Microbiota as a biomarker of tuberculosis treatment response: association with clinical history, inflammation and outcome

Abstract Tuberculosis requires prolonged multidrug therapy. Whether this disrupts the microbiota in ways reflecting previous treatment exposure and outcome is unknown. We performed 16S rRNA gene sequencing on sputum and stool from 107 people with tuberculosis (cases) in a longitudinal cohort, sampled at baseline, month 2, end-of-treatment (month 6), and months 12 and 18, and ≤2 close contacts per case (n=180) at baseline, months 6 and 18. Serum cytokines were quantified in cases (baseline, months 6 and 18) and contacts (baseline). Treatment outcome was defined by month 2 culture-conversion and month 6 culture/smear status. Longitudinal microbial time-series were clustered using LoClust and linked to clinical data. Cases' α-diversity declined initially, recovering by month 6 in sputum and month 12 in stool. Stool β-diversity reductions, driven by Dialister-enrichment and Clostridiales-depletion, persisted to month 18; close contacts remained stable throughout. Unfavourable outcomes showed sputum Pseudomonas- and Mycobacterium gordonae-enrichment and Porphyromonas endodontalis-depletion, and stool Clostridiales-enrichment and Dialister-depletion. Clustering identified five α-diversity trajectories, with declining-diversity trajectories enriched for previous TB. Microbiota-inflammation associations persisted beyond treatment, with Clostridiales independently predicting IP-10 and TNF-α, and Dialister predicting C-reactive protein. Microbiota recovery trajectories are linked to previous tuberculosis, and specific taxa to outcome and inflammation, highlighting biomarker potential.