANOVA). Categorical variables were summarized as percentages and compared with the χ² test. Continuous data were summarized as means (with SD) and compared with two-sample t test, ANOVA, or summarized as medians (with interquartile range) and compared with the Wilcoxon rank-sum test with multiple comparisons corrected for by the Benjamini-Hochberg procedure. To evaluate the association of buccal taxa with physiological variables, we used univariable and multivariable linear regression analyses. Covariates that could plausibly influence microbiota and lung function were adjusted for in the multivariable linear regression model and included: age, sex, smoking history, coronary artery disease history, and antifibrotic use (29–36). For the survival analysis, we initially used Cox additive models and penalized splines. The models were adjusted for age, sex, smoking history, baseline FVC% predicted, and Dl_CO% predicted, coronary artery disease history, and antifibrotic use. Plots from the additive Cox model were visualized to identify potential inflection points where the slope changes and used to categorize patients and generate Kaplan-Meier curves. We initially examined this in the overall CleanUP-IPF cohort and then stratified by treatment assignment. We used the same approach for genus relative abundance that was significantly associated with disease severity as the primary independent variable in the Cox regression models. R version 4.2.2 (R Foundation for Statistical Computing) and vegan, survival, ggpubr, and pspline packages were used for the analyses.

Results

Characteristics of Patients

Buccal swabs were available for analysis in 511 CleanUP-IPF patients. All 511 baseline clinical characteristics are reported in the online supplement (Table E1). Baseline FVC data were available for 490 and baseline Dl_CO data for 483 patients with matched buccal microbiota data, and clinical characteristics are reported in Table 1. Buccal microbial communities were characterized by 16S rRNA amplicon sequencing of DNA extracted from buccal swabs. Total 16S rRNA counts and descriptive statistics are provided (Figures E1 and E2).

Table 1.

Baseline Characteristics of Patients

Characteristic	Overall
No. participants	490
Age, yr, mean (SD)	71 (7)
Male sex	385 (79)
Smoking history
Never-smoker	176 (36.0)
Former smoker	31 (63.4)
Current smoker	3 (0.6)
Race
White	490 (100)
Asian	0
African American	0
Other	0
Spirometry
FVC, L/min	2.76 (0.82)
FVC% predicted	70.7 (17.9)
FEV1, L/min	2.22 (0.62)
FEV1% predicted	77.9 (18.9)
DlCO, ml/min/mm Hg	11.67 (4.52)
DlCO% predicted	40.4 (15.4)
Comorbidities
Coronary artery disease	138 (28)
Sleep apnea	148 (31)
GERD	294 (61)
Diabetes	83 (17)
Pulmonary hypertension	38 (8)
Medications
Nintedanib	185 (38)
Pirfenidone	305 (62)

Definition of abbreviations: GERD =  gastroesophageal reflux disease history at time of study enrollment.

Data for continuous variables are given as median and interquartile range unless otherwise noted. Categorical variables are listed by number and percentage of total per group indicated. Percentage may not always add to 100% because of rounding.

Buccal Microbiota Associate with Disease Severity in IPF

Percentage predicted FVC (ppFVC) is a validated marker of disease severity in IPF (37). For our initial analysis of associations with disease severity, we stratified ppFVC by tertiles. We then examined associations between baseline ppFVC and buccal community composition. We found significant variation by ppFVC tertile groups using PCA and ordination (PERMANOVA P < 0.001) (Figure 1A). Pairwise analysis showed that ppFVC high tertile is different in buccal community composition to the intermediate (PERMANOVA P = 0.008) and low tertile groups (PERMANOVA P < 0.001). Statistical significance was maintained (P = 0.003) after adjusting for age, sex, study site, and smoking status. We used a biplot configuration to understand the directionality of relationships between ppFVC and buccal microbiota (Figure 1B), which highlights an association between the genus Streptococcus and higher ppFVC. At the genus level, the community composition displayed by rank abundance demonstrates that the buccal microbiome in IPF is dominated by Streptococcus, with a mean relative abundance across clinical samples of 49.43% (±24.30% SD) (Figure 1C). Amplicon sequence variant level ranked abundance is reported in the online supplement (Figure E3). In a univariate model, there was a positive correlation between Streptococcus abundance and ppFVC (P = 0.008; r = 0.12) (Figure 1D). This association persisted in a multivariate regression model (mean difference, 0.80; 95% confidence interval [CI], 0.16–1.43% predicted per 10% increase) (Table 2). To further evaluate taxon dominance, we examined the average maximum abundance of the most abundant genus within each ppFVC tertile. The high ppFVC tertile is significantly associated with a higher maximum average abundance compared with the intermediate (maximum abundance 59.13% ± 20.28% vs. 52.13% ± 9.52%; P = 0.002) and low tertiles (maximum abundance 59.13% ± 20.28% vs. 51.63% ± 19.87%; P = 0.002), supporting an association between genus dominance and higher ppFVC (Figure 1E). We report increased relative abundance of Streptococcus in the high FVC tertile compared with the low FVC tertile (53.93% ± 25.62% vs. 45.87% ± 24.27%; P = 0.008) and intermediate FVC tertile (53.93% ± 25.62% vs. 48.50% ± 22.35%; P = 0.04) (Figure E4).

Figure 1.

Streptococcus genus dominance of the buccal mucosa associates with preserved lung function and reduced disease severity in idiopathic pulmonary fibrosis (IPF). A total of 511 patients enrolled in the CleanUP-IPF (Study of Clinical Efficacy of Antimicrobial Therapy Strategy Using Pragmatic Design in IPF) trial provided buccal swabs for DNA extraction. Buccal microbial communities were characterized by 16 s rRNA sequencing. For initial evaluation, patients were stratified into groups based on percentage predicted FVC (ppFVC) tertiles. (A) Principal component analysis (PCA) and ordination of 16S rRNA gene data from patients with recorded baseline FVC (n = 490) demonstrates significant variation based on ppFVC tertiles by permutational multivariable ANOVA (PERMANOVA) (P = 0.0009). Data points are weighted by ppFVC with a range of minimum 22% to maximum 125% and colored by ppFVC tertile. (B) Biplot of genus ordination data demonstrating directionality of association between genera and FVC with the Streptococcus genus associating with higher FVC tertile clustering in the PCA. (C) Rank relative abundance of all clinical samples demonstrates domination of buccal communities by the genus Streptococcus, with other commonly reported buccal genera present. (D) Linear regression of Streptococcus genus and FVC (Spearmen regression line with 95% confidence intervals) with significant positive association. (E) Taxon domination is associated with higher FVC tertile cohort (pairwise Wilcoxon rank-sum with Benjamini Hochberg correction for multiple comparisons).

Table 2.

Associations of Buccal Microbiota Diversity and the Genus Streptococcus with Lung Function

Microbiota Variable	Mean Difference in ppFVC (95% CI)	P Value	Mean Difference in ppFEV1 (95% CI)	P Value	Mean Difference in ppDlCO (95% CI)	P Value
Shannon diversity index	−3.60 (−5.92 to −1.29)	0.002	−3.22 (−5.65 to −0.79)	0.01	−0.55 (−2.57 to 1.47)	0.59
Streptococcus relative abundance	0.80 (0.16 to 1.43)	0.01	0.64 (−0.03 to 1.31)	0.06	−0.24 (−0.80 to 0.31)	0.40
Bacterial burden	0.14 (−1.76 to 2.05)	0.88	0.07 (−1.92 to 2.07)	0.94	1.41 (−0.25 to 3.07)	0.10

Definition of abbreviations: CI = confidence interval; pp = percent predicted.

Linear regression models adjusted for age, sex, smoking history, coronary artery disease history, and antifibrotic usage. Results reported per 1-unit increment in Shannon diversity index and 10% increment in Streptococcus relative abundance. Results reported per 1-unit increment in log₁₀-transformed bacterial burden.

We next measured the buccal microbial diversity (α-diversity) using the Shannon diversity index (SDI). The SDI was negatively correlated with the Streptococcus genus (R = −0.78; P < 0.0001) (Figure 2A). We used a linear regression model to examine relationships between microbial diversity and ppFVC. In univariable analysis, there was a significant negative correlation between SDI and ppFVC (R² = −0.16; P = 0.0004) (Figure 2B). This association persisted in a multivariable regression model (mean difference in ppFVC of −3.60 per 1-unit increment in SDI, 95% CI, −5.92 to −1.29) (Table 2). There was a significant negative association between diversity and FEV₁ (Table 2). Buccal diversity was significantly reduced in the high ppFVC group compared with intermediate ppFVC (mean, 1.43 ± 0.64 vs. 1.69 ± 0.68; P = 0.002) and the low ppFVC group (mean, 1.43 ± 0.64 vs. 1.69 ± 0.67; P = 0.001) (Figure E5). We did not find a significant association between diversity and Dl_CO% predicted (Table 2 and Figure E6). We also assessed the clinical relevance of bacterial burden in buccal swabs, measured by droplet digital PCR of the 16S rRNA gene. We did not find an association between bacterial burden and ppFVC by univariable analysis (R = 0.01; P = 0.81) (Figure 2C) or in a linear regression model (mean difference, 0.14 per 1-unit increment in log₁₀ transformed bacterial burden; 95% CI, −1.76 to 2.05) (Table 2). However, there was a significant association between bacterial burden and SDI in buccal swabs, with increasing bacterial burden associated with lower SDI (Figure E7). We do not report statistically significant differences in SDI or buccal community composition by age, sex, smoking status, comorbidity, or ethnicity (Tables E2 and E3).

Figure 2.

Lower microbial diversity of the buccal mucosa associates with preserved lung function in idiopathic pulmonary fibrosis (IPF). A total of 511 patients enrolled in the CleanUP-IPF (Study of Clinical Efficacy of Antimicrobial Therapy Strategy Using Pragmatic Design in IPF) trial provided buccal swabs for DNA extraction. Buccal microbial communities were characterized by 16 s rRNA sequencing. A total of 490 patients had complete data available for analysis of associations between FVC and buccal microbial diversity. Microbial diversity was measured using the Shannon diversity index (SDI). Bacterial burden in swabs was calculated by droplet digital PCR of the 16S rRNA gene. (A) Strong negative correlation between Streptococcus genus relative abundance and microbial diversity, with increasing Streptococcus abundance associated with reduced SDI and higher percentage predicted FVC (ppFVC) (data points weighted by ppFVC with a range of minimum 22% to maximum 125% predicted) (Spearmen regression with line and 95% confidence intervals [CIs]). (B) Linear regression of SDI and ppFVC demonstrating a significant negative association (Spearmen regression with line and 95% CIs). (C) No association between bacterial burden in buccal swabs and ppFVC was observed (log₁₀ transformation of 16S rRNA copies per μl of buccal DNA) (Spearmen regression with line and 95% CIs).

Buccal Microbiota Predict Disease Progression and Death in IPF

We next determined the association between death and buccal microbiota in the usual care cohort (no long-term antimicrobial exposures). Event summary data are shown in Table 3. The range of time-to-censored event was 0.2–35.4 months. In the usual care cohort with available FVC and Dl_CO data points (n = 250), an additive Cox model showed an association between SDI and relative death rate that plateaued at an approximate SDI threshold of 1.0 unit (Figure 3A). We used Kaplan-Meier analysis to study this microbial diversity threshold (SDI < 1 or ⩾1). In this cohort, greater diversity was associated with a significantly greater risk of death than lower diversity (log-rank P = 0.042) (Figure 3B). In a multivariable Cox regression model, greater buccal microbial diversity was associated with a greater relative rate of death (hazard ratio [HR], 1.73 per 1-unit increment in SDI; 95% CI, 1.03–2.90) (Table 3). In the usual care cohort, we found a significant association with lower relative death rate and higher buccal Streptococcus abundance (Figure 3C), which associated with a Streptococcus abundance threshold of approximately 70%. Patients with buccal Streptococcus >70% relative abundance had significantly improved survival in the usual care cohort (log-rank P = 0.02) (Figure 3D). Using a multivariable Cox regression model, higher Streptococcus abundance was associated with a reduced relative rate of death (HR, 0.85; 95% CI, 0.73–0.99) in the usual care cohort (Table 3). There was no association between the relative rate of death and buccal bacterial burden in the usual cohort (HR, 0.84; 95% CI, 0.55–1.29) (Table E4). To identify the species of Streptococcus associated with these observations, we used shotgun metagenomics in a sample of patients (n = 20) (see methods section in the online supplement). The Streptococcus genus was confirmed as the most abundant (Figures E8–E10). At the species level, Streptococcus mitis dominated the buccal mucosa (Figure 3E).

Table 3.

Associations of Buccal Microbiota Diversity and the Genus Streptococcus with Survival

Cohort	Hazard Ratio (95% CI)	P Value
Shannon diversity index
Overall cohort (n = 483)	1.18 (0.82–1.68)	0.38
Usual care group (n = 250)	1.73 (1.03–2.90)	0.04
Doxycycline group (n  = 115)	0.54 (0.21–1.37)	0.19
Cotrimoxazole group (n = 118)	1.13 (0.56–2.28)	0.73
Streptococcus relative abundance
Overall cohort (n = 483)	0.94 (0.85–1.05)	0.27
Usual care group (n = 250)	0.85 (0.73–0.99)	0.04
Doxycycline group (n = 115)	1.12 (0.88–1.44)	0.36
Cotrimoxazole group (n = 118)	0.99 (0.83–1.19)	0.93

Definition of abbreviation: CI = confidence interval.

Cox regression models adjusted for age, sex, smoking history, coronary artery disease history, antifibrotic usage, and baseline FVC and Dl_CO. Results reported per 1-unit increment in Shannon diversity index and 10% increment in Streptococcus relative abundance. There were three participants in the usual care group, one in the doxycycline group, and three in the cotrimoxazole group missing baseline Dl_CO measurements. Bold values indicate statistical significance.

Figure 3.

Key features of the buccal microbiome predict survival in idiopathic pulmonary fibrosis (IPF). Buccal swabs from 253 patients enrolled in the CleanUP-IPF (Study of Clinical Efficacy of Antimicrobial Therapy Strategy Using Pragmatic Design in IPF) trial and not exposed to long-term antimicrobials. Buccal microbial communities were characterized by 16S rRNA sequencing, and microbial diversity was calculated by Shannon diversity index (SDI). (A) Additive Cox model plot of SDI demonstrates a clear association between lower diversity and reduced relative death rate, with a threshold at approximately 1 unit SDI (dashed red line). (B) Kaplan-Meier plots of SDI stratified by diversity threshold of ⩾1 or <1. (C) Additive Cox model plot showing relationship between relative abundance of Streptococcus and relative death rate demonstrating a threshold abundance (approximately 70% relative abundance) of Streptococcus. (D) Kaplan-Meier plots of buccal Streptococcus abundance stratified at 70% and mortality. (E) Shotgun metagenomics allowing for species-level identification of microbial taxa, with Streptococcus mitis species dominating buccal communities.

Clinical Outcomes and the Buccal Microbiome in Antimicrobial Treatment Cohorts

We next tested the association between key features of the buccal microbiome (microbial diversity and Streptococcus abundance) with survival in the long-term antimicrobial treatment arms of the trial. Here we show initially using PCA and ordination that there was no difference in community composition at trial enrollment between groups assigned different treatments (PERMANOVA P = 0.87) (Figure 4A) or by relevant clinical variables (Table E3). We report no difference in SDI between assigned treatment groups at baseline (usual care–assigned group SDI mean, 1.55 ± 0.59 SD, cotrimoxazole-assigned group SDI mean, 1.47 ± 0.59 SD, doxycycline-assigned group SDI mean, 1.53 ± 0.58 SD) (P = 0.43) (Figure 4B). We found no difference in the relative abundance of the Streptococcus genus across assigned treatment groups (usual care mean, 49.0 ± 24.0 SD, cotrimoxazole-assigned group mean, 50.9 ± 24.2 SD, doxycycline-assigned group mean, 48.1 ± 24.3 SD) (P = 0.61) (Figure 4C). In the long-term cotrimoxazole treatment cohort there was no significant association between buccal microbial diversity and survival by Kaplan-Meier studies (Figure 4D) or significant association between buccal diversity and relative rate of death in Cox regression models (HR, 1.13; 95% CI, 0.56–2.28) (Table 3). Similarly, there was no association between buccal abundance of Streptococcus and survival by Kaplan-Meier studies (Figure 4E) and survival in Cox regression models (HR, 0.99; 95% CI, 0.83–1.19) (Table 3). In the long-term doxycycline treatment cohort, no significant association was observed between buccal microbial diversity by KM studies (Figure 4F) or Cox regression models (HR, 0.54; 95% CI, 0.21–1.37) (Table 3). We also did not observe a significant association between buccal Streptococcus abundance and survival in the doxycycline-assigned cohort (Figure 4G and Table 3). Finally, we found no association between buccal bacterial burden and the relative rate of death in the doxycycline-assigned treatment group (HR, 1.29; 95% CI, 0.61–2.72) or the cotrimoxazole-assigned treatment group (HR, 0.81; 95% CI 0.45–1.48) in Cox regression models (Table E4). In summary, compared with the usual care group, associations between buccal microbiota and survival were attenuated in patients who received long-term antibiotics.

Figure 4.

Clinical outcomes and the buccal microbiome in assigned antimicrobial treatment cohorts. Buccal swabs from patients enrolled in the CleanUP-IPF (Study of Clinical Efficacy of Antimicrobial Therapy Strategy Using Pragmatic Design in IPF) trial and exposed to either long-term cotrimoxazole (n = 120) or doxycycline (n = 116). Buccal microbial communities were characterized by 16S rRNA sequencing, and microbial diversity was calculated by Shannon diversity index (SDI). (A) All patients at baseline trial enrollment when stratified by assigned treatment groups (i.e., usual care, doxycycline, or cotrimoxazole), principal component analysis (PCA), and ordination demonstrate no difference in overall community composition. (B) No difference in SDI across assigned treatment groups at baseline. (C) No difference in Streptococcus relative abundance across all treatment groups at baseline. (D) Kaplan-Meier plot of diversity stratified by threshold of ⩾1 SDI in the cotrimoxazole cohort. (E) Kaplan-Meier plot of Streptococcus abundance stratified by threshold of ⩾70% relative abundance in the cotrimoxazole cohort. (F) Kaplan-Meier plot of diversity stratified by threshold of ⩾1 SDI in the doxycycline cohort. (G) Kaplan-Meier plot of Streptococcus abundance stratified by threshold of ⩾70% relative abundance in the doxycycline cohort.

Discussion

In the CleanUP-IPF trial, a large-scale pragmatic multicenter study of patients with IPF, we found that lower buccal microbial diversity and commensal S. mitis dominance are associated with preserved lung function and increased survival in patients not exposed to long-term antibiotics. However, these relationships between buccal microbiota and IPF became dissociated after treatment with long-term antimicrobial therapy. The reasons for this are uncertain. We speculate that antimicrobial therapy altered microbiota, resulting in a loss of this relationship.

Microbial diversity is a key feature of microbial communities and linked to immune regulation (38). Lower gut microbial diversity occurs in advanced urbanized populations (39) and promotes poorer clinical outcomes across multiple diseases (2, 20). In IPF, lower microbial diversity of lung microbiota associates with increased alveolar fibrotic cytokines (6). However, our study identifies lower buccal diversity as protective and lowering the risk of mortality in IPF. It is not clear why contiguous mucosal surfaces of the respiratory tract would demonstrate varied associations between diversity and disease. Differences may reflect changes in fibrosis-related systemic inflammation and associated comorbidities or treatment. Oral dysbiosis is associated with changes in diversity and implicated in the pathogenesis of systemic disease and chronic lung disease (24, 40, 41). Our work supports further investigation of oral disorders such as periodontitis and their impact on oral and pulmonary microbial diversity and immune dysregulation in lung disease (42).

Our findings highlight key aspects of buccal microbial homeostasis as protective in IPF. Healthy humans’ micro-aspirate oropharyngeal bacteria sufficient to shape pulmonary microbial communities (15, 43). Pulmonary T-cell functionality can correlate with oral microbiota (44). In addition, lung microbial communities that resemble oral communities may lead to better clinical outcomes (45, 46). Thus, the oral microbiota play a critical role in shaping the lung microbiota and local immune tone in the lung. Notably, S. mitis, an important oral commensal, may help prevent oral colonization with potentially harmful bacteria (47). S. mitis inhibits pathogenic bacterial expansion through hydrogen peroxide production, stimulates epithelial cells to release antimicrobial peptides, and attenuates host inflammatory signaling (48–50). However, experimental oral aspiration with S. mitis spp can promote a persistent Th17 pulmonary environment, a cytokine system associated with fibrogenesis (16). Streptococcus produces toxins that are fibrogenic in the lung (16, 51). Previous studies using BAL fluid demonstrate increased relative abundance of Streptococcus operational taxonomic units predict disease progression and death (7). All these considerations would offer additional avenues for intervention in IPF. Loss of important oral commensals may reflect evolving periodontal or oral pathology, increased pulmonary micro-aspiration, or systemic inflammation. Our observations suggest the potential for a critical pathogenic oral and pulmonary interconnection for Streptococcus in IPF and the immune response. Randomization of patients to antimicrobial strategies based on an a priori assessment of microbial communities may facilitate trial enrichment and improved therapeutics.

The absence of any data pertaining to lifestyle factors and behaviors that impact oral microbiota or oral disease is a study limitation. The pragmatic design of CleanUP-IPF precluded BAL samples, disallowing any associative studies between habitats. We cannot rule out that IPF may increase the risk of micro-aspiration through progressive restrictive lung physiology. Dl_CO was not significantly associated with the oral microbiota. Dl_CO values were inconsistently collected, a function of pragmatic trial design, and the Dl_CO maneuver can be more difficult to perform with increasing disease severity. Bacterial burden in buccal swabs was also not associated with disease severity or outcomes. However, as a singular habitat measure, the buccal burden may not accurately reflect overall oral bacterial burden, which may be a crucial determinant of lung bacterial burden. Further studies characterizing the bacterial burden in multiple oral habitats are required. As Streptococcus abundance is robustly correlated with microbial diversity, the association between disease severity, Streptococcus abundance, and microbial diversity are linked and not independent. Although our approach to identify S. mitis was comprehensive, the number of samples studied was limited in size. Future work examining the presence of S. mitis and the functional implications for the oral microbiome are required. Key features of the oral microbiome are associated with disease severity and risk of death. However, the data suggest that other unknown factors are also contributing to the effect, supporting interconnections between oral microbiota, lung microbiota, and inflammation. Further study is required to address these questions. Last, associations with death in both antimicrobial treatment arms may be impacted by antimicrobial effects on host-mediated inflammation, and the unexplored gut microbiome could confound direct effects on buccal diversity and composition. The dissociation of buccal diversity and buccal S. mitis abundance with death in the antimicrobial treatment arms will require further study.

In conclusion, we report novel and intriguing associations between key features of the buccal microbiome and disease severity and death in a multicenter IPF clinical trial. Microbiota in the oral habitat may prove causal in lung disease, and directed oral therapies would be an attractive treatment strategy. Our work also highlights the need for greater consideration of antibiotic selection and more knowledge and targeted studies of host microbial interactions with antibiotic therapy. Predicting disease progression and severity through buccal swabs would represent a significant advance in our ability to care for patients with IPF. Future work ascertaining the relationship between the buccal microbiome and immune dysregulation, a critical driver of IPF pathophysiology, will enable the development of novel targeted immunomodulatory therapies.