Predicted Metabolic Function of the Gut Microbiota of Drosophila melanogaster

Abstract

An important goal for many nutrition-based microbiome studies is to identify the metabolic function of microbes in complex microbial communities and their impact on host physiology. This research can be confounded by poorly understood effects of community composition and host diet on the metabolic traits of individual taxa. Here, we investigated these multiway interactions by constructing and analyzing metabolic models comprising every combination of five bacterial members of the Drosophila gut microbiome (from single taxa to the five-member community of Acetobacter and Lactobacillus species) under three nutrient regimes. We show that the metabolic function of Drosophila gut bacteria is dynamic, influenced by community composition, and responsive to dietary modulation. Furthermore, we show that ecological interactions such as competition and mutualism identified from the growth patterns of gut bacteria are underlain by a diversity of metabolic interactions, and show that the bacteria tend to compete for amino acids and B vitamins more frequently than for carbon sources. Our results reveal that, in addition to fermentation products such as acetate, intermediates of the tricarboxylic acid (TCA) cycle, including 2-oxoglutarate and succinate, are produced at high flux and cross-fed between bacterial taxa, suggesting important roles for TCA cycle intermediates in modulating Drosophila gut microbe interactions and the potential to influence host traits. These metabolic models provide specific predictions of the patterns of ecological and metabolic interactions among gut bacteria under different nutrient regimes, with potentially important consequences for overall community metabolic function and nutritional interactions with the host.IMPORTANCE Drosophila is an important model for microbiome research partly because of the low complexity of its mostly culturable gut microbiota. Our current understanding of how Drosophila interacts with its gut microbes and how these interactions influence host traits derives almost entirely from empirical studies that focus on individual microbial taxa or classes of metabolites. These studies have failed to capture fully the complexity of metabolic interactions that occur between host and microbe. To overcome this limitation, we reconstructed and analyzed 31 metabolic models for every combination of the five principal bacterial taxa in the gut microbiome of Drosophila This revealed that metabolic interactions between Drosophila gut bacterial taxa are highly dynamic and influenced by cooccurring bacteria and nutrient availability. Our results generate testable hypotheses about among-microbe ecological interactions in the Drosophila gut and the diversity of metabolites available to influence host traits.

Keywords: Drosophila; competition; constraint-based modeling; cross-feeding; microbiome; mutualism.

FIG 1

Bacterial growth dynamics on media of different nutrient content. Growth dynamics displayed as biomass formation rate predicted for growth nutrient-rich medium (A), base medium (B), and minimal medium (C).

FIG 2

Ecological interactions in simulated communities of different diversity. (A) Impact of coculture and medium on the sign of interactions (+, beneficial; −, antagonistic; 0, neutral) between bacteria in the three test media. No commensal interactions were observed in any media type. (B) Overlapping metabolites consumed by bacteria in the 31 simulated communities. Significantly different (P < 0.05) groups by Tukey’s HSD post hoc test are indicated by different letters.

FIG 3

Metabolite use patterns associated with competitive, parasitic, and mutualistic growth outcomes. Each point represents the frequency of a metabolite use pattern associated between all pairwise interactions across all communities from 2 to 5 members. The relative frequency of metabolite use is calculated by dividing the number of times a metabolite is used in a particular pattern (single-use, co-consumed, cross-fed, single-produced, or coproduced) by the total number of times the metabolite is produced or consumed in the 31 simulated communities. Black bars indicate the median frequency of occurrence for each metabolite use pattern. Significantly different (P < 0.05) groups by Tukey’s HSD post hoc test are indicated by different letters. The panel at the left summarizes the five different types of metabolite use (triangle, metabolite; open and closed circles, co-occurring Microbe A and Microbe B, respectively).

FIG 4

Metabolite use patterns for metabolite classes of amino acids, carbon, nucleotides, vitamins, and cofactors. Tick marks on the x axis indicate the relative frequency of the consumption or production of a metabolite and range from 0 to 1 at 0.25 increments. The relative frequency of metabolite use is calculated by dividing the number of times a metabolite is used in a particular pattern (single-use, co-consumed, cross-fed, single-produced, or coproduced) by the total number of times the metabolite is produced or consumed in the 31 simulated communities. Black circles demarcate metabolites that are initially present in each medium.

FIG 5

Metabolic roles of individual bacteria. Predicted metabolite production and consumption profiles for metabolite classes nucleotides, vitamins, and cofactors (A) and amino acid and carbon (B). The two-letter abbreviations at the top of each plot represent individual bacteria: AF, Acetobacter fabarum; AP, Acetobacter pomorum; AT, Acetobacter tropicalis; LB, Lactobacillus brevis; LP, Lactobacillus plantarum. Tick marks on the x axis indicate the relative frequency of the consumption or production of a metabolite and range from 0 to 1 at 0.25 increments.

FIG 6

Metabolic function of simulated microbial taxa under different conditions. (A) Effect of community size on metabolite richness. Metabolite richness is calculated as the number of metabolites either consumed or released by a given taxon in each microbial treatment and medium combination. Effect of community size for each medium is indicated with the estimated marginal mean (open circles or diamonds) and standard error (SE) from ANOVA models. Letters indicate results from post hoc Tukey’s test, which was conducted separately for each medium. Closed, colored circles indicate individual metabolite richness values for each taxon under each condition. Effect test results are displayed in Table S4D in the supplemental material. (B) Global effect of species identity on metabolite richness for the number of compounds consumed and released. The conditional mean and standard deviation are displayed for the best linear unbiased prediction. Dotted lines indicate the grand mean for metabolite richness across all species. (C) Principal-component analysis (PCA) correlating metabolite consumption or release rates with community size, medium type, and microbial presence. Black arrows indicate metabolite type scores, and colored arrows display the correlation vectors for microbial presence (only significant vectors are plotted). The percent variance explained by each axis is shown in parentheses.