Muribaculaceae Genomes Assembled from Metagenomes Suggest Genetic Drivers of Differential Response to Acarbose Treatment in Mice

Abstract

The drug acarbose is used to treat diabetes and, by inhibiting α-amylase in the small intestine, increases the amount of starch entering the lower digestive tract. This results in changes to the composition of the microbiota and their fermentation products. Acarbose also increases longevity in mice, an effect that has been correlated with increased production of the short-chain fatty acids propionate and butyrate. In experiments replicated across three study sites, two distantly related species in the bacterial family Muribaculaceae were dramatically more abundant in acarbose-treated mice, distinguishing these responders from other members of the family. Bacteria in the family Muribaculaceae are predicted to produce propionate as a fermentation end product and are abundant and diverse in the guts of mice, although few isolates are available. We reconstructed genomes from metagenomes (MAGs) for nine populations of Muribaculaceae to examine factors that distinguish species that respond positively to acarbose. We found two closely related MAGs (B1A and B1B) from one responsive species that both contain a polysaccharide utilization locus with a predicted extracellular α-amylase. These genomes also shared a periplasmic neopullulanase with another, distantly related MAG (B2) representative of the only other responsive species. This gene differentiated these three MAGs from MAGs representative of nonresponding species. Differential gene content in B1A and B1B may be associated with the inconsistent response of this species to acarbose across study sites. This work demonstrates the utility of culture-free genomics for inferring the ecological roles of gut bacteria, including their response to pharmaceutical perturbations. IMPORTANCE The drug acarbose is used to treat diabetes by preventing the breakdown of starch in the small intestine, resulting in dramatic changes in the abundance of some members of the gut microbiome and its fermentation products. In mice, several of the bacteria that respond most positively are classified in the family Muribaculaceae, members of which produce propionate as a primary fermentation product. Propionate has been associated with gut health and increased longevity in mice. We found that genomes of the most responsive Muribaculaceae showed signs of specialization for starch fermentation, presumably providing them a competitive advantage in the large intestine of animals consuming acarbose. Comparisons among genomes enhance existing models for the ecological niches occupied by members of this family. In addition, genes encoding one type of enzyme known to participate in starch breakdown were found in all three genomes from responding species but none of the other genomes.

Keywords: competition; gut microbiome; longevity.

FIG 1

Comparison of novel and previously described Muribaculaceae genomes. Novel MAGs (labeled B1A through B8) are combined with publicly available genomes and MAGs as well as 30 MAGs constructed in reference that are hypothesized to reflect three polysaccharide utilization guilds (marker colors). (A) MAGs reconstructed in this study (highlighted in black) are placed in a phylogenetic context using an approximate maximum-likelihood concatenated gene tree based on an amino acid alignment of 36 shared, single-copy genes. The tree is rooted by Porphyromonas gingivalis ATCC 33277 (not shown), and four additional Bacteroidales genomes are included as an outgroup. Nodes with less than 70% confidence are collapsed into polytomies, and topological support greater than 95% is indicated (black dots on internal branches). Branch lengths indicate an estimate of expected substitutions per site. A version of this panel with GenBank accessions for all publicly available genomes is available as Fig. S1 at https://doi.org/10.5281/zenodo.4450697. (B and C) Functional comparisons are visualized by plotting the first two principal components (PCs) of an ordination based on counts of predicted proteins annotated with GH and CBM domains either aggregated by CAZy family (B) or possessing a signal peptide (C) and aggregated by OPF. Purple arrows indicate the contributions of the labeled features, and axes are labeled with the fraction of all variation accounted for by that PC. Novel MAGs (black triangles) are labeled, as are 7 previously described cultivar genomes and high-quality MAGs: “Candidatus Homeothermus arabinoxylanisolvens” (Ha), Muribaculum intestinale (Mi), Duncaniella muris (Dm), Duncaniella freteri (Df), Duncaniella dubosii (Dd), Paramuribaculum intestinale (Pi), “Candidatus Amulumruptor caecigallinarius” (Ac).

FIG 2

Diagrams of PULs plausibly reflecting activity on starch and/or dextran in B1A, B2, B3, B. thetaiotaomicron, and B. ovatus. Regions are labeled with the genome name and the interval of gene numbers. For B1A PULs, the matching gene numbers in B1B are noted in parentheses. ORFs are depicted as arrows pointed 5′ to 3′ along the coding sequence, and colors indicate homology to genes and domains known to participate in either starch or dextran utilization. ORF outlines indicate predicted localization based on the presence of an N-terminal signal peptide and nearby residues. Matching numbers indicate homology based on OPF clustering and are arbitrarily assigned (1, Opf01209; 2, Opf02007; 3, Opf02000; 4, Opf00042; 5, Opf01405; 6, Opf02584; 7, Opf01765; 8, Opf00431; 9, Opf09589; 10, Opf15294; 11, Opf14773; 12, Opf01209; 13, Opf03138; 14, Opf04347; 15, Opf04327; 16, Opf16791).

FIG 3

Visualization of differential gene content in two OTU-1 populations. Heatmaps depict relative mapping coverage of metagenomes against putative protein-coding genes in MAGs B1A (left of gray line) or B1B (right). Rows represent one or more pooled libraries for each mouse included in the study, and columns represent individual genes. Alternating black and white spans over heatmap columns indicate contig boundaries in each MAG; the orientation and order of contigs is arbitrary. All coverage values are normalized to the median coverage of that genome’s features within each mouse. The site at which each mouse was housed is indicated by colored spans on the left (UT, dark green; UM, dark blue), and mice identified as unambiguous representations of each population are indicated (B1A, light blue; B1B, light green; uncertain, white). Rows are ordered based on a hierarchical clustering by cosine distance, depicted in the tree on the left.