Fig. 2: Gut microbiome characterization in COVit-2 trial participants.
a, Stool samples from 88 participants were collected at baseline (week (W) 0), during intervention (weeks 2 and 4; nicotinamide (NAM) or placebo) and at follow-up (week 6). Cohort 1 included 35 participants per group (NAM: 25 females, 10 males; placebo: 24 females, 11 males), and cohort 2 included 9 participants per group (NAM: 4 females, 5 males; placebo: 5 females, 4 males). Samples underwent 16S rRNA (n = 280) and shotgun metagenomics (n = 72) sequencing. b, α-diversity analysis (Shannon index at amplicon sequence variant level) of 16S rRNA data showed no significant (n.s.) differences across intervention groups or timepoints (two-sided Wilcoxon rank-sum test; likelihood ratio test on linear mixed-effect models; n per group is depicted below each box plot). Box plots show the median (centre line), interquartile range (IQR, box), 1.5 × IQR (whiskers) and outliers (points). c, Microbiota shifts (Aitchison distance, 16S rRNA) were examined using constraint-based principal coordinates analysis in participants with key COVID-19-related symptoms. Significant differences emerged between nicotinamide and placebo at week 2 and week 4 (n = 45 per intervention; PERMANOVA, R2 = 0.015, Fxy = 1.43, false discovery rate (FDR) = 0.002) but not at baseline (week 0) (NAM: n = 24; placebo: n = 23; PERMANOVA, R2 = 0.018, Fxy = 0.82, FDR = 0.99; Supplementary Fig. 4). Dots indicate individual samples, and arrows represent trajectories (baseline → week 2 → week 4). Ellipses show sample distributions per intervention group (solid line: 70% confidence; dashed line: 80% confidence; assuming multivariate normality). Black arrows show the impact of key COVID-19-related symptoms, intervention and age on microbiota dissimilarity, proportional to their correlation. Placebo and key COVID-19-related symptoms had similar effects. FDR: Benjamini–Hochberg-corrected P values. d, Variance partition analysis of the top 20 microbial genera that show highest variation at week 2 and week 4 in the 16S data (n = 67 per intervention). The bar plot shows the mean variance explained for the top 20 microbial genera, with variance attributed to covariates including age (light green), body mass index (BMI) (yellow), key COVID-19-related symptoms (red), fever at baseline (orange), sex (dark green), intervention (purple) and residuals (grey). Prefixes in genus labels denote higher taxonomic ranks: f_, family; p_, phylum. Only samples with at least 5,000 reads were included. For additional metagenomics-based variance partition analyses at the taxonomical level, see Supplementary Fig. 7 and Supplementary Table 21.
