Improved gut microbiome recovery following drug therapy is linked to abundance and replication of probiotic strains

Abstract

Probiotics have been used for decades to alleviate the negative side-effects of oral antibiotics, but our mechanistic understanding on how they work is so far incomplete. Here, we performed a metagenomic analysis of the fecal microbiota in participants who underwent a 14-d Helicobacter pylori eradication therapy with or without consumption of a multi-strain probiotic intervention (L. paracasei CNCM I-1518, L. paracasei CNCM I-3689, L. rhamnosus CNCM I-3690, and four yogurt strains) in a randomized, double-blinded, controlled clinical trial. Using a strain-level analysis for detection and metagenomic determination of replication rate, ingested strains were detected and replicated transiently in fecal samples and in the gut during and following antibiotic administration. Consumption of the fermented milk product led to a significant, although modest, improvement in the recovery of microbiota composition. Stratification of participants into two groups based on the degree to which their microbiome recovered showed i) a higher fecal abundance of the probiotic L. paracasei and L. rhamnosus strains and ii) an elevated replication rate of one strain (L. paracasei CNCMI-1518) in the recovery group. Collectively, our findings show a small but measurable benefit of a fermented milk product on microbiome recovery after antibiotics, which was linked to the detection and replication of specific probiotic strains. Such functional insight can form the basis for the development of probiotic-based intervention aimed to protect gut microbiome from drug treatments.

Keywords: Antibiotics; L. paracasei CNCM I-1518; fermented milk product; gut microbiome recovery; probiotics; replication.

Figure 1.

Clinical study design.

Figure 2.

Strain-level detection and replication of product strains in the gut microbiota. a. Relative abundance of Test product strains were measured by mapping reads from each metagenome to the concatenated scaffolds of the seven product strains. Multi-mapped reads were excluded, and unique mapped reads were sum-scaled by the total reads in that sample to calculate percent abundance of product strains b. Percent unique mapped reads were further scaled by flow-cytometry based microbial cell counts for cytometry-scaled quantification of product strains. c. Rate of replication assessed by strain-level metagenomic estimation of replication rate (SMEG) scores d. Prevalence of strains detected based on replication rate. * denotes significant difference in test group between D14 and D28 timepoints (FDR < 0.05).

Figure 3.

Species composition and diversity across trial. a. Inverse Simpson’s Index. Both groups showed significant changes in species-based alpha diversity (Wilcoxon rank-sum, FDR: 0.003 to <0.0001) throughout the trial, albeit with no significant differences between Test and Control groups. b. Within-subject beta diversity to baseline (D0) composition at D42 (Bray-Curtis dissimilarity, Wilcoxon rank-sum, FDR = 0.03). c. Within-group (between all Test, or between all Control subjects) beta diversity differences (Bray-Curtis dissimilarity, Wilcoxon rank-sum, FDR = 0,041) d. Principal Coordinates Analysis (PCoA) of Bray-Curtis dissimilarity shows a taxonomic shift immediately post-therapy (D0-D14), followed by significant differences between Test and Control composition at D14 and D28 (PERMANOVA, FDR = 0.004, 0.008 respectively), with a gradual but incomplete return toward a baseline-like state in both groups at D42 (overall D0-D42 difference in PERMANOVA, FDR = 0.001) E. Smoothed ROC curves for the classification of intervention (Control versus Test) at D0, D14, D28 and D42, showing very strong classification performance at D14 and D28, and a reduced degree of classification at D42.

Figure 4.

Product consumption alters functional diversity and composition, through both the addition of product strains and the enrichment of function in the total gut microbiome. a. Inverse Simpson’s Index of metabolic pathways. Alpha-diversity was reduced after Hp therapy and (D0 versus D28/D42; Wilcoxon test, FDR = 0.02). b. Within-subject beta-diversity to baseline shows a decrease in dissimilarity (i.e., increase in similarity) at timepoint D14-D42, however these trends were not significantly different between Test and Control (Bray-Curtis dissimilarity, Wilcoxon rank-sum, FDR = 0.8–1.0). c. Within-group beta-diversity dissimilarities at each timepoint (PERMANOVA), showing metabolic composition to differ significantly at D14 (FDR = 0.004). d. Grey: decreases (left) and increases (right) in total pathway relative abundances due to impact of Hp eradication therapy, at D14 with reference to D0 (FDR < 0.05); Colored bars: changes significantly associated with Test product consumption, contributed both directly from the product species (L. paracasei: red;L. rhamnosus: orange; S. thermophilus: green), as well as from the wider gut microbiome (‘pooled effect’ in blue; all FDR < 0.05). Apparent decreases in pooled pathway abundance represent stochastic differences in starting abundance (D0) between Test and Control, rather than decreases as a result of product consumption. e. Significant differences in total functional pathway abundance associated with Test product species at timepoints D14 (L. paracasei: red;L. rhamnosus: orange; S. thermophilus: green; pooled community functions: blue). Similar effects (panels D & E) are seen at D28 (Fig S6).

Figure 5.

Identification and characterization of gut microbiota recovery following Hp eradication therapy. a. Dynamics of alpha-diversity in participants stratified into “recovery” and “non-recovery” groups based on ‘median boundary’ species-level Inverse Simpson Index at D42. b. Within-subject dissimilarity to baseline (Bray-Curtis) was significantly lower in the “recovery” group throughout the duration of the trial c. Microbial load (Log 10 cell count/g fecal samples, flow cytometry) was significantly lower in the non-recovery group. Amounts of caproate (d) and (e) valerate (Log10 mmol/g dry fecal weight) were more abundant in Recovery subjects at D28 & D42 and D14 & D42, respectively. Abbreviations: FDR: false discovery rate; **: FDR < 0.01; ****: FDR <0.0001.

Figure 6.

Recovery was linked to consumption of the Test product, abundance of probiotic strains, and in vivo replication of L. paracasei CNCM I-1518. a. Partitioning of recovery/non-recovery based on D42 alpha diversity shows unevenly distributed subjects across Test and Control groups, with a higher number of recovered subjects in the Test group although not significant (p = 0.1). b. Number of subjects from Test and Control in the recovery/non-recovery groups. c. Flow cytometry scaled abundances of Test product strains. L. paracasei and L. rhamnosus strains abundance was higher in recovery compared non-recovery groups. d. Flow cytometry scaled abundances of yogurt strains e. Replication rates of L. paracasei CNCM I-1518 was higher in the recovery (*: FDR < 0.05;**: FDR < 0.01; ****: FDR < 0.0001; p < 0.05: §).