The Life of Commensal Escherichia coli in the Mammalian Intestine

TYRELL CONWAY,¹ KAREN A. KROGFELT,² AND PAUL S. COHEN^3*

[SECTION EDITOR: JAMES P. NATARO]

Posted December 29, 2004

¹Department of Botany and Microbiology, University of Oklahoma, Norman, OK 73019, ²Department of Gastrointestinal Infections, Statens Seruminstitut, 2300 Copenhagen S, Denmark, and ³Department of Cell and Molecular Biology, University of Rhode Island, Kingston, RI 02881

^∗ Corresponding author. Phone: 401-874-5920, Fax: 401-874-2202, E-mail: This e-mail address is being protected from spambots. You need JavaScript enabled to view it

INTRODUCTION

A wide variety of bacteria colonize the large intestines of humans and animals. The bacterial populations involved are relatively stable, consisting of hundreds of species that interact in such a way as to achieve a relatively constant numerical balance (27). In human adults, the obligate anaerobes make up greater than 99.9% of the cultivable bacteria (81). Prominent genera include Bacteroides, Bifidobacterium, Clostridium, Eubacterium, Fusobacterium, Lactobacillus, Peptococcus, Peptostreptococcus, and Veillonella (70, 101). Escherichia coli is the predominant facultative anaerobe in the gastrointestinal tracts of mammals (27). Its genome was one of the first to be completed, and E. coli is arguably the best understood of all model organisms (8). Yet the essence of how E. coli colonizes the mammalian large intestine is not understood. Evidence is mounting that commensal strains of E. coli reside in the mucus layer of the large intestine. We will discuss evidence suggesting that various surface components of E. coli (e.g., capsule, fimbriae, flagella, and lipopolysaccharide) either aid or inhibit penetration of the mucus layer and consequently influence the establishment of a stable microflora. We will also discuss evidence that the mucus layer contains receptors for bacterial surface components that can either aid or inhibit the colonization process. Lastly, we will discuss what is known about the specific nutrients present in mouse cecal mucus that support E. coli colonization.

THE MAMMALIAN LARGE INTESTINE

In contrast to the small intestine, which contains relatively few bacteria, the majority of the intestinal biota resides in the large intestine. Approximately 10¹¹ bacteria can be cultivated per gram of adult human colonic contents (101). The large intestine, which constitutes the terminal part of the digestive system, is divided into three main sections: the cecum, the colon, and the rectum. Since nutrients derived from food are, for the most part, absorbed in the small intestine, the main function of the large intestine is to reabsorb water and inorganic salts. As water is reabsorbed during passage of intestinal contents through the large intestine, the contents are compacted and then eliminated as feces. Unlike the small intestine, the surface of the large intestinal mucosa is smooth, i.e., there are no intestinal villi. The large intestine epithelium consists of goblet cells and enterocytes. The crypts of Lieberkuhn, in which epithelial cells proliferate, are longer and more uniform than those in the small intestine and are also lined with goblet cells and enterocytes. In humans, the colonic epithelium is renewable and it is estimated that as many as 2 × 10⁶ to 5 × 10⁶ epithelial cells are shed per minute (16). The relatively thick (up to 800 μm) mucus layer that overlies the epithelium consists of mucin, a 2-MDa gel-forming glycoprotein and a large number of smaller glycoproteins, proteins, glycolipids, lipids, and sugars (1, 7, 29, 55, 92, 95, 105, 108). Presumably, shed epithelial cells are a source of many of the smaller mucus components (92, 95). The mucus layer itself is in a dynamic state, constantly being synthesized and secreted by the mucin-secreting, specialized goblet cells and is degraded to a large extent by the indigenous intestinal microbes (51, 85). Degraded mucus components are shed into the intestinal lumen, forming a part of the luminal content that is excreted in the feces (51).

Freter lists four microhabitats within each section of the intestine (34). The first is on the surface of the epithelial cells. Many indigenous and pathogenic bacteria, including pathogenic strains of E. coli, colonize this site via attachment to specific receptors (11, 54, 58, 61, 62, 89). The second microhabitat is the deep mucus layer of the crypts in the ileum, cecum, and colon. Mainly motile spiral-shaped bacteria inhabit this site, including Borrelia, Treponema, and Spirillum (61, 62). The third site is the mucus layer that covers the epithelial cells throughout the intestinal tract. Commensal strains of E. coli appear to reside in this microhabitat in the cecum and colon of the mammalian intestine (80, 93, 94, 96) along with indigenous populations that probably reflect, at least in part, the composition of the approximately 500 indigenous cultivable species (27, 119). The contents and microbiota of the first three microhabitats are sloughed into the fourth habitat, the lumen.

Growth of E. coli Occurs in the Mucus Layer of the Intestine

A large and growing body of evidence indicates that commensal E. coli strains grow in the intestine on nutrients acquired from mucus. Fluorescence in situ hybridization of mouse intestinal thin slices showed that E. coli BJ4, a rat commensal isolate, is dispersed in the mucus layer but is not associated with the epithelium (93). The same appears to be true of the human commensal strain, E. coli MG1655 (80). The in situ growth rate of E. coli BJ4 cells embedded in mucus is rapid (generation time of 40 to 80 min), while cells in the luminal contents are static (94). In vitro, rapid growth in intestinal mucus (20 to 30 min generation time), but not in luminal contents, is in keeping with the hypothesis that E. coli and Salmonella enterica serovar Typhimurium grow on nutrients acquired from mucus (64, 75, 87, 114). Among mutants unable to colonize the mouse large intestine are those that fail to penetrate mucus (63, 73, 75, 80), have difficulty surviving in mucus (87), or have difficulty growing in mucus (87, 108, 109). Thus, the ability of E. coli to grow and survive in mucus appears to be critical for intestinal colonization. As discussed later, at least seven mucus-derived sugars, the most important of which is gluconate, contribute with varied degrees to colonization by the commensal strain E. coli MG1655.

Colonization and Colonization Resistance

Colonization in the present context is defined as the indefinite persistence of a bacterial population in the intestine of an animal in stable numbers without repeated introduction of the bacterium to that animal. Colonization resistance refers to the ability of a complete intestinal microflora to resist colonization by an invading bacterium (113). For example, when healthy human volunteers are fed E. coli strains isolated from their own feces, those strains do not colonize (3). Experimental evidence suggests that colonization resistance to an invading bacterium can, in part, be explained by a lag phase caused by short-chain fatty acids and hydrogen sulfide that are the metabolic end products of metabolism of the indigenous microflora (35, 44). Thus, if the numbers of invaders are small they may be completely eliminated from the intestine before exiting the lag phase. However, even when the numbers of invaders are high, the complete intestinal microflora, in most instances, still prevents their establishment (34, 36, 113).

An analogy can be drawn between the mammalian intestine and a chemostat (36, 62). Two different microorganisms with a preference for the same growth-limiting nutrient cannot coexist in chemostats; the one that utilizes it slightly more efficiently will eventually out-compete the other (32). However, if two microorganisms utilize different growth-limiting nutrients, they can coexist and maintain stable populations in a chemostat (32, 112). Work with continuous flow cultures (CF cultures) in chemostats designed to mimic the intestine (34, 36) led to the theory that being physically attached to the intestinal wall allows a species to remain in the intestine despite growing at a rate lower than the washout rate from the intestine. Moreover, this theory predicts that two bacterial strains competing for the same limiting nutrient can coexist in the intestine if the metabolically less efficient one is attached to the intestinal wall (36). In addition, the data obtained from CF cultures showed that if an established bacterium and an invading one are equally fit in competing for the same limiting nutrient, the established bacterium will eliminate the invading bacterium if it is attached to the intestinal wall, because large wall populations can reduce the limiting nutrient concentration to the point that an invader will not be able to multiply in the lumen of the intestine at a rate fast enough to resist washout (33). Thus, the mammalian intestine can be thought of as a chemostat in which several hundred species of bacteria are in equilibrium, competing for resources from a mixture of limiting nutrients. According to the theory, the only way for a bacterial species to survive in the intestine without being physically attached to the host intestinal wall is to out-compete all other organisms for a particular nutrient or nutrient mixture, since failure to do so will inevitably lead to displacement from the host. Despite the appeal of this explanation of colonization resistance, there appears to be a continuous succession of E. coli strains in the mammalian intestine. Some strains are present for months to years, while others persist only transiently, for a few days. An average of five E. coli biotypes have been found in feces of individual humans (6), indicating that diversity exists among the normal commensal E. coli strains in the intestine and strongly suggesting that different E. coli strains possess different capacities for utilizing growth-limiting nutrients.

Layers within the Mucus Layer

The mucus layer of the conventional mouse intestine turns over about every 2 h (96). If commensal E. coli strains do not attach to intestinal epithelial cells and growth takes place predominantly in the mucus layer (80, 93), it seems reasonable that, to maintain a stable population, the bacterial growth rate must keep pace with the turnover rate of the mucus layer. However, the situation is complicated by the finding that there are, in fact, two components of the mucus layer in the rat intestinal tract, a loosely adherent suction-removable layer closest to the lumen of the intestine and an adherent layer firmly attached to the mucosa (7). In the rat colon, the thickness of the adherent layer is about 116 μm and that of the suction-removable layer is about 714 μm (7). As the authors point out, it is possible that the more stable adherent layer protects the epithelial cells and that the suction-removable layer, which is replaced rapidly when it is physically removed, acts as a lubricant (7). Although not proven conclusively, it may very well be that the adherent layer turns over slowly and that the suction-removable layer turns over rapidly (7). If so, commensal strains of E. coli could bind to receptors present in the firmly adherent layer and avoid rapid washout, whereas unattached progeny that enter the suction-removable layer as it is formed could be washed out of the intestine in feces. In this event, commensal E. coli strains could grow in the adherent mucus layer at a rate slower than the washout rate of the suction-removable layer from the intestine and still colonize the intestine. Strains that only grow in the suction-removable layer would, of course, have to grow at a rate equal to its turnover rate to maintain a stable population.

IN VIVO COLONIZATION MODELS

The advantages and disadvantages of three in vivo models of intestinal colonization are considered here. The first is the conventional animal, which was discussed above. The advantage of this first model is that the conventional animal has a complete microflora and, therefore, the interaction between an invading microorganism and the established microflora can be studied. The disadvantage is that invaders have difficulty colonizing because of colonization resistance.

The second model of intestinal colonization is the germfree or gnotobiotic animal. The advantage of this model is that there is no indigenous microflora and most microorganisms colonize readily; however, this is also a distinct disadvantage of the model in that nutrients may be accessible that are not normally accessible or nutrients that are normally accessible may not be so. On a practical level, compared with conventional animals, gnotobiotic animals are considerably more difficult to handle and more expensive to purchase and house. In addition, the intestines of gnotobiotic animals differ physiologically from conventional animals in several respects; for example, gnotobiotic mice have poorly developed lamina propria, shallow crypts of Lieberkuhn in the ileum, thin and villous mucosa in the cecum, decreased intestinal mobility, and higher intestinal oxygen tension (110). Nevertheless, the gnotobiotic animal is proving to be excellent for identifying potentially important molecular mechanisms of the colonization process that may be relevant in conventionalized animals. For example, Bacteroides thetaiotaomicron senses l-fucose availability in the monoassociated gnotobiotic mouse intestine and coordinately induces a system that enhances synthesis of fucosyl glycans by intestinal enterocytes together with a pathway for l-fucose catabolism (48, 50). Therefore, members of the microflora appear capable of inducing the synthesis of the nutrients they prefer for growth. In addition, B. thetaiotaomicron influences the establishment of the intestinal mucosal barrier in ex-gnotobiotic mice by inducing the expression of a Paneth cell protein, Ang4, which is bactericidal for several gram-positive pathogens (48, 121). Therefore, the concept that host-microbe interactions in the intestine are more symbiotic than commensal in nature is being explored successfully in ex-gnotobiotic mice (24, 47, 48, 49, 121).

The third model of intestinal colonization is the streptomycin-treated mouse. In 1954, it was reported that mice that were extremely resistant to Salmonella enteritidis infection became infected after oral administration of streptomycin (9). It was subsequently shown that streptomycin treatment altered the microecology of the cecal contents, including a decrease in the populations of the facultative anaerobic enterococci, streptococci, and lactobacilli and the anaerobic lactobacilli and bifidobacteria. Accompanying these changes in microflora was a general decrease in the concentration of volatile fatty acids, which, as discussed above, may play a role in colonization resistance (45), and a minor increase in pH (6.42 to 6.73) (45). Nevertheless, populations of the genera Bacteroides and Eubacterium in the cecal contents of streptomycin-treated mice were unchanged. Moreover, the numbers of strict anaerobes in the cecal contents of streptomycin-treated and conventional mice were essentially identical (1 × 10⁹ to 2 × 10⁹ CFU/g of contents) (45). Therefore, while the streptomycin-treated mouse model is not perfect, invading microorganisms must compete for nutrients with a large number of strict anaerobes in the intestine. The streptomycin-treated mouse model is relatively inexpensive and has been used extensively to study E. coli and serovar Typhimurium colonization of the mouse large intestine, as will be discussed below. Colonization of this model is simple, requiring only that streptomycin sulfate (5 g/liter) be administered in drinking water and that the microorganisms fed to the mice be streptomycin resistant.

STAGES OF COLONIZATION

Streptomycin-treated mice produce approximately 100 mg of feces per hour (P. S. Cohen, unpublished work). By using the streptomycin-treated-mouse model, it has been shown that when 10⁵ CFU of either E. coli, serovar Typhimurium, or Klebsiella pneumoniae strains is fed to streptomycin-treated mice, the organisms grow from small numbers at 5 h after feeding (10⁵ CFU/g of feces) to large numbers (10⁸ to 10⁹ CFU/g of feces) by 24 h after feeding (26, 86, 87). Therefore, when small numbers of invading E. coli, serovar Typhimurium, or K. pneumoniae strains reach the streptomycin-treated-mouse large intestine, they appear to have a nonlimiting source of nutrients to initiate the colonization process, presumably caused by the elimination of the streptomycin-sensitive facultative microflora (45). The initiation phase lasts anywhere from 1 to 3 days after feeding, following which E. coli strains reach a stable population of between 10⁷ and 10⁸ CFU/g of feces which then lasts indefinitely. This stage is referred to as the maintenance stage. The maintenance stage is more akin to persistent colonization of the intestine. During maintenance, to achieve a stable population the growth rate of colonized strains must match the turnover rate of the mucosal contents (33) and persistence most likely depends on nutrient(s) that become limiting after E. coli has grown to high numbers.

Initiation and maintenance are separable stages of colonization. E. coli K-12 leuX mutants cannot make the leucyl-tRNA specific to the 5'-TTG-3' codon but can use the leucyl-tRNA specific to the 5'-TTA-3' codon by wobble (57). An E. coli K-12 leuX mutant grows as well as its parent in mouse cecal mucus in vitro during the exponential phase, but survives poorly in mucus relative to the wild-type strain during the stationary phase (87). When the leuX mutant is the only E. coli strain fed to mice, it grows as well as the wild-type strain from small to large numbers in the mouse intestine during the initiation stage but then is rapidly eliminated (87). Complementationof the leuX mutant with the wild-type leuX gene restores the ability of the mutant to maintain itself in the intestine (87). Initiation and maintenance also seem separable nutritionally, i.e., E. coli MG1655 appears to utilize different nutrients for growth during maintenance than it uses during initiation, as will be discussed below.

RATE OF E. COLI GROWTH IN THE INTESTINE

Several studies have been designed to determine the growth rate of E. coli in the intestine. In most of these studies the intestine has been treated as one compartment with perfect mixing. Therefore, in each case, the calculations reflect the average growth rate of the entire E. coli population within the entire intestine. Depending on the study, the average generation time of E. coli in the conventional mammalian intestine (i.e., with a complete microflora) has been estimated at between and 4 and 40 h (28, 38, 39, 102). Freter et al. (36) also examined the average generation time of E. coli in the gnotobiotic mouse intestine by introducing about 10⁴ CFU into the stomachs of several animals and at various times removing the entire intestines, homogenizing them, and determining total CFU. In this way, the generation time in the germfree mouse intestine was determined to be about 20 min, i.e., approximately the same as in a rich broth culture in the laboratory.

Recently, the growth rate of E. coli in the mouse cecum has been examined. By using various sole carbon sources in minimal medium to grow E. coli (e.g., glucose, succinate), it was shown decades ago that its growth rate is directly proportional to its cellular ribosome content (83). E. coli ribosomal RNA (rRNA) content and consequently the E. coli growth rate can be estimated by quantitative in situ hybridization using E. coli-specific rRNA target probes and epifluorescence microscopy (94). By using this method, it has been shown that a commensal strain of E. coli has an rRNA content that correlates with a generation time of between 40 and 80 min in the cecal mucus layer of germfree mice (96) and in the cecal mucus layer of streptomycin-treated mice (94); in the latter experiment the intestinal anaerobic population remains largely intact (44). When streptomycin-treated mice or germfree mice were conventionalized by feeding them cecal contents of conventional mice and removing streptomycin water (from the streptomycin-treated mice), the in vivo generation time increased in both cases to about 120 min (96). In addition, it was shown that E. coli in the cecal luminal contents, both in vitro and in vivo, but not in cecal mucus, accumulated precursor 16S rRNA much as it does in the presence of chloramphenicol, suggesting that an antimicrobial agent present in mouse cecal luminal contents inhibits E. coli growth in that site (64). In summary, the available evidence suggests that E. coli commensal strains grow within the cecal mucus layer of the conventional mouse large intestine by doubling about every 2 h and are sloughed into luminal contents where growth stops, perhaps due to the presence of at least one antimicrobial agent, prior to leaving the intestine in feces.

NONNUTRITIONAL ASPECTS OF COLONIZATION: MOTILITY AND ADHESION TO AND PENETRATION OF THE MUCUS LAYER

When E. coli and serovar Typhimurium enter the mouse cecum they initially encounter cecal contents. Since E. coli and serovar Typhimurium fail to grow in cecal contents but grow rapidly in cecal mucus (64, 75, 87, 114), they must be able to enter the mucus layer and grow using available nutrients. Neither chemotaxis nor motility is required for this process in streptomycin-treated mice, i.e., nonmotile and nonchemotactic mutants of E. coli F-18, a human commensal strain, and avirulent serovar Typhimurium colonize as well as their wild-type parents in cofeeding experiments (74, 75). Furthermore, during growth in cecal mucus in vitro, both E. coli F-18 and the avirulent serovar Typhimurium strains become nonmotile but retain flagella (74, 75). Suspending Luria broth-grown cultures in mouse cecal mucus does not result in immediate loss of motility (74, 75). Therefore, mucus viscosity is not responsible for the loss of motility. Since motility is not required, yet bacteria enter the mucus layer anyway, it seems likely that E. coli and serovar Typhimurium enter the mucus layer from cecal contents as water is reabsorbed in the large intestine. This mechanism of bacterial penetration of the cecal mucus layer may be similar to that responsible for the entry of polystyrene beads (37).

In contrast to the large intestine, it appears that motility and chemotaxis do play a role in salmonellae penetrating the mucus layer in the ileum (99, 120). Moreover, it appears that motility and/or chemotaxis is required for Campylobacter jejuni colonization of the suckling mouse intestine (111), Helicobacter pylori colonization of both the mouse and gnotobiotic piglet stomachs (30), and rapid penetration of rabbit ileal mucus by Vibrio cholerae (37).

It is interesting to consider why motility/chemotaxis appears to be required for pathogenesis in the small intestine but unnecessary for colonization of the large intestine.The peristaltic action of the small intestine propels luminal contents at a rate that is far greater than that in the cecum and colon (34). Therefore, to avoid washout from the small intestine, pathogens must bind to specific receptors present in mucus and then penetrate the mucus layer extremely rapidly, presumably via chemotaxis and motility. After penetration, these strains stably bind to specific glycoprotein and glycolipid receptors present on epithelial cells and initiate pathogenesis (12). Whether adhesion of pathogenic E. coli strains to specific receptors present in mucus reduces their rate of penetration through the mucus layer to the epithelial cell surface is still an open question. It should be noted, however, that receptors present in mucus, depending on concentration, appear to participate in either promoting or preventing intestinal disease. As an example, glycoprotein and glycolipid E. coli K88ab-specific receptors in piglet ileal mucus (53) are present in far greater concentrations in preweaned K88ab-resistant 35-day-old piglets than in neonatal, K88ab-sensitive piglets (14). Moreover, 35-day-old piglet ileal mucus (3 mg of protein/ml) has been shown to prevent adhesion of an E. coli K88ab strain to underlying intestinal epithelial cells in vitro, whereas an equivalent concentration of newborn piglet ileal mucus was unable to do so (14). It therefore seems likely that the low concentration of K88ab-specific receptors in newborn ileal mucus is sufficient to allow E. coli K88ab adhesion to the ileal mucus layer but is insufficient to prevent subsequent adhesion of K88ab to the underlying epithelial cells. The high concentration of K88ab-specific receptors in the mucus of 35-day-old piglets is apparently sufficient to prevent adhesion to underlying epithelial cells, thus preventing disease. Similar results have been reported for E. coli 987P (18), E. coli K99 (82), E. coli RDEC-1 (21), and Yersinia enterocolitica (90).

More recently, two additional mechanisms that most likely facilitate rapid penetration of enteropathogens through the mucus layer have been described. The first mechanism is exemplified by enteroaggregative E. coli (EAEC) strains that secrete a 10.2-kDa anti-aggregation protein (Aap), encoded by the aap gene, which remains noncovalently bound to the outer surface of the EAEC cells. EAEC aap mutants form much larger cell aggregates than the wild-type strain and penetrate a layer of mucus in vitro more slowly (41, 103). Since Aap limits the size of aggregates, it has been called a dispersin (88, 103). It will be of great interest to determine whether Aap allows rapid penetration of EAEC through the mucus layer to the epithelial cell surface in vivo. Furthermore, it will be of great interest to determine whether dispersin synthesis stops once the epithelial surface is reached since EAEC is characterized by aggregative adherence (103).

The second recently described mechanism of rapid mucus penetration is exemplified by pathogenic E. coli and Shigella species that secrete a variety of high-molecular-weight serine proteases (42). While these enzymes appear to have several functions, including cleavage of pepsin, human coagulation factor V, and erythroid spectrin (10, 22), two of them have been shown to be mucinolytic (22, 41). Thus, EAEC secretes Pic (protein involved in intestinal colonization), a 109.8-kDa serine protease with in vitro mucinolytic activity on bovine submaxillary mucus and crude mouse mucus substrates. Pic is encoded by the pic gene located on the EAEC chromosome (41). Shigella flexneri (41) and the uropathogenic strain E. coli CFT073 (117) also contain a chromosomal pic gene. Tsh, a 106-kDa serine protease, encoded on a ColV-like plasmid by the tsh gene (temperature-sensitive hemolysin) and secreted by avian pathogenic E. coli (106), also degrades bovine submaxillary mucus (22). It is therefore possible that the mucinolytic activity of Pic and Tsh facilitate penetration through intestinal mucus by EAEC and avian pathogenic E. coli, respectively. Vibrio cholerae produces a zinc-dependent metalloprotease (Hap) encoded by the chromosomal hapA gene that also has mucinolytic activity (104). Moreover, it has been shown in this case that transcription of the hapA gene is significantly enhanced by either porcine stomach mucin or bovine submaxillary gland mucin but not by bovine serum albumin. Furthermore, a V. cholerae hapA mutant failed to penetrate a layer of porcine stomach mucin gel in vitro as rapidly as its wild-type parent (104). In addition, Vibrio anguillarum, a fish enteropathogen, produces a metalloprotease (EmpA) encoded by the empA gene (19). EmpA production is induced ninefold by Atlantic salmon gastrointestinal mucus (19). Whether EmpA has mucinolytic activity and whether it facilitates mucus penetration have not been tested.

Penetration of the mucus layer not only may facilitate binding to the underlying epithelial cells but also may allow commensal strains of E. coli, as well as enteropathogens, to be closer to the source of nutrients in the intestine, e.g., nutrients released by lysed exfoliated epithelial cells. As might be expected, bacterial cell surface components appear to regulate the rate of mucus layer penetration. Two lines of evidence support this statement. E. coli F-18 contains the fim operon and makes type 1 fimbriae (56, 72). Type 1 fimbriae confer on the great majority of E. coli strains the ability to bind to a variety of eukaryotic cell surfaces (56) and to specific glycoprotein receptors present in mouse cecal mucus (115). The expression of type 1 fimbriae is subject to phase variation owing to the stochastic inversion of a 314-bp fragment containing the fimA promoter (56). Inversion of the fimA promoter is regulated by the relative concentrations of the recombinases encoded by the fimB and fimE genes (56). An excess of fimB product locks the fimA promoter in the "on" position (56). Only 30% of the E. coli F-18 cells grown in Luria broth have type 1 fimbriae, whereas 100% of the E. coli F-18 cells containing an extra fimB gene on a plasmid (phase-locked "on" E. coli F-18) have type 1 fimbriae. Phase-locked "on" E. coli F-18 bind almost threefold better to mouse cecal mucus than E. coli F-18 in vitro, grows as well as F-18 in mouse cecal mucus in vitro, but penetrates a mucus layer model in vitro far more slowly than F-18 (73). In competition with laboratory-grown F-18 in the streptomycin-treated-mouse intestine, phase-locked "on" F-18 fails to initiate and is eliminated by day 3 after feeding (<10² CFU/g of feces), whereas the laboratory-grown F-18 colonizes indefinitely at a level of about 10⁸ CFU/g of feces (73). In contrast, when phase-locked "on" E. coli F-18 are allowed to colonize for 10 days prior to feeding the laboratory-grown F-18, it colonizes indefinitely thereafter at 10⁸ CFU/g of feces (73). Therefore, it appears that, if given enough time to penetrate the mucus layer, phase-locked "on" F-18 competes well with laboratory-grown F-18 in the mouse intestine. These results suggest that after growth in Luria broth, it is the F-18 cells without type 1 fimbriae (70%) that penetrate the cecal mucus layer in vivo more rapidly than type 1 fimbriated F-18 cells (30%) and, in doing so, utilize nutrients at the mucus/epithelial cell interface, thereby starving type 1 fimbriated cells at the mucus/lumen interface for nutrients, and consequently eliminating them from the intestine. Surprisingly, E. coli F-18 cells isolated from mouse cecal mucus at 10 days after feeding all appear to be type 1 fimbriated (59), suggesting that, once the nonfimbriated cells penetrate the mucus layer, something in mucus induces type 1 fimbria synthesis. It may be that leuX is involved, since, as discussed above, the leucyl-tRNA specific to the 5'-TTG-3' codon is required for maintenance of E. coli in the mouse intestine (87). Since the leucyl-tRNA encoded by leuX is increased relative to other tRNAs under such conditions (20) and since the fimB gene has twice the number of 5'-TTG-3' codons as the fimE gene (98) and both genes are equal in size, the fimA promoter, being in the "on" position, would be favored and, consequently, synthesis of type 1 fimbriae would ensue (98).

Rapid penetration of the mucus layer by serovar Typhimurium in the mouse large intestine is also facilitated by wild-type lipopolysaccharide (LPS) containing the O-polysaccharide. A serovar Typhimurium rfb mutant, unable to make O-polysaccharide, grows as well as its wild-type parent in mouse cecal mucus in vitro (75) and binds fourfold better to mouse cecal mucus than does its parent in vitro but penetrates a mucus layer both in vivo (viewed in cecal section) and in vitro far more slowly than does its parent (63, 75). In competition with its wild-type parent in the streptomycin-treated-mouse intestine, the rfb mutant is eliminated by day 11 after feeding (<10² CFU/g of feces), whereas its parent colonizes indefinitely at a level of about 10⁷ CFU/g of feces (86). In contrast, when the rfb mutant is allowed to colonize for 8 days prior to feeding the wild-type strain, it colonizes indefinitely thereafter at 10⁷ CFU/g of feces (86). Therefore, if given enough time to penetrate the mucus layer and presumably gain access to nutrients closer to the epithelial cell surface, the rfb mutant competes well with the wild-type strain in the mouse intestine. It therefore appears that the LPS O-polysaccharide facilitates penetration of serovar Typhimurium through the mucus layer. An O-polysaccharide-deficient serovar Typhimurium mutant has also been reported to be a poor colonizer of the broiler chick intestine relative to its wild-type parent (15).

Capsular polysaccharide may also facilitate penetration of bacteria through the intestinal mucus layer and enhance subsequent colonization. The importance of capsule in intestinal colonization is exemplified by K. pneumoniae, which is involved in extraintestinal infections but which initially appears to colonize the intestinal tract (68). A K. pneumoniae capsule mutant, constructed by allelic exchange and unable to make the K35 capsular polysaccharide, binds two- to ninefold better to a variety of mucus nonsecreting eukaryotic cell lines than its parent in vitro but binds 3-fold lower than its parent to colonic epithelial cells of a mucus-secreting human cell line isolated from a patient with a colon carcinoma (25). When fed to streptomycin-treated mice, the K. pneumoniae wild-type strain colonized indefinitely at a level of 10⁸ CFU/g of feces, whereas the capsule mutant, despite growing as well as its parent for 1 day in the intestine, decreased continuously such that by 20 days after feeding it declined to a level of only 10⁴ CFU/g of feces (26). The spatial distributions of the wild-type and capsule mutant were viewed in situ in the mouse colon at 1 day and 20 days after feeding. The wild-type strain was found distributed throughout the colonic mucus layer as predominantly single cells on days 1 and 20 after feeding (26). In contrast, although the capsule mutant was found in mucus primarily as single cells at 1 day after feeding, at 20 days it was found predominantly in clumps ranging from 3 to 20 cells (26). These results suggest that formation of clumps of capsule mutant cells in mucus prevents penetration and makes the strain prone to elimination as the mucus layer turns over. An E. coli K-12 LPS mutant that is unable to make both O-polysaccharide and core polysaccharide has also been reported to form clumps of cells in mouse cecal mucus in vivo and is unable to colonize the streptomycin-treated mouse large intestine even when it is the only strain fed to mice (80).

In summary, several lines of evidence suggest that penetration through the mucus layer toward the underlying epithelial cells is required for colonization by gram-negative commensal strains and enteropathogens. It is also important to emphasize that changes in the bacterial cell surface that lead to increased adhesion to eukaryotic cells in vitro do not necessarily result in increased colonizing ability or increased pathogenic potential in vivo. As described above, the increase in binding to epithelial cells may be accompanied by an inability to traverse the mucus layer and consequent elimination from the intestine.

NUTRITIONAL ASPECTS OF COLONIZATION: SIGNATURE-TAGGED MUTAGENESIS, IN VIVO EXPRESSION TECHNOLOGY (IVET), AND DNA MICROARRAYS

As stated above, the large intestine contains 400 to 500 different cultivable bacterial species (27, 119). This diversity of microorganisms is believed to reflect their ability to occupy different ecological niches. To reiterate, the theory developed by Freter and coworkers (33, 34, 36) postulates that numerous ecological niches within the intestine are defined by nutrient availability. According to this hypothesis, individual species have a preference for one, or very few, of the growth-limiting nutrients in the intestine that come from ingested food, epithelial and bacterial cell debris, and the mucus layer above the epithelium. The diversity of available nutrients is thought to result in a balanced ecosystem in which each nutrient-defined niche is occupied by an individual strain or species, with individual population sizes determined by the available concentration of the preferred nutrient.

At this point, it is important to discuss further what might constitute a preferred nutrient. The majority of intestinal bacteria require a fermentable carbohydrate for growth and this mode of metabolism is assumed to be the nutritional basis for intestinal colonization by most species (100). The total carbohydrate content of colonic mucus is approximately 50% (2). The dominant glycoprotein in the intestine is mucin, which makes mucus highly viscous and which is about 80% polysaccharide and 20% protein (2). Intestinal mucin oligosaccharides contain five major sugars: N-acetylgalactosamine, N-acetylglucosamine, N-acetylneuraminic acid (sialic acid), l-fucose, and d-galactose (2). At first glance, it might appear that each of these sugars would be utilized as a preferred nutrient by a different bacterial species for growth in the intestine. However, as an example, it is conceivable that two different bacterial species could utilize N-acetylglucosamine as a major carbon source if one species were able to bind a mucin oligosaccharide containing the sugar and thereafter degrade it to monosaccharides within the cell, while the second species relied on free N-acetylglucosamine for growth, released from mucin by extracellular glycosidases secreted by other intestinal bacteria. According to this scenario, N-acetylglucosamine mutants of each of the two species would colonize the intestine poorly relative to its parent and each parent would be utilizing N-acetylglucosamine for growth in the intestine, but each parent would be utilizing a different preferred limiting nutrient. In theory, there could be several members of the intestinal flora that utilize N-acetylglucosamine as a major carbon source. Each of these might bind a unique N-acetylglucosamine-containing oligosaccharide better than any other member of the intestinal flora. Intestinal bacteria do bind oligosaccharides; e.g., B. thetaiotaomicron, a gram-negative obligate anaerobe, which comprises as much as 25% of the entire bacterial population of the human colon (81), does, in fact, bind and degrade polysaccharides for growth by using cell-associated enzymes (4, 5). Several Bacteroides spp. grow almost as well on free N-acetylhexosamines as on hexoses (100). In addition, members of the genera Bifidobacterium and Ruminococcus secrete a variety of glycoside hydrolases that are active on both mucins and glycosphingolipids (52, 102). The monosaccharides released could be utilized by members of the intestinal flora that are unable to metabolize oligosaccharides, such as E. coli (52).

Mucus contains not only mucin but also smaller proteins, glycoproteins, lipids, and glycolipids (1, 29, 55, 92, 95, 105). The total content of the hexuronates glucuronate and galacturonate in mouse cecal mucus, including both free and polymerized forms, is about 0.6% by weight (108). Mouse cecal mucus also contains arabinose, mannose, and ribose (31), as well as a small amount of gluconate as a free monosaccharide (0.7 mM), which presumably arises from bacterial oxidation of glucose or dephosphorylation of 6-phosphogluconate released by exfoliated epithelial cells (91). If the essence of the Freter hypothesis is correct, then the population size of an individual species in the large intestine would depend on the available concentration of its preferred nutrient, and the relatively small population of E. coli in the intestine, despite its rapid growth rate, indicates that the concentration of its preferred monosaccharide(s) is low.

To our knowledge, other than work on the nutritional basis of E. coli MG1655 mouse intestinal colonization to be discussed below, there are no published studies that have systematically investigated the nutritional basis for growth in the intestine of any member of the intestinal biota. However, there are some data relating to this issue, obtained from signature-tagged mutagenesis studies, IVET studies, and DNA microarray studies, in which the nutritional basis of virulence or colonization was not of primary concern.

Genetic screens based on signature-tagged mutagenesis identify individual mutants that are lost from pools of mutants used to infect suitable animals (43). In general, individual mutants that are lost have defects in genes that are important for virulence or colonizing ability in vivo. In one study in which E. coli K1 essential in vivo genes were screened for infection of the infant rat descending colon, none involved in specific carbon source utilization were identified (71). The same was true for a similar study involving identification of genes required for K. pneumoniae colonization of the mouse intestine (107). However, in another study, a signature-tagged K. pneumoniae glpP mutant was identified, putatively deficient in the ability to degrade stored glycogen, that failed to colonize the mouse colon (69). A search for genes required for Haemophilus influenzae systemic infection of the infant rat revealed a putative lctP gene that encodes l-lactate permease, suggesting that transport of l-lactate promotes virulence (46). In another study, signature-tagged V. cholerae mutants with reduced ability to colonize the small intestines of infant mice were isolated (78). Several of the mutants had insertions in genes involved in energy and central carbon metabolism, suggesting that both glycolytic (maltose and galactose) and gluconeogenic (acetate) pathways might be involved in colonization (78).

IVET uses the host to select for bacterial genes that are expressed at high levels in animal tissues but are expressed poorly or not at all under laboratory conditions (66). While the method clearly works extremely well, many genes that are identified, when mutated, do not reduce infectivity in animal models (40). Very few IVET data pertaining to nutrients that might be important for infectivity in the intestine have been reported. However, in vivo expression of the fadB gene in serovar Typhimurium isolated from the spleens of infected BALB/c mice suggests that β-oxidation of fatty acids might be important, but this observation has not been pursued (67). IVET analysis of K. pneumoniae in the spleens of mice did not reveal any metabolism genes (60). However, IVET of Lactobacillus reuteri genes expressed in the mouse gastrointestinal tract included xylA, encoding xylose isomerase, implicating xylose catabolism during colonization (116).

The use of microarrays to examine in vivo gene expression in pathogens has recently been reviewed (13). The expression profile resulting from a properly conceived experiment can disclose the metabolic and biosynthetic pathways that are affected in a change from one growth condition to another, e.g., from in vitro to in vivo growth. DNA array analysis of Borrelia burgdorferi in the rat peritoneum (97) and V. cholerae in rice water stools (77) revealed genes generally known to be involved in metabolism, but no specific mention of carbon nutrient metabolism was made. The results of recent additional microarray searches for V. cholerae genes expressed in vivo during growth in the rabbit ileal loop for 8 h (122) and for serovar Typhimurium genes following murine macrophage infection have also been published (23). The Vibrio data suggest that glycerol-3-phosphate, fructose, and maltose are among the nutrients used anaerobically for growth in the rabbit small intestine. The Salmonella data showed induction of a limited number of metabolism genes after macrophage infection, including genes involved in gluconate, glucuronate/galacturonate, and galactonate transport. Also of interest is the apparent role of the glyoxylate pathway and hence of C₂ compounds (e.g., acetyl coenzyme A from fatty acid degradation) as nutrients for survival and persistence of microbes in host macrophages (reviewed in reference 65) and, in the case of Mycobacterium tuberculosis, in persistence in the mouse lung (76).

Much remains to be learned about the in vivo nutrition and physiology of pathogenic E. coli strains that colonize their mammalian host to initiate the disease process. Current pathogenesis research focuses predominantly on individual virulence determinants. Clearly, beyond genes encoding virulence determinants, host-associated bacteria use cellular processes, which are essential for in vivo survival but are not virulence determinants per se. These essential in vivo genes are likely to encode functions required for growth in vivo, such as metabolic pathways for the scavenging of nutrients and growth factors from the host (13). Since many cellular processes, including stress-response systems and metabolic pathways, are likely to be essential for all bacteria to live in association with their host, it seems reasonable to extend the concept of essential in vivo genes to include those required by commensal strains for colonization. Since these essential in vivo processes contribute to co-colonization of the intestine by hundreds of different species and perhaps constitute a first line of defense against infection by pathogenic strains, they certainly warrant further examination.

NUTRITIONAL ASPECTS OF E. COLI MG1655 AND EDL933 INTESTINAL COLONIZATION

Studies of the human commensal strains E. coli F-18 and K-12 strongly implicated the nutrient gluconate and the Entner-Doudoroff pathway as being important for colonization of the streptomycin-treated-mouse large intestine (109). Since E. coli colonization of the mouse intestine appears to require the ability to grow in mucus, these results and the emergence of DNA arrays as a powerful tool focused attention on the possibility that identification of genes induced by growth in mouse cecal mucus relative to growth in minimal medium containing glucose as the carbon source might lead to identifying nutrients necessary for maximum E. coli mouse large intestine colonizing ability. E. coli MG1655, a human commensal K-12 strain, was chosen as the strain to be tested since it had been completely sequenced (8).

The gene expression profiles of cecal mucus-grown E. coli MG1655 identified genes involved in catabolism of N-acetylglucosamine, N-acetylneuraminic acid, l-fucose, d-ribose, d-galactose, d-glucuronate, d-galacturonate, d-gluconate, and maltose (11a). There was also evidence for catabolism of glycerol-3-phosphate and amino acids (11a), which require the tricarboxylic acid (TCA) cycle and gluconeogenesis for E. coli MG1655 to grow. These data, suggesting that E. coli MG1655 consumes several nutrients found in cecal mucus, were the starting point for testing the roles of individual nutrients in supporting E. coli MG1655 colonization of the mouse large intestine.

Based on the gene expression profile of E. coli MG1655 cells grown in cecal mucus in vitro, mutants with deletions in genes corresponding to induced metabolic pathways were constructed by allelic replacement (17) and were tested for their ability to compete with the corresponding wild-type MG1655 strain when simultaneously fed to mice in low numbers (10⁴ to 10⁵ CFU/mouse). Fecal samples were tested over a 15-day period for numbers of each strain (CFU/gram of feces). The data obtained indicated that colonization of the mouse intestine by E. coli MG1655 appears to depend on its ability to grow on a combination of at least seven fermentable monosaccharides that are released from mucus-derived polysaccharides by enzymes secreted by the dominant anaerobic microflora (with the exception of gluconate, which cannot be polymerized). Gluconate appears to be a major carbon source used by E. coli MG1655 to colonize, having an impact on both the initiation and maintenance stages; i.e., in the presence of the wild-type strain, the mutant defective in gluconate catabolism (Δ edd) failed to grow in the intestine during the first 3 days after feeding, by 7 days dropped to a level of only 10³ CFU/g of feces, and remained at that level thereafter, whereas the wild-type strain colonized at a level of 10⁸ CFU/g of feces (11a). N-Acetylglucosamine and N-acetylneuraminic acid appeared to be involved in initiation, but not maintenance, i.e., in the presence of the wild-type strain, both the N-acetylglucosamine mutant (Δ nagE) and the N-acetylneuraminic acid mutant (Δ nanAT) grew poorly in the intestine during the first 3 days after feeding (∼50-fold lower than the wild type), but then stabilized at a level of about 100-fold lower than the wild type thereafter (11a). These sugars appear to be those that are preferred by E. coli MG1655 in vitro as well as in vivo (11a). Glucuronate (Δ uxaC), mannose (Δ manA), fucose (Δ fucK and Δ fucAO), and ribose (Δ rbsK) appeared to be involved in maintenance (∼100-fold lower than the wild type by day 15 after feeding), but not initiation (11a). Of further interest is the fact that each of the mutants, when fed to mice alone, colonized at normal levels. This result further indicates that E. coli MG1655 cometabolizes several sugars simultaneously in the intestine, allowing it to achieve its maximal rate of growth throughout both initiation and maintenance.

Despite several additional genes being induced when grown on mucus, their corresponding nutrients did not appear to be used by E. coli MG1655 in the mouse intestine, i.e., initiation and maintenance levels of the mutants and the wild-type strain were identical in cofeeding experiments (11a, 79). The nutrients apparently not utilized in the intestine by E. coli MG1655 included ethanolamine (Δ eutBC), glycerol-3-phosphate released from phospholipids (Δ glpTQ), glycerol (Δ glpK), fructuronate (Δ gntP), glucuronate released from oligosaccharides by β-glucuronidase (Δ uidA), galacturonate (Δ uxaB), tryptophan (Δ tnaA), and gluconeogenic nutrients (Δ ppsA Δ pckA).

Of great interest is the finding that E. coli MG1655 and E. coli EDL933, an O157:H7 strain that causes enterohemorrhagic colitis (118), appear to utilize different nutrients for growth when in competition in the streptomycin-treated-mouse large intestine (79). Thus, E. coli EDL933 was able to grow from small numbers (10⁵ CFU/g of feces) to large numbers (10⁸ CFU/g of feces) in the presence of large numbers of MG1655 (10⁹ CFU/g of feces) during initiation, using only as yet unidentified sugars for growth, but it was eliminated slowly during maintenance as it switched, in part, to growth on presently unidentified gluconeogenic nutrients (79). In the absence of E. coli MG1655, E. coli EDL933 used only sugars for growth in the mouse intestine, both during initiation and maintenance, suggesting that during maintenance E. coli MG1655 outcompetes E. coli EDL933 for one or more sugars it normally uses for successful maintenance in the absence of MG1655, thereby forcing E. coli EDL933 to use gluconeogenic nutrients in an attempt to remain in the intestine (79).

CLOSING COMMENTS

It is clear that the systematic analysis of E. coli MG1655, together with the initial characterization of strain EDL933 that has just begun, suggests the testable hypothesis that there is a nutritional basis for intestinal colonization which explains the steady progression of commensal strains in human hosts (6). One prediction of this hypothesis is that different commensal strains consume a subset of the available nutrients in the intestine and leave other nutrients available for additional E. coli strains to colonize. It may soon be possible to explain how various commensal strains have an impact on nutrient availability, in some cases providing an open niche for infecting pathogens and, in others, preventing infection. Hence, individual differences in the E. coli microflora could be the reason for differences in human susceptibility to gastroenteritis (84). Future research will hopefully provide a nutritional framework for understanding whether and how commensal E. coli strains serve as the first line of defense against pathogenic E. coli intestinal infections.

ACKNOWLEDGMENTS

This work was supported by the U. S. Public Heath Service (T.C. and P.S.C.) and the Danish Medical Research Council (K.A.K.).