Microbiota restoration reduces antibiotic-resistant bacteria gut colonization in patients with recurrent Clostridioides difficile infection from the open-label PUNCH CD study

Abstract

Background: Once antibiotic-resistant bacteria become established within the gut microbiota, they can cause infections in the host and be transmitted to other people and the environment. Currently, there are no effective modalities for decreasing or preventing colonization by antibiotic-resistant bacteria. Intestinal microbiota restoration can prevent Clostridioides difficile infection (CDI) recurrences. Another potential application of microbiota restoration is suppression of non-C. difficile multidrug-resistant bacteria and overall decrease in the abundance of antibiotic resistance genes (the resistome) within the gut microbiota. This study characterizes the effects of RBX2660, a microbiota-based investigational therapeutic, on the composition and abundance of the gut microbiota and resistome, as well as multidrug-resistant organism carriage, after delivery to patients suffering from recurrent CDI.

Methods: An open-label, multi-center clinical trial in 11 centers in the USA for the safety and efficacy of RBX2660 on recurrent CDI was conducted. Fecal specimens from 29 of these subjects with recurrent CDI who received either one (N = 16) or two doses of RBX2660 (N = 13) were analyzed secondarily. Stool samples were collected prior to and at intervals up to 6 months post-therapy and analyzed in three ways: (1) 16S rRNA gene sequencing for microbiota taxonomic composition, (2) whole metagenome shotgun sequencing for functional pathways and antibiotic resistome content, and (3) selective and differential bacterial culturing followed by isolate genome sequencing to longitudinally track multidrug-resistant organisms.

Results: Successful prevention of CDI recurrence with RBX2660 correlated with taxonomic convergence of patient microbiota to the donor microbiota as measured by weighted UniFrac distance. RBX2660 dramatically reduced the abundance of antibiotic-resistant Enterobacteriaceae in the 2 months after administration. Fecal antibiotic resistance gene carriage decreased in direct relationship to the degree to which donor microbiota engrafted.

Conclusions: Microbiota-based therapeutics reduce resistance gene abundance and resistant organisms in the recipient gut microbiome. This approach could potentially reduce the risk of infections caused by resistant organisms within the patient and the transfer of resistance genes or pathogens to others.

Trial registration: ClinicalTrials.gov, NCT01925417 ; registered on August 19, 2013.

Keywords: Antibiotic resistance; Clostridioides difficile; Fecal microbiota transplantation; Metagenomics; Microbiome; Multidrug resistance.

Conflict of interest statement

Rebiotix, Inc. provided access to study specimens and data and reviewed the manuscript prior to submission, but was not involved in this study’s design, specimen processing, data analysis, or interpretation. Washington University investigators had the final say on published content. Erik R. Dubberke is a consultant for Sanofi, Pfizer, Synthetic Biologics, BioK+, and Rebiotix with grants from Pfizer. The remaining authors declare no other competing interests.

Figures

Fig. 1

Sampling schematic. Patients were given RBX2660 (green square) after vancomycin oral therapy (left panel). Stools (labeled as maroon circles) provided were sequenced and used for subsequent analyses. If a patient had CDI recurrence (red triangle), they were offered a second dose of RBX2660 (green square) with subsequent stools provided after the second study drug (right panel). Any antibiotic treatment during the trial is labeled as yellow diamonds. Patient IDs colored red failed first treatment and received antibiotics or second dose and constitute the RI group (n = 17). Patients who had no recurrence of symptoms or received antibiotics were considered successes (SI group, n = 12). All subsequent figures utilize data after the first dose. Data after second RBX2660 is used only for Figs. 5a–e and 6d. Three stool samples that failed sequencing were excluded from this figure and downstream analyses

Fig. 2

Microbiota composition similarity to the donor at 7 days is predictive of treatment outcome. The donor and recipient microbiota compositions were assessed via 16S rRNA gene sequencing followed by DADA2, and their similarity to the donor product was quantified by weighted UniFrac at each timepoint. a Gray lines represent individuals successfully treated with one administration (SI group) while red lines are patients who needed further treatment (RI group, a second product or antibiotics). N = 28 patients and 130 samples. b Plot demonstrating average distance from the donor at timepoints 0 and 7 days after treatment. N = 28 total patients and 44 samples. Box subsumes 75% of the data with a horizontal bar at the median. *p < 0.05, Mann-Whitney U test

Fig. 3

Taxonomy and microbial functional pathways converge after therapy receipt. a, b Principal component analysis (PCA) of patient microbiome taxonomic composition from 16S data (a) and of functional pathway abundances from whole metagenomic sequencing (b) in the SI group. Each colored dot represents an individual fecal sample after the first intervention with the circle representing 95% confidence interval with non-intersecting circles therefore statistically significant. Panel a shows 96 samples from all twelve patients with successful treatment and all four donors, while panel b shows 52 samples from eight successful patients and four donors (all of those who passed shotgun sequencing quality filters). c PCA from timepoint 7 samples after first study treatment only, colored by the SI or RI group (46 samples from all donors and all patients with day 7 samples; patient N = 25; donor N = 4). Each sample is connected to the centroid of its outcome group by a segment of the same color. d Taxonomy biplot shows the vectors of influence from taxa in distinguishing day 7 samples. The input samples, axes, and origin are the same as in c

Fig. 4

Taxa significantly associated with distance from the donor and successful response to RBX2660. A heatmap demonstrating the relative abundance over time after first RBX2660 is shown for donors and the SI group. These taxa are the top 11 identified by the PCA in Fig. 3d as significantly associated with successful treatment. Dark blue corresponds to 0.001% relative abundance with lighter blue 0.1% relative abundance. Each column represents a sample from a patient over time from left to right with donor samples at the right. N = 109 samples from 12 subjects from the SI group and 4 donors

Fig. 5

Antibiotic-resistant organisms cultured from patient and donor stools and the corresponding ASVs from species were tracked over time. a–e Antibiotic susceptibility profiles for each cultured organism from any sample from donor and patient with the corresponding phylogenetic tree. All breakpoints in antibiotic concentration were determined by CLSI 2016 criteria. Taxonomic labels are derived from DADA2 ASV assignments, with Enterobacter being further specified from family level based on metaphlan2 and MALDI-TOF taxonomy assignments. The designation A indicates recipient and D indicates donor. The following number indicates the study ID number followed by timepoint of isolation. A and E connote single colonies on separate plates. f–j Each of the ASVs corresponding to the species in a–e are shown in relative abundance over time in 131 fecal samples from 28 patients and 4 donors after the first study treatment. *p < 0.05, **p < 0.01 for relative abundance differences 7 days after therapy between SI and RI groups using Mann-Whitney U test. TMP-SMX, trimethoprim-sulfamethoxazole

Fig. 6

Antibiotic resistance gene abundance correlates with distance from the donor. a ARGs were quantified in metagenomic sequences (N = 21 patients and 4 donors) and summarized by mechanism. All ARG counts were transformed by log (ARG + 1) for visibility. b Two gene families within the β-lactamase class show opposite trajectories (N = 21 patients and 4 donors). c Patient-origin ARGs shown over time after RBX2660. d ARG abundance is plotted versus 1-(distance from donor) using weighted UniFrac. A generalized mixed effects log normal regression model of the formula ARGs ~ DFD + (1| PatientID) is shown, where DFD was significantly predictive of and correlated with ARG count compared to the null model (Chisq = 72.28, d.f.(full) = 1, p < 2.2 × 10−16). For c and d, all patients of both outcome groups were included for 153 total samples with patient N = 25 and donor N = 4. a–c Significance was determined by pairwise Wilcoxon with Benjamini-Hochberg correction. *p < 0.05; **p < 0.001