Abstract:
BACKGROUND : Long-term azithromycin (AZM) treatment reduces the frequency of acute respiratory exacerbation in children and adolescents with HIV-associated chronic lung disease (HCLD). However, the impact of this treatment on the respiratory bacteriome is unknown. METHOD : African children with HCLD (defined as forced expiratory volume in 1 s z-score (FEV1z) less than − 1.0 with no reversibility) were enrolled in a placebo-controlled trial of once-weekly AZM given for 48-weeks (BREATHE trial). Sputum samples were collected at baseline, 48 weeks (end of treatment) and 72 weeks (6 months post-intervention in participants who reached this timepoint before trial conclusion). Sputum bacterial load and bacteriome profiles were determined using 16S rRNA gene qPCR and V4 region amplicon sequencing, respectively. The primary outcomes were within-participant and within-arm (AZM vs placebo) changes in the sputum bacteriome measured across baseline, 48 weeks and 72 weeks. Associations between clinical or socio-demographic factors and bacteriome profiles were also assessed using linear regression. RESULTS : In total, 347 participants (median age: 15.3 years, interquartile range [12.7–17.7]) were enrolled and randomised to AZM (173) or placebo (174). After 48 weeks, participants in the AZM arm had reduced sputum bacterial load vs placebo arm (16S rRNA copies/μl in log10, mean difference and 95% confidence interval [CI] of AZM vs placebo − 0.54 [− 0.71; − 0.36]). Shannon alpha diversity remained stable in the AZM arm but declined in the placebo arm between baseline and 48 weeks (3.03 vs. 2.80, p = 0.04, Wilcoxon paired test). Bacterial community structure changed in the AZM arm at 48 weeks compared with baseline (PERMANOVA test p = 0.003) but resolved at 72 weeks. The relative abundances of genera previously associated with HCLD decreased in the AZM arm at 48 weeks compared with baseline, including Haemophilus (17.9% vs. 25.8%, p < 0.05, ANCOM ω = 32) and Moraxella (1% vs. 1.9%, p < 0.05, ANCOM ω = 47). This reduction was sustained at 72 weeks relative to baseline. Lung function (FEV1z) was negatively associated with bacterial load (coefficient, [CI]: − 0.09 [− 0.16; − 0.02]) and positively associated with Shannon diversity (0.19 [0.12; 0.27]). The relative abundance of Neisseria (coefficient, [standard error]: (2.85, [0.7], q = 0.01), and Haemophilus (− 6.1, [1.2], q < 0.001) were positively and negatively associated with FEV1z, respectively. An increase in the relative abundance of Streptococcus from baseline to 48 weeks was associated with improvement in FEV1z (3.2 [1.11], q = 0.01) whilst an increase in Moraxella was associated with decline in FEV1z (-2.74 [0.74], q = 0.002). CONCLUSIONS : AZM treatment preserved sputum bacterial diversity and reduced the relative abundances of the HCLD-associated genera Haemophilus and Moraxella. These bacteriological effects were associated with improvement in lung function and may account for reduced respiratory exacerbations associated with AZM treatment of children with HCLD.
Description:
AVAILABILITY OF DATA AND MATERIAL : The 16S rRNA amplicon sequences, anonymised clinical, laboratory and sociodemographic
data and analysis codes that support the findings of this study
are available in Sequence Read Archive (accession number PRJNA 769290)
and GitHub repository (https:// github. com/ Regin aEsin amAbo tsi/ BREAT HE_
Sputum_ bacte riome_ analy sis).
ADDITIONAL FILE 1: FIGURE S1. A bar plot of the taxa and their relative abundance of the extraction and sequencing mock controls compared to manufacturer profiles. FIGURE S2. A scatterplot showing the correlation between samples repeated within a run (WR, n = 74) and between runs (BR, n=28). FIGURE S3. A scatterplot showing the spread of biological samples (n=960) and the negative controls (primestore, n=43) 16S copies vs final number of reads (A1 and A2), Shannon alpha diversity index (B1 and B2) and age of participant in years (C1 and C2). FIGURE S4. Ordination plots of showing the spread of biological samples (n=960) and the negative controls (primestore, n=43) coloured by their 16S copies. FIGURE S5. Ordination plots showing the spread of biological samples (n=960) and the negative controls (primestore, n=43) coloured by their number of reads. FIGURE S6. Ordination plots showing the spread of biological samples (n=960) and the negative controls (primestore, n=43) coloured by the age of the participant. FIGURE S7. Rarefaction curves showing number of ASVs detected and 16S copies of samples. FIGURE S8. Rarefaction curves showing number of ASVs detected and number of reads of samples. FIGURE S9. Bar plot showing the profiles of biological samples with <100 16S copies (n=2) in comparison to Primestores profiles (n=43). FIGURE S10. Bar plot showing the profiles of biological samples with >100 to <1000 16S copies (n=10) in comparison to Primestores profiles (n=43). FIGURE S11. Ordination plots showing the profiles of a subset of biological samples with low 16S copies and the negative controls. FIGURE S12. Ordination plots showing the profiles of a subset of biological samples with low reads and the negative controls. FIGURE S13. Ordination plots showing the spread of biological samples (n=960) and the negative controls (primestore, n=43) coloured by the run in which the sample was processed. FIGURE S14. Ordination plots showing the spread of biological samples (n=960) and the negative controls (primestore, n=43) coloured by the country of sampling. FIGURE S15. Ordination plots showing the spread of biological samples (n=960) and the negative controls (primestore, n=43) coloured by visit. FIGURE S16. Ordination plots showing the spread of biological samples (n=960) and the negative controls (primestore, n=43) coloured by the age at sampling. FIGURE S17. Output from decontamination analysis using the DECONTAM R package. FIGURE S18. Boxplot of Shannon alpha diversity index between trial arms at each visit (A) and between study visits in AZM (B) and Placebo (C) arms. FIGURE S19. Violin boxplot comparing two beta diversity metrics between samples collected from participants in the AZM arms at baseline and 48 weeks, 48 and 72 weeks and baseline and 72 weeks. FIGURE S20. Violin boxplot comparing two beta diversity metrics between samples collected from participants in the Placebo arms at baseline and 48 weeks, 48 and 72 weeks and baseline and 72 weeks. FIGURE S21. Principal Coordinates Analysis of Atchison (A) and Bray-Curtis (B) [on unrarefied ASV counts] distance matrixes between trial arms at each visit. FIGURE S22. Barplot of the relative abundances of the top 10 most prevalent phyla in all samples. FIGURE S23. Barplot of the relative abundances of the top 12 most prevalent genera in all samples. FIGURE S24. Heatmap displaying the q values of the genera detected as differentially abundant between AZM and placebo arms at 48 weeks by 10 statistical methods. FIGURE S25. Heatmap displaying the q values of the genera detected as differentially abundant within the AZM arm between baseline and 48-week samples by 10 methods. FIGURE S26. Heatmap displaying the q values of the genera detected as differentially abundant within the AZM arm between 48- and 72-week samples by 10 methods. TABLE S1. The taxonomy of the ASVs in the extraction and sequencing control. TABLE S2. List of 70 ASVs detected by the DECONTAM R package as potential contaminants based on comparison between biological samples and negative controls. TABLE S3. The association between bacterial load (16S rRNA copies) and selected variables using linear mixed effects modelling. TABLE S4. The association between Shannon diversity indices and selected variables using linear mixed effects modelling. TABLE S5. Results of differential abundance testing of bacterial taxa from AZM and Placebo samples from 48 weeks using 10 methods. TABLE S6. Results of differential abundance testing of bacterial taxa from AZM and Placebo samples from 72 weeks using DESeq2. TABLE S7. Results of differential abundance testing of bacterial taxa from AZM and Placebo samples from 72 weeks using Ancom-II. TABLE S8. Results of differential abundance testing of bacterial taxa from samples from the AZM arm at baseline and 48 72 weeks using 10 methods. TABLE S9. Results of differential abundance testing of bacterial taxa from samples from the AZM arm at 48 and 72 weeks using 10 methods. TABLE S10. Results of differential abundance testing of bacterial taxa from Placebo samples from 48 and 72 weeks using DESeq2. TABLE S11. Results of differential abundance testing of bacterial taxa from Placebo samples from baseline and 72 weeks using DESeq2. TABLE S12. Contributions of top genera to overall dissimilarity between AZM and Placebo arms at 48 weeks and, within the AZM arm, between Baseline and 48-week samples- SIMPER analysis. TABLE S13. Univariate linear regression analysis of withinparticipant Aitchison distance (outcome) and within-participant change in lung function metrics (FVCz and FEV1z) between visits.
