EHS
EHS

Moderate and transient impact of antibiotic use on the gut microbiota in a rural Vietnamese cohort


Study cohort, study design and fecal sample collection

This study was conducted within the context of an existing prospective household-based community cohort in Ha Nam in northern Vietnam, which was established to quantify the burden of influenza and to gain insights into the influenza virus transmission in a tropical setting. A full description of the demographics and lifestyle of the Ha Nam cohort has been published previously30.

To examine the impact of antibiotic use on shifts in the gut microbiota of residents of this community, we conducted a longitudinal study over a period of six months from November 2014 to June 2015. The study design encompassed weekly interviews to collect information on health issues and consumption of antibiotic use in the preceding week and monthly fecal samples for microbial profiling. The interviews included questions to indicate the indications for which antibiotics were used in the preceding week (including having respiratory symptoms (cough, breathless, sore throat)).

From 80 households (HHs), we selected 11 HHs in which at least one household member used antibiotics during the six months follow-up. In total this resulted in 41 participants of whom 32 used at least one course of antibiotics.

In addition, we randomly selected a control group from 11 HHs (22 participants) where none of the household members used antibiotics during the study period. We used randomizer (www.randomizer.org) for the selection.

To investigate the effects of antibiotic use, samples were divided into one of three categories: samples collected within two weeks after last antibiotic use (E2), samples collected within two to four weeks after last antibiotic exposure (E4), and samples collected without any antibiotic exposure in the preceding four weeks (Fig. 1).

Figure 1
figure 1

Study design and classification of samples into categories based on the number of days of exposure to antibiotics prior to sampling. The blue boxes indicate HHs, subjects and samples are part of the assessment. The light grey boxes indicate HHs of the study cohort and part of participants whose samples did not be included in the assessment.

To compare the microbial diversity between our cohort and a cohort with less antibiotic use, we included a random selection of healthy Dutch adults (≥ 18 years) who participated in the Carriage Of Multiresistant Bacteria After Travel (COMBAT) Study. This study was conducted to evaluate the travel-associated risk factors on the acquisition, persistence and transmission of antimicrobial resistance (AMR) in gut microbiota of healthy people. The detailed information of demographic characterization of the cohort has been reported previously31. For the purpose of the present study, we only included fecal samples collected at baseline (before travel) of 106 subjects with no history of antibiotic use during the past year. Differences in Shannon diversity and observed species richness in fecal samples of Dutch and Vietnamese individuals were examined by linear regression analyses with adjustment for age and sex as potential confounding factors.

We used the term “microbial resistance to pulse perturbations” to indicate the resistance of the microbiota to change upon exposure to antibiotic use. Microbial resilience, on the other hand, reflects the recovery speed upon a shift in the microbiota. In this study, to examine the microbial resistance to pulse perturbations, we stratified Vietnamese individuals who used antibiotics according to their microbial resistance. For this purpose, we calculated the Aitchison distance between the samples of each subject collected before and immediately after antibiotic use. Subjects were classified as having a low microbial resistance when the Aitchison distance was above the median, whereas subjects with an Aitchison distance below the median were classified as having a high microbial resistance to pulse perturbations changes.

We also examined if the number of different antimicrobial resistance genes (ARGs) encoding extended-spectrum beta-lactamases, carbapenemases and plasmid-mediated colistin resistance in fecal samples was associated with the microbial community structure.

DNA extraction, 16S –rRNA amplicon sequencing and determination of antimicrobial resistance genes

Fecal samples were transported within 4 h after defecation to the National Institute of Hygiene and Epidemiology (NIHE) in cold conditions (4–8 °C). Within 3 h after transport, samples were stored at − 80 °C until DNA extraction. Microbial DNA was manually extracted from frozen feces according to protocol Q of the International Human Microbiome Standards consortium32. Briefly, fecal samples were firstly subjected to mechanical lysis by repeated Bead-Beating (RBB) followed by column-based purification.

We performed amplicon libraries preparation and sequencing as described previously33. In brief, we amplified the V4 region of the 16S rRNA gene from each DNA sample in triplicate using the 515 f./806r primer pair34. PCR amplicons of the triplicate reactions were pooled; next products were purified using AMPure XP purification (Agencourt, Massachusetts, USA) according to the manufacturer’s instructions and eluted in 25 µl 1 × low TE (10 mM Tris–HCl, 0.1 mM EDTA, pH 8.0). Subsequently, we quantified DNA concentration by Quant-iT PicoGreen dsDNA reagent kit (Invitrogen, New York, USA) using a Victor3 Multilabel Counter (Perkin Elmer, Waltham, USA). The purified amplicons were mixed in equimolar concentrations to ensure equal representation of each sample and, were sequenced on an Illumina MiSeq instrument using the V3 reagent kit (2 × 250 cycles).

Bioinformatics

After demultiplexing raw reads, we visualized the quality profiles of forward and revered reads by uploading the Fastq files into R using package DADA2 (v1.16.0)35. We set maximum number of expected errors (E_max) at 0.5 to remove low-quality reads (with expected errors (E) > E_max) from demultiplexed reads. We then remained good-quality regions of reads for further steps by truncation at positions 200 of forward and 140 of reverse reads. Samples with total read counts below 12,000 were not further analyzed. Merged reads were used to generate amplicon sequence variants (ASV) using DADA2 (V1.16.0)35. Identified ASVs were aligned to construct a phylogeny with FastTree236. Taxonomy was assigned using the RDP v16 (Ribosomal Database Project—SILVA 1.38 v2) with the set formatted for DADA2 package37.

ASVs that were present in less than 5% of all samples as well as reads with a total relative abundance < 0.01% were removed from downstream analysis. At ASV level of taxonomic aggregation, we calculated the alpha-diversity (observed richness and Shannon index) and beta-diversity (Aitchison and Bray–Curtis distances) as measures of within- and between-sample microbial diversity, respectively. To examine the change in alpha-diversity from baseline (M1) to the subsequent time points (M2, M3, M4, M5, M6), we next calculated the delta of the alpha diversity (\(d\Delta\)) using the formula:

$${\text {delta alpha diversity}} \left( {\text d\Delta } \right) = {\text {alpha diversity of the following, time point – alpha diversity at M1}} $$

Statistical analysis

All statistical and computational analysis and visualizations were performed in R v4.1.1 (2021–08-10) using the following packages: reshape2 (v.1.4.4), purrr (v.0.3.4), readr (v.1.3.1), tidyr (v.1.1.2), tibble (v.3.0.4), tidyverse (v.1.3.0)38, RColorBrewer (v.1.1–2), ggthemes (v.4.2.0), circlize (v.0.4.10), ComplexHeatmap (v.2.0.0)39, ggplot2 (v.3.3.2)38, lme4 (v.1.1–27.1), glmmTMB (v.1.1.2.3), ggstatsplot (v.0.9.0)40,dplyr (v.1.0.2)38, phyloseq (v.1.28.0), microViz (0.7.10.9004)41.

Linear mixed effects models were used to test for the effects of antibiotic use, age, and time points on alpha diversity (IBM SPSS Statistics, v 27.0, IBM, Armonk (NY), USA). The model included exposure to antibiotics, age and time point of sample collection as fixed effects and participant ID as a random effect to control for repeated measures. The variance inflation factors (VIF) were calculated prior to model fitting to assess correlated predictor parameters, none were excluded on the basis of collinearity. In addition, the residuals were plotted and checked for homoscedasticity. We next used generalized linear mixed effects models using a negative binomial distribution to test the effect of antibiotic use on the relative abundance of individual bacterial genera (glmmTMB package, R). For each genus, we fitted the model Genus (Counts) ~ Exp_ABx_3codes + Age + Time_point + Age *Exp_ABx_3codes + offset (log (Total_counts)) + (1|ParticipantID). In the model, the variance ‘Exp_ABx_3codes’ is a logical values including three categories that was grouped by number of days since the last antibiotic use prior to sampling as described in the study design.

To examine differentially abundant genus-level taxa between Dutch and Vietnamese individuals, we conducted Analysis of Compositions of Microbiomes with Bias Correction (ANCOM-BC). Differentially abundant genera (p < 0.05 upon Bonferroni correction) are presented42.

Ethical approval

The research was approved by the Oxford University Tropical Research Ethics Committee (OxTREC, 49–14), the National Institute of Hygiene and Epidemiology, Vietnam (NIHE) institutional review board and National Hospital for Tropical Diseases, Vietnam.

All methods were performed in accordance with the relevant guidelines and regulations.

We confirm that informed consent was obtained from all participants and/or their legal guardians.



Source link

EHS
Back to top button