Functional heterogeneity in the fermentation capabilities of the healthy human gut microbiota

Abstract

The human gut microbiota is known for its highly heterogeneous composition across different individuals. However, relatively little is known about functional differences in its ability to ferment complex polysaccharides. Through ex vivo measurements from healthy human donors, we show that individuals vary markedly in their microbial metabolic phenotypes (MMPs), mirroring differences in their microbiota composition, and resulting in the production of different quantities and proportions of Short Chain Fatty Acids (SCFAs) from the same inputs. We also show that aspects of these MMPs can be predicted from composition using 16S rRNA sequencing. From experiments performed using the same dietary fibers in vivo, we demonstrate that an ingested bolus of fiber is almost entirely consumed by the microbiota upon passage. We leverage our ex vivo data to construct a model of SCFA production and absorption in vivo, and argue that inter-individual differences in quantities of absorbed SCFA are directly related to differences in production. Though in vivo studies are required to confirm these data in the context of the gut, in addition to in vivo read outs of SCFAs produced in response to specific fiber spike-ins, these data suggest that optimizing SCFA production in a given individual through targeted fiber supplementation requires quantitative understanding of their MMP.

Fig 1. Microbial metabolic phenotype varies significantly in the healthy population.

(a) Schematic of assay setup and sampling frequency. (b) 24h time traces of butyrate concentration over time in response to inulin, pectin, cellulose and control, in two different participants. Participant C only produces butyrate from inulin, while participant G produces it from pectin as well. Both participants have linear production regimes in the 0-4h window used to calculate production rate. (c) Acetate, propionate and butyrate production rates from inulin and pectin in different participants, presented as Z-scores computed across all participants. Each row represents measurements from a single sample. SCFA production rates were measured ex vivo in mM/h for each participant in response to each condition. Production rates were computed between timepoints 2h and 4h for each condition and production rates from the control condition (no spike-in) were subtracted. Cellulose timepoints were indistinguishable from control and therefore were not presented.

Fig 2. SCFA production capacity can be predicted from individual bacterial OTUs.

(a) AUC values for different RFCs trained either to predict high or low SCFA content in stool at baseline, or high or low SCFA production rate ex vivo in response to specific dietary fibers. High and low production is defined according to the z-score across all participants in the study. (b) Relationship between propionate and butyrate production rates, and the relative abundance of an unassigned OTU of the Lachnospiraceae family, showing a relationship between its relative abundance and butyrate production in response to inulin, specifically. (c) Similar relationship but specific to the relative abundance of a Prevotella copri OTU and propionate production in response to inulin. (d) Feature importances from RFCs trained to predict high or low SCFA production in response to inulin from bacterial family abundances.

Fig 3. General features of an individual’s MMP are stable over time.

(a) Continuous production rates of each SCFA in response to inulin and pectin for two timepoints separated by at least six months, expressed as z-scores relative to the population in the dataset. (b) The same data, but collapsed to high or low producers of a given SCFA in response to a given fiber (red = high, blue = low).

Fig 4. Phenomenological model of in vivo SCFA production predicts that inter-individual differences affects quantities absorbed.

(a) Schematic of model parameters. (b) Predicted quantities of butyrate absorbed versus excreted in the stool in participant H as a function of the colonic epithelial absorption rate constant, assuming a transit time of 12 hours. Values of the rate constant measured by the dialysis bag and CaCo monolayer approaches discussed in the text are shown explicitly. (c) Phase diagram of predicted quantity of butyrate absorbed as a function of the epithelial absorption rate constant and transit time (participant H). Values of the rate constant measured by the dialysis bag and CaCo monolayer approaches discussed in the text are shown explicitly. (d) Predicted amount of each SCFA absorbed (in mmol) using a transit time of 12 hours for each subject and the dialysis bag rate constant parameters.