Initial Gut Microbial Composition as a Key Factor Driving Host Response to Antibiotic Treatment, as Exemplified by the Presence or Absence of Commensal Escherichia coli

Abstract

Antibiotics are important for treating bacterial infection; however, efficacies and side effects of antibiotics vary in medicine and experimental models. A few studies have correlated microbiota composition variations with health outcomes in response to antibiotics; however, no study has demonstrated causality. We had noted variation in colonic expression of C-type lectins, regenerating islet-derived protein 3β (Reg3β) and Reg3γ, after metronidazole treatment in a mouse model. To investigate the effects of specific variations in the preexisting microbiome on host response to antibiotics, mice harboring a normal microbiota were allocated to 4 treatments in a 2-by-2 factorial arrangement with or without commensal Escherichia coli and with or without metronidazole in drinking water. E. coli colonized readily without causing a notable shift in the microbiota or host response. Metronidazole administration reduced microbiota biodiversity, indicated by decreased Chao1 and Shannon index values, and altered microbiota composition. However, the presence of E. coli strongly affected metronidazole-induced microbiota shifts. Remarkably, this single commensal bacterium in the context of a complex population led to variations in host responses to metronidazole treatment, including increased expression of antimicrobial peptides Reg3β and Reg3γ and intestinal inflammation indicated by tumor necrosis factor alpha levels. Similar results were obtained from 2-week antibiotic exposure and with additional E. coli isolates. The results of this proof-of-concept study indicate that even minor variations in initial commensal microbiota can drive shifts in microbial composition and host response after antibiotic administration. As well as providing an explanation for variability in animal models using antibiotics, the findings encourage the development of personalized medication in antibiotic therapies.IMPORTANCE This work provides an understanding of variability in studies where antibiotics are used to alter the gut microbiota to generate a host response. Furthermore, although providing evidence only for the one antibiotic, the study demonstrated that initial gut microbial composition is a key factor driving host response to antibiotic administration, creating a compelling argument for considering personalized medication based on individual variations in gut microbiota.

Keywords: Escherichia coli; gut microbiota; metronidazole; regenerating islet-derived protein 3β (Reg3β); regenerating islet-derived protein 3γ (Reg3γ).

FIG 1

Experimental protocol. (A) E. coli resuspended in PBS at a concentration of 2.0 × 10⁸ CFU/ml was given to mice (0.1 ml/each mouse). Metronidazole was given at 750 mg/liter in drinking water. Body weight was recorded weekly. Mice were sacrificed on day 4 (A1) or day 14 (A2). (B) Enumeration of E. coli in mouse feces before metronidazole treatment and 4 days after metronidazole/water administration. Dots represent individual mice, and lines depict the mean values. (C) Body weight change during the E. coli treatment and after 4 days of metronidazole/water treatment. For all treatment groups, n = 8. Data are shown as means ± SEM. Means that do not share a letter (a or b) are significantly different. α = 0.05.

FIG 2

(A) Alpha diversity analysis of bacterial communities in colon contents of mice. All the colonic contents were harvested after 4 days of metronidazole/water administration. Data are shown as means ± SEM. Means that do not share a letter (a, b, or c) are significantly different. α = 0.05. (B) Bar chart indicating microbial community profiles between groups, summarized down to the genus level. Microbial compositions of the four groups before experimental treatment are labeled CON_PRE, MET_PRE, EC_MET_PRE, and EC_PRE, respectively. uncl, unclassified.

FIG 3

(A) Principal-component analysis (PCA) plots of the bacterial communities based on the weighted UniFrac distance matrix. Each plot point represents an individual mouse. (B) Box plots show selective levels of bacterial abundance in different treated groups at the family level. Colonic contents were collected after 4 days of metronidazole/water treatment. For all treatment groups, n = 8. Data are shown as means ± SEM. Means that do not share a letter (a, b, or c) are significantly different. α = 0.05.

FIG 4

Quantitative RT-PCR results of (A) Reg3β, (B) Reg3γ, (C) MUC2, and (D) IL-22 expression in the colon of untreated and E. coli- and metronidazole-treated mice. Colonic tissue samples were harvested after 4 days of metronidazole/water administration. For all treatment groups, n = 8. Data are shown as means ± SEM. Means that do not share a letter (a or b) are significantly different. α = 0.05.

FIG 5

Results of cytokine analysis of (A1) TNF-α, (B1) IL-1β, (B2) IL-6, and (B3) IL-10 production in the colon. Colonic tissue samples were collected after 4 days of metronidazole/water treatment. For all treatment groups, n = 8. Data are shown as means ± SEM. Means that do not share a letter (a or b) are significantly different. α = 0.05. (A2) Correlation of colonic E. coli bacterial load with TNF-α expression levels in EC-MET group. Spearman's correlation coefficient (r values) and significance P values are shown.

FIG 6

Distal colon sections from CON, MET, EC, and EC-MET mice 4 days after metronidazole/water treatment were stained with hematoxylin and eosin. There was no significant inflammation evidence in any of the treatments, including inflammation and damage to the lumen, surface epithelium, mucosa, and submucosa, as well as the number of goblet cells. Original magnification and bars: left, ×40 and 1,000 μm; right, ×400 and 100 μm.

FIG 7

Enumeration of E. coli in mouse feces and colonic gene expression of mice colonized with (A) commensal E. coli isolates, (B) a wild mouse E. coli isolate, and (C) a rat E. coli isolate. For enumeration of E. coli in mouse feces, samples were taken before metronidazole treatment and after 14 days (A1) or 4 days (B1 and C1) of metronidazole/water treatment. Dots represent individual mice, and lines depict the mean values. Reg3β and Reg3γ expression in the colon of untreated and E. coli- and metronidazole-treated mice was detected by quantitative RT-PCR. Colonic tissue samples were harvested after 14 days (A2 and A3) or 4 days (B2, B3, C2, and C3) of metronidazole/water administration. For all treatment groups, n = 5. Data are shown as means ± SEM. Means that do not share a letter (a or b) are significantly different. α = 0.05.