Gut microbiota modulation with long-chain corn bran arabinoxylan in adults with overweight and obesity is linked to an individualized temporal increase in fecal propionate

Abstract

Background: Variability in the health effects of dietary fiber might arise from inter-individual differences in the gut microbiota's ability to ferment these substrates into beneficial metabolites. Our understanding of what drives this individuality is vastly incomplete and will require an ecological perspective as microbiomes function as complex inter-connected communities. Here, we performed a parallel two-arm, exploratory randomized controlled trial in 31 adults with overweight and class-I obesity to characterize the effects of long-chain, complex arabinoxylan (n = 15) at high supplementation doses (female: 25 g/day; male: 35 g/day) on gut microbiota composition and short-chain fatty acid production as compared to microcrystalline cellulose (n = 16, non-fermentable control), and integrated the findings using an ecological framework.

Results: Arabinoxylan resulted in a global shift in fecal bacterial community composition, reduced α-diversity, and the promotion of specific taxa, including operational taxonomic units related to Bifidobacterium longum, Blautia obeum, and Prevotella copri. Arabinoxylan further increased fecal propionate concentrations (p = 0.012, Friedman's test), an effect that showed two distinct groupings of temporal responses in participants. The two groups showed differences in compositional shifts of the microbiota (p ≤ 0.025, PERMANOVA), and multiple linear regression (MLR) analyses revealed that the propionate response was predictable through shifts and, to a lesser degree, baseline composition of the microbiota. Principal components (PCs) derived from community data were better predictors in MLR models as compared to single taxa, indicating that arabinoxylan fermentation is the result of multi-species interactions within microbiomes.

Conclusion: This study showed that long-chain arabinoxylan modulates both microbiota composition and the output of health-relevant SCFAs, providing information for a more targeted application of this fiber. Variation in propionate production was linked to both compositional shifts and baseline composition, with PCs derived from shifts of the global microbial community showing the strongest associations. These findings constitute a proof-of-concept for the merit of an ecological framework that considers features of the wider gut microbial community for the prediction of metabolic outcomes of dietary fiber fermentation. This provides a basis to personalize the use of dietary fiber in nutritional application and to stratify human populations by relevant gut microbiota features to account for the inconsistent health effects in human intervention studies.

Trial registration: Clinicaltrials.gov, NCT02322112 , registered on July 3, 2015. Video Abstract.

Keywords: Arabinoxylan; Dietary fiber; Gut microbiota; Inter-individual variability; Overweight adults; Short-chain fatty acids.

Conflict of interest statement

JW has received research funding and consulting fees from industry sources involved in the manufacture and marketing of fibers, and is a co-owner of Synbiotics Health, a developer of synbiotic products. All other authors declare no conflict of interest.

Figures

Fig. 1

Study design. Shaded study week blocks indicate a scheduled clinic visit. The “X” indicates the task was completed during the study week. C-DHQ II, Canadian diet history questionnaire II; stool characteristics, self-reported stool consistency and bowel movement frequency

Fig. 2

Arabinoxylan alters the global composition of fecal bacterial communities and induces distinct shifts in taxa. a Non-metric multidimensional scaling (NMDS) plot based on Euclidean distance metrics of arabinoxylan and microcrystalline cellulose groups at each time point (inter-subject β-diversity) showing changes in the distance between subjects over time. Euclidean distances b between fecal microbiotas of subjects at each study time point (inter-subject) and c between each subject’s fecal microbiota at baseline and during W1 and W6 of treatment (intra-subject). d α-Diversity (displayed as Shannon index and total OTUs) of the fecal microbiotas of subjects at each time point. e Absolute change (ΔW6–BL) in relative abundance of bacterial taxa affected by the dietary intervention. Data analyzed using PERMANOVA for a, GEE models (with Bonferroni correction) for b and d, and Mann–Whitney tests for c. For e, data were analyzed using either Wilcoxon tests to assess within-group changes relative to baseline, or Mann–Whitney tests to assess between-group changes (i.e., AX vs. MCC; with FDR correction). β-diversity and compositional data were reported as mean ± SD, and centered log-ratio transformed prior to the statistical analyses. BL baseline, OTU operational taxonomic unit, W1 week 1, W6 week 6

Fig. 3

Identification of co-abundance response groups (CARGs) during arabinoxylan supplementation. a Heatmap shows the change (ΔW6–BL) in relative abundance of 41 OTUs affected by arabinoxylan (p < 0.1, Wilcoxon test). The hierarchical dendrogram shows clustering of centered log-ratio (CLR) transformed OTUs (rows) based on Spearman’s correlation distances by the complete-linkage clustering algorithm, and then grouped on the dendrogram into seven CARGs by PERMANOVA (p < 0.05). Subjects (columns) clustered based on Euclidean distances. Colors from blue to red indicate the direction and magnitude of change. b Co-response network analysis. Each node represents an OTU, where the size is proportional to the change (ΔW6–BL) in relative abundance, the shape indicates direction of change (positive: circle; negative: square), and the color references the respective CARG to which it was clustered. Lines between nodes represent significant positive (red line) or negative (blue line) Spearman’s correlations (rs values ≥ 0.5 or ≤ − 0.5 and q values < 0.05). BL baseline, OTU operational taxonomic unit, W6 week 6

Fig. 4

Temporal and individualized responses of the OTUs and CARGs affected by arabinoxylan and microcrystalline cellulose. a Plots show the temporal response of the ten most abundant OTUs (detected in > 25% of subjects) and the seven CARGs. Centered log-ratio transformed data were analyzed by Friedman’s test (with Dunn’s correction) to assess within-group changes between time points (i.e., ΔW1–BL and ΔW6–W1). b Bubble plot shows individualized differences (ΔW6–BL) in relative proportions of the ten most abundant OTUs (percentage of total microbiota composition) and CARGs (sum of OTUs) detected after 6 weeks of arabinoxylan and microcrystalline cellulose supplementation. The size of the bubble is proportional to the change in abundance relative to baseline, while the color of the bubble represents the direction of the change (red: increase; black: decrease). The “X” indicates that the OTU was either undetected or the change was < 0.02% relative abundance. BL baseline, CARG co-abundance response group, OTU operational taxonomic unit, W1 week 1, W6 week 6

Fig. 5

Temporal and individualized output of fecal SCFAs in response to arabinoxylan and microcrystalline cellulose supplementation. a Line plots show the temporal response of acetate, propionate, and butyrate; reported as mean ± SD. b Individualized temporal propionate response of W6-responders (red) and W6-nonresponder (black) (grouped based on ΔW6–W1). Data analyzed for a and b using Friedman test (with Dunn’s correction) to assess within-group changes between time points, and for b using Mann–Whitney tests to assess differences between-group at each time point. BL baseline, CARG co-abundance response group, OTU operational taxonomic unit, SCFA short-chain fatty acid, W1 week 1, W6 week 6

Fig. 6

The individualized temporal propionate response to arabinoxylan associates with compositional responses in the fecal microbiota. Principal component analysis plots based on Euclidean distance comparing the relative abundance of fecal microbiota, both at baseline and arabinoxylan-induced shifts (ΔW6–baseline), between W6-responders (red) and W6-nonresponders (black). Microbiota variables (i.e., OTU or CARG) that contributed the most to inter-subject variation were shown as vectors on the plot when statistical significances were determined by PERMANOVA (p < 0.05). CARG co-abundance response group, OTU operational taxonomic unit, W1 week 1, W6 week 6

Fig. 7

Individualized arabinoxylan-induced propionate responses could be explained by baseline gut microbiota composition and microbiota shifts. a Heatmap shows the linear associations between the individualized propionate response (ΔW6–BL; dependent variable; columns) and microbiota profiles (BL, ΔW1–BL, ΔW6–BL; predictors; rows). Cells represent individual multiple linear regression models (with FDR correction) that assess whether the predictors explain the individualized propionate response. Multivariate microbiota data were simplified into principal component (PC) variables PC1, PC2, and PC3 prior to analysis. Each model contained the best one or two predictors of PCs, individual CARGs, or significant OTUs selected by stepwise regression. All models were adjusted by fiber dose/sex. Colors from white to red indicate relative AICc (corrected Akaike information criterion) values calculated by AICc valueHighest AICc valuex100. Lower AICc values (red) indicate higher quality models. b Scatter plots show the linear relationship between propionate responses (ΔW6–BL) and either the baseline contribution of all OTUs to PC1 or the shifts of CARG1. Color and size of each point indicate propionate response magnitude and the shaded area specifies the 95% confident interval. The top six OTUs that contributed the most to either PC1 of all OTUs or CARG1 are further provided. AX arabinoxylan, BL baseline, CARG co-abundance response group, MCC microcrystalline cellulose, OTU operational taxonomic unit, W1 week 1, W6 week 6

Fig. 8

Relationship between propionate responses to arabinoxylan and proposed primary degraders, secondary fermenters, and metabolite utilizers. a Individual multiple linear regression models determine OTU responses (ΔW6–BL) that predict the fecal propionate response (ΔW6–BL). Y-axis shows the β-coefficient for each predictor, as in the average propionate response when OTU relative abundance increases 1%. X-axis shows the p value for each predictor. All models were adjusted by fiber dose/sex, where bubble size represents the adjusted-R2. b Proposed model of bacterial cross-feeding in the gut during degradation of complex, soluble arabinoxylans. OTU operational taxonomic unit