Longitudinal microbial and molecular dynamics in the cystic fibrosis lung after Elexacaftor-Tezacaftor-Ivacaftor therapy

Abstract

Background: Cystic fibrosis (CF) is a genetic disorder causing poor mucociliary clearance in the airways and subsequent respiratory infection. The recently approved triple therapy Elexacaftor-Tezacaftor-Ivacaftor (ETI) has significantly improved lung function and decreased airway infection in persons with CF (pwCF). This improvement has been shown to occur rapidly, within the first few weeks of treatment. The effects of longer term ETI therapy on lung infection dynamics, however, remain mostly unknown.

Results: Here, we applied 16S rRNA gene amplicon sequencing, untargeted metabolomics, and neutral models to high-resolution, longitudinally collected sputum samples from pwCF on ETI therapy (162 samples, 7 patients) and compared to similarly collected data set from pwCF not taking ETI (630 samples, 9 patients). Because ETI reduces sputum production, samples were collected in freezers provided in the subject's homes at least 3 months after first taking ETI, with those on ETI collecting a sample approximately weekly. The lung function (%ppFEV1) of those in our longitudinal cohort significantly improved after ETI (6.91, SD = 7.74), indicating our study cohort was responsive to ETI. The daily variation of alpha- and beta-diversity of both the microbiome and metabolome was higher for those on ETI, reflecting a more dynamic microbial community and chemical environment during treatment. Four of the seven subjects on ETI were persistently infected with Pseudomonas or Burkholderia in their sputum throughout the sampling period while the total bacterial load significantly decreased with time (R = - 0.42, p = 0.01) in only one subject. The microbiome and metabolome dynamics on ETI were personalized, where some subjects had a progressive change with time on therapy, whereas others had no association with time on treatment. To further classify the augmented variance of the CF microbiome under therapy, we fit the microbiome data to a Hubbell neutral dynamics model in a patient-stratified manner and found that the subjects on ETI had better fit to a neutral model.

Conclusion: This study shows that the longitudinal microbiology and chemistry in airway secretions from subjects on ETI has become more dynamic and neutral and that after the initial improvement in lung function, many are still persistently infected with CF pathogens.

Keywords: Cystic fibrosis; Elexacaftor–Tezacaftor–Ivacaftor; Lung pathogens; Metabolome; Microbiome; Neutral models; Pseudomonas aeruginosa.

Fig. 1

Sampling collection data through time and lung function response to ETI treatment among participants. A Schematic of longitudinal samples collected for this study. Each black line represents a sample collection day. Black arrowheads represent the day in which subjects started on ETI treatment. B Lung function variation for highest ppFEV1% predicted pre- and post-ETI treatment per subject on each group and the absolute delta variation in the change in ppFEV1% predicted per subject post-ETI treatment within each group. The significance was determined through the DM t-test and Welch’s t-test respectively. C Lung function as ppFEV1% predicted of individual subjects through time (days) before and after ETI. Colors represent the treatment status (pre- and post-ETI) in which lung function was recorded and its corresponding significance (T-test) between the two different periods

Fig. 2

The average microbiome and metabolome diversity of subjects in this study. A Average change in the alpha-diversity (Shannon index) of lung microbiome and metabolome of pwCF off/on ETI treatment per day. Square shapes represent individuals who did not take ETI, while circular shapes are associated with those who did take ETI. B Average of the beta-diversity of the lung microbiome and metabolome of pwCF before and after ETI per day. Weighted UniFrac distance was used for microbiome data and Bray–Curtis dissimilarity for metabolome data. P-values obtained from the Wilcoxon test are displayed. Scatterplots of the predicted vs. observed C microbiome and D metabolome association with the time in days since pwCF started on ETI treatment. This data was obtained from RF regression analysis of each subject as well as their Pearson correlation tests

Fig. 3

Microbiome dynamics of subjects on ETI therapy through time of sampling. A Bar plots representing the microbiome’s relative abundance of longitudinally collected sputum samples from seven subjects on ETI. The sample collection time point (beginning and end of days) per subject is displayed, while gaps across sampling are not shown. The results of bacterial cultures in the clinics are shown for P. aeruginosa and S. aureus. The genus level and its classification as a CF classic pathogen as well as their oxygen tolerance are also presented. B Line-dot plots represent the pathogen:anaerobe log-ratios and C log rRNA gene copies in the sputum of each subject on ETI treatment collected through time.

Fig. 4

Metabolome dynamics of subjects on ETI therapy through time of sampling. A Bar plots representing the metabolome’s normalized peak area of longitudinally collected sputum samples from seven subjects on ETI. The sample collection time point (beginning and end of days) per subject is displayed, while gaps across sampling are not shown. B Line plots represent the GLP:small peptide log-ratios. C Molecular networking displaying Pseudomonas-like molecules from sputum samples of subjects on ETI. Nodes in yellow denote those molecular features annotated by GNPS. D Scatter plot representing the molecular dynamics between selected Pseudomonas metabolites, the time in days since subjects 011 and 015 started on ETI, while the size represents the relative abundance of Pseudomonas at the time of sampling over time

Fig. 5

Modulator therapy is associated with increased community neutrality and immigration. A Model comparison between a simplified neutrality model with dispersal limitation and a stochastic binomial distribution. B–D Fitting statistics of stratified 16S data to neutrality model. E Estimated dispersal in communities with and without modulator therapy. F Time evolution of predicted dispersal in communities with and without modulator therapy, 0.95 confidence interval of regression is plotted