Detection of bile acids in bronchoalveolar lavage fluid defines the inflammatory and microbial landscape of the lower airways in infants with cystic fibrosis

Abstract

Background: Cystic Fibrosis (CF) is a genetic condition characterized by neutrophilic inflammation and recurrent infection of the airways. How these processes are initiated and perpetuated in CF remains largely unknown. We have demonstrated a link between the intestinal microbiota-related metabolites bile acids (BA) and inflammation in the bronchoalveolar lavage fluid (BALF) from children with stable CF lung disease. To establish if BA indicate early pathological processes in CF lung disease, we combined targeted mass spectrometry and amplicon sequencing-based microbial characterization of 121 BALF specimens collected from 12-month old infants with CF enrolled in the COMBAT-CF study, a multicentre randomized placebo-controlled clinical trial comparing azithromycin versus placebo. We evaluated whether detection of BA in BALF is associated with the establishment of the inflammatory and microbial landscape of early CF lung disease, and whether azithromycin, a motilin agonist that has been demonstrated to reduce aspiration of gastric contents, alters the odds of detecting BA in BALF. We also explored how different prophylactic antibiotics regimens impact the early life BALF microbiota.

Results: Detection of BA in BALF was strongly associated with biomarkers of airway inflammation, more exacerbation episodes during the first year of life, increased use of oral antibiotics with prolonged treatment periods, a higher degree of structural lung damage, and distinct microbial profiles. Treatment with azithromycin, a motilin agonist, which has been reported to reduce aspiration of gastric contents, did not reduce the odds of detecting BA in BALF. Culture and molecular methods showed that azithromycin does not alter bacterial load or diversity in BALF. Conversely, penicillin-type prophylaxis reduced the odds of detecting BAs in BALF, which was associated with elevated levels of circulating biomarkers of cholestasis. We also observed that environmental factors such as penicillin-type prophylaxis or BAs detection were linked to distinct early microbial communities of the CF airways, which were associated with different inflammatory landscapes but not with structural lung damage.

Conclusions: Detection of BA in BALF portend early pathological events in CF lung disease. Benefits early in life associated with azithromycin are not linked to its antimicrobial properties. Video Abstract.

Keywords: Azithromycin; Bile acids; Cystic fibrosis; Gut-lung axis; Inflammation; Lung microbiota; Neutrophils.

Fig. 1

Detection of bile acids in BALF associates with airway inflammation and clinical outcomes. A-H Box plots overlaid with density curves (violin plots, blue) representing the proportion of neutrophils (A) and the levels of IL8 (B), the proportion of the lungs with structural disease (C), the number of pulmonary exacerbations (D), the number of oral/inhaled antibiotic rounds (E), the average number of days (Log10 transformed) treated with oral/inhaled antibiotics (F), with respect to the detection of BAs in BALF. Data in C was obtained from CT scan images evaluated with the PRAGMA-CF scoring system [18]. In D-F, the number of events during the first year of life is represented. The data represented in F includes the total duration of any treatment started before the collection of the BALF sample. Individual data points (red) with jitter are depicted on the top of each box plot. Groups were compared using the Wilcoxon rank-sum test: *, p < 0.05; **, p < 0.01; ***, p < 0.001; ****, p < 0.0001; n.s., p > 0.05

Fig. 2

Early microbial assemblies associate with contrasting inflammatory outcomes. A Principal component analysis shows the linear projection of the 16S-based compositional profiles onto the first two components of the model. Each sample is labelled based on the DMM model-based cluster membership. B The heatmap summarises the 16S taxonomic profiles grouped by Dirichlet component (metacommunity/cluster) (cluster 1, n = 39; cluster 2, n = 39; cluster 3, n = 35). Rows represent OTUs, which are ordered from top to bottom based on their contribution to each metacommunity. For simplification, only the top 10 OTUs are represented. The intensity of the colour in the heatmap is proportional to the square-root of the relative abundance of each OTU. For each Dirichlet component, narrow columns represent BALF samples, and wide columns represent the component mean abundance of each OTU. C-E Box plots overlaid with density curves (violin plots, blue) representing the proportion of neutrophils (C) and the levels of IL8 (D) in BALF, and the proportion of the lung with structural disease (E), with respect to the different metacommunities. Groups were compared using the Wilcoxon rank-sum test and p-values corrected using the Bonferroni method: *, p < 0.05; ***, p < 0.001; n.s., p > 0.05

Fig. 3

Detection of BA in BALF links to distinctive microbial profiles. Box plots overlaid with density curves (violin plots, blue) representing the bacterial burden (A) and bacterial diversity (B) with respect to the detection of BA in BALF. Individual data points (red) with jitter are represented on the top of each box plot. Notches in the boxplot represent 95% confidence interval for the median. Groups were compared using the Wilcoxon rank-sum test: ****, p < 0.0001. C. Differential abundance analysis between BALF samples with and without BA detection. Coefficients from the ANCOM-BC log linear model with pointwise 95% Bonferroni-corrected confidence intervals are plotted. Only statistically significant features are represented. Taxonomic entities enriched in BALF samples with or without BA detection are indicated with positive and negative fold change values respectively. Family P5D1-392 represents unculturable microorganisms from the order Lactobacillales

Fig. 4

Azithromycin does not influence detection of BA in BALF. A-F Marginal effect of treatment arm (A, C, E) or azithromycin dosage (B, D, F), on the odds of detecting BA with pointwise 95% confidence intervals, calculated from logistic regression models. C-F Predicted probabilities after controlling for pancreatic insufficiency (C-D) or p.F508del homozigosity (E–F)