Colonic Butyrate-Producing Communities in Humans: an Overview Using Omics Data

Abstract

Given the key role of butyrate for host health, understanding the ecology of intestinal butyrate-producing communities is a top priority for gut microbiota research. To this end, we performed a pooled analysis on 2,387 metagenomic/transcriptomic samples from 15 publicly available data sets that originated from three continents and encompassed eight diseases as well as specific interventions. For analyses, a gene catalogue was constructed from gene-targeted assemblies of all genes from butyrate synthesis pathways of all samples and from an updated reference database derived from genome screenings. We demonstrate that butyrate producers establish themselves within the first year of life and display high abundances (>20% of total bacterial community) in adults regardless of origin. Various bacteria form this functional group, exhibiting a biochemical diversity including different pathways and terminal enzymes, where one carbohydrate-fueled pathway was dominant with butyryl coenzyme A (CoA):acetate CoA transferase as the main terminal enzyme. Subjects displayed a high richness of butyrate producers, and 17 taxa, primarily members of the Lachnospiraceae and Ruminococcaceae along with some Bacteroidetes, were detected in >70% of individuals, encompassing ~85% of the total butyrate-producing potential. Most of these key taxa were also found to express genes for butyrate formation, indicating that butyrate producers occupy various niches in the gut ecosystem, concurrently synthesizing that compound. Furthermore, results from longitudinal analyses propose that diversity supports functional stability during ordinary life disturbances and during interventions such as antibiotic treatment. A reduction of the butyrate-producing potential along with community alterations was detected in various diseases, where patients suffering from cardiometabolic disorders were particularly affected. IMPORTANCE Studies focusing on taxonomic compositions of the gut microbiota are plentiful, whereas its functional capabilities are still poorly understood. Specific key functions deserve detailed investigations, as they regulate microbiota-host interactions and promote host health and disease. The production of butyrate is among the top targets since depletion of this microbe-derived metabolite is linked to several emerging noncommunicable diseases and was shown to facilitate establishment of enteric pathogens by disrupting colonization resistance. In this study, we established a workflow to investigate in detail the composition of the polyphyletic butyrate-producing community from omics data extracting its biochemical and taxonomic diversity. By combining information from various publicly available data sets, we identified universal ecological key features of this functional group and shed light on its role in health and disease. Our results will assist the development of precision medicine to combat functional dysbiosis.

Keywords: butyrate; cardiometabolic disease; ecology; functional stability; gut microbiota.

FIG 1

Characterization of the butyrate-producing community in samples derived from healthy individuals of nine metagenomic studies (I to IX; n = 826). (A) Overview of butyrate-forming pathways, including all major genes involved. (B) Abundances of bacteria exhibiting respective pathways as percentages of total bacteria. (C) Mean number of observed taxa associated with each pathway in these samples. (D) Relative abundances of taxa comprising all pathways. Only taxa that were detected in >70% of all individuals are shown (along with Butyrivibrio crossotus, which showed high abundances in several samples); all taxa shown are known butyrate producers. Bacteria exhibiting the acetyl-CoA pathway are arranged according to their terminal enzymes, butyryl-CoA:acetate CoA transferase (but) and butyrate kinase (buk); bacteria exhibiting but and buk (both) or lacking both enzymes (alternative [alt.]) are also indicated. Members of the Lachnospiraceae are indicated in bold, Ruminococcaceae are indicated in italics, and Bacteroidetes are underlined. Black bars in violin plots (B) represent means. Error bars represent standard deviations. *, taxa detected in >90% of individuals. gct, glutaconate-CoA transferase (α, β subunit); hgCoAd, 2-hydroxyglutaryl-CoA dehydratase (α, β, and γ subunit); gcd, glutaconyl-CoA decarboxylase (α, β subunit); thl, acetyl-CoA acetyltransferase (thiolase); bhbd, β-hydroxybutyryl-CoA dehydrogenase; cro, crotonase; bcd, butyryl-CoA dehydrogenase (including electron transfer protein α, β subunit); kamA, lysine-2,3-aminomutase; kamDE, β-lysine-5,6-aminomutase (α, β subunit); kdd, 3,5-diaminohexanoate dehydrogenase; kce, 3-keto-5-aminohexanoate cleavage enzyme; kal, 3-aminobutyryl-CoA ammonia-lyase; abfH, 4-hydroxybutyrate dehydrogenase; abfD, 4-hydroxybutyryl-CoA dehydratase and vinylacetyl-CoA 3,2-isomerase (same gene); 4hbt, butyryl-CoA:4-hydroxybutyrate CoA transferase; ato, butyryl-CoA:acetoacetate CoA transferase (α, β subunit).

FIG 2

Correlation between abundances of butyrate-forming pathways (A), of terminal enzymes linked to the acetyl-CoA pathway (B), and of all individual taxa harboring a pathway (C). Analyses were performed on data derived from eight data sets (studies I to VI, VIII, and IX), where line width represents correlation strength, defined as the number of data sets displaying a correlation (P and Q < 0.05 and Spearman’s ρ > 0.4; a minimum of three correlations was required for connecting individual nodes; the dashed line in panel B represents correlations in two data sets). Node sizes reflect mean abundances (n = 813). Members of the Lachnospiraceae are indicated in bold, Ruminococcaceae are indicated in italics, and Bacteroidetes are underlined. Only taxa that were detected in >50% of samples were considered for analysis. but, butyryl-CoA:acetate CoA transferase; buk, butyrate kinase.

FIG 3

Succession of the butyrate-producing community after birth. Samples from mothers (n = 100) and their infants (1 week [n = 98], 4 months [n = 100], and 12 months [n = 100] after birth) were analyzed. (A) Abundances of bacteria exhibiting respective pathways as percentages of total bacteria (results for the glutarate pathway in newborns were manually shifted down to fit the plot layout). (B) Abundances of acetyl-CoA pathway groups, i.e., “enzyme” (cumulative abundance of all taxa exhibiting distinct terminal enzymes; enzy.) and “family” (cumulative abundance of all taxa of respective taxonomic families), as well as abundances of major individual taxa. Pathway affiliations of taxa are indicated by the color bars; members of the acetyl-CoA pathway are arranged on the family level. Significant differences (P < 0.05; *) and trends (P < 0.1; +) between mothers and their 12-month-old infants based on FDR-corrected pairwise Student t tests (pathway abundances) and Wilcoxon signed-rank tests (B) are illustrated; taxa enriched in 12-month-old infants are highlighted in bold. Black bars in violin plots represent mean values. Ruminococ., Ruminococcaceae; Por./ot., Porphyromonadaceae/other families.

FIG 4

Alterations of the butyrate-producing potential in disease. (A and B) Abundances (defined as amounts of bacteria exhibiting respective pathways as percentages of total bacteria) of total pathways (A) as well as acetyl-CoA pathway groups, i.e., “enzyme” (cumulative abundance of all taxa exhibiting distinct terminal enzymes; enzy.) and “family” (cumulative abundance of all taxa of respective taxonomic families), and of individual taxa of all pathways (B) in diseased individuals relative to healthy controls (as percent; red, decrease; blue, increase). Eight data sets encompassing type 2 diabetes (T2D; studies IV, n = 43/53; VIII, n = 185/182; and XII, n = 293/75), obesity (III, n = 123/169; IV, n = 36/7; V, n = 57/7), type 1 diabetes (T1D; XII, n = 293/31), cardiovascular disease (CVD; VII, n = 13/12), liver cirrhosis (IX, n = 114/123), inflammatory bowel disease (IBD—ulcerative colitis [UC] and Crohn’s disease [CD]; III, n = 14/21/4, respectively), and colorectal cancer (CRC; II, n = 52/52; V, n = 66/91) were analyzed (n = x/y refers to sample sizes of healthy controls/patients, respectively). Values for total pathway abundance differences (A) that are highlighted in white, bold, and italic font represent significant changes to controls (P < 0.05; FDR corrected) from linear regression analysis, whereas simple white fonts show results that tended to be different (P < 0.1; FDR corrected). In panel B, significant differences (*, P < 0.05; +, P < 0.1) based on FDR-corrected Mann-Whitney U tests of acetyl-CoA pathway groups and of individual taxa are indicated (for the IBD data set, the bootstrapped version of the test was used due to small sample sizes). Pathway affiliations of taxa are indicated by the color bars; members of the acetyl-CoA pathway are arranged on the family level. Mf+/− refers to metformin-treated or untreated (+/−) samples (IV, n = 20/33; VIII, n = 15/56; XII, n = 58/17). The symbol Σ represents results of meta-analyses for metformin treatment (ΣMf+/−; IV, VIII, and XII) and obesity (ΣObese; III, IV, and V). Table 1 has the key to Roman numerals referring to individual data sets. Por./other, Porphyromonadaceae/other families; Glut., glutarate; 4-Ami., 4-aminobutyrate.

FIG 5

The butyrate-producing community during disturbance. (A) Abundances of bacteria exhibiting respective pathways as percentages of total bacteria in metagenomic data (pathway abundance) before (day 0) and after 7 days of antibiotic treatment [day 7 (Cefpr.); n = 18, study XIII] as well as 83 days after treatment (day 90). (B) Expression of butyrate-producing pathways based on metatranscriptomic data (relative to the mean expression of three housekeeping genes) during a dietary intervention study (XI) that subjected 10 individuals to a sequential plant (plant)- and animal-product (animal)-based diet is shown. Results of samples from individual subject-specific baseline diets (base 1 [before plant-based diet] and base 2 [before animal-product-based diet]) are included. (C) Pathway abundances (study XIV) derived from metagenomic data before (day 0) and during high-fiber, low-protein dietary interventions in Chinese children suffering from genetic obesity (Prader-Willi syndrome, n = 17) and diet-related “simple” obesity (n = 21) who were sampled before interventions (day 0) and after 30 days (day 30). Prader-Willi syndrome patients were additionally sampled at days 60 (day 60) and 90 (day 90). (D) Heat map showing abundance/expression changes of acetyl-CoA pathway groups, i.e., “enzyme” (cumulative abundance of all taxa exhibiting distinct terminal enzymes; enzy.) and “family” (cumulative abundance of all taxa of respective taxonomic families), and of major individual taxa during interventions. Pathway affiliations of taxa are indicated by the color bars; members of the acetyl-CoA pathway are arranged on the family level. In panel D, only abundance changes of taxa at day 7 (cefprozil treatment) compared with day 0 (d7/0) are shown for study XIII, whereas changes in gene expression associated with respective taxa between the plant-based diet and (i) either the first baseline diet (pl./b1) or (ii) the animal-product-based diet (an./pl.) are displayed for study XI. For study XIV, results from all time points (compared with day 0) are displayed. Significant differences (P < 0.05; *) and trends (P < 0.1; +) based on FDR-corrected pairwise Student t tests (A to C) and Wilcoxon signed-rank tests (D) are illustrated. Individual violin plots were manually shifted vertically to fit the plot size; black bars represent means. Lachno., Lachnospiraceae; Rumino., Ruminococcaceae; o., other families.

FIG 6

Influence of Ac pathway diversity on temporal stability (untreated individuals) and during interventions. Data derived from the Human Microbiome Project (study I), antibiotic-treated individuals (XIII) and their untreated controls (XIII_C), and dietary interventions in Chinese children suffering from Prader-Willi syndrome (XIV_PWS) and diet-related “simple” obesity (study XIV_SO) as well as from autologous (XV_C) and allogenic (XV) transplant patients were included in the analyses. Calculations were based on abundance differences of taxa between the initial time point and all subsequent time points in each subject (I, n = 53; XIII, n = 18; XIII_C, n = 6; XIV_PWS, n = 51; XIV_SO, n = 20; XV_C, n = 20; XV, n = 20). (A) Percentage of taxa that decreased in abundance over time within each individual. Results from random communities (R, gray violin plots), constructed by 20 random samplings (function sample in R; range, −100 to +100; n = 1,000), are included. A schematic representation from highest discordance (HD; half of the taxa decrease, whereas the other half increase) to lowest discordance/highest concordance (LD; all members change in the same direction) is indicated. Variances, i.e., deviations from 50%, that were significantly higher than those in random communities are indicated (*, P < 0.05; Student’s t tests). (B) “Absolute changes,” i.e., cumulative absolute abundance changes of individual taxa between two time points including the direction (decrease/increase), which represents the overall abundance change of the entire pathway (directions of final cumulative abundance changes were omitted; hence, all values were ≥0), with “total changes,” i.e., cumulative absolute abundance change of individual taxa disregarding the direction. In the panel, results of “absolute changes” are followed by results of “total changes” for each data set. The percentages of “absolute changes” from “total changes” (means ± standard deviations) are printed for all data sets. For explanations, see the text. Black bars in violin plots represent mean values.