Spatial analysis of murine microbiota and bile acid metabolism during amoxicillin treatment

Abstract

Antibiotics cause collateral damage to resident microbes that is associated with various health risks. To date, studies have largely focused on the impacts of antibiotics on large intestinal and fecal microbiota. Here, we employ a gastrointestinal (GI) tract-wide integrated multiomic approach to show that amoxicillin (AMX) treatment reduces bacterial abundance, bile salt hydrolase activity, and unconjugated bile acids in the small intestine (SI). Losses of fatty acids (FAs) and increases in acylcarnitines in the large intestine (LI) correspond with spatially distinct expansions of Proteobacteria. Parasutterella excrementihominis engage in FA biosynthesis in the SI, while multiple Klebsiella species employ FA oxidation during expansion in the LI. We subsequently demonstrate that restoration of unconjugated bile acids can mitigate losses of commensals in the LI while also inhibiting the expansion of Proteobacteria during AMX treatment. These results suggest that the depletion of bile acids and lipids may contribute to AMX-induced dysbiosis in the lower GI tract.

Keywords: CP: Metabolism; CP: Microbiology; antibiotics; bile acids; microbiome.

Figure 1.. Amoxicillin treatment disrupts microbial communities across the length of the mouse GI tract

(A) Experimental schematic. Created with BioRender.com. (B) Bacterial load across GI compartments quantified by qPCR using universal primers targeting the 16S-V3 gene region (*q < 0.05; Mann-Whitney with Holm-Šídák correction for AMX-treated vs. untreated within each GI site). (C) α-Diversity metrics across GI compartments calculated from 16S amplicon sequencing data (*q < 0.05; **q < 0.01; Mann-Whitney with Holm-Šídák correction). (D) PCA plots of GI sample β-diversity using unweighted UniFrac distances calculated from 16S amplicon sequencing data. (E) Species-level abundance plots based on shotgun-metagenomic reads aligned to the MGBC Kraken2/Bracken database and normalized by total bacterial load (16S copies/mg lumenal contents) per sample (based on 16S-V3 qPCR). Vehicle-treated control samples on top and AMX-treated samples below. The 38 species represented in the legend were selected by combining all species that were among the 12 most abundant within any individual GI compartment. “Other” represents the sum of relative abundances for all other species. For all data, n = 6.

Figure 2.. Disruption of microbial communities correlates with shifts in metabolites along length of GI tract

(A) Heatmap indicating log₂-fold changes of metabolites (rows) that displayed altered abundance within at least one GI compartment upon AMX treatment, identified by FI-MS. Columns represent individual GI compartments or whole blood (differential abundance defined as adjusted p < 0.01 for AMX vs. untreated samples; Welch’s t test with Benjamini-Hochberg correction; n = 6). Ions that could not be matched to any known metabolites based on chemical formula were excluded (see method details for annotation strategy). (B) PCA of untargeted FI-MS metabolomic data from different GI compartments of AMX-treated and untreated mice (n = 6). (C) Correlation network for SI samples indicating positive (red edges) and negative correlations (blue edges) between abundant bacterial species (taxa with mean relative abundance >0.1% and detected in at least half of all samples) and all differentially abundant metabolites. High abundance species are defined as being among the 20 most abundant within at least one GI site. Correlation values determined by Spearman rank correlation and Holm correction (cutoff for inclusion: p < 0.05; n = 36 total SI samples).

Figure 3.. AMX treatment depletes SI-specific BSH activity

(A–C) Relative levels of putative primary BAs taurocholate/tauromuricholate (C₂₆H₄₅NO₇S) (A), cholate/muricholate (C₂₄H₄₀O₅) (B), and taurine (C₂H₇NO₃S) (C) in indicated GI sites from untreated (black) and AMX-treated (red) mice quantified by untargeted FI-MS. Mean and SEM shown (**p < 0.01; ****p < 0.0001; two-way ANOVA with Šídák’s multiple comparisons test; n = 6). (D) BSH activity levels within lumenal samples collected from indicated GI sites in untreated (top) and AMX-treated (bottom) mice quantified as freed taurine upon incubation with taurocholate substrate. Activity is normalized by sample mass and was measured after a 24-h incubation at 37°C. Mean and SD shown (**p < 0.01; ****p < 0.0001; two-way ANOVA with Šídák’s multiple comparisons test; n = 6). (E) BSH gene expression within metatranscriptomic reads from indicated GI sites stratified by contributing species from untreated (top) and AMX-treated mice (bottom). CoPM, copies per million reads; HUMAnN 3; n = 6). Mean and SEM shown. (F and G) Correlation plots for L. johnsonii abundance vs. cholate/muricholate (C₂₆H₄₅NO₇S) (F) and taurine (C₂H₇NO₃S) (G) abundances across all SI samples (r_s = Spearman correlation coefficient, **p < 0.01; ****p < 0.0001; n = 36).

Figure 4.. AMX treatment depletes secondary and unconjugated BAs in SI and FAs in the lower GI tract, and elevates simple sugars in the cecum and colon

(A–D) Heatmaps indicating fold change in BA (A), FA (B), acylcarnitines (C), and simple sugar metabolites (D) across each GI site (columns) based on LC-MS/MS (*p < 0.05; Welch’s t test and <5% false discovery rate; n = 6). Asterisks in compound names indicate compounds for which identity has not been confirmed using a purified chemical standard.

Figure 5.. GI microbial communities elicit location-dependent transcriptomic responses to AMX treatment and elevate FA metabolism across all GI sites

(A) Community-level transcript abundances of the 10 most abundant metabolic pathways within SI (left side) and LI (right side) samples from untreated (top) and AMX-treated (bottom) mice. Abundances determined by quantification of metatranscriptomic reads mapped to MetaCyc pathways using the HUMAnN 3 pipeline. Bars represent mean and SD of abundances per group, n = 6. (B and C) Scatterplots displaying AMX treatment coefficients from linear modeling of log₂-transformed abundances of MetaCyc reactions (B) and pathways (C) within SI (x axis; n = 36) and LI (y axis; n = 24). Bar plots (right side) indicate total number of identified reactions (B) or pathways (C), total number of reactions/pathways differentially expressed only in the SI (blue in scatterplots), the LI (red in scatterplots), and the number differentially expressed in both regions (purple in scatterplots). (D) Compiled bar plots indicate AMX treatment coefficient values and standard error from linear modeling of individual MetaCyc pathways within each of the six GI locations sampled. Pathways included represent a compilation of the 10 most differentially expressed pathways based on AMX treatment coefficient values within each of the six individually sampled GI sites. “SI-specific” pathways (top) were only differentially expressed in SI sites, “LI-specific” pathways (center) were only differentially expressed in LI sites, and “Shared” pathways (bottom) were differentially expressed in both upper and lower GI sites. Linear modeling performed using MaAsLin2 with normalized reaction/pathway abundance data generated with HUMAnN 3 (differential expression defined as q < 0.05 (red bars); MaAsLin2; n = 6 per group).

Figure 6.. γ-Proteobacteria drive increase in FA metabolism pathway expression across the GI tract under AMX treatment

(A–D) Line plots indicate relative expression (HUMAnN 3) of FA utilization (A), FA salvage (B), saturated FA biosynthetic (C), and unsaturated FA biosynthetic pathways (D) from metatranscriptomic reads from SI and LI sites (black, untreated mice; red, AMX treated, line indicates mean expression level per GI site; *q < 0.05; **q < 0.01; ***q < 0.001 for AMX treatment coefficients, MaAsLin2; n = 6). Pie charts indicate relative contributions of individual species to pathway expression within each site in AMX-treated mice. Relative contribution to pathway expression was calculated as mean CoPM per reaction from each species divided by total mean CoPM per reaction across all species (HUMAnN 3 species-stratified output). (E) Relative proportion of metagenomic reads aligning to each species per GI site in AMX-treated mice (Bracken2). (F) Heatmap indicates Spearman correlation coefficient values for gene expression of saturated FA pathways (rows) with putative saturated FA molecules(columns); values indicate Spearman correlation coefficients (bold values with gold border indicate statistically significant correlations; p < 0.05 Spearman rank correlation; n = 36 [SI, top], and 24 [cecum, bottom]).

Figure 7.. Cholate supplementation can prevent expansion of Proteobacteria and reduce collateral damage of AMX treatment on the GI microbiota

(A) Ratio of taurine-C/UC CA (left) and total BAs (right) quantified by Metabolon with targeted LC-MS/MS (BAs included in panel C/UC: CA, α-muricholic acid, β-muricholic acid, deoxycholic acid, lithocholic acid, and ursodeoxycholic acid, *p < 0.05; one-way ANOVA with Tukey’s multiple comparisons test; n = 12). Mean and SD shown. (B) Bacterial load within SI (left) and cecum (right) across treatment groups quantified by qPCR using universal primers targeting the 16S-V3 gene region(*p < 0.05; **p < 0.01; ***p < 0.001; Kruskal-Wallis test and Dunn’s multiple comparisons test per GI site; n = 8). Lines indicate median. (C) α-Diversity within SI (left) and cecum (right) across treatment groups (Shannon’s index; p > 0.05 for all comparisons; Kruskal-Wallis test and Dunn’s multiple comparisons test per GI site; n = 12). Lines indicate median. (D) PCA plot of Bray-Curtis β-diversity distances calculated from 16S amplicon sequencing data of SI and cecum samples. (E) Violin plots showing all vs. all distances within untreated samples and between untreated samples and other treatment groups (left) and within AMX-treated and between AMX-treated and other treatment groups (right). Bray-Curtis distances; *q < 0.05; **q < 0.01; ***q < 0.001; QIIME2:PERMANOVA and Benjamini-Hochberg correction; n = 12 per group. (F) Relative abundance plots generated from 16S amplicon sequencing data aligned at the class level (n = 12). Bars indicate mean and SEM. (G) Relative abundance of γ-Proteobacteria in each treatment group (*q < 0.10; **q < 0.01; ***q < 0.001; MaAsLin2; n = 12). (H) Bar charts for SI (left) and cecum (right) showing AMX + cholate treatment coefficients for bacterial taxa, demonstrating statistically significant changes in abundance in AMX + cholate vs. AMX treatment conditions (q < 0.1, with AMX treatment as baseline; MaAsLin2; n = 12). Bars indicate model coefficient and standard error.