The Lung Microbiome Predicts Mortality and Response to Azithromycin in Lung Transplant Recipients with Chronic Rejection

Abstract

Rationale: Chronic lung allograft dysfunction (CLAD) is the leading cause of death after lung transplant, and azithromycin has variable efficacy in CLAD. The lung microbiome is a risk factor for developing CLAD, but the relationship between lung dysbiosis, pulmonary inflammation, and allograft dysfunction remains poorly understood. Whether lung microbiota predict outcomes or modify treatment response after CLAD is unknown. Objectives: To determine whether lung microbiota predict post-CLAD outcomes and clinical response to azithromycin. Methods: Retrospective cohort study using acellular BAL fluid prospectively collected from recipients of lung transplant within 90 days of CLAD onset. Lung microbiota were characterized using 16S rRNA gene sequencing and droplet digital PCR. In two additional cohorts, causal relationships of dysbiosis and inflammation were evaluated by comparing lung microbiota with CLAD-associated cytokines and measuring ex vivo P. aeruginosa growth in sterilized BAL fluid. Measurements and Main Results: Patients with higher bacterial burden had shorter post-CLAD survival, independent of CLAD phenotype, azithromycin treatment, and relevant covariates. Azithromycin treatment improved survival in patients with high bacterial burden but had negligible impact on patients with low or moderate burden. Lung bacterial burden was positively associated with CLAD-associated cytokines, and ex vivo growth of P. aeruginosa was augmented in BAL fluid from transplant recipients with CLAD. Conclusions: In recipients of lung transplants with chronic rejection, increased lung bacterial burden is an independent risk factor for mortality and predicts clinical response to azithromycin. Lung bacterial dysbiosis is associated with alveolar inflammation and may be promoted by underlying lung allograft dysfunction.

Keywords: chronic lung allograft dysfunction (CLAD); lung microbiome; lung transplantation.

Figure 1.

Conceptual model of relationships between lung dysbiosis, lung inflammation, and chronic lung allograft dysfunction (CLAD). The natural history of lung transplantation (black arrows) is the development of CLAD followed by disease progression resulting in death caused by progressive respiratory failure. Prior studies (gray arrows) have demonstrated that lung transplantation results in lung dysbiosis relative to healthy controls and that both lung dysbiosis and lung inflammation predict CLAD development. Additionally, it has been shown that azithromycin can halt the progressive lung function decline in at least some patients with CLAD. In the present study (blue arrows), we demonstrate that lung dysbiosis at CLAD onset predicts progression-free survival and post-CLAD mortality and moderates the effect of azithromycin on post-CLAD survival. We further show that lung dysbiosis may promote lung inflammation and that both CLAD and lung inflammation may potentiate lung dysbiosis.

Figure 2.

Increased lung bacterial burden predicts shorter survival and shorter progression-free survival among recipients of lung transplants with newly diagnosed chronic lung allograft dysfunction (CLAD). (A) Lung bacterial burden measured in acellular BAL samples (n = 76) collected within 90 days of CLAD onset, color-coded by bacterial burden. (B) Patients with the lowest bacterial burden (i.e., less than approximately 10,000 16S ribosomal RNA gene copies per milliliter of BAL fluid) had improved survival after CLAD onset as compared with patients with higher bacterial burden. (C) Similarly, patients with the lowest bacterial burden had improved progression-free survival after CLAD onset as compared with patients in the middle and highest bacterial burden tertiles. Hypothesis testing was performed using the log-rank test.

Figure 3.

Lung microbiota and post–chronic lung allograft dysfunction (CLAD) outcomes in patients with and without restrictive allograft syndrome (RAS)–like opacities. To determine whether the association of bacterial burden and post-CLAD survival differed between CLAD phenotypes, we stratified patients into those with or without computed tomography scan opacities suggestive of RAS. (A) Among patients without RAS-like opacities (n = 48), patients in the lowest bacterial burden tertile had better post-CLAD survival than those with higher bacterial burden. (B) In contrast, among patients with RAS-like opacities (n = 28), there was no relationship between lung bacterial burden and post-CLAD survival. Hypothesis testing was performed using the log-rank test.

Figure 4.

Comparison of alveolar neutrophilia, lung bacterial burden, and clinical culture results. Patients in the lowest and middle lung bacterial burden tertiles had similar levels of BAL neutrophilia, whereas patients with the highest bacterial burden had markedly higher BAL neutrophilia. It is important to note that BAL neutrophilia was not associated with BAL bacterial culture results (P = 0.84). Lines indicate medians and interquartile range. Hypothesis testing was performed using the Kruskal–Wallis test.

Figure 5.

Relationship between lung bacterial burden and clinical efficacy of azithromycin in chronic lung allograft dysfunction (CLAD). Leveraging temporal differences in clinical practices, we investigated whether the mortality risk associated with lung bacterial burden was modified by azithromycin (Azithro) treatment. For patients in the (A) lowest and (B) middle lung bacterial burden tertiles, treatment with azithromycin after CLAD onset had no significant relationship with post-CLAD survival. However, among patients in (C) the highest lung bacterial burden tertile, azithromycin was associated with a dramatic improvement in post-CLAD survival. Hypothesis testing was performed using the log-rank test.

Figure 6.

The lung microbiome community composition significantly differs between patients in the lowest, middle, and highest bacterial burden tertiles. The overall community composition for the acellular BAL specimens collected at chronic lung allograft dysfunction (CLAD) onset is presented in (A) a principal component analysis, with (B) a biplot to assist in interpretation. Patients who are in the lowest bacterial burden tertile, indicated in blue, tend to have communities dominated by genera that are commonly found in low-biomass specimens (e.g., Flavobacterium), whereas patients in the highest bacterial burden tertile, indicated in red, tend to have communities dominated by genera that are commonly associated with the oropharynx (e.g., Prevotella, Veillonella, and Streptococcus). Relative abundance comparison (mean ± SD), ranked by the most abundant taxa in (C) the lowest bacterial burden tertile and (D) the highest bacterial burden tertile further confirm this trend. After accounting for multiple comparisons, we found that Flavobacterium, an unclassified Verrucommicrobiaceae, and Pelomonas—taxa commonly observed in low-biomass specimens—significantly differed on the basis of lung bacterial burden tertile. *P ⩽ 0.05, **P ⩽ 0.01, and ***P ⩽ 0.001. Hypothesis testing was performed using PERMANOVA in (A) and mvabund in (A), (C), and (D). NOS = not otherwise specified; PC = principal component.

Figure 7.

To evaluate whether lung bacterial burden had a subclinical effect on alveolar immune signals, we identified patients who had low versus high lung bacterial burden from among a cohort of asymptomatic recipients of lung transplants with no evidence of rejection or infection undergoing per-protocol 1-year posttransplant surveillance bronchoscopy. It is important to note that these groups were similar with regard to underlying lung function and did not have a significant difference in BAL neutrophil percentage. BAL concentrations of chemokines/cytokines were measured using a high-sensitivity magnetic bead immunoassay. Even after accounting for multiple comparisons, the patients with higher bacterial burden had higher concentrations of (A) TNF-α, (B) IL-1α, (C) IL-1β, (D) IL-1Ra, (E) IL-8, and (F) macrophage inflammatory protein-1α/chemokine (C-C motif) ligand 3 (MIP-1α/CCL3). Lines indicate medians and interquartile range. Hypothesis testing was performed using Wilcoxon rank-sum test, with P values corrected to account for multiple comparisons with the Benjamini–Hochberg method. LLD = lower limit of detection; TNF-α = tumor necrosis factor α.

Figure 8.

We inoculated sterilized BAL fluid collected from 41 lung transplant bronchoscopies with 10⁶ colony-forming units (CFUs) of P. aeruginosa and measured the CFUs after 24 hours. (A) Bacterial growth was lowest in BAL collected during per-protocol surveillance (i.e., patients with normal posttransplant lung function, no pulmonary symptoms, and no evidence for infection or rejection) and higher in patients with respiratory infection and chronic lung allograft dysfunction (CLAD). (B) The BAL neutrophil percentage in these samples showed a similar pattern, with the highest neutrophil percentages observed in the specimens collected during respiratory infections and CLAD. (C) Bacterial growth was positively associated with alveolar neutrophilia in univariate analyses (Spearman’s ρ = 0.55, P < 0.001), but in multivariable testing that included the clinical status at the time of BAL testing, the significance of the correlation was attenuated (multivariable P = 0.14). Lines indicate medians and interquartile range. Hypothesis testing was performed using an ANOVA (on log-adjusted values), with multiple pairwise comparisons evaluated using the Tukey method in (A) and (B). In (C), the correlation between P. aeruginosa growth and BAL neutrophil percentage was evaluated using Spearman’s correlation and linear regression for univariate and multivariable analysis, respectively.