Dietary selection of metabolically distinct microorganisms drives hydrogen metabolism in ruminants

Volatile fatty acid production and absorption is modulated by diet

Animals were adapted to a starch-rich diet by gradually increasing dietary concentrate content from 50 to 90% over the three 100-d experimental periods (Fig. 1A). With both diets, rumen structure and epithelial morphology were robust and rumen pH remained above 6.0 (Fig. 1A, D Additional Table S4), indicating a healthy rumen function throughout the entire experiment.

Cattle with high fiber intake exhibited greater fiber digestibility, while cattle adapted to the starch-rich diet had greater organic matter and starch digestibility but lower fiber digestibility in vivo and in the in vitro experiment 1 (Fig. 1B, C, Additional Table S4, p < 0.001). Cattle adapted to the starch-rich diet exhibited greater rates of carbohydrate degradation, leading to increased total VFA concentration in periods 1 and 2 in vivo (Additional Table S4), and in the in vitro experiment 1 (Fig. 1C, p < 0.001). The lower VFA concentration (p < 0.01) observed in vivo at period 3 with the starch-rich diet could have been the result of a greater increase in the rate of VFA absorption than production, as suggested by the increased copy number of some of the genes related to VFA absorption and intracellular pH regulation in the ruminal epithelium (Fig. 1F), namely the limiting enzyme in ketone formation 3-hydroxy-3-methylglutaryl-CoA lyase (HMGCL), 3-hydroxy-3-methylglutaryl-CoA synthase isoform 2 (HMGCS-2), Na+/H+ exchanger 2 (NHE-2), and Na+/H+ exchanger 3 (NHE-3) [34,35,36]. Changes in VFA absorption between diets are also suggested by the alternation of rumen organ structure, with a lower ratio of rumen organ to carcass and shorter rumen papilla in the starch-rich diet (Fig. 1D, E).

Acetate molar percentage and the acetate to propionate molar ratio were higher in cattle adapted to the fiber-rich diet (p < 0.001). Such differences in VFA profiles agree with the in vitro ruminal fermentation results (Fig. 1B, C, Additional Table S4). The starch-rich diet increased (p < 0.001) ruminal dissolved H2 (dH2, 2.5-fold higher) and reduced CH4 production (p = 0.003). Altogether, these results suggest that the two diets differentially influence VFA production, H2 metabolism, and methanogenesis, likely due to the promotion of distinct microbial activities.

Different microbial taxa were enriched by the two diets

16S rRNA gene sequencing was used to examine rumen microbial community composition. Across the 24 samples, 3028 amplicon sequence variants (ASVs) were observed. Microbial composition clearly clustered by diet, based on unweighted UniFrac dissimilarity (Fig. 2A), indicating that carbohydrate type drove marked changes in the microbial community composition. In addition, the random forest was further applied to classify groups and selected the 20 most important bacterial genera ranked by mean decrease accuracy of classification of bacterial communities by diet (Fig. 2B, Additional File S1). Among them, Ruminobacter and Succinivibronaceae UCG-002 were the two most representative genera for the starch-rich diet, whilst Fibrobacter was the most representative genus in case of the fiber-rich diet. Furthermore, correlation network analysis of these representative bacterial genera revealed strong positive correlations within each treatment, and negative correlations between the treatments (p < 0.05, Spearman’s | | > 0.5, Additional Fig. S3).

Fig. 2: Fiber-rich and starch-rich treatments select distinct microbial communities.
figure 2

A principal coordinate analysis (PCoA) profile of ruminal bacterial community based on unweighted UniFrac dissimilarity matrix at the taxa (ASVs) level (PERMANOVA, p = 0.001, R2 = 0.19); B random forest analysis of rumen bacterial community. The y-axis, from top to bottom, displays the genera ranked by their relative importance based on mean decrease accuracy in the classification of groups; C relative abundance of the top five phyla; D relative abundance of genus Fibrobacter, Ruminobacter, and Treponema; E the relationships of feed digestibility and volatile fatty acids with bacterial community at genus level (average percent > 0.5%). No rank means there is no specific taxonomic information at the genus level. Only Spearman’s significance levels p < 0.05 are shown, yellow and blue indicate negative and positive correction respectively. OM organic matter; NDF neutral detergent fiber. Significance was tested using independent two-group Wilcoxon rank-sum tests. Data with error bars are expressed as mean ± standard error. **p < 0.01, ***p < 0.001, n = 12/group.

At the phylum level, Bacteroidota (Bacteroidetes) and Fibrobacterota (Fibrobacteres) were significantly enriched with the fiber-rich diet, whereas Proteobacteria and Spirochaetota (Spirochaetes) were more abundant with the starch-rich diet (Fig. 2B, p < 0.001). It should be noted that, due to the specificity of the primer pairs used (targeting the V3–V4 primer), Spirochaetota and archaea are potentially underrepresented in this analysis [37, 38]. At the genus level, Fibrobacter (37.3-fold higher), Christensenellaceae R-7 group (2.6-fold higher), and Bacteroidales RF16 group (4.0-fold higher) were enriched with the fiber-rich diet, while Ruminobacter (12.0-fold higher), Treponema (6.9-fold higher), and Succinivibrionaceae UCG-002 (21.8-fold higher) were enriched with the starch-rich diet (Fig. 2D, Additional Fig. S2, p < 0.001). Four of these genera were also among those with the highest discriminatory power based on the random forest analysis (Fig. 2B). In addition, Fibrobacter, Bacteroidales RF16 group, and Christensenellaceae R-7 group were positively correlated with acetate concentration and neutral detergent fiber (NDF) digestibility, while Ruminobacter, Succinivibrionaceae UCG-002, and Treponema were positively correlated with propionate concentration and starch digestibility (Fig. 2E). Differences in microbial composition are thus associated with distinct substrate preferences and degradation abilities.

Distinct pathways of carbohydrate degradation are selected by two diets

After removing the reads assigned to the host, we obtained 503 Gb of paired-end sequencing data, which averaged 21.0 Gb (ranging from 17.4 to 29.1 Gb) per sample. We first investigated the metabolic capacities of the rumen microbiomes by using the Kyoto Encyclopedia of Genes and Genomes (KEGG) [39]. There were 71.4 and 33.9% genes assigned to KEGG Orthology (KO) database and KEGG pathways, respectively. Principal coordinate analysis (PCoA) of all KO genes showed that the fiber-rich and starch-rich diets selected for different metabolic functions (Fig. 3A), with “carbohydrate metabolism pathways” being the most abundant category that was significantly different between diets (Additional Fig. S4).

Fig. 3: Fiber-rich and starch-rich treatment exhibits distinct genes of CAZymes and KEGG enzymes enriched in rumen microbiome.
figure 3

A PCoA profile of all KO genes (PERMANOVA, P < 0.001, R2 = 0.70); B relative abundance of total of CAZymes genes; C PCoA profile of the GH family (PERMANOVA, P < 0.001, R2 = 0.64); D gene abundance of the top five GH family enzymes, and their phylogenetic distribution enriched per phylum (AA auxiliary activity, CBM carbohydrate-binding module, CE carbohydrate esterase, GH glycoside hydrolase, GT glycosyltransferase, PL polysaccharide lyase); E acetate, butyrate, propionate and methanogenesis pathways of KEGG enzymes expressed as ratios of alignments of fiber-rich versus starch-rich treatment (log2 ratio); and pie charts show the phylogenetic distribution of the pathways enriched in each treatment at phylum. Ace-P acetate production pathway, Pyr-But-P pyruvate to butyrate production pathway, Ace-But-P acetate to butyrate production pathway, Pro-Lac-P propionate (lactate) pathway, Pro-Suc-P propionate (succinate) pathway, Met methanogenesis pathway. All KO genes in enriched pathways were assigned to the identified phylum. Details of all KO genes together with identified genus and pathways are in additional files S4 and S5. Significance was tested using independent two-group Wilcoxon rank-sum tests. Data with error bars are expressed as mean ± standard error. Asterisks denote significant adjusted p values: **p < 0.01, ***p < 0.001, n = 12/group.

We screened for carbohydrate-active enzymes (CAZymes) in the assembled metagenomic contigs (Additional File S2). High starch intake increased the abundance of total CAZymes (p = 0.01), including auxiliary activities (AAs), carbohydrate-binding module (CBMs), and glycosyltransferases (GTs) (Fig. 3B). Furthermore, the rank of the top five carbohydrate-active enzyme classes and top ten assigned phyla or genera differed between the two treatments (Additional Fig. S5). The GHs had the highest relative abundance among the six CAZyme classes, and the abundance of GH subfamilies was significantly different between the two dietary treatments (Fig. 3C). The fiber-rich diet selected for a greater abundance of β-xylosidase GH43 (p = 0.001, 1.3-fold higher), which was assigned to Prevotella, Bacteroides, and Fibrobacter at the genus level (Additional Fig. S6). Abundance of α-amylase GH13 was greater with the starch-rich diet (Fig. 3D, Additional File S3, p < 0.001, 1.4-fold higher) and was assigned to Prevotella, Ruminobacter, and Clostridium (Additional Fig. S6). Thus, both dietary treatments promoted bacterial communities with distinct capacities to digest carbohydrates.

We then analyzed the abundance of genes encoding for enzymes related to the pathways of acetate, propionate, butyrate, and methanogenesis (Additional File S4). The fiber-rich diet enriched for the pathways of acetate production (pta gene), acetate to butyrate production (acs gene), propionate production via succinate as intermediate (MUT and sdhA genes), and methanogenesis (mcrA and mcrB genes), while the starch-rich diet enriched for the acrylate pathway of propionate production (ldh and pct genes) (Fig. 3E, Additional Fig. S7, p < 0.05). Bacteroidota, Firmicutes, and Proteobacteria were the major phyla assigned to the acrylate pathway in the starch-rich diet. With the fiber-rich diet, Fibrobacterota was prominent in fermentative acetate and propionate production, and as expected, Firmicutes was the major phylum for acetate to butyrate production pathway, and Methanobacteriota (Euryarchaeota) was the assigned phylum for methanogenesis (Fig. 3E, Additional File S5).

Distinct hydrogen production and incorporation pathways are selected by two diets

Given the significant differences observed in VFA profiles, dH2 concentrations, and CH4 production (Fig. 1B, C), we screened for the genes encoding for the catalytic subunits of H2-producing and H2-consuming enzymes in the assembled contigs. Of the 2,686 genes identified, 82%, 17% and 1.2% were annotated as [FeFe]-, [NiFe]- and [Fe]-hydrogenases respectively. Hydrogenases were taxonomically assigned to 149 genera, including Bacteroides, Clostridium, Oscillibacter, Methanobrevibacter, and Ruminococcus (Additional File S6). As the two diets exhibited distinct hydrogenase composition (Additional Fig. S8), we then classified hydrogenases into subgroups. The trimeric group A3 [FeFe]-hydrogenases, which mediate the process of electron confurcation during fermentative carbohydrate degradation leading to H2 production [40], were the most abundant hydrogenases with both treatments (Fig. 4A). The diaphorase subunit (HydB) of these hydrogenases was highly enriched with the fiber-rich diet (p < 0.01, 1.8-fold higher), and was primarily encoded by Firmicutes and Bacteroidota at the phylum level (Additional File S6).

Fig. 4: Fiber-rich and starch-rich treatment resulted in different hydrogenase and terminal reductases level in rumen microbiome.
figure 4

A Hydrogenase genes distributions assigned by phylum; B genes of associated terminal reductases distributions assigned by phylum. Fermentative hydrogenases (group B, A1 and A2 FeFe-hydrogenases), electron-bifurcating hydrogenases (group A3 and A4 FeFe-hydrogenases), energy-converting hydrogenases (bidirectional; group 4a, 4c, 4d, 4e, 4f and 4g NiFe-hydrogenases), methanogenic hydrogenases (Fe-hydrogenases, group 3a, 3c, 4h, 4i and 1k NiFe-hydrogenases), respiratory hydrogenases (group 1a, 1b, 1c, 1d, 1k, 2a), sensory hydrogenases (group C FeFe-hydrogenases). HydB hydrogenase-associated diaphorase. NifH nitrogenase. H2 uptake pathways can be coupled to fumarate reduction (FrdA fumarate reductase), nitrate ammonification (NrfA, ammonia-forming nitrite reductase; NarG, dissimilatory nitrate reductase; NapA, periplasmic nitrate reductase), sulfate and sulfite reduction (AprA adenylylsulfate reductase; AsrA alternative sulfite reductase; DsrA, dissimilatory sulfite reductase), dimethyl sulfoxide and trimethylamine N-oxide reduction (DmsA DMSO and TMAO reductase), reductive acetogenesis (AcsB, acetyl-CoA synthase), aerobic respiration (CydA cytochrome bd oxidase), and methanogenesis (McrA methyl-CoM reductase). No rank means there is no specific taxonomic information at the phylum level. Only genes of average relative abundance > 1 were shown, while other were shown in additional file S6. Significance was tested using independent two-group Wilcoxon rank-sum tests. Data with error bars are expressed as mean ± standard error. **p < 0.01, ***p < 0.001, n = 12/group.

The fiber-rich diet enriched genes for methanogenic hydrogenases (group 3a [NiFe]-hydrogenases and [Fe]-hydrogenases) encoded by Methanobacteriota (Fig. 4A, p < 0.05, 1.5-fold higher). The starch-rich diet enriched genes encoding for fermentative group B [FeFe]-hydrogenases (p < 0.01, 2.2-fold higher), energy-converting group 4e and 4g [NiFe]-hydrogenases, respiratory group 1d [NiFe]-hydrogenases, and sensory group C3 [FeFe]-hydrogenases (Fig. 4A, p < 0.05), which were taxonomically assigned mainly to Firmicutes and Spirochaetota (Additional File S6). These results suggest the starch-rich diet resulted in enhanced H2 flow, which was consistent with greater fermentation and higher rumen dH2 concentration (Fig. 1B).

We further analyzed signature genes that support hydrogenotrophic growth, including methanogenesis, acetogenesis, fumarate reduction, sulfidogenesis, nitrate reduction, and aerobic respiration [5]. The fiber-rich diet selected for signature genes associated with methanogenesis, including methyl-CoM reductase (mcrA, Fig. 4B, p = 0.009, 1.6-fole higher), which were mainly affiliated with Methanobrevibacter (Additional File S6). The starch-rich diet selected for a higher abundance of group 1d [NiFe]-hydrogenases (p < 0.05) encoded by Firmicutes, a membrane-bound type of enzymes known to support hydrogenotrophic respiration, along with the signature gene for sulfate reduction (aprA, p = 0.002, 3.0-fold higher). An unexpected result was that the marker gene for hydrogenotrophic acetogenesis (acetyl-CoA synthase, acsB, p < 0.01) was threefold more abundant with the fiber-rich diet (Fig. 4B). Acetyl-CoA synthase is a reversible enzyme, for example mediating acetate utilization in various sulfate-reducing bacteria [41], but most reads (about 75%) were most closely related to those of acetogenic Firmicutes.

Genome mapping and in vitro fermentation verifies microbial functions

The Hungate1000 collection of assembled genomes, which includes virtually all of the bacterial and archaeal species that have been cultivated from the rumen of diverse ruminant species [30, 42], was used to gain a stronger strain-level understanding of the findings of this study. One million random reads from each sample were extracted and aligned to the Hungate1000 genomes, resulting in an average mapping rate of 20.2% (Additional Fig. S9). We found that 257 and 168 genomes were more enriched by the fiber-rich and starch-rich diets, respectively (Additional Fig. S10). Genomes encoding for GH43 were more enriched with the fiber-rich diet and assigned to Bacteroides, Prevotella, and Fibrobacter, while genomes encoding for GH13 were enriched with the starch-rich diet and assigned to Ruminobacter and Clostridium. The fiber-rich diet selected 8 out of 10 putatively methanogenic genomes, which harbor unique hydrogenases (group 3a, 3c, 4h, 4i [NiFe]-hydrogenases and [Fe]-hydrogenases) and the signature gene mcrA, as well as putatively 5 out of 8 hydrogenotrophic acetogenic bacterial genomes encoding electron-bifurcating hydrogenases (group A3, A4 [FeFe]-hydrogenases) or the signature gene acsB. The starch-rich diet selected for 67 out of 130 genomes encoding fermentative hydrogenases (group A1, A2, and B [FeFe]-hydrogenases), assigned to Lachnospirales and Selenomonadales, and 17 out of 33 genomes encoding for respiratory hydrogenases (group 1 and 2a [NiFe]-hydrogenases) assigned to Selenomonadales. These genomes with terminal sulfate and sulfite reductases were more enriched with the starch-rich diet, and mainly assigned to Lachnospirales (Additional File S7).

We performed an analysis of ranks of attributes based on differential abundance between two treatments. Fibrobacter succinogenes subsp. elongatus strain HM2 and Lachnospiraceae bacterium AC2028, both of which have functions in fiber degradation [30], were the most representative genomes enriched with the fiber-rich diet (p < 0.001). Both Ruminobacter sp. RM87 and Succinimonas amylolytica DSM 2873, known to degrade starch [30], were the most representative genomes with the starch-rich diet (Fig. 5A, p < 0.001, Additional Fig. S11). We validated by q-PCR that 16S rRNA gene copies of F. succinogenes and Ruminobacter amylophilus were the more abundant in the fiber-rich and starch-rich treatments, respectively (Additional Fig. S12, p < 0.001). Enhanced fiber degradation was further verified by in vitro experiment 2, which indicated that inoculating microbiome of fiber-rich diet resulted in greater NDF degradation (Fig. 5B, p = 0.026).

Fig. 5: Metabolic features enriched in microbiome of fiber-rich and starch-rich diets in relation to function.
figure 5

A metabolic features of representative microorganisms. Reads from each sample were aligned to sequenced genomes of cultured rumen microorganisms in the Hungate1000 collection using the burrows-wheeler alignment tool. The hydrogenase function, substrate utilization, and metabolite production of each microorganism based on the known growth characteristics [5, 30] are colored in blue, orange, and green, respectively. The ratios between alignments of fiber-rich/starch-rich samples to each genome are presented, and data were expressed as log2 (ratio). Genomes attributes: fibrolytic bacteria, Fibrobacter succinogenes subsp. elongatus strain HM2 and Lachnospiraceae bacterium AC2028; amylolytic bacteria, Ruminobacter sp. RM87 and Succinimonas amylolytica DSM 2873; acetate producer, Lachnospiraceae bacterium G41 and Fibrobacter succinogenes subsp elongatus strain HM2; butyrate producer, Lachnospiraceae bacterium AC2028 and Butyrivibrio sp. LB2008; lactate utilizer and propionate producer, Megasphaera elsdenii strain J1 and Anaerovibrio lipolyticus LB2005; hydrogenotrophic methanogen, Methanobrevibacter thaueri strains DSM 11995 and sp. YE315; acetogens, Acetitomaculum ruminis DSM 5522. Detailed information of other genomes is shown in additional file S7; B methane (CH4) production, neutral detergent fiber (NDF) degradation, pH and volatile fatty acid (VFA) profile of in vitro rumen experiment 2 with rice straw fermentation by inoculating fiber-rich or starch-rich selected microbiome. Data with error bars are expressed as mean ± standard error. **p < 0.01, ***p < 0.001, n = 12/group.

We selected the two most differential abundant genomes of microorganisms inferred to produce acetate, propionate, or butyrate (Additional File S7). The microbiome of the fiber-rich diet favored acetate and butyrate production while being enriched with Lachnospiraceae bacterium G41 and AC2028, F. succinogenes subsp. elongatus strain HM2, and Butyrivibrio sp. LB2008 (p < 0.01), while the microbiome of the starch-rich diet favored propionate production with lactate as an intermediate along with enrichment of Megasphaera elsdenii strain J1 and Anaerovibrio lipolyticus LB2005 (p < 0.001, Additional Fig. S11). These results were further verified by in vitro experiment 2 by incubating rice straw inoculated with fiber-rich or starch-rich selected microbiomes (Fig. 5B). Accordingly, five of six representative genomes encoded fermentative, electron-bifurcating and energy-converting hydrogenases, while one genome (i.e., Megasphaera elsdenii strain J1) had the capacity to use H2 as a substrate for VFA production (Fig. 5A). These results further upheld that dietary carbohydrate selected for microbiomes with distinct pathways of VFA production and H2 metabolism.

We selected genomes encoding marker genes for hydrogenotrophic methanogenesis (mcrA) and acetogenesis (acsB). The most differentially abundant genomes were Methanobrevibacter thaueri strains DSM 11995 and sp. YE315, as well as the hydrogenotrophic acetogen Acetitomaculum ruminis DSM 5522 [43], which were all enriched with the fiber-rich diet (Fig. 5A and Fig. S11, p < 0.01, Additional File S7). Both M. thaueri strains DSM 11995 and sp. YE315 encode methanogenic hydrogenases to use H2 to reduce CO2 to CH4, while A. ruminis DSM 5522 encodes electron-bifurcating group A3 [FeFe]-hydrogenases predicted to use H2 to reduce CO2 to acetate (Fig. 5A). The acsB gene is also encoded in some butyrate-producing bacteria (e.g., Eubacterium limosum), which produces acetate through the Wood–Ljungdahl pathway and converts it to butyrate [44, 45]. In turn, in vitro experiment 2 where rice straw was the sole substrate indicated that inoculation of the fiber-rich selected microbiome resulted in lower CH4 and greater acetate and butyrate production compared to the starch-rich selected microbiome (Fig. 5B, p < 0.05). This suggests the possibility that the fiber-rich diet may have selected for hydrogenotrophic acetogens, which may use some of the H2 that would otherwise be used for methanogenesis to produce acetate.

Read more here: Source link