The number of microbes associated with our gut likely exceeds our total number of somatic and germ cells. Despite their numbers, almost nothing is known about the molecular mechanisms that determine whether the interaction between a microbial species and its host will be beneficial. Recent results obtained from in vivo models have revealed critical roles for glycoconjugates in helping define the outcome of two such host–microbial relationships. In one case, attachment of Helicobacter pylori to fucosylated or sialylated glycans produced by various gastric epithelial lineages and their progenitors skews the destiny of colonization toward pathogenicity. In the second case, a molecular dissection of how Bacteroides thetaiotaomicron, a normal inhabitant of the distal small intestine, is able to communicate with intestinal epithelial cells has revealed a novel role for host fucosylated glycans in forging a mutually beneficial relationship. These observations lend support to the hypothesis that the capacity to synthesize diverse carbohydrate structures may have arisen in part from our need to both evade pathogenic relationships and to coevolve symbiotic relationships with our nonpathogenic resident microbes.
Accepted on November 29, 2000;
Humans are in constant contact with microbes throughout their lives. The total number of microorganisms that colonize our mucosal surfaces has been estimated to exceed our total number of somatic and germ cells by at least an order of magnitude (Savage, 1977). The relationships we have with components of our microflora span a continuum from mutually beneficial (symbiotic) to benefiting one partner without necessarily being detrimental to the other (commensal) to benefiting one partner while producing significant loss of fitness in the other (pathogenic). Some of the factors that drive a particular host–microbial relationship to one end of this continuum or the other are now being elucidated at the molecular level. It is becoming clear that one of these factors is the repertoire of glycans expressed by the host. In this review, we emphasize how intestinal glycoconjugates play critical roles in host–microbial interactions and how they may help define where a particular relationship falls on the symbiotic-commensal-pathogenic continuum.
The mammalian gut is more densely populated with microbes than any other organ. Its resident microbial society (microflora) is exceedingly complex, being composed of more than 400 bacterial species (Savage, 1977). This society exhibits an astonishing degree of spatial and temporal complexity. Bacterial species have characteristic geographic distributions along the length of the gastrointestinal tract, with certain species preferentially colonizing the stomach (e.g., Lactobacilli), and others, such as Escherichia coli, preferentially colonizing the colon. Furthermore, the composition of the microflora changes as the host develops, with different species colonizing the gut at characteristic times. For example, there is a dramatic shift in species composition during the weaning period (Savage, 1977).
The carbohydrate structures elaborated on the surfaces of intestinal epithelial cells likewise show a great deal of diversity, varying as a function of cell lineage, cellular location along the cephalocaudal axis, and developmental stage. For example, structures recognized by Bauhinia purpurea lectin (e.g., Galβ3GalNAc) are not evident in the small intestine of one widely used inbred strain of mice until the conclusion of weaning, whereas structures recognized by Sambucus nigra lectin (e.g., NeuAcα2,6Gal/GalNAc) are detectable in members of the mucus-producing goblet cell lineage early in postnatal life but not during adulthood (Falk et al., 1994). The regional and developmental specificity in glycan production suggests that the expression patterns of glycans may be functionally linked to the spatial and temporal complexity of the intestinal microbiota.
The adherence of pathogenic bacteria to cellular targets in mammalian tissues is a crucial step in the pathogenesis of many infections. For a number of Gram-negative pathogens, attachment is mediated by bacterial adhesin–host glycan interactions that define the host range and tissue tropism of the organism. Although several of these interactions have been modeled using cultured cell lines, there are relatively few cases where the impact of carbohydrate-mediated bacterial binding on pathogenesis has been well characterized in vivo (reviewed in Karlsson, 1999). For example, the FimH adhesin expressed by uropathogenic strains of E. coli (UPEC) mediates the binding of this organism to mannosylated glycans synthesized by the superficial facet (or umbrella) cells of the bladder urothelium (Wu et al., 1996). The FimH-mediated binding of UPEC is a prerequisite for exfoliation and death of the superficial cells of the urothelium, as well as for bacterial invasion into deeper layers of the breached urothelial barrier (Mulvey et al., 2000). Some of the structural details of how this adhesin mediates these interactions have been revealed by X-ray crystallographic studies. These studies disclosed that FimH contains an N-terminal adhesin domain with a carbohydrate-binding pocket capable of accommodating a D-mannose residue (Choudhury et al., 1999).
Recent results obtained from in vivo mouse models have indicated that the relationships between gastrointestinal bacterial species and their hosts may depend on the availability of specific epithelial cell surface glycans. Below, we illustrate this concept using examples of how two gut microbes exploit specific host glycans in different ways to evolve either a pathogenic or a mutually beneficial relationship.
Host glycans as mediators of a pathogenic host–microbial relationship: the case of Helicobacter pylori
H. pylori is a Gram-negative bacterium found in the stomachs of more than half of the human population. Despite its prevalence, it produces significant pathology in only a subset of hosts. H. pylori colonization is commonly associated with an active gastritis (Marshall and Warren, 1984). However, in some humans this gastritis may progress to chronic atrophic gastritis, duodenal ulcer, gastric adenocarcinoma, or mucosa-associated lymphoid tissue (MALT) lymphoma (Cover and Blaser, 1999; Blaser, 2000). Because its ability to elicit pathology is restricted to a minority of colonized individuals, some have suggested that H. pylori may exist primarily as a commensal and that it only emerges as a pathogen from a concurrence of host, microbial, and environmental factors (Blaser, 1997; Blaser and Kirschner, 1999b). In vivo studies indicate that the location of H. pylori in the commensal–pathogen continuum may be determined in part by the repertoire of glycans expressed in gastric ecosystem of its host and by the microbe’s ability to express the corresponding adhesins.
H. pylori binding to Lewisb glycans
Although H. pylori persists mainly in the gastric mucus layer of colonized individuals, it can interact directly with gastric epithelia expressing glycans terminating with the Lewisb (Leb) blood group antigen (Boren et al., 1993). Approximately 70% of humans produce this histo-blood group antigen. H. pylori adherence is highly selective for Leb structures (see Figure 1), because bacteria do not bind to Leb antigen substituted with a terminal GalNAcα1,3 residue (representing the blood group A determinant; see Figure 1) (Boren et al., 1993). These observations have been invoked to explain the higher prevalence of ulcers associated with H. pylori colonization of individuals with blood type O compared to people with blood types A or B (Boren et al., 1994). Epithelial attachment of H. pylori to Leb structures appears to be mediated by the blood group antigen-binding adhesin (BabA; Ilver et al., 1998).
These observations led to the hypothesis that Leb-mediated binding to the epithelium determines whether colonization of the stomach will more likely result in a pathogenic relationship between microbe and host. This hypothesis is supported by observations made in a transgenic mouse model. Leb expression in humans is limited to mucus-producing pit cells within the gastric epithelium (Boren et al., 1993). Unlike humans, mice do not normally synthesize Leb. Some inbred strains, such as FVB/N, make lacto-N-fucopentaose I-like structures (Fucα1,2Galβ1,3GlcNAcβ) that can serve as acceptors for the human Lewis enzyme, α1,3/4 fucosyltransferase (α1,3/4 FT). Therefore, the model was created by expressing human α1,3/4 FT in the pit cell lineage of FVB/N mice, resulting in the production of Leb-containing glycans starting in the first week of postnatal life and continuing throughout adulthood (Falk et al., 1995).
Leb-mediated binding of H. pylori clinical isolates to the gastric epithelium significantly alters the outcome of infection (Guruge et al., 1998). The overall efficiency and density of colonization by H. pylori is equivalent in Leb transgenic mice and in their normal counterparts. In both groups of mice, the majority of bacteria are associated with the luminal mucus. However, bacteria bind directly to the epithelium only in Leb-positive transgenic animals. This binding is associated with an enhanced cellular immune response and more severe gastritis. H. pylori attachment to Leb-positive pit cells also elicits production of antibodies that react with Lex-containing structures associated with both bacterial lipopolysaccharide (LPS) and hosts’ acid-producing parietal cells. Higher titers of these antibodies are associated with greater parietal cell loss, which is one pathologic outcome of H. pylori infection seen in a subset of colonized individuals (those with chronic atrophic gastritis).
This enhanced humoral immune response, precipitated by Leb-mediated attachment of the microbe, may arise from a form of molecular mimicry (Chan et al., 1995; Sherburne and Taylor, 1995) involving bacterial and host Lewis x glycans (see Figure 1 for structure). The LPS O antigen in 85% of H. pylori clinical isolates contains Lex (and/or Ley) immunodeterminants (Aspinall et al., 1996; Simoons-Smit et al., 1996). Lex glycans are also expressed by parietal cells in FVB/N transgenic (and normal) mice irrespective of whether they are colonized by H. pylori. Because clinical isolates express surface epitopes that mimic those that are normally present in the host (e.g., Monteiro et al., 2000), the Leb mouse model suggests that the parietal cell loss that occurs in H. pylori infected humans with chronic atrophic gastritis (Genta, 1997) may occur through an autoimmune process. In this model, the autoimmune response is elicited by bacterial surface carbohydrates and by the ability of H. pylori to bind to the gastric epithelium.
Variations in the expression of LPS-associated Lex- and Ley-type structures may also help explain why some hosts evolve a pathologic and others a relatively benign response to H. pylori colonization. The H. pylori genome encodes α1,2- and α1,3-fucosyltransferases that are responsible for synthesis of these Lewis structures (Ge et al., 1997; Martin et al., 1997; Tomb et al., 1997). Expression of the fucosyltransferases is subject to phase variation, which refers to a reversible on- and off-switching of surface epitope production (Appelmelk et al., 1998). The phase variation is thought to involve a series of tandem oligonucleotide repeats located at the 5′ ends of the H. pylori fucosyltransferase genes (Ge et al., 1997; Martin et al., 1997; Tomb et al., 1997). This entails a novel mechanism, known as slipped-strand base mispairing, which results in replication errors and thus a high frequency of spontaneous variation in repeat number. The result is generally a translational or transcriptional switch that facilitates phase-variable expression (see van Belkum et al., 1998 for details).
Phase variation may function to provide a pool of bacterial phenotypes that confers an evolutionary advantage to a bacterial population by facilitating selection of host-adapted subpopulations. In the case of H. pylori, structural similarities between microbial epitopes and epitopes normally expressed in host cell lineages (such as parietal cells) may under certain circumstances, lead to self (host)-directed immunity and a pathologic host–microbial relationship. Under other circumstances, the adoption of these similarities may be a mechanism by which a subpopulation evades the provoked immune response. An explicit example of this phenomenon comes from studies of Haemophilus influenzae, a Gram-negative bacterium that causes meningitis and respiratory tract infections. Surface-exposed carbohydrate structures contained in the LPS of H. influenzae are critical for pathogenesis. Phase variation results in strains with or without Galα1,4Gal surface epitopes. H. influenzae variants that produce this epitope show increased virulence and immune system evasion in vivo compared to those that lack this carbohydrate structure (Roche and Moxon, 1995).
H. pylori binding to NeuAcα2,3Galβ1,4 glycans
In 1994, the International Agency for Research on Cancer designated H. pylori as a human carcinogen (IARC, 1994). The association between this organism and cancer has been supported by several case-control epidemiologic studies (Nyren, 1998). One risk factor for development of cancer appears to be the development of chronic atrophic gastritis, a condition associated with loss of two of the principal epithelial cell types of the stomach: parietal cells and pepsinogen-producing zymogenic cells. Multifocal atrophic gastritis correlates with gastric ulcer disease and gastric cancer (Schrager et al., 1967; Sipponen 1992; Lechago and Correa, 1993; Eidt and Stolte, 1994; Kuipers et al., 1995; Lin et al., 2000).
Although development of atrophic gastritis may be an important event in H. pylori–associated carcinogenesis (reviewed in Genta, 1997), the mechanism has remained obscure. However, another transgenic mouse model has revealed that a second host glycan structure can mediate H. pylori adherence to the gastric epithelium. However, this glycan provides new insights about the mechanisms that may affect initiation and progression of neoplasia in humans who lose parietal cells during the course of their persistent H. pylori colonization.
The model was created using parietal cell–specific transcriptional regulatory elements from the mouse H+/K+-ATPase β subunit gene to drive expression of an attenuated diphtheria toxin fragment A (tox176) (Li et al., 1996). Expression of tox176 causes the elimination of parietal cells and two secondary effects: a block in the differentiation of zymogenic cells plus enhanced proliferation of the multipotent gastric stem cell and its immediate committed daughters. The amplification of these progenitors presumably reflects the loss of parietal cell signals that normally serve to restrict their proliferative activity (see Syder et al., 1999 for discussion). Over time, the progenitors progressively increase in number until they become a predominant population in the gastric epithelium (Syder et al., 1999).
Several of the histopathologic features seen in the stomachs of tox176 transgenic mice resemble those encountered in patients with chronic atrophic gastritis, including increased proliferation of epithelial progenitor cells and loss of parietal and zymogenic cells (Syder et al., 1999). In addition, parietal cell loss in humans with chronic atrophic gastritis is associated with augmented production of sialylated glycans (Farinati et al., 1988). Consistent with this, the amplified population of gastric progenitor cells in tox176 mice express glycans that bind Maackia amurensis agglutinin (MAA) (Syder et al., 1999), which recognizes sequences terminating with NeuAcα2,3Galβ1,4.
The MAA-reactive glycans expressed by the expanded progenitor population mediate binding of a clinical isolate of H. pylori to the gastric epithelium (Syder et al., 1999). Competition studies using purified soluble neoglycoconjugates demonstrated that bacterial binding to the gastric epithelium of tox176 animals is inhibited by 3′-sialyllactose and by sialyl-Lex, both of which contain the terminal dissacharide NeuAcα2,3Galβ (Figure 1). In contrast, binding is not competed with 6′-sialyllactose, sialyl-Lea, or the nonsialylated Lewis structures (Figure 1; Syder et al., 1999). Thus, H. pylori appears to express an adhesin that recognizes 3′ but not 6′ sialic acid–containing structures. The lack of competition by nonsialylated Lewis structures indicates that sialic acid is an absolute requirement for binding. However, because sialyl-Lex but not sialyl-Lea inhibits binding, the sialic acid must be attached to a β1,4-linked Gal, indicating that the minimal unit of recognition is NeuAcα2,3Galβ1,4.
Together, these findings suggest that H. pylori expresses at least two distinct adhesins. One of these (BabA) mediates binding to Leb-containing structures; the other, not yet identified, recognizes terminal NeuAcα2,3Galβ1,4. Presumably, which of these adhesins mediates binding to the gastric epithelium depends both on their expression by the colonizing strain and on the availability of the cognate host epithelial glycan(s).
As in the Leb mouse, binding of H. pylori to NeuAcα2,3Galβ1,4-glycans expressed by the amplified epithelial progenitors in the tox176 mouse stomach has a marked impact on the outcome of colonization. Colonized tox176 mice show a more pronounced gastritis than do nontransgenic mice. Furthermore, there is an enhanced cellular immune response, similar to that associated with Leb-mediated adherence. In vivo binding to NeuAcα2,3Galβ1,4-glycans also elicits an enhanced humoral immune response, characterized by higher titers of circulating anti-Lex than infected normal animals. As in the Leb mice, these antibodies react with bacterial LPS-associated Lex immunodeterminants and cross-react with Lex-positive parietal cells present in normal mice (Syder et al., 1999). Because tox176 animals lack parietal cells and do not contain Lex glycans in their remaining gastric epithelial populations, this finding provides direct evidence that glycan-mediated attachment of H. pylori can elicit a host humoral response to bacterial LPS immunodeterminants.
A model of H. pylori infection
These studies of Leb- and NeuAcα2,3Galβ-mediated binding of H. pylori suggest a model of how colonization may lead to chronic atrophic gastritis and initiation of neoplasia. According to the model (summarized in Figure 2), colonization of a host expressing one or more carbohydrate structures (such as Leb) that are recognized by a bacterial adhesin (or adhesins) leads to epithelial binding. Attachment enhances both cellular and humoral immune responses. Molecular mimicry between LPS carbohydrate structures (e.g., Lex), associated with at least a subpopulation of colonizing organisms, and structures produced by parietal cells leads to parietal cell loss. Parietal cell depletion results in the amplification of epithelial progenitors expressing terminal NeuAcα2,3Galβ1,4. These sialylated structures act as additional targets for bacterial adhesins, thereby favoring persistent infection in a changing gastric ecosystem (see Kirschner and Blaser, 1995; Blaser and Kirschner, 1999 for mathematical modeling studies of the impact of H. pylori attachment on persistent colonization). The ability of various strains of H. pylori to bind directly to multipotential gastric stem cells and their immediate daughters may have a profound effect on initiation and/or progression of gastric cancer. It is interesting to note that NeuAcα2,3Galβ1,4 glycans are expressed not only in human gastric adenocarcinoma but also in its postulated precursors—chronic atrophic gastritis, gastric dysplasia, and metaplasia (Farinati et al., 1988).
In summary, these transgenic mouse models have allowed an in vivo exploration of how glycans mediate, in a complex way, H. pylori’s interaction with its host. The host glycan repertoire may skew the relationship away from commensalism and toward pathology. An interesting question is whether H. pylori adapts its pattern of adhesin production in vivo to coincide with available epithelial glycan structures and/or whether it directly regulates the pattern of glycan expression in the epithelial cells of its host.
A role for host glycans in commensal relationships: the case of Bacteroides thetaiotaomicron
As noted above, H. pylori provides an example in which host glycans may influence the progression of a host-microbial relationship toward a pathogenic interaction. Analysis of another mouse model has demonstrated that specific host glycans are also involved in establishing and maintaining a nonpathogenic, mutually beneficial relationship with at least one indigenous intestinal microbe.
It is becoming increasingly clear that the normal microflora play critical roles in shaping host cellular differentiation pathways in the intestinal mucosa (Falk et al., 1998). One manifestation of this phenomenon is seen in the ability of components of the microflora to modify host glycoconjugate production. For example, adult germ-free BALB/c mice do not produce fucosyl-asialo-GM1 in their small intestinal epithelium. However, transient expression of this glycolipid was observed when these mice were colonized with a complete normal microflora or a component of the normal microflora known as the segmented filamentous bacteria (SFB, Umesaki et al., 1995). The appearance of this structure was associated with increased activity of a fucosyltransferase that adds fucose to asialo-GM1 (Umesaki et al., 1982).
The NMRI inbred strain of mice provides another example of a microbe-dependent induction of a fucosylated glycan (and a corresponding fucosyltransferase) in the intestinal epithelium. These glycans are detectable with Ulex europeaus lectin, which recognizes structures terminating with Fucα1,2Galβ. In NMRI mice that acquire a microflora from birth (i.e., are conventionally raised), expression of Fucα1,2Galβ-glycans is evident in the distal small intestine (ileum) beginning at postnatal day 21 (P21). However, at P21 production of these glycans is restricted to epithelial cells overlying rare isolated villi (see Figure 3A). Expression increases during the next 7 days of development, resulting in generalized expression of Fucα1,2Galβ-glycans in all ileal villi. This pattern is maintained throughout adulthood (Figure 3A). The initial detection of Fucα1,2Galβ-glycans at P21 corresponds to the beginning of the weaning period, a time in development that is associated with a dramatic change in the composition of the intestinal microflora from a predominance of facultative anaerobes to a predominance of obligate anaerobes (Savage, 1977). This developmental pattern of Fucα1,2Galβ-glycan expression in conventionally raised NMRI mice suggested that components of the evolving normal intestinal microflora might play a role in eliciting host production of specific carbohydrate structures.
The hypothesis was tested by examining fucosylated glycan expression in germ-free NMRI mice. As in conventionally raised animals, germ-free mice exhibit limited expression of ileal Fucα1,2Galβ-glycans at P21. In contrast to conventionally raised mice, this expression is completely extinguished by P28 (Figure 3A). A bacterial role in sustaining host production of this glycan in the postweaning period was confirmed by showing that introduction of a complete distal intestinal microflora into adult germ-free mice elicits a generalized pattern of Fucα1,2Galβ expression in the ileum. Furthermore, these changes are accompanied by induction of a host α1,2-fucosyltransferase (Bry et al., 1996). Together, the results demonstrate that host–microbial cross-talk, involving one or more components of the normal intestinal microflora, is responsible for sustained intestinal production of this glycan.
B. thetaiotaomicron is a strict anaerobe and one of the most abundant normal inhabitants of the distal small intestine and colon of mice and humans (Moore and Holdeman, 1974; Ushijima et al., 1983). When adult germ-free mice are colonized with B. thetaiotaomicron, ileal expression of Fucα1,2Galβ-glycans is induced, recapitulating the effect of a complete microflora (Bry et al., 1996). The ability of B. thetaiotaomicron to evoke production of these fucosylated glycans is dependent on the density of colonizing organisms in the intestinal lumen and is not associated with direct bacterial binding to the epithelium. These two observations suggest that the induction of ileal epithelial Fucα1,2Galβ-glycans may be mediated by a soluble bacterial factor. Furthermore, the induction is not a general effect of colonization with any Gram-negative bacterium. For example, two other prominent components of the normal microflora (Peptostreptococcus micros and Bifidobacterium infantis) do not elicit glycan expression when either is inoculated into adult germ-free NMRI mice (Bry et al., 1996).
Because B. thetaiotaomicron is genetically manipulatable (Salyers et al., 2000), some of the molecular details of how this bacterium influences host glycan expression are now understood. B. thetaiotaomicron uses a novel mechanism to induce the host to synthesize a specific glycan structure that the microbe can then tap as a supply of monomeric fucose to be taken up and utilized as a carbon and energy source.
A functional connection between B. thetaiotaomicron fucose utilization and stimulation of host Fucα1,2Galβ-glycan production was initially revealed by studying a mutant B. thetaiotaomicron strain, Fu-4. Fu-4 carries a chromosomal transposon insertion that renders it unable to utilize L-fucose as a carbon source (Salyers and Pajeau, 1989). The Fu-4 strain is also unable to elicit host Fucα1,2Galβ-glycan production in germ-free NMRI mice, unlike the wild-type strain (Bry et al., 1996; Hooper et al., 1999). Cloning of the transposon insertion locus in Fu-4 revealed disruption of a gene encoding L-fucose isomerase (fucI). FucI is part of a cluster of five genes encoding proteins required for fucose utilization. In addition to the isomerase, these include an aldolase, kinase, and permease (Hooper et al., 1999). The first open reading frame in the bacterial gene cluster encodes a repressor, designated FucR. Genetic and biochemical studies revealed that monomeric fucose functions, through FucR, as the inducer of the isomerase, kinase, and aldolase genes, as well as the gene encoding FucR itself (Hooper et al., 1999). In the absence of fucose, FucR represses transcription of the fucose pathway genes by binding to sequences in the promoter located directly upstream of the gene cluster (Figure 3B).
Colonizations of germ-free mice with mutants harboring disruptions of the fucose pathway genes revealed that FucR is the critical link between bacterial fucose utilization and bacterial signaling the host to produce Fucα1,2Galβ-glycans (Hooper et al., 1999). Mutants with disruptions of fucI are unable to signal, in contrast to mutants with disruptions of genes encoding the other enzymes involved in fucose breakdown. However, a strain containing a disruption of fucR that also blocks FucI production does elicit host fucosylated glycan synthesis. This surprising result indicated that, in addition to regulating transcription of the fucose pathway enzymes, FucR also mediates transcriptional repression of a second genetic locus distinct from the fucose utilization gene cluster. The product of this second locus plays a direct or indirect role in production of a bacterial signal that elicits glycan production. This locus, designated control of signal production (csp; see Figure 3B), has not yet been directly identified, although its existence is inferred from the genetic experiment.
FucR’s ability to coordinate fucose consumption by the microbe and fucosylated glycan production by the host appears to occur through a distinctive regulatory mechanism. A model of the mechanism is presented in Figure 3B and is consistent with the observed fucose utilization and host signaling phenotypes of the various mutant B. thetaiotaomicron strains. In the model, fucose acts through FucR both as an inducer of transcription of the fucose utilization operon and as a corepressor at csp. This explains why mutants that lack the isomerase (FucI) are unable to signal. Because this enzyme catalyzes the first enzymatic step in the breakdown of fucose, its absence leads to an accumulation of fucose in the bacterium. Fucose accumulation increases the proportion of FucR with bound fucose and thus a silencing of csp. The model predicts that lowering intracellular fucose levels will shift FucR to its fucose-unbound form and release csp from inhibition so that signaling can occur. In fact, disruption of the constitutively expressed fucose permease gene (fucP) reduces import of fucose into B. thetaiotaomicron and promotes signaling to the host. The model also predicts that if the buildup of fucose that ensues because of FucI deficiency could be prevented, then a signaling phenotype would be restored. This prediction is confirmed by the fact that a nonsignaling isomerase deficient strain is converted to a signaling strain by disrupting the permease gene (Hooper et al., 1999).
The model suggests that FucR acts as a molecular sensor that coordinates fucose supply and demand. When the bacteria have ample fucose, fucose-bound FucR allows expression of bacterial genes required for breakdown of the available fucose and turns off production of the csp-dependent signal that tells intestinal cells to synthesize more fucosylated glycans. Conversely, when fucose is needed, FucR in its unbound form allows transcription of csp to occur, resulting in a “request” to the host for more fucose. Whether fucose is ample or needed is defined by the affinity of FucR for fucose. This value (Kd = 10 µM in vitro; Hooper et al., 1999) represents a threshold concentration of intracellular fucose that determines the bacterial response.
Engineering production of its own nutrient source in the intestinal epithelium makes sense for such an organism as B. thetaiotaomicron. Like other strict anaerobes, it colonizes the intestine at weaning, when this ecosystem is already densely populated with an entrenched preweaning microflora that limits the availability of nutrient sources. The preformed pool of Fucα1,2Galβ-glycans seen in scattered ileal villi at the beginning of weaning (P21) may provide sufficient hydrolyzable fucose to promote initiation and early progression of B. thetaiotaomicron colonization. Once a critical density of organisms is attained, the population can signal the host to provide a sustained supply of Fucα1,2Galβ-glycans. By using this strategy, the bacterium is able to conserve energy by signaling only when fucose supplies are low. The host, in turn, has to synthesize only as much glycan as is necessary to support the nutritional needs of at least one of its commensals. The host benefits by gaining some control over the composition of its microflora, particularly during potentially destabilizing transitional periods, such as weaning.
This postulated role for Fucα1,2Galβ-glycans as a microbial energy source that helps negotiate a mutually beneficial host–microbial relationship would represent a novel biological function for a mammalian glycan. Fucosylated glycans represent a logical and economical choice from the perspective of both the bacterium and the host. Fucose is well represented in intestinal glycoconjugates, where it is almost always a terminal α-linked sugar (Bjork et al., 1987; Finne et al., 1989). Therefore, a secreted bacterial α-fucosidase should be sufficient to harvest L-fucose for import and metabolic processing. Consistent with this conceptualization, B. thetaiotaomicron secretes an α-fucosidase activity that can cleave fucose residues present in these glycans (Hooper et al., 1999). Moreover, fucose is commonly linked to core oligosaccharide structures that are constitutively synthesized by mammalian cells. Thus, a bacterial “request” to the host to manufacture fucosylated glycans may only require that the host activate transcription of one or more of its fucosyltransferase genes.
Host–microbial relationships and glycan diversity: an evolutionary perspective
A remarkable feature of mammalian glycoconjugates is their tremendous structural complexity and diversity. Much of the structural heterogeneity arises, fundamentally, from the nontemplate, combinatorial nature of carbohydrate synthesis and modification. The inner core structures of many glycans, such as those N-linked to proteins, are highly conserved from yeast to mammals and mediate critical intracellular functions (Ellgaard et al., 1999). In contrast, the outer chains of oligosaccharides can vary widely, depending on the protein or lipid to which they are linked, and the tissue and cell type in which they are expressed. Several distinctive outer chain structures play essential roles in important biological processes. For example, N-linked glycans terminating with mannose-6-phosphate direct proteins to the lysosomal compartment (Dahms et al., 1989), while structures terminating with GalNAc-4-SO4 mediate rapid receptor-mediated clearance of glycoprotein hormones from the circulation (reviewed in Hooper et al., 1996). Similarly, O-glycans terminating with sialyl-Lex play important roles in regulating leukocyte trafficking and adhesion (McEver and Cummings, 1997). However, these examples represent the exceptions rather than the rule; the vast majority of structures described to date have not been associated with clearly delineated functions. Nevertheless, the fact that many of these glycan structures are synthesized in a cell type–specific and developmentally regulated manner strongly suggests that they play important, albeit unappreciated, physiological roles (Dennis et al., 1999).
Why do higher organisms invest energy in generating this vast array of outer branch chain structures? Why is there such a variety of glycosyltransferases that act to synthesize these chains? Several very elegant studies of knockout mice that lack enzymes responsible for producing outer chain structures have revealed relatively subtle phenotypes (Maly et al., 1996; Hennet et al., 1998). A theme emerging in the recent literature is that one source of selective pressure to create the capacity for glycan diversity (besides the need to satisfy subtle endogenous functions) could be the need to evade pathogenic microbes (Gagneux and Varki, 1999; Dennis et al., 1999). For example, a bacterial species could initiate pathogenic interactions with their hosts by first binding to glycan structures. As an evasive tactic, the host could eliminate expression of the structure, either by inactivation of glycosyltransferases that mediate its production (e.g., see Xu et al., 1999) or by activation of a glycosyltransferase that utilizes the structure as an acceptor, thereby rendering the glycan unrecognizable by the microbe.
In addition to evading pathogens, we would like to propose that a major source of selective pressure for evolving the capacity for glycan diversity is the need to create and sustain relationships with our indigenous societies of symbionts. B. thetaiotaomicron’s ability to induce its host to produce Fucα1,2Galβ-containing structures that provide a harvestable source of monomeric fucose illustrates how intestinal glycan structures may function to benefit members of the nonpathogenic microflora. At the same time, because a stable intestinal microflora enhances the fitness of its host, there should be selective pressure for expression of such glycans.
Given the vast numbers and extraordinary complexity of the mammalian intestine’s microbial community, it seems reasonable to speculate that these types of selective pressures would be particularly active in the gut. B. thetaiotaomicron’s ability to influence host glycan synthesis may epitomize a general strategy used by other members of the commensal microflora involving other types of carbohydrate structures. The mammalian intestine and its resident microbes may adapt to one another in part by coevolution of diverse host carbohydrate structures and the corresponding microbial glycosidases. This coadaptation may help explain some of the observed spatial and developmental patterns of intestinal glycan production (Falk et al., 1994). The capacity to generate structural diversity in glycans through nontemplate-based mechanisms should also facilitate the host’s ability to accommodate dynamic fluxes in the composition of its gut microflora and the rapid evolution of individual species. The challenge we now face is to create vastly simplified experimental model systems so that we can begin to understand the factors that govern the complex marriage of microbes to host carbohydrates and its meaning to both partners.
The complexities of the microflora and of host glycans seem overwhelming. However, there is every reason to believe that great progress will be made in understanding their interrelationships in the near future. Recent technological advances have produced an unprecedented opportunity to create in vivo models for examining the impact of host glycans on host–microbial interactions. For example, the development of systems for knocking out genes at any time during or after completion of mouse intestinal development (Saam and Gordon, 1999; Wong et al., 2000) makes it possible to manipulate production of specific glycan structures in the host. The explosion in functional genomics, including the availability of high-density DNA arrays representing a large fraction of the mouse genome, permits comprehensive profiling of changes in host gene expression. The increasing availability of facilities for raising mice under germ-free conditions allows generation of in vivo models where the impact of one or a few microbial species on host glycan production can be examined.
Combining these technologies provides a powerful way of addressing some of the issues raised in this review. By introducing H. pylori into germ-free transgenic mice that express Leb- or NeuAcα2,3Galβ1,4-glycans in their gastric epithelium and then using DNA microarrays to monitor gene expression in their stomachs and in the stomachs of normal mice, it should be possible to begin to decipher the molecular mechanisms that underlie attachment-mediated pathogenesis. Likewise, because B. thetaiotaomicron is genetically manipulable, it is possible to generate strains that do or do not elicit production of specific host glycan structures, introduce them into germ-free mice, and use DNA arrays to begin to define the mechanisms by which a commensal is able to manipulate glycan production (as well as other physiological functions). Inducing Cre-mediated knockouts of glycosyltransferases (or components of the synthetic apparatus that produce donors for these enzymes) in adult germ-free mice will set the stage for examining the impact of various carbohydrate structures on specific host–microbial interactions. At the same time, the capacity to rapidly sequence microbial genomes creates an opportunity to take a genetically manipulable organism, such as B. thetaiotaomicron, and use arrays containing bacterial genes to broadly survey changes in microbial gene expression during the course of colonization of germ-free mice. This type of analysis should yield testable hypotheses about which microbial genes participate in regulating host glycan production. The impact of genetically engineered manipulations of host glycan production on microbial gene expression can also be examined by combining functional genomics, gnotobiotics, and Cre-based techniques.
Together, these types of experiments, which are generalizable to various microbes, tissues, and host species, represent the start of a very challenging journey to begin to appreciate the origins and significance of the incredible diversity in carbohydrate structures and how these structures help define host–microbial relationships.
LPS, lipopolysaccharide; MAA, Maackia amurensis agglutinin; MALT, mucosa-associated lymphoid tissue; SFB, segmented filamentous bacteria; tox176, attenuated diphtheria toxin fragment A; UPEC, uropathogenic E. coli.
To whom correspondence should be addressed