Gut microbiota disturbance during antibiotic therapy: a multi-omic approach

Abstract

Objective: Antibiotic (AB) usage strongly affects microbial intestinal metabolism and thereby impacts human health. Understanding this process and the underlying mechanisms remains a major research goal. Accordingly, we conducted the first comparative omic investigation of gut microbial communities in faecal samples taken at multiple time points from an individual subjected to β-lactam therapy.

Methods: The total (16S rDNA) and active (16S rRNA) microbiota, metagenome, metatranscriptome (mRNAs), metametabolome (high-performance liquid chromatography coupled to electrospray ionisation and quadrupole time-of-flight mass spectrometry) and metaproteome (ultra high performing liquid chromatography coupled to an Orbitrap MS(2) instrument [UPLC-LTQ Orbitrap-MS/MS]) of a patient undergoing AB therapy for 14 days were evaluated.

Results: Apparently oscillatory population dynamics were observed, with an early reduction in Gram-negative organisms (day 6) and an overall collapse in diversity and possible further colonisation by 'presumptive' naturally resistant bacteria (day 11), followed by the re-growth of Gram-positive species (day 14). During this process, the maximum imbalance in the active microbial fraction occurred later (day 14) than the greatest change in the total microbial fraction, which reached a minimum biodiversity and richness on day 11; additionally, major metabolic changes occurred at day 6. Gut bacteria respond to ABs early by activating systems to avoid the antimicrobial effects of the drugs, while 'presumptively' attenuating their overall energetic metabolic status and the capacity to transport and metabolise bile acid, cholesterol, hormones and vitamins; host-microbial interactions significantly improved after treatment cessation.

Conclusions: This proof-of-concept study provides an extensive description of gut microbiota responses to follow-up β-lactam therapy. The results demonstrate that ABs targeting specific pathogenic infections and diseases may alter gut microbial ecology and interactions with host metabolism at a much higher level than previously assumed.

Keywords: Antibiotic Therapy; Colonic Microflora; Gastrointestinal Function; Gene Expression; Meta-Analysis.

Figures

Figure 1

Total and active bacterial composition based on 16S rDNA and 16S rRNA analyses, respectively, in the follow-up study. Samples FS-0, FS-3, FS-6, FS-11 and FS-14 correspond to the materials collected on days 0, 3, 6, 11 and 14 of antibiotic (AB) treatment, respectively. The FS-40 sample corresponds to the materials collected 40 days after cessation of the AB treatment.

Figure 2

(A) Correspondence analysis of the expressed genes in each sample. (B) Clustering of the samples based on the type and abundance of expressed genes, applying the Bray–Curtis distance.

Figure 3

Taxonomic assignments of mRNAs for each sample according to the lowest common ancestor algorithm.

Figure 4

Partial least-squares discriminant analysis score plots and clustering analysis of metabolite profiles after different comparisons. (A) The whole dataset (8600 features) with the prediction for quality control (QC) samples, seven components, R2=0.989, Q2=0.670; the robustness of the analytical procedure was demonstrated by the tight clustering of the QC samples. (B) Discriminant variables identified by comparing samples in a pairwise fashion (382 discriminant features from 988 of the 4349 initial variables that were present in all three replicates of samples from any group), four components, R2=0.978, Q2=0.928. (C) Statistically significant variables identified in the METLIN database (49 features), four components, R2=0.968, Q2=0.915. (D) The effect of antibiotics on the human gut microbiota, as determined by a two-way hierarchical clustering analysis of the metabolite profiles. Hierarchical clustering was performed with a matrix of the total masses that passed the filtering and statistical treatments for each sample. Less abundant masses in a given community are shown in blue, whereas more abundant masses are shown in red. Note: sample FS-3 was discarded from the analysis due to the presence of faecal material in the cell extracts.

Figure 5

Number of quantified proteins showing either high (black bars) or low (white bars) abundance levels relative to the proteins identified in sample FS-0. Only proteins with values ≥1.5 or ≤-1.5 log2 ratios were considered.

Figure 6

(A) Correspondence analysis of the expressed proteins in each sample. (B) Clustering of the samples based on the type and abundance of expressed proteins, with Pearson's correlation applied to calculate the distances. The two axes, CA1 and CA2, in (A) bundle 33% and 28% of the total observed variation, respectively.

Figure 7

Graphical representation of the high-abundance and low-abundance proteins in different pathways, according to the clusters of orthologous group (COG) number assigned to each protein. (A) Rubrerythrin/ferritin COGs: COG1592 (rubrerythrin) and COG1528 (ferritin-like protein). (B) Translation factor and translation enzyme COGs: COG0193 (peptidyl-tRNA hydrolase), COG0264 (translation elongation factor Ts), COG0050, COG0532 (GTPases—translation elongation factors), COG0216 (protein chain release factor A), COG0480 (translation elongation factors (GTPases)) and COG0231 (translation elongation factor P/translation initiation factor 5A). (C) Glycolysis, pyruvate, glutamate and other related COGs: COG0126 (3-phosphoglycerate kinase), COG0205 (6-phosphofructokinase), COG0148 (enolase), COG0076 (glutamate decarboxylase and related proteolipid protein-dependent proteins), COG1830 (DhnA-type fructose-1,6-bisphosphate aldolase and related enzymes), COG0334 (glutamate dehydrogenase/leucine dehydrogenase), COG1053 (succinate dehydrogenase/fumarate reductase, flavoprotein subunit), COG0588 (phosphoglycerate mutase 1), COG0479 (succinate dehydrogenase/fumarate reductase, Fe-S protein subunit), COG0191 (fructose/tagatose bisphosphate aldolase), COG0149 (triosephosphate isomerase), COG0166 (glucose-6-phosphate isomerase), COG0057 (glyceraldehyde-3-phosphate dehydrogenase/erythrose-4-phosphate dehydrogenase) and COG0469 (pyruvate kinase). (D) Antimicrobial transporters, multidrug efflux pumps and other transporter COGs: COG0841 (cation/multidrug efflux pump), COG2825 (outer membrane protein), COG3292 (predicted periplasmic ligand-binding sensor domain), COG3264 (small-conductance mechanosensitive channel), COG1538 (outer membrane protein) and COG1629 (outer membrane receptor proteins). Only proteins with values ≥1.5 or ≤-1.5 log2 ratios were considered.

Figure 8

The ‘presumptive’ model related to the follow-up effect of antibiotics (ABs) on the microbial and metabolic composition of the human gut. The model is based on the combination of experimental multi-omics data. The biliary excretion of ABs triggers a cascade of metabolic events. At the earlier stages of AB therapy, the bacteria respond by promoting systems to avoid the antimicrobial effects of the drugs (expressing beta-lactamases, antimicrobial peptide transporters and multidrug efflux pumps and producing glycero(lyso)phospholipids—G(L)PL) and to cope with an intermittent nutrient supply while decreasing polysaccharides and lipopolysaccharide (LPS) production. Genes involved in cell envelope biosynthesis and the degradation of peptidoglycan-like components are increasingly expressed until the end of AB treatment but with a time delay compared with other drug-detoxifying mechanisms. Finally, the bacterial metabolism of the bile acid, hormones and cholesterol synthesised in the liver and pancreas is attenuated by AB therapy, thus possibly affecting entero-hepatic recirculation and systemic lipid metabolism, that is, the emulsification, absorption and transport of dietary fats; however, after treatment cessation, the metabolism of these factors improved significantly. Similarly, the pool of vitamins that are directly synthesised by gut bacteria was significantly improved after treatment cessation. The nutrient supply mechanisms, such as glycolysis, pyruvate decarboxylation, tricarboxylic acid (TCA) cycle, glutamate metabolism, and iron uptake, that are induced at earlier stages (day 6) become attenuated during the late stages of the therapy and become significantly attenuated after treatment cessation, suggesting that the entero-hepatic recirculation system may contain a lower amount of iron, sugars, branched amino acids, short organic acids and pyruvate produced or transported by colonic bacteria. At the active bacterial structure level, an apparently oscillatory population dynamic was further observed, with the initially predominant active Bacteroidaceae becoming replaced by Burkholderiaceae after treatment cessation. The broken line indicates the overall trend in each of the gut bacteria components during the follow-up treatment.