Complex carbohydrate utilization by gut bacteria modulates host food consumption

Abstract

The gut microbiota interacts with dietary nutrients and can modify host feeding behavior, but underlying mechanisms remain poorly understood. Gut bacteria digest complex carbohydrates that the host cannot digest and liberate metabolites that serve as energy sources and signaling molecules. Here, we use a gnotobiotic mouse model to examine how gut bacterial fructose polysaccharide metabolism influences host intake of diets containing these carbohydrates. Two Bacteroides species ferment fructans with different glycosidic linkages: B. thetaiotaomicron ferments levan with β2-6 linkages, whereas B. ovatus ferments inulin with β2-1 linkages. We find that mice eat relatively more diet containing the carbohydrate that their gut bacteria cannot ferment compared to the fermentable ones: mice colonized with B. thetaiotaomicron consume more inulin diet, while mice colonized with B. ovatus consume more levan diet. Knockout of bacterial fructan utilization genes attenuates this difference, whereas swapping the fermentation ability of B. thetaiotaomicron to inulin confers increased consumption of levan diet. Bacterial fructan fermentation and host feeding behavior are associated with neuronal activation in the arcuate nucleus of the hypothalamus. These results reveal that bacteria nutrient metabolism modulates host food consumption through sensing of differential energy extraction, which contributes to our understanding of determinants of food choice.

Fig. 1. Colonization with differential fructan-utilizing bacteria increases host consumption of diets containing the non-fermentable fructan.

a Fructan chemical structures. Inulin has β2-1 linkages between fructose subunits, while levan has β2-6 linkages between fructose subunits. b Growth curves of B. thetaiotaomicron (Bt) and B. ovatus (Bo) in minimal media containing 0.5% inulin, pure levan, or dietary levan. Each dot represents mean of 3 biological replicates, error bars represent SEM. OD600 optical density at 600 nm. c Growth curves of Bt and Bo in minimal media containing 5% inulin diet (ID) or levan diet (LD). Each dot represents mean of 3 biological replicates, error bars represent SEM. CFU colony forming unit. d Acetate produced in supernatants of bacteria grown for 24 h in minimal media containing 5%ID or LD. (n = 3 each). p-values (from left to right) = <0.0001, <0.0001. ANOVA interaction p-value < 0.0001. e Experimental schematic for sequential feeding post-colonization. Germ-free mice were habituated with both diets for 7 days before being colonized with Bt or Bo. The mice ate one diet ad libitum for 11 days and then the other diet for 10 days. The mice were then provided both diets in their cage overnight. f Average daily diet intake post-colonization for mice colonized with Bt and Bo. p-values (from left to right) = 0.0301, 0.5389. ANOVA interaction p-value = 0.0159. g Average daily diet intake post-colonization for mice colonized with Bt and Bo by diet order. p-values (from left to right) = <0.0001, 0.8281, 0.0623, <0.0001. ANOVA interaction p-value < 0.0001. h Overnight diet intake from days 30–31 for mice colonized with Bt and Bo. p-values (from left to right) = 0.0030, 0.1981. ANOVA interaction p-value = 0.0009. i Overnight diet intake from days 30–31 for mice colonized with Bt and Bo by diet order. p-values (from left to right) = <0.0001, 0.9999, 0.9798, 0.0963. ANOVA interaction p-value < 0.0001. Error bars for (d, f–i) represent mean ± SEM. Data from (f–i) are combined from three independent experiments. For (f and h), n = 22 for Bt, 23 for Bo. For (g and i), n = 11 for Bt NF → F, 11 for Bt F → NF, 12 for Bo F → NF, 11 for Bo NF → F. For (d, f–i), 2-way ANOVA with matched column measures, comparing means across rows, and Sidak’s corrections were performed. ns = p-value > 0.10; * = p-value < 0.05; **** = p-value < 0.0001. ID inulin diet, LD levan diet, F fermentable diet, NF non-fermentable diet.

Fig. 2. Bacterial fructan utilization drives host consumption of diets containing the non-fermentable fructan.

a Growth curves of wild-type B. thetaiotaomicron (Bt), mutant B. thetaiotaomicron lacking susCD (Bt^ΔsusCD), and inulin-utilizing B. thetaiotaomicron (Bt^Bo-susCD) in minimal media containing 0.5% fructose, inulin, levan (pure), or levan (purified for dietary formulation). Each dot represents mean of 3 biological replicates, error bars represent SEM. OD600 optical density at 600 nm. b Growth curves of Bt, Bt^ΔsusCD, and Bt^Bo-susCD in minimal media containing 5% inulin diet (ID) or levan diet (LD). Each dot represents mean of 3 biological replicates, error bars represent SEM. CFU colony forming unit. c Experimental schematic for sequential feeding post-colonization. d Average daily diet intake post-colonization for mice colonized with Bt, Bt^ΔsusCD, and Bt^Bo-susCD. p-values (from left to right) = 0.0002, 0.0115, 0.4467. ANOVA interaction p-value = 0.0004. e Average daily diet intake post-colonization for mice colonized with Bt, Bt^ΔsusCD, and Bt^Bo-susCD by diet order. p-values (from left to right) = 0.0016, 0.0174, 0.1318, 0.0805, 0.1244, 0.0001. ANOVA interaction p-value < 0.0001. f Overnight diet intake from days 30–31 for mice colonized with Bt, Bt^ΔsusCD, and Bt^Bo-susCD. p-values (from left to right) = 0.0220, 0.7600, 0.8595. ANOVA interaction p-value = 0.0556. Error bars for (d–f) represent mean ± SEM. Data from (d–f) are combined from three independent experiments. For (d and f), n = 24 for Bt, 23 for Bt^ΔsusCD, and 24 for Bt^Bo-susCD. For (e), n = 12 for Bt NF → F, 12 for Bt F → NF, 12 for Bt^ΔsusCD ID → LD, 11 for Bt^ΔsusCD LD → ID, 12 for Bt^Bo-susCD F → NF, and 12 for Bt^Bo-susCD NF → F. For (d–f), 2-way ANOVA with matched column measures, comparing means across rows, and Sidak’s corrections were performed. ns = p-value > 0.10; * = p-value < 0.05; ** = p-value < 0.01; *** = p-value < 0.001. ID inulin diet, LD levan diet, F fermentable diet, NF non-fermentable diet.

Fig. 3. Colonization with fructan-utilizing gut bacteria increases host consumption of fructan-free diet.

a Growth curves of wild-type B. ovatus (Bo) and B. ovatus susCD KO (Bo^ΔsusCD) in minimal media containing 0.5% fructose or inulin. Each dot represents mean of 3 biological replicates, error bars represent SEM. OD600 optical density at 600 nm. b Growth curves of Bo and Bo^ΔsusCD in minimal media containing 5% cellulose diet (CD) or inulin diet (ID). Each dot represents mean of 3 biological replicates, error bars represent SEM. CFU colony forming unit. c Experimental schematic for sequential feeding post-colonization. d Average daily diet intake post-colonization for mice colonized with Bo and Bo^ΔsusCD. p-values (from left to right) = 0.0402, 0.3021. ANOVA interaction p-value = 0.4461. e Average daily diet intake post-colonization for mice colonized with Bo and Bo^ΔsusCD by diet order. p-values (from left to right) = 0.0114, 0.9948, 0.1490, >0.9999. ANOVA interaction p-value = 0.0971. f Overnight diet intake from days 30–31 for mice colonized with Bo and Bo^ΔsusCD. p-values (from left to right) = 0.7558, 0.2584. ANOVA interaction p-value = 0.1337. Error bars for (d–f) represent mean ± SEM. Data from (d–f) are combined from two independent experiments. For (d and f), n = 15 for Bo, 16 for Bo^ΔsusCD. For (e), n = 8 for Bo NF → F, 7 for Bo F → NF, 8 for Bo^ΔsusCD CD → ID, and 8 for Bo^ΔsusCD ID → CD. For (d–f), 2-way ANOVA with matched column measures, comparing means across rows, and Sidak’s corrections were performed. ns = p > 0.10; * = p-value < 0.05. CD cellulose diet, ID inulin diet, F fermentable diet, NF non-fermentable diet.

Fig. 4. Bacterial fructan utilization by B. thetaiotaomicron promotes diet choice-induced neuronal activation in the arcuate nucleus of the hypothalamus.

a Schematic of experimental set up: 1–1.5 h after the fasting-induced, mice were sacrificed, and their brains collected. Brains were sectioned coronally for the arcuate nucleus of the hypothalamus (ARH), and medial ARH sections (around Bregma −1.82) were stained for neurons with NeuN (magenta) and early activation marker, cFos (cyan). b Representative images of medial ARH section in mice colonized with Bt, Bt^ΔsusCD, and Bt^Bo-susCD. Scale bar for all panels is 100 µm. c Representative images of cFos-positive neurons in mice colonized with Bt, Bt^ΔsusCD, and Bt^Bo-susCD. Scale bar for all panels is 50 µm. d Percentage of cFos-positive neurons out of total neurons in ARH of mice colonized with Bt, Bt^ΔsusCD, and Bt^Bo-susCD. p-values (from left to right) = 0.0394, 0.1081. Error bars (d) for represent mean ± SEM. Data are combined from two independent experiments with n = 9 for Bt, 9 for Bt^ΔsusCD, and 7 for Bt^Bo-susCD. From the first experiment (n = 3–4 per group), each dot represents the average of three technical replicates; from the second experiment (n = 4–5 per group), each dot represents one technical replicate. Unpaired t-tests (2-tailed) with Welch’s correction were performed. ns = p-value > 0.10; * = p-value < 0.05. F fermentable diet, NF non-fermentable diet.