Xylan utilisation promotes adaptation of Bifidobacterium pseudocatenulatum to the human gastrointestinal tract

Abstract

Dietary carbohydrates impact the composition of the human gut microbiota. However, the relationship between carbohydrate availability for individual bacteria and their growth in the intestinal environment remains unclear. Here, we show that the availability of long-chain xylans (LCX), one of the most abundant dietary fibres in the human diet, promotes the growth of Bifidobacterium pseudocatenulatum in the adult human gut. Genomic and phenotypic analyses revealed that the availability of LCX-derived oligosaccharides is a fundamental feature of B. pseudocatenulatum, and that some but not all strains possessing the endo-1,4-β-xylanase (BpXyn10A) gene grow on LCX by cleaving the xylose backbone. The BpXyn10A gene, likely acquired by horizontal transfer, was incorporated into the gene cluster for LCX-derived oligosaccharide utilisation. Co-culturing with xylanolytic Bacteroides spp. demonstrated that LCX-utilising strains are more competitive than LCX non-utilising strains even when LCX-derived oligosaccharides were supplied. In LCX-rich dietary interventions in adult humans, levels of endogenous B. pseudocatenulatum increased only when BpXyn10A was detected, indicating that LCX availability is a fitness determinant in the human gut. Our findings highlight the enhanced intestinal adaptability of bifidobacteria via polysaccharide utilisation, and provide a cornerstone for systematic manipulation of the intestinal microbiota through dietary intervention using key enzymes that degrade polysaccharide as biomarkers.

Fig. 1. Comparative analysis of CAZyme composition among human-resident bifidobacteria.

The heatmap shows copy numbers of the CAZyme gene per genome of 486 strains belonging to B. pseudocatenulatum (102 strains), B. adolescentis (49 strains), B. longum subspecies longum (151 strains), B. breve (99 strains) and B. bifidum (85 strains). Each column represents a strain, and each row represents a CAZyme gene detected from at least one strain in this data set. CAZymes marked in bold and underlined represent xylan degradation-related CAZymes. Black circles indicate the gene with the highest average copy number in B. pseudocatenulatum, among other species (Mann–Whitney U test; P < 0.01). Five of these CAZymes were also shown in separated plots with mean ± standard error. See Supplementary Table S2 for the actual copy number of each CAZyme gene.

Fig. 2. LCX utilisation by the strains that possess a GH10 (BpXyn10A) gene.

Growth curves of our isolates and the type strain in the medium supplemented with XOS (a), arabinoxylan (b) or xylan (c) as the only carbohydrate source. In panel c, total organic acid production in the supernatant is also shown. ***P value of Mann–Whitney U test <0.001. d The association between the presence/absence of predicted GH10 gene and growth phenotype. + indicates an exponential increase in turbidity of >0.1. e Endo-xylanase activity of purified recombinant GH10 analysed using thin-layer chromatography. A sample was collected over time after mixing each substrate, and the enzyme was applied. X1: xylose. X2: xylobiose. X3: xylotriose. X4: xylotetraose. X5: xylopentaose. X6: xylohexaose. f Domain organisation of BpXyn10A identified using dbCAN2 [29] and the PSORT web application (http://psort.hgc.jp/form.html); GH10, catalytic module of glycoside hydrolase 10; CBM9, carbohydrate-binding module 9; TM, transmembrane region. g Localisation of endo-1,4-β-xylanase activity of the LCX-utilising and non-utilising strains. h Growth of the non-LCX-utilising strain (YIT 4072^T) in the arabinoxylan medium supplemented with multiple concentrations of purified recombinant BpXyn10A.

Fig. 3. Genetic locus and strain distribution of BpXyn10A.

a The gene arrangement around the BpXyn10A gene. Nucleotide identity of each gene was visualised using GenomeMatcher software [49]. Gene annotation details are also shown in Fig. 4a using the YIT 11952 as an example. b A phylogenetic tree based on the alignment of core genes in 35 isolates and the type strain. Strains with the BpXyn10A gene are shown in blue. Local support values at the branch nodes were computed using the Shimodaira–Hasegawa test with the default parameter settings of fasttree [50].

Fig. 4. The LCX utilisation system of B. pseudocatenulatum.

a Expression profiles of genes around the BpXyn10A gene from strain YIT 11952. Heatmap depicts log₂ fold changes of the expression on xylose (X), XOS, xylan (XY) and arabinoxylan (AX) relative to lactose. Annotation information is from Prokka [27]. Asterisks indicate functional proteins that were not annotated in Prokka, but were found to have >95% amino acid homology by Blastp searches against the NCBI nr database. Locus ID YIT11952_xxxxx are abbreviated with the last numbers after the underscore. b LCX utilisation model using arabinoxylan as an example. Initial degradation of LCX occurs by both cell-anchored and extracellularly released BpXyn10A. Degradation products were imported to the cytoplasm and further degraded by the AXH utilisation system as described previously [39]. Imported degradation products induce the expression of the BpXyn10A and AXH utilisation system.

Fig. 5. Co-culture of B. pseudocatenulatum and other HGM species in medium with arabinoxylan.

a Cell numbers of mono- and co-cultures of B. pseudocatenulatum strains with or without the BpXyn10A gene. P^a: P value of Wilcoxon signed-rank test, P^b: P value of Mann–Whitney U test. b Cell numbers of Ba. ovatus co-cultured with B. pseudocatenulatum strains with or without the BpXyn10A gene. P^b: P value of Mann–Whitney U test. c Cell number of B. longum subsp. longum co-cultured with B. pseudocatenulatum strain with (YIT 11952) or without (YIT 4072^T) the BpXyn10A gene. Data are expressed as mean of quadruplicate experiments ± standard deviation.

Fig. 6. Effect of the presence or absence of the BpXyn10A gene in the human intestine on the population of B. pseudocatenulatum and genus Bifidobacterium during LCX-rich cereal intervention.

a Study design. Participants consumed a 60 g bowl of LCX-rich food per day (see Methods). b B. pseudocatenulatum cells with BpXyn10A and total B. pseudocatenulatum, and relative abundance of genus Bifidobacterium on the participants grouped based on the presence or absence of B. pseudocatenulatum and BpXyn10A gene. Cell numbers were determined using the quantitative value of qPCR targeting the BpXyn10A and 16S rRNA genes of B. pseudocatenulatum. Relative abundance was obtained from 16S rRNA gene amplicon analysis. **P value of Wilcoxon signed-rank test <0.01; *<0.05. n.s.: not statistically significant. c PERMANOVA exploring the differences in microbial composition between groups based on unweighted Unifrac distance. d Comparison of Shannon index between groups. P values of Kruskal–Wallis test are shown.