Metagenomic Characterization of Gut Microbiota of Carriers of Extended-Spectrum Beta-Lactamase or Carbapenemase-Producing Enterobacteriaceae Following Treatment with Oral Antibiotics and Fecal Microbiota Transplantation: Results from a Multicenter Randomized Trial

Background: The R-GNOSIS (Resistance in Gram-Negative Organisms: Studying Intervention Strategies) WP3 study was the first multicenter randomized clinical trial systematically investigating fecal microbiota transplantation (FMT) for intestinal decolonization of extended-spectrum beta-lactamase-producing Enterobacteriaceae (ESBL-E) or carbapenemase-producing Enterobacteriaceae (CPE). Here, we characterized the temporal dynamics of fecal microbiota changes in a sub-cohort of the R-GNOSIS WP3 participants before and after antibiotics/FMT using whole metagenome shotgun sequencing. Methods: We sequenced fecal DNA obtained from 16 ESBL-E/CPE carriers having received oral colistin/neomycin followed by FMT and their corresponding seven donors. Ten treatment-naïve controls from the same trial were included. Fecal samples were collected at baseline (V0), after antibiotics but before FMT (V2) and three times after FMT (V3, V4 and V5). Results: Antibiotic treatment transiently decreased species richness and diversity and increased the abundance of antibiotic resistance determinants (ARDs). Bifidobacterium species, together with butyrate- and propionate-producing species from Lachnospiraceae and Ruminococcaceae families were significantly enriched in post-FMT microbiota of treated carriers. After FMT, the proportion of Enterobacteriaceae was lower compared to baseline but without statistical significance. Conclusions: Combined antibiotic and FMT treatment resulted in enrichment of species that are likely to limit the gut colonization by ESBL-E/CPE.

For four individuals of the FMT-treated group (Pa-R14, TA-R1, Ut-R4 and Ut-R5) and two from the treatment-naïve group (Ut-R3 and TA-R2), the sample from time point V5 was missing. Feces were also not sampled from TA-R2 at V4.
For donors, we sequenced DNA from aliquots of native feces and/or of FMT preparation.

Antibiotic treatment and fecal microbiota transplantation
Antibiotic treatment consisted of administration of colistin sulphate (2 million international units 4×/day per os) and neomycin sulphate tablets (350 mg of neomycin base 4×/day per os) for 5 days. On the second calendar day after the end of antibiotic treatment, FMT was performed without prior bowel lavage. Two centers (Ge, Pa) used oral, capsulized FMT (15 capsules on two consecutive days), the other two centers (Ut, TA) performed FMT via nasogastric tube (80 mL of FMT preparation on a single day).

Preparation and storage of samples from ESBL-E/CPE carriers
Stools from carriers were collected at the study sites or by the patients at home, conserved at room temperature and transported to the microbiology lab in maximum 72 hours from their emission. Feces were homogenized with a BagMixer 400P (Interscience, SaintNom-la-Bretèche, France), 1/10 diluted in saline and frozen in 500-µL aliquots at -80°C.

Donor sample preparation and storage
Fecal suspensions for nasogastric tube procedure were prepared as follows: homogenized stool was suspended in sterile non-bacteriostatic normal saline (5 mL per g of stool), filtered and centrifuged. Fecal concentrate was suspended in one-third the original volume in saline containing 10% glycerol to obtain a final volume of about 80 mL per 40 g of stool. For capsule preparation, fecal concentrate, obtained from the initial stool suspension (6 mL saline per g of stool), was suspended in one-tenth the original volume in saline containing 80% glycerol. Both types of fecal suspensions and native stools were kept frozen at -80 °C. For details see Huttner et al. 2019 [1].

DNA extraction from stool samples and sequencing
DNA extraction was performed as described by Frossard et al. [2] with some modifications. Briefly, 200 µL of frozen fecal suspension or native stools and 400 µL Buffer GT (RBC Bioscience, New Taipei City, Taiwan) were added to a Nucleospin Bead Tube containing ceramic beads (Macherey-Nagel, Düren, Germany). After shaking for 20 min at maximum speed on a Vortex-Genie 2, RNA was removed by incubating the mixture for 5 min with 1 µL of 50 mg/mL RNAse A (Roche, Basel, Switzerland). The lysate was then centrifuged for 1 min at 11,000 × g and 400 µL of supernatant loaded into a MagCore Sample Tube of a MagCore Genomic DNA Tissue Kit (RBC Bioscience). DNA extraction was then performed on a MagCore HF16 Automated Nucleic Acid Extractor (RBC Bioscience) using the program 401. DNA was eluted in 100 µL of Tris 10 mM pH8. The concentration of DNA was measured by QuBit dsDNA BR Assay Kit (Life Technologies, Carlsbad, CA, USA). DNA libraries were prepared with NextSeq 500 High-Output v2 kit (Illumina, San Diego, CA, USA) and sequenced (2 × 150) on a NextSeq 500 (Illumina) at LGC Genomics GmbH (Berlin, Germany). In total, we sequenced 152 samples: 21 from 7 donors, 76 from the 16 FMT-treated carriers, 47 from the 10 treatment-naïve carriers and 8 negative extraction controls (DNA extracts obtained by adding no sample).

Pre-processing, taxonomic classification and relative abundance computation
The quality of raw reads was inspected with the FastQC tool (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Raw reads were adapter-trimmed and scanned for quality filtering with Trimmomatic v0.36 [3] (sliding window size = 20, average quality = 28, minimum length = 100). Duplicated reads were removed with in-house perl scripts and those mapping by Kraken2 [4] to human genome (GRCh38.p7) were removed. The remaining reads were then assigned to bacteria, viruses and fungi by Kraken2 using a confidence score of 80%. The reference genome database was constructed as described by Kirstahler [5]. We added to the reference database 367 genomes available (as of September 2018) from Gastrointestinal Human Microbiome Project (GI HMP) (The Human Microbiome Project [6]). In total, 13,537 genomes were used: 5949 from bacteria, 7532 from viruses and 56 from fungi. The relative abundance of reads mapping to a given bacterial species was expressed as a percentage of the total number of reads assigned to bacterial phyla. The relative abundance of viral families and fungal genera was computed by taking into account the total number of microbial (bacterial, viral and fungal) reads.

Sequencing data availability
Quality-filtered dereplicated sequencing reads were deposited as FASTQ files at the European Nucleotide Archive (ENA) under the project PRJEB35816. Reads assigned to human genome (GRCh38.p7) by Kraken2 were removed before data submission.
For sequences with multiple best hits to ARDs, we kept the hit corresponding to the reference sequence with the highest number of matches for the entire dataset. To compute the ARD content of a sample, the sum of read counts mapping to a ResFinder gene or antimicrobial class was divided by the sum of reads mapping to the 16S rRNA gene. The obtained ratios were multiplied by 1000. Prediction of ESBL-E/CPE carriage was performed by querying the ResFinder databse for blaCTX-M, blaNDM, blaOXA, blaSHV, blaTEM and blaKPC genes.

Filtering of bacterial species
First, we filtered out the reads corresponding to bacterial species that had higher relative abundance in negative extraction controls (n = 8) than in test samples (n = 144). Those species were likely to result from contaminants present in DNA extraction reagents or from sample handling. We then removed species with a mean relative abundance < 0.001%, resulting in a dataset containing 402 species.

Statistical analyses
Sample selection. For donors, we only analyzed samples that were used for FMT transplantation (via oral capsule or nasogastric route).

Pairwise comparisons of global microbiota profiles.
Bacterial communities were tested for similarities and differences in relation to predictor variables with permutational analysis of variance (pairwise PERMANOVA, with 9999 permutations, unrestricted permutation of raw data and Type III sums of squares) based on the Bray-Curtis similarity matrices of square-root transformed species relative abundance. PERMANOVA tests were performed in PRIMER v7 (PRIMER-E Ltd, Plymouth, UK).
The feces of one carrier (TA-R2) was not collected at V4 and therefore all samples from this individual were excluded when comparing decolonized and persistently colonized groups. TA-R2, initially assigned to intervention group after randomization, received no treatments; this subject was therefore re-assigned to the treatment-naïve group.
Ecological indexes. Before alpha-diversity analyses (species diversity and richness), the counts of reads classified at the species level were rarefied to 40,000 with the rarefy function in R (vegan package). Shannon diversity was computed with the R diversity function (vegan). For calculating richness, a species was considered present in a sample if represented by at least one read.

Identification of differentially abundant species.
We performed two-sided Wilcoxon signed rank to evaluate the significance of intra-individual changes of species abundance between different time points, otherwise two-sided Wilcoxon rank sum test was used. All tests were done in R software.

Species shared with the donor
To compute the number of species shared between an FMT-treated carrier of ESBL-E/CPE at a given time point and a corresponding donor, we selected species with a relative abundance > 0.1% in both the carrier and the donor. Species shared between carriers and donors were identified by using the venn function implemented in R package gplots v3.0.1. We then determined the frequency of each shared species over the total number of patients analyzed at each time point. Of note, at V5, stools were collected from 12 out of 16 FMT-treated carriers. Figure S1. The abundance of ARDs from different anitimicrobial classes. Boxplots report the counts of reads assigned to a given ARD class normalized to 1000 16S rRNA gene reads counts. Figure S2. Differentially abundant species between FMT-treated and treatment-naïve carriers. For each of the 60 selected species, we report the corresponding phylum and family (middle), the significance of differences in the relative abundance between FMT-treated and treatmentnaïve carriers at a given time point (left) and the relative abundance in donor, FMT-treated and treatment-naïve carriers (right). We selected species associated with a significant Wilcoxon rank sum test (p < 0.05), a fold change ≥ 1.5 in the relative abundance and a mean relative abundance ≥ 0.1% in at least one of the compared groups. Significant differences (p < 0.05) associated with a ≥ 1.5 fold change in the relative abundance are represented as black shaded cells; white cells denote other cases. Relative abundances are square-root-transformed and color-scaled as indicated in the legend.   Tables  Table S1. Summary table describing