Report

Glycan Foraging in Vivo by an Intestine-Adapted Bacterial Symbiont

See allHide authors and affiliations

Science  25 Mar 2005:
Vol. 307, Issue 5717, pp. 1955-1959
DOI: 10.1126/science.1109051

Abstract

Germ-free mice were maintained on polysaccharide-rich or simple-sugar diets and colonized for 10 days with an organism also found in human guts, Bacteroides thetaiotaomicron, followed by whole-genome transcriptional profiling of bacteria and mass spectrometry of cecal glycans. We found that these bacteria assembled on food particles and mucus, selectively induced outer-membrane polysaccharide-binding proteins and glycoside hydrolases, prioritized the consumption of liberated hexose sugars, and revealed a capacity to turn to host mucus glycans when polysaccharides were absent from the diet. This flexible foraging behavior should contribute to ecosystem stability and functional diversity.

The adult human body is a composite of many species. Each of us harbors ∼10 times as many microbial cells as human cells (1). Our resident microbial communities provide us with a variety of metabolic capabilities not encoded in our genome, including the ability to harvest otherwise inaccessible nutrients from our diet (2). The intestine contains an estimated 10 trillion to 100 trillion microorganisms that are largely members of Bacteria but include representatives from Archaea and Eukarya (1, 3). Changes in diet appear to produce only modest effects on the species composition of the adult human colonic microbiota, although its metabolic activity may change considerably (4). The genomic foundations of these metabolic adjustments have yet to be defined in vivo.

Fermentable carbohydrates are the principal energy source for human colonic anaerobes (4). Recent studies of Escherichia coli, a facultative anaerobe with relatively minor representation in the microbiota, revealed that genetic restriction of its ability to metabolize various monosaccharides differentially compromised its ability to colonize the intestines of mice that had undergone antibiotic “knockdown” of their microbiota (5). The stability of an ecosystem apparently correlates with the ability of its members to mount diverse responses to environmental fluctuations; a corollary is that adaptive foraging behavior stabilizes ecosystems (6). Here, we use a simplified gnotobiotic mouse model of the human intestinal ecosystem to show that Bacteroides thetaiotaomicron, a highly abundant obligate anaerobe found in the colonic microbiota of most normal adult humans (7, 8), redirects its carbohydrate-harvesting activities from dietary to host polysaccharides according to nutrient availability.

B. thetaiotaomicron is a glycophile that can break down a broad array of dietary polysaccharides in vitro [reviewed in (3)]. This capacity is reflected in its genome (9), which has the largest repertoire of genes involved in acquisition and metabolism of polysaccharides among sequenced microbes, including 163 paralogs of two outer-membrane proteins that bind and import starch, 226 predicted glycoside hydrolases, and 15 polysaccharide lyases (10). In contrast, the human genome contains 98 known or putative glycoside hydrolases (10). More than half of the carbohydrate-degrading enzymes produced by B. thetaiotaomicron are predicted to be secreted into the periplasmic or extracellular space and thus, in principle, are capable of liberating oligo- and monosaccharides from undigested dietary polysaccharides and host mucus for consumption by B. thetaiotaomicron, other members of the microbiota, or the host.

Adult germ-free male mice were maintained on a standard autoclaved chow diet rich in plant polysaccharides. Gas chromatography–mass spectrometry (GC-MS) (11) established that glucose, arabinose, xylose, and galactose were the predominant neutral sugars present in this chow (mole ratio = 10:8:5:1). Seven-week-old mice were colonized with a single inoculum of B. thetaiotaomicron and killed 10 days later (a period that spans two to three cycles of turnover of the intestinal epithelium and its overlying mucus layer). Colonization density, in terms of colony-forming units per milliliter, ranged from 107 to 109 in the distal small intestine (ileum) to 1010 to 1011 in the cecum and proximal colon. Scanning electron microscopy revealed that B. thetaiotaomicron was attached to small food particles and embedded in mucus (Fig. 1).

Fig. 1.

Scanning electron microscope images showing distribution of B. thetaiotaomicron within its intestinal habitat. (A) Low-power view of the distal small intestine of B. thetaiotaomicron– monoassociated gnotobiotic mice, showing a villus (arrow) viewed from above. (B to D) Progressively higher power views showing B. thetaiotaomicron associated with luminal contents (food particles, shed mucus) [arrows in (B) and (C)] and embedded in the mucus layer overlying the epithelium [boxed region in (C), larger image in (D)]. Scale bars, 50 μm (A), 5 μm [(B) and (C)], 0.5 μm (D).

The cecum is an anatomically distinct structure, located between the distal small intestine and proximal colon, that is a site of great microbial density and diversity in conventionally raised mice (12). Nutrient use by B. thetaiotaomicron in the cecum was defined initially by whole-genome transcriptional profiling. Cecal contents, including the mucus layer, were removed immediately after killing of nonfasted mice (n = 6), and RNA was extracted (11). The B. thetaiotaomicron transcriptome was characterized with the use of custom GeneChips containing probe pairs derived from 4719 of the organism's 4779 predicted genes (table S1) (11). The results were compared to transcriptional profiles obtained from B. thetaiotaomicron grown from early log phase to stationary phase in a chemostat containing minimal medium plus glucose (MM-G) as the sole fermentable carbohydrate source (fig. S1).

Unsupervised hierarchical clustering of the GeneChip data sets disclosed remarkable uniformity in the in vivo transcriptional profiles of B. thetaiotaomicron harvested from individual gnotobiotic mice (fig. S2A). A total of 1237 genes were defined as significantly up-regulated (11) in vivo relative to their expression in MM-G. The functions of these up-regulated genes were classified by clusters of orthologous groups (COG) analysis. The largest up-regulated group belonged to the “carbohydrate transport and metabolism” COG, whereas the largest group of genes down-regulated in vivo belonged to the “amino acid transport and metabolism” COG (fig. S3A).

The starch utilization system (Sus) proteins SusC and SusD are components of a B. thetaiotaomicron outer-membrane complex involved in binding of starch and malto-oligosaccharides for subsequent digestion by outer-membrane and periplasmic glycoside hydrolases (13). Thirty-seven SusC and 16 SusD paralogs were up-regulated in vivo by a factor of ≥10 relative to bacteria growing in MM-G (fig. S4A).

The indigestibility of xylan-, pectin-, and arabinose-containing polysaccharides in dietary fiber reflects the paucity of host enzymes required for their degradation. The human genome contains only one putative glycoside hydrolase represented in the nine families of enzymes known in nature with xylanase, arabinosidase, pectinase, or pectate lyase activities, whereas the mouse genome has none (10). In contrast, B. thetaiotaomicron has 64 such enzymes (table S2) (10), many of which were selectively up-regulated in vivo by factors of 10 to 823. These included five secreted xylanases, five secreted arabinosidases, and a secreted pectate lyase (Fig. 2, A to C) (fig. S4B).

Fig. 2.

Carbohydrate foraging by B. thetaiotaomicron. (A) B. thetaiotaomicron gene expression during growth from log phase to stationary phase in minimal medium containing 0.5% glucose (MM-G) or 0.5% maltotriose (a simplified starch composed of three α1-4–linked glucose residues; MM-M) versus the ceca of monoassociated gnotobiotic mice fed a polysaccharide-rich diet. Predicted operons (21) are shown together with their component gene products. All genes listed were significantly up-regulated in vivo relative to MM-G according to criteria defined in (11). Note that during growth in MM-G versus MM-M, only 13 of the 4719 genes queried exhibit a factor of ≥10 difference in their expression. Eight of these genes constitute a Sus operon (22): Its three Sus α-amylases are the only ones among 241 B. thetaiotaomicron glycoside hydrolases and polysaccharide lyases whose expression changes by a factor of ≥10 as a result of exposure to maltotriose, underscoring the specificity of the organism's induced responses to the glycosidic linkages that it must process [e.g., compare α- and β-glucosidases in (B) and data in (C)]. (B and C) Selective induction of glycoside hydrolases in vivo. (B) Induction of expression of groups of glycoside hydrolases in the cecum relative to MM-G and MM-M (see table S4 for a list of genes; the number of genes in each group is indicated in parentheses; summed GeneChip signals for B. thetaiotaomicron transcripts called “present” for individual samples within an experimental group were averaged to calculate the aggregate mean signal ± SEM). (C) Biochemical evidence of B. thetaiotaomicron's “preparedness” for degrading glycans. Lysates were generated from bacteria during late log-phase growth in MM-G. The organism produces a portfolio of hydrolases capable of processing a wide variety of glycosides, even when exposed to a single fermentable monosaccharide. Mean values ± SD of triplicate assays are shown. (D) GC-MS of neutral and amino-sugars in cecal contents from germ-free versus B. thetaiotaomicron–colonized mice [n = 4 animals per group; mean values ± SEM are shown; asterisks indicate a statistically significant difference compared to germ-free (P < 0.01, Student's t test)].

GC-MS analysis of total cecal contents harvested from fed germ-free mice (11) revealed that xylose, galactose, arabinose, and glucose were the most abundant monosaccharide components (Fig. 2D). After 10 days of colonization by B. thetaiotaomicron, significant reductions in cecal concentrations of three prominent hexoses (glucose, galactose, and mannose) were observed. There were no significant decreases in pentoses or amino-sugars (Fig. 2D). The selective depletion of hexoses likely reflects the combined effects of microbial and host utilization. B. thetaiotaomicron colonization increased expression of the principal sodium/glucose transporter Sglt1 in the intestinal epithelium, reflecting an enhancement of host utilization of liberated monosaccharides (14). Moreover, of the 1237 bacterial genes up-regulated in vivo, 310 were assignable (11) to Enzyme Commission (EC) numbers in metabolic maps in the Kyoto Encyclopedia of Genes and Genomes (KEGG) (15). The results of this metabolic reconstruction were consistent with active delivery of mannose, galactose, and glucose to the glycolytic pathway, and arabinose and xylose to the pentose phosphate pathway [fig. S5; see (16) for maps of all B. thetaiotaomicron genes with EC assignments that exhibited changed expression in vivo versus MM-G].

Host mucus provides a “consistent” endogenous source of glycans in the cecal habitat that could offer alternative nutrients to the microbiota during periods of change in the host's diet. B. thetaiotaomicron embeds itself in this mucus layer (Fig. 1D). GeneChip analysis provided evidence that the bacterium harvests glycans from mucus. For example, in vivo, B. thetaiotaomicron exhibited significant up-regulation (by factors of 2 to 10; P < 0.05) of (i) an operon (BT0455-BT0461) that encodes a sialidase, sialic acid–specific 9-O-acetylesterase, mannosidase, and three β-hexosaminidases (Fig. 2A); (ii) a mucin-desulfating sulfatase (BT3051); and (iii) a chondroitin lyase (BT3350). Fucose in host glycans is an attractive source of food: It typically occupies a terminal α-linked position and is constitutively produced in the cecal mucosa (17). We found that two secreted α-fucosidases (BT1842, BT3665) and a five-component fucose utilization operon (BT1272-BT1277) were induced (Fig. 2A). Operon induction, which occurs through the interaction of l-fucose with a repressor encoded by its first open reading frame (ORF) (18), is indicative of bacterial import and utilization of this hexose.

To determine whether the absence of fermentable polysaccharides in the diet increases foraging on mucus glycans, we compared B. thetaiotaomicron gene expression in the ceca of two groups of age- and gender-matched adult gnotobiotic mice. One group received the standard polysaccharide-rich chow diet from weaning to killing. The other group was switched to a diet devoid of fermentable polysaccharides but rich in simple sugars (35% glucose, 35% sucrose) 14 days before colonization. All mice were colonized with B. thetaiotaomicron for 10 days, and bacterial gene expression was defined in each of their ceca at the time of killing.

The presence or absence of polysaccharides in the diet did not produce a significant effect on the density of cecal colonization (19). Using the transcriptional profiles of 98 B. thetaiotaomicron genes from the “replication, recombination, and repair” COG as biomarkers, we found that the cecal bacterial populations clustered most closely to cells undergoing log-phase growth in vitro, irrespective of the diet (fig. S2B and table S3).

The simple-sugar diet evoked a B. thetaiotaomicron transcriptional response predominated by genes in the “carbohydrate transport and metabolism” COG (fig. S3B). Glycoside hydrolase and polysaccharide lyase genes that were up-regulated in vivo by a factor of ≥2.5 relative to MM-G cultures segregated into distinct groups after unsupervised hierarchical clustering (Fig. 3) (fig. S6). The group of 24 genes most highly expressed on the simple-sugar diet encoded enzymes required for degradation of host glycans (e.g., eight hexosaminidases, two α-fucosidases, a sialidase) and did not include any plant polysaccharide–directed arabinosidases or pectin lyases. In addition, all components of the fucose utilization operon (BT1272-BT1277) were expressed at higher levels in mice fed the simple-sugar diet (average induction greater than MM-G by a factor of 12) than in mice fed the polysaccharide-rich diet (average induction greater than MM-G by a factor of 6). The sialylated glycan degradation operon (BT0455-BT0461) exhibited a comparable augmentation of expression on the simple-sugar diet.

Fig. 3.

Diet-associated changes in the in vivo expression of B. thetaiotaomicron glycoside hydrolases and polysaccharide lyases. Unsupervised hierarchical clustering yields the following groups of genes up-regulated in vivo by a factor of ≥2.5 on average, relative to their average level of expression at all growth phases in MM-G: group 1, highest expression on a simple-sugar diet, includes activities required for degradation of host glycans; group 2, equivalent expression on both diets; group 3, highest on a polysaccharide-rich standard chow diet; includes enzymes that degrade plant glycans. Average relative differences in expression in vivo versus in vitro (MM-G) are shown in fig. S6. Predicted enzyme substrate specificities unique to the group are listed at the right.

A similar cluster analysis revealed two distinct groups of genes encoding carbohydrate binding/importing SusC/SusD paralogs: a group of 61 expressed at highest levels in B. thetaiotaomicron from the ceca of mice fed a polysaccharide-rich diet, and a group of 21 expressed at highest levels with a simple-sugar diet (fig. S7). Thirteen of the SusC/SusD paralogs expressed at highest levels on a polysacchariderich diet are components of predicted operons that also contain ORFs specifying glycoside hydrolases and polysaccharide lyases. Five pairs of the SusC/SusD paralogs expressed at highest levels on a simple-sugar diet are part of predicted operons. No SusC/SusD paralogs from one diet group are found in operons containing up-regulated glycoside hydrolase genes from the other diet group (fig. S8). Together, the data indicate that subsets of B. thetaiotaomicron's genome are dedicated to retrieving either host or dietary polysaccharides (depending on their availability), although it appears that when both sources are available, harvesting energy from the diet is preferred.

Diet-associated changes in glycan-foraging behavior were accompanied by changes in the expression of B. thetaiotaomicron's capsular polysaccharide synthesis (CPS) loci (fig. S9). Relative to growth in MM-G, CPS3 was down-regulated in vivo irrespective of host diet, CPS4 was up-regulated in the ceca of mice fed a polysaccharide-rich diet, and CPS5 was up-regulated with a high-sugar diet (fig. S9). The other five CPS loci did not manifest significant differences in their expression during growth in vitro versus in vivo, nor with diet manipulation. These findings suggest that B. thetaiotaomicron is able to change its surface carbohydrates in response to the nutrient glycan environment that it is accessing and perhaps also for evading a host immune response (17).

A schematic overview of how B. thetaiotaomicron might scavenge for carbohydrates in the distal intestine is shown in fig. S10. Bacterial attachment to food particles, shed mucus, and exfoliated epithelial cells is directed by glycan-specific outer-membrane binding proteins (exemplified by SusC/SusD paralogs) (20). B. thetaiotaomicron contributes to diversity and stability within the gut by adaptively directing its glycan-foraging behavior to the mucus when polysaccharide availability from the diet is reduced. Hence, host genotype and diet intersect to regulate the stability of the microbiota. Co-evolution of glycan structural diversity in the host, together with an elaborate repertoire of nutrient-regulated glycoside hydrolase genes in gut symbionts, endows the system with flexibility in adapting to changes in diet. Although our study has focused on the glycan-foraging behavior of B. thetaiotaomicron in monoassociated germ-free mice, similar analyses can now be used to assess the impact of other members of the gut microbiota on B. thetaiotaomicron and on one another. The results should help to define the molecular correlates of behaviors that underlie the assembly and maintenance of microbial communities in dynamic nutrient environments. They should also provide a framework for developing effective ways to manipulate these communities to promote health or treat various diseases.

Supporting Online Material

www.sciencemag.org/cgi/content/full/307/5717/1955/DC1

Materials and Methods

Figs. S1 to S10

Tables S1 to S4

References

References and Notes

View Abstract

Navigate This Article