Butyrate producing colonic Clostridiales metabolise human milk oligosaccharides and cross feed on mucin via conserved pathways

Abstract

The early life human gut microbiota exerts life-long health effects on the host, but the mechanisms underpinning its assembly remain elusive. Particularly, the early colonization of Clostridiales from the Roseburia-Eubacterium group, associated with protection from colorectal cancer, immune- and metabolic disorders is enigmatic. Here, we describe catabolic pathways that support the growth of Roseburia and Eubacterium members on distinct human milk oligosaccharides (HMOs). The HMO pathways, which include enzymes with a previously unknown structural fold and specificity, were upregulated together with additional glycan-utilization loci during growth on selected HMOs and in co-cultures with Akkermansia muciniphila on mucin, suggesting an additional role in enabling cross-feeding and access to mucin O-glycans. Analyses of 4599 Roseburia genomes underscored the preponderance and diversity of the HMO utilization loci within the genus. The catabolism of HMOs by butyrate-producing Clostridiales may contribute to the competitiveness of this group during the weaning-triggered maturation of the microbiota.

Fig. 1. Growth of Roseburia and Eubacterium spp. on HMOs and upregulation of HMOs utilization loci in Roseburia.

Growth curves of R. hominis (a) and R. inulinivorans (b) on glucose, LNT, GNB, LNB, and/or purified HMOs from mother’s milk compared to no-carbon source controls over 24 h. c Growth levels of R. inulinivorans on LNT, LNB, GNB and of E. ramulus on LNT within 24 h including glucose and a no-carbon source controls. d, Growth of R. hominis, R. inulinivorans and E. ramulus on lactose, 2′FL, 3FL, 3′SL and 6′SL as well as on monosaccharides from HMOs and mucin after 24 h including a non-carbon source control. Growth analyses (a–d) on media supplemented with 0.5 % (w/v) carbohydrates (for R. inulinivorans on 1% (w/v) and 4% (w/v) purified HMOs from mothers milk) were performed in independent biological triplicates. The growth data are presented as mean values with the error bars representing the standard deviations (SD) for a–c. e HMO and mucin oligomeric growth substrates in a–d. The HMO utilization loci in R. hominis (f) and R. inulinivorans (g) identified from proteomic analyses of cells growing on LNT and HMOs from mother’s milk, respectively, relative to glucose. Genes are denoted by their protein products: transcriptional regulator (Trans. R.); ABC transporter solute binding protein (RhLNBBP (f) and RiLe^a/bBP (g)); ABC transporter permease protein (PP); hypothetical proteins (HP); Glycoside hydrolase 136 (RhLnb136_I, RhLnb136_II (f) and RiLe^a/b136_I, RiLe^a/b136_II (g)); Glycoside hydrolase 112 (RhGLnbp112 (f) and RiGLnbp112 (g)); Glycoside hydrolase 29 (RiFuc29 (g)); Glycoside hydrolase 95 (RiFuc95 (g)) and histidine kinase sensory protein (His. K.) The proteomic analyses (f–g) were in biological triplicates and the log₂-fold change from the label free quantification of upregulated gene products is shown. Glycan structures presentation according to Symbol Nomenclature for Glycans (SNFG) (https://www.ncbi.nlm.nih.gov/glycans/snfg.html). Source data are provided as a Source data file labelled with the corresponding figure number and panel definition.

Fig. 2. Specificities of GH136 enzymes that mediate the HMO degradation.

a Activity of RiLe^a/b136 on fucosylated HMOs. b Activity of RhLnb136 on LNT. c Activity of ErLnb136 on LNT. a–c The hydrolysates were analysed by MALDI-ToF MS without (b, c) or with a permethylation. a Masses of methylated sugars are in parentheses and the ion peaks correspond to the Na⁺ adducts of the methylated sugars. a–c relative intensity (percentage intensity) is shown. The MALDI-ToF MS analyses (a–c) were performed from independent triplicates (one analysis from each biological enzymatic reaction replicate) and all analyses yielded similar results.

Fig. 3. Crystal structure of the GH136 lacto-N-biosidase from E. ramulus (ErLnb136).

a–c Overall structure and a semitransparent surface of ErLnb136 consisting of an N-terminal domain (ErLnb136_I, cyan-blue) and a C-terminal β-helix domain (ErLnb136_II, green). The enzyme is shown in a a view orthogonal to the C-terminal β helix domain, b the view of a rotated 180° and c a view along the axis of C-terminal β helix domain, to highlight the interaction of ErLnb136_I and ErLnb136_II. d A molecular surface top view of the active site and a close up view e to illustrate the contribution of the ErLnb136_I domain to the active site architecture, especially the tyrosine (Y145, magenta) that contributes to substrate affinity. f The weighted mF_o-DF_c omit electron density map (contoured at 4.0 σ) of the LNB unit (yellow sticks) in the active site. The water (red sphere) mediated and direct hydrogen bonds that recognize the LNB are the yellow dashed lines. d–f The catalytic nucleophile (D575) and catalytic acid/base residue (D568) are labelled in red. a–c Disordered regions (residues 180–199 and 225–241) are shown as orange dotted lines.

Fig. 4. Roseburia transport proteins mediate capture of HMOs and related host derived oligosaccharides.

a Binding analysis of HMOs and host derived oligosaccharides to RhLNBBP and RiLe^a/bBP. b, c Growth and uptake preference of R. inulinivorans on an equimolar mixture of Le^b tetraose, Le^a triose and H triose type I. b Growth level of R. inulinivorans on an equimolar mixture of Le^b tetraose, Le^a triose and H triose type I within 24 h including a no-carbon control. c Time course of the relative percentages of Le^b tetraose, Le^a triose and H triose type I in culture supernatants from b based on HPAEC-PAD analyses presented in d. d Representative HPAEC-PAD chromatograms showing time course analysis of culture supernatants of R. inulinivorans grown on YCFA media supplemented with 1.5 mM Leb tetraose, 1.5 mM Le^a triose and 1.5 mM H triose type I. Binding affinities a of RhLNBBP were determined by isothermal titration calorimetry (ITC) while binding affinities of RiLe^a/bBP were determined by surface plasmon resonance (SPR) due to low availability of the ligands and justified by the comparability of binding constants from these techniques^,. Both analyses were in independent duplicates (n = 2) and the K_D values are reported with error bars representing the error of the fit to the binding isotherms. Growth experiments b were performed as independent biological triplicates (n = 3) and triplicate HPAEC-PAD analyses c, d were performed (one analysis/per biological replicate) whereby all HPAEC-PAD analyses yielded similar results.

Fig. 5. The conservation and structure of HMO utilization loci in Roseburia.

a Global abundance of GH112, GH136_I, GH136_II and GH10 xylanase genes in 4599 Roseburia genomes illustrating the broad occurrence and conservation of the HMO utilization apparatus. b Heat map showing the segregation of GH112-containing genomes from a into different species-level genome bins (SGBs) and the corresponding relative abundance patterns of HMO utilization genes within each SGB. This data shows the frequent co-occurrence of GH136 and GH112 genes, although some Roseburia strains encode only the GNB/LNB degrading GH112. c Principal coordinate analysis of 818 Roseburia gene-landscapes defined stringently based on ≥70% identity to the GH112 and GH136 with any of the five references Roseburia genomes displayed in Supplementary Fig. 1 and including 10 proteins up- and downstream of the GH112. d The most frequently occurring gene landscapes in each Roseburia SGB, as anchored by aligning at the 3′ terminal of GH112 genes. The gene landscape analyses provide a signature for the HMO utilization loci that are defined by at least one GH112, a GH136, an ABC-transporter, and a transcriptional regulator.

Fig. 6. Model for HMOs and related host glycan utilization by Roseburia and other Lachnospiraceae.

In R. hominis, LNT, LNB and the mucin derived GNB are captured by RhLNBBP for uptake into the cytoplasm and LNT is subsequently hydrolysed to LNB. Both LNB and GNB are phosphorolyzed by RhGLnbp112 into α-d-galactose-1-phosphate and the corresponding N-acetylhexosamines GlcNAc and GalNAc, respectively. Lactose is likely hydrolysed by a canonical β-galactosidase. In R. inulinivorans, initial hydrolysis of HMOs or O-glycans from glyco-lipids/proteins occurs at the outer cell surface by RiLe^a/b136, which has two C-terminal putative galactose-binding domains. The import of degradation products is mediated by the RiLe^a/bBP-associated ABC transporter. Fucosyl decorations are removed by the concerted activity of RiFuc95 and RiFuc29 before RhGLnbp112 phosphorolyzes the resulting LNB or imported GNB into monosaccharides, as described in R. hominis. Galactose and galactose-1-phosphate products are converted via the Leloir pathway to glucose-6-phosphate and N-acetylhexosamine sugars are converted to GlcNAc-6-phosphate before entering glycolysis. The pyruvate generated from glycolysis is partly converted to butyrate. Roseburia inhabits the outer mucus layer together with A. muciniphilia. R. inulinivorans cross-feeds on sialic acid and accesses β-(1 → 4)-linked blood group A and B oligosaccharides from mucin and glyco- lipids/proteins via RiGH98. Black solid arrows show enzymatic steps established or confirmed in this study. Black dotted arrows indicate steps based on literature. Grey dotted arrows indicate butyrate production by R. hominis and R. inulinivorans from mucin in co-culture with A. muciniphilia. The glycan structure key is the same as in Fig. 1.