Interspecies cross-feeding orchestrates carbon degradation in the rumen ecosystem

Abstract

Because of their agricultural value, there is a great body of research dedicated to understanding the microorganisms responsible for rumen carbon degradation. However, we lack a holistic view of the microbial food web responsible for carbon processing in this ecosystem. Here, we sampled rumen-fistulated moose, allowing access to rumen microbial communities actively degrading woody plant biomass in real time. We resolved 1,193 viral contigs and 77 unique, near-complete microbial metagenome-assembled genomes, many of which lacked previous metabolic insights. Plant-derived metabolites were measured with NMR and carbohydrate microarrays to quantify the carbon nutrient landscape. Network analyses directly linked measured metabolites to expressed proteins from these unique metagenome-assembled genomes, revealing a genome-resolved three-tiered carbohydrate-fuelled trophic system. This provided a glimpse into microbial specialization into functional guilds defined by specific metabolites. To validate our proteomic inferences, the catalytic activity of a polysaccharide utilization locus from a highly connected metabolic hub genome was confirmed using heterologous gene expression. Viral detected proteins and linkages to microbial hosts demonstrated that phage are active controllers of rumen ecosystem function. Our findings elucidate the microbial and viral members, as well as their metabolic interdependencies, that support in situ carbon degradation in the rumen ecosystem.

Fig. 1. Phylogeny and genomic sampling of 77 rumen MAGs.

Maximum likelihood tree of the ribosomal protein rpsC (S3) with reference genomes (3,140), genomes from other rumen MAG studies (345) and genomes recovered here (68). Branches are marked with coloured lines by the rumen data set where the genome originated (centre legend). Coloured circles on the outside of the tree highlight the highest-level taxonomic classification that genomes recovered in this study represent the sampling of (top left legend). The asterisks indicate MAGs containing a partial 16S rRNA gene sequence (>300 bp). The tree is rooted by Euryarchaeota (Eur). Phylum-level groups are outlined in shades of blue and are labelled on the inside circle (Tener, Tenericutes; Fi, Fibrobacteres; Sa, Saccharibacteria (TM7); Spir, Spirochaetes; L, Lentisphaera; Prot, Proteobacteria). Named groups have grey shading behind the circles, including genomes belonging to known genera. The full tree in Newick format is provided in Supplementary Dataset 11. Note, the red lines are missing for the RC9 gut group genomes because they did not contain rpsC proteins; however, the placement of these genomes was confirmed by concatenated ribosomal protein analyses (Fig. 3).

Fig. 2. Metabolic reconstruction of all 77 unique near-complete genomes in this study.

a, Maximum likelihood tree of 16 concatenated ribosomal proteins from all 77 MAGs recovered in this study. Branches are coloured by phyla. Full tree in newick format is provided in Supplementary Dataset 12. b, Heat map showing the presence of genes or pathways (listed on the right) found in each MAG (bottom). The presence of a gene or pathway is denoted by a box, coloured by taxonomic assignment. Genes or pathways that were not detected in that MAG are represented with a black box. For pathway completion, 60% of the pathway needed to be present. The functional category is denoted on the left-hand side. BP, bisphosphate; BPG, bisphosphoglycerate; G-3-P, glyceraldehyde-3-phosphate; P, phosphate; PG, phosphoglycerate; PPO, polyphenol oxidase; PPP, pentose phosphate pathway; TMAO, trimethylamine-N-oxide; PEP, phosphoenolpyruvate; rnf, Ferredoxin:NAD⁺-oxidoreductase; ech, ech-type hydrogenase.

Fig. 3. Detection and expression of PULs across known and previously undescribed Bacteroidales members.

a, Maximum likelihood phylogenetic tree of 16 concatenated ribosomal proteins from all 32 Bacteroidetes MAGs recovered here. These genomes span at least ten families within the Bacteroidales order. Lines connecting the tree to the genome name are coloured by taxonomic family assignment. Circles represent nodes with bootstrap support greater than 70, out of 100 bootstraps. Tree scale defines branch length. The full tree in Newick format is provided in Supplementary Dataset 13. b, The number of PULs encoded (grey) and expressed (coloured) from genomes. c, Expressed PULs organized by substrate. The coloured boxes indicate that at least one gene within the PUL was detected in proteomics, with two or more unique peptides. Full descriptions of PULs and detected proteins are given in Supplementary Dataset 1, Supplementary Table 5. MLGs, mixed-linked glucans.

Fig. 4. Biochemical confirmation of PUL predictions.

a, Gene organization of a predicted mannan PUL identified in Prevotella sp. PREV32, which was selected for in-depth biochemical characterization. Gene abbreviations: epim., mannobiose-2-epimerase; GPH, glycoside-pentoside-hexuronide transporter; hypo., hypothetical. b, Purified proteins incubated with six plant polymers highlighted the differential use of plant polymers by glycoside hydrolases. The bars represent an average of three replicates (white diamonds), are coloured by plant polymer and represent reducing ends (glucose equivalents) recovered. c, Incubation of purified enzymes with carob galactomannan (CGM) identified peaks matching elution times of standards, including mannose (M), β-1,4-manno-oligosaccharides (M2–M3) and manno-oligosaccharides substituted with (X) galactose residues (Gal(X)M(X)). The spectra represent one run of three replicates. ‘CGM neg’ is a negative control with CGM and no enzymes. nC, nanocoulombs. d, A hypothetical model, based on predicted protein locations^, and biochemical data, depicting a process in which galactomannan is bound via outer membrane lipoproteins (SusCD), hydrolysed via GH5B to shorter manno-oligomers (with and without galactosylations) and imported into the periplasm. Galactose residues could be removed by a GH36 (detected in proteomics and located in another part of the genome), providing manno-oligomers as substrate for the GH26 enzyme. Hydrolysed galactomannan products (mannobiose and mannose) could be further transported into the cytoplasm by the GPH transporter (trans). In the cytoplasm, the GH130 enzyme can process either mannobiose or mannosyl-glucose generated from the epimerase in the PUL. Based on the substrates identified for GH5A, it could have a role in paving the way for GH5B access to mannans in complex substrates. Other enzymes in this PUL (indicated by the asterisks) include CE7, which could contribute to the degradation of acetylated mannans, and GH26B, which could be complementing GH26A.

Fig. 5. Network analysis of plant carbon degradation.

a, CoMPP value of detected carbon polymers in winter rumen fluid, representing the relative abundance of polymers in the winter diet sample. b, Concentration of 5C and 6C sugars measured by ¹H NMR. c, Network nodes represent carbon substrates (rectangles) and MAGs (circles; 47 total). The abbreviations and antibodies used for carbon substrates are provided in Supplementary Dataset 1, Supplementary Table 7. MAGs are sized by total coverage and coloured by connectivity. Nodes are connected if proteins for degrading the substrate were detected in metaproteomics and were unique to the genome. Polymers are at the top in dark grey and sugars are at the bottom in light grey. The red stars indicate that proteins for SCFA production were detected from that MAG in metaproteomics. Edges from highly connected (more than six connections) genome nodes are outlined in red. MAGs are labelled by taxonomic assignment, with established genera named by genus and previously undescribed genomes labelled by phylum. BACT; Bacteroidetes; BUTY, Butyrivibrio sp.; FIBRO, Fibrobacter sp.; FIRM, Firmicutes; GPROT, Gammaproteobacteria; PREV, Prevotella sp.; RUM, Ruminococcus sp.; SEL, Selenomonas sp.; TREPO, Treponema sp. Names and abbreviations are also provided in Supplementary Dataset 1, Supplementary Table 1.

Fig. 6. Host–viral interaction network.

Viral genomes (hexagons) are connected to predicted microbial host genomes (circles) by edges. All genomes (viral and microbial) are sized by abundance across data sets. Viral genomes or contigs are coloured by taxonomic assignment. Microbial host MAGs are coloured by connectivity noted on the bottom right. An edge was drawn if a confident link could be established by tetranucleotide frequency (grey edges) or protospacers within the CRISPR–CRISPR-associated protein (Cas) systems (black edges). The length of the edge has no meaning. Viral genomes or contigs with proteins detected in proteomics are outlined in red.