Unraveling the pectinolytic function of Bacteroides xylanisolvens using a RNA-seq approach and mutagenesis

The human genome does not encode enzymes for the breakdown of dietary carbohydrates other than sucrose, lactose and a portion of starch [1, 2]. All other complex carbohydrates form dietary fibres that by-pass the absorptive regions of the small intestine and provide a major source of nutrients for the resident gut microbiota. Symbiotic microorganisms that reside in the human large intestine have the capacity to utilize and convert dietary fibres into simple molecules such as the short chain fatty acids acetate, propionate and butyrate that provide an important energy source to the host and additional health benefits [3–7]. Among the main dietary fibres (resistant starches, cellulose, hemicelluloses and pectin), resistant starches have received the greatest scientific focus with nutritional studies aimed at determining the effect of resistant starch consumption on the intestinal microbiota and host health [6, 8, 9]. On the other hand, plant cell wall (PCW) polysaccharide components are much more difficult to study because of their molecular complexity and the difficulty preparing homogenous and pure fractions. This is particularly true for pectin, which is composed of as many as 17 different monosaccharides and more than 20 different linkages [10]. Pectin is made of three major pectic polysaccharides, all containing D-galacturonic acid (GalA). The first one, homogalacturonan (HG), is a linear polymer consisting of 1,4-linked α-D-GalA that may be both methyl-esterified and acetylated. The second polysaccharide, rhamnogalacturonan I (RGI), consists of the repeating disaccharide [/4)- α-D-GalA-(1/2)- α-L-Rha-(1/] to which a variety of different glycan chains (principally α-L-arabinan and arabinogalactan) are attached to the L-rhamnose residues. The third polysaccharide bears the confusing name of rhamnogalacturonan II (RGII), although it is a backbone of HG rather than RG, with complex side chains of rare sugars attached to the GalA residues. Structurally, RGI and HG constitute the ‘backbone’ of pectic polymers. However, another structure has recently been proposed in which HG is a long side chain of RGI [11]. RGI is the most heterogeneous of these three pectic polysaccharides because of its diverse sugar composition and the variation in length of sugar side chains whereas RGII is thought to have a highly conserved structure [11].

It has been reported that around 4 to 5 g of pectins are consumed each day in a normal western diet [11] and that approximately 90 % reach the colon [12]. Yet, there is no clear-cut observation on the impact of pectin on the intestinal microbiota [13] although dietary pectin consumption has been shown to provide health benefits to the host lowering cholesterol and serum glucose level [14, 15], with a possible functional role of intestinal microbiota in these benefits [16]. Likewise, little is known on human intestinal pectin degraders and on the enzyme systems involved in pectin breakdown [13]. Several bacterial isolates belonging to the Bacteroides and Faecalibacterium genera have been described to be able to grow on pectin as the sole carbon source [17–19]. Because of the molecular complexity of pectin, these bacteria must produce an arsenal of carbohydrate active enzymes (CAZymes; http://www.cazy.org/; [20]). Only one study has investigated pectinolytic enzyme systems in human intestinal bacteria and this was done on Bacteroides thetaiotaomicron and Bacteroides ovatus using a transcriptomic microarray approach. This study underlined the complexity of this fibrolytic function since several different Polysaccharide Utilization Loci (PUL) were identified in the two species [17]. Furthermore, recent studies have also investigated hemicellulose degradation by Bacteroides species and have shown the existence of one and two PUL(s) dedicated to xyloglucan and xylans, respectively [21, 22].

Consequently, the objective of the present study was to further investigate the PCW polysaccharide degradation by the human gut microbiota by focusing on the pectin enzyme system(s) of Bacteroides xylanisolvens XB1A, a xylanolytic species belonging to the human core microbiome [23] that exhibits pectinolytic activities.

This study was undertaken to elucidate the pectinolytic function of B. xylanisolvens XB1A, knowing that pectin is a natural dietary component found in fruits and vegetables and citrus and apple pectins are widely used in food industry [11].

Citrus pectin has been shown to be rich in GalA (>70 % total sugars) and to be highly methylated, indicating that it consists predominantly of HG [27, 28]. Its major neutral sugars are D-galactose, L-arabinose and L-rhamnose, indicating the presence of RGI branched with neutral sugar side chains like α-L-arabinan and arabinogalactan [28]. The monosaccharide composition as well as the degree of methylation of the commercial citrus pectin used in this study confirmed this description (Additional file 2: Table S1). B. xylanisolvens must then deploy enzymes targeting primarily HG and RGI. CAZymes involved in pectin degradation must disrupt esterified groups (mainly methyl groups) and neutral sugar side chains from the backbone, and also cleave the glycosidic bonds within the GalA-containing polysaccharide backbones (HG, RG-I, RG-II) and within the neutral sugar side chains [13].

RNA-seq transcriptomic analysis showed that PULs 2, 13, 49, 50, 51 and 68 play a major role in citrus pectin breakdown (Fig. 1). Taking glucose substrate as reference condition, PUL 49 (BXY_31910 to BXY_32010) and 50 (BXY_32120 to BXY_32390) were the most up-regulated with citrus pectin. PUL 49 composition in CAZyme genes (two PL1, two CEs) suggests that it targets homogalacturonan (Table 1). Homologous PULs were found to be up-regulated on purified HG in Bacteroides thetaiotaomicron VPI-5482 (PUL 75) and Bacteroides ovatus ATCC 8483 (PUL 100), hence confirming this hypothesis (Fig. 7; http://www.cazy.org/PULDB/) [17]. B. xylanisolvens PUL 50 encodes CAZymes targeting L-rhamnose-containing (GH28, GH105, GH106, PL11) and D-galactose-containing oligosaccharides (GH2, GH3, GH42), suggesting that this PUL could be dedicated to RGI. The organization of this PUL presents similarities with a portion of PUL 77 in B. thetaiotaomicron VPI-5482 shown to be overexpressed on purified RGI [17]. PUL 51 (BXY_32430 to BXY_32730) was less overexpressed on citrus pectin than PUL 49 and PUL 50. Its composition in CAZyme genes suggests that it encodes enzyme that attack both the HG and RGI pectin backbone (Table 1). PUL 13 (BXY_15570 to BXY_15720) and PUL 68 (BXY_45310 to BXY_45350) were particularly overexpressed at early stages (mid-log phase) of B. xylanisolvens XB1A growth while PUL 2 (BXY_06320 to BXY_06540) was upregulated at late stages (late-log phase) of growth. Although it is not possible to propose a role for PUL 68 in pectin breakdown, PUL 13 encodes CAZymes potentially active on neutral sugar side chains since the three GH43 and two GH51 enzymes may target arabinose-containing oligosaccharides. In B. thetaiotaomicron VPI-5482, PUL 7 (http://www.cazy.org/PULDB/) whose organization is almost identical to PUL 13 was indeed shown to be induced on arabinans [17]. The fact that PUL 13 (and maybe PUL 68) are overexpressed at early stages of bacterial growth suggests that the targeted oligosaccharides are the most accessible pectin moieties and/or the easiest to hydrolyze. Finally, we assume that PUL 2 targets the most resistant pectin polysaccharide RGII. This assumption is based on the fact that PUL 2 is homologous to PUL 89 which is up-regulated on purified RGII in B. thetaiotaomicron VPI-5482 [17]. Actually, PUL 2 is overexpressed at late stages of B. xylanisolvens growth, once the easy hydrolysable pectic molecules i.e. HG and RGI have been consumed and rendered less available. Also, PUL 2 encodes a GH95 enzyme putatively able to attack the complex side chains of rare sugars in RGII (Table 1), enzyme that is not present in the other PULs.

Like citrus pectin, apple pectin contains more than 70 % of highly-methylated HG [29], and this was confirmed by the analysis of the commercial citrus and apple pectins used in this study. The difference between the two pectins may reside in the rhamnogalacturonans, with longer RGI regions in apple pectin and higher content in neutral sugars [30]. Also, several articles on apple pectin have reported contamination by co-extracted starch [30, 31]. A contamination with 4.4 % starch was found in the apple pectin used in this study and may explain the fact that PUL 71 was the most overexpressed PUL by B. xylanisolvens on this substrate. Indeed, this PUL must encode a starch utilization system considering its homology with the starch utilization locus (PUL 66, http://www.cazy.org/PULDB/) described in B. thetaiotaomicron [25]. Otherwise, PUL 49 and 50 were also up-regulated with apple pectin, as observed with citrus pectin, although the intensity of the response was lower on apple pectin, probably because the bacterium also consumed starch.

Once treated with amylases, the composition of apple pectin was very similar to that of citrus pectin if we consider their content in GalA, rhamnose, galactose and the degree of methylation. Nevertheless, apple pectin contained two-fold less arabinose, three-fold more glucose and five-fold more xylose than citrus pectin, suggesting a lower extent in arabinan side chains and a residual contamination with starch and PCW polysaccharide moieties. Using this amylase-treated apple pectin, and through the analysis of the relative expression of susC-like genes, we validated the induction of five PULs over the six PULs identified with citrus pectin and confirmed their involvement in pectin-utilization by B. xylanisolvens XB1A.

Overall, this transcriptomic study using both RNA-seq and RT-qPCR data obtained with two pectin sources and at two stages of bacterial growth, pointed to two overexpressed PULs i.e. 49 (BXY_31910 to BXY_32010) and 50 (BXY_32120 to BXY_32390) that we assumed targeted HG and RGI, respectively. Considering that HG is the most abundant moiety within pectin, PUL 49 must therefore play a major role in pectin deconstruction. To confirm this hypothesis, PUL 49 was mutated on the susC-like gene, which encodes a membrane TonB-dependent transporter allowing internalization of pectic oligomers within the bacterial periplasm, and hence the mutation is expected to block substrate utilization by the bacterium [25]. The analysis of the mutant in comparison to the wild type strain confirmed our hypothesis since the growth of the mutant on citrus pectin and amylase-treated apple pectin was strongly affected, both in rate and density. However growth was not completely abolished, suggesting that the bacterium could still utilize pectin moieties by means of the other enzyme systems possibly encoded by the other PULs highlighted in our study (PULs 2, 13, 50, 51 and 68). Performing deletion mutagenesis of all putative pectinolytic PULs identified in this study would have been ideal to better comprehend their role but this was not possible because B. xylanisovens XB1A was particularly recalcitrant to genetic manipulation. With untreated apple pectin as substrate, the mutant growth was only affected during stationary phase, possibly because it preferentially used contaminating starch and started to hydrolyze pectin only when the medium was depleted in starch. This observation highlights the adaptive capacity of B. xylanisolvens in utilizing mixtures of plant polysaccharides. Nevertheless, starch contamination prevented us from comparing thoroughly the RNA-seq transcriptomic profiles of B. xylanisolvens on apple and citrus pectins. As a consequence, caution must be taken when using so called “purified” apple pectins when the objective is to screen for pectinolytic bacteria and to analyze pectinolytic function.

To better understand the mutant phenotype, we analyzed PUL 49 (BXY_31910 to BXY_32010) gene expression in the mutant grown on citrus pectin in comparison to the wild type strain. Plasmid insertion into the susC-like gene (BXY_31990) blocked the expression not only of that gene but also of the two downstream genes i.e. susD-like (BXY_31980) and FnIII-like genes (BXY_31970), suggesting that these three genes form an operon. This is supported by short (<20pb) intergenic distances and operon prediction using Rockhopper v. 1.30 [32]. Moreover, except for the gene encoding the sensor/regulator protein (HTCS, BXY_31960), the other PUL 49 genes were also slightly repressed in the mutant. This overall repression is probably due to the fact that the sensor/regulator protein, usually located in the cytoplasmic membrane, was not activated by HG oligosaccharides that could not enter the periplasm through the SusC-like transporter.

Finally, this study showed the architectural complexity of homogalacturonan-targeting PULs in Bacteroides which highlights important inter- and intra-species genetic rearrangements. Nevertheless, within the B. xylanisolvens species, two distinct PUL architectures were identified with strains harboring a short PUL analogous to PUL 49 (BXY_31910 to BXY_32010), and other strains harboring a long PUL consisting of a PUL 49-like segment and of another PUL segment, as in B. ovatus. A question is whether the acquisition or loss of this PUL segment has functional repercussions in terms of efficiency of pectin utilization and/or in terms of adaptability to structurally-different pectins. Only interaction studies between Bacteroides species harboring different PULs in response to well characterized (and pure) pectin from various plants will allow us to answer this question. Of course, wider functional and ecological studies are needed to understand how dietary fibers and especially PCW polysaccharides drive the composition and metabolism of the fibrolytic community within the intestinal microbial ecosystem. In the long term, such studies would provide valuable information aiming at maintaining a healthy gut microbiota or rebalancing a disturbed microbiota through diets enriched in specific PCW polysaccharides.

This study shows the existence of six PULs involved in pectin degradation by B. xylanisolvens XB1A, one of them being particularly important in this function. This species deploys a very complex enzymatic machinery that probably reflects the structural complexity of this dietary fibre. Our findings highlight the metabolic plasticity of B. xylanisolvens towards non-starch dietary polysaccharides which contributes to its competitive fitness within the human gut ecosystem.

The data sets supporting the results of this article are included within the article and its additional supplementary file. The transcriptome data are available in GEO datasets at NCBI (http://www.ncbi.nlm.nih.gov/) under the accession number GSE74379.