Better living through microbial action: the benefits of the mammalian gastrointestinal microbiota on the host

Introduction

The gastrointestinal tract (GI-tract) of mammals is densely populated by microorganisms, commonly referred to as the gastrointestinal microbiota. Bacteria are predominant, but Archaea, yeasts, fungi and protozoa are also part of the microbiota. The gastrointestinal microbiota consists of more than 10¹⁴ bacteria, which is 10 times the number of somatic cells in the human body (Luckey, 1972). The host/microbiota relationship in a healthy individual is a homeostatic symbiosis, in which the host provides the microbiota with nutrients and a stable environment. In return, the microbiota assists the host in maturing the GI-tract and priming the immune system early in life, and provides the host with nutritional contributions. Monogastric mammals may obtain a substantial part of their energy requirement from microbial fermentation of otherwise indigestible food items, while ruminants are totally dependent on this input. Finally, the microbiota offers some protection against harmful microbial species. While the anatomy of the GI-tract differs substantially between mammal species, the function is to provide energy and nutrients through food ingestion, digestion and absorption. The intestinal epithelium is specialized to ensure optimal absorption of nutritional compounds, yet at the same time to exclude and neutralize or detoxify harmful components of the intestinal contents including microorganisms. The epithelial function is influenced by direct host/microbiota interactions, and through microbial activities on the diet, accordingly, a microbiota ensuring the best possible operation of the epithelium may be considered advantageous.

Acquisition and establishment of the microbiota

Bacteria start colonizing the sterile infant gut within hours after birth followed by a bacterial succession until an adult microbiota has been established post weaning. In two large studies the dominant bacterial groups in the infant GI-tract were found to be Proteobacteria, Bacteroidetes, Firmicutes, Actinobacteria and Verrucomicrobia (Favier et al., 2002; Palmer et al., 2007). Seemingly, the gut microbiota develops in a chaotic progression during the first months of life depending on bacterial exposure from the surrounding environment. At 6 months of age the human faecal microbiota is dominated by Bacteroidetes and Firmicutes, common occurrence of Verrucomicrobia, and very low abundance of Proteobacteria and aerobic Gram-negative bacteria in general (Palmer et al., 2007).

Predominance of Firmicutes and Bacteroidetes in mammals has been found in several large-scale 16S rRNA sequence-based studies. More than 80% of the identified phylotypes belong to these two phyla in human gut biopsies and faecal samples from a wide range of mammalian species (Leser et al., 2002; Eckburg et al., 2005; Ley et al., 2008a). Although the mammalian gut microbiota is dominated by only a few phyla, the genera and species diversity is huge. Four hundred to 1000 phylotypes, roughly corresponding to bacterial species, have been estimated to inhabit a healthy human intestine by 16S rRNA cloning and sequencing (for a review see Rajilic-Stojanovic et al., 2007).

A recent study showed that the bacterial community structure is more similar within mammalian species than between different species, and further, that the microbiota overall composition reflects gut anatomy and diet type. Thus, the microbiota of herbivore, omnivore and carnivore groups of mammals cluster discretely (Ley et al., 2008a,b). The faecal microbiota of adult human individuals is unique and highly stable through time (Zoetendal et al., 1998; Tannock et al., 2004; Jernberg et al., 2007), and the composition is at least to some extent determined by host genetics (Zoetendal et al., 2001; Stewart et al., 2005).

The gastrointestinal habitat

The complexity of the microbial GI-tract habitat varies between mammalian species and among locations in the GI-tract on a both longitudinal and horizontal axis. The main microbial exposure is through the ingestion of food and drinks. The diet is therefore of major importance as a reservoir of microorganisms and as a substrate to the intestinal microbes. Mechanical and enzymatic digestion of food begins in the mouth and continues in the stomach. The anatomy of the stomach varies considerably among mammal species, with the largest difference between simple stomached (monogastric) animals and ruminants. In ruminants food is fermented by microorganisms in the rumen before digestion in the small and large intestines. Some monogastric animals including rodents, horses and pigs have a non-glandular part of the stomach which is densely populated by Lactobacillus species (Savage et al., 1968; Barrow et al., 1980; Yuki et al., 2000). This area is characterized by the lack of mucus covering and by the direct adhesion of bacteria to epithelial cells (Tannock, 2005). In monogastric animals the food passes from the oesophagus or the non-glandular part of the stomach to the glandular, acid secreting part. The low pH (1–2 in adult humans) is detrimental to most microorganisms and is one of the body's most effective defences against pathogens. Some microbial species can survive the conditions in the stomach at low population levels. The most ‘famous’ stomach bacterium is Helicobacter pylori, which causes gastritis. Fifty per cent of the world's human population is estimated to be H. pylori infected, but the majority have no indication of disease (Maaroos, 1995; Lee, 1999). Non-H. pylori species have been associated with gastritis in animals like dogs, pigs, sheep and cattle (Cattoli et al., 1999a,b; De Groote et al., 1999a,b; Dore et al., 2001). Sarcina ventriculi is another species found in human stomach contents, mainly from vegetarians (Aries et al., 1969; Crowther, 1970). In rodents, which have a less acidic stomach (pH 3–5) due to the constant presence of food, a denser microbiota is present consisting of acid-resistant lactobacilli inhabiting the gastric surface (Lee, 1999).

The small intestine comprises the duodenum, jejunum and ileum (Fig. 1). Most of the digestion and absorption of food take place in the small intestine. The small intestine is lined with simple columnar epithelial tissue, covered by a mucus layer, and has a large surface area due to the villi and microvilli. When food passes into the duodenum, pH and bacterial load are low. Food is blended with bile, bicarbonate and digestive enzymes in the duodenum and when the intestinal content (chyme) reaches the large intestine it has been converted to a neutral-alkaline pH (Kaunitz and Akiba, 2006). In healthy humans, chyme passes from the duodenum to the end of the ileum in 1–4 h and the bacterial content increases from approximately 10⁴ bacteria to 10⁸ bacteria per ml of intestinal content (Laux et al., 2005) (Fig. 2A and B). This implies a high microbial activity, and a competition between host and microbiota for easily digestible nutrients. The small intestine is the place where bacteria have been reported to deconjugate bile, and to produce vitamins, and amino acids, essential to the host (Tannock et al., 1989; Conly and Stein., 1992; Tannock and McConnell, 1994; Tannock et al., 1994; Torrallardona et al., 2003a,b). Whereas the transition time is short for bacteria populating chyme, it can be prolonged if bacteria colonize the intestinal mucus layer. Two mucus layers have been identified in the GI-tract of rats: a layer firmly attached to the epithelium and a loosely adherent layer facing the intestinal lumen. The mucus layers function as a mechanical barrier separating luminal bacteria from the epithelium. Although bacteria colonize the outer layer (100–200 μm thick in duodenum and jejunum and up to 500 μm in thickness in ileum), the inner firmly attached layer is devoid of bacteria (Atuma et al., 2001; Johansson et al., 2008). In the small intestine, the inner mucus layer is thin (15 μm) and discontinuous and individual tips of the villi can be devoid of mucus (Atuma et al., 2001). Shear forces and the constant shedding of epithelial cells and the attached mucus enhance uptake of nutrients from the chyme and reduce the number of mucus-associated bacterial cells. Upon sloughing off, mucus-associated bacteria are mixed with the intestinal content and can therefore be detected in this. The microbiota of the human small intestinal content and the mucus layer is not well characterized because of the difficulties in sampling these locations. In a study by Ahmed and colleagues (2007), using fluorescent in situ hybridization (FISH) and DAPI staining to visualize the microbial population colonizing healthy terminal ileum samples, it was found that bacteria were present as heterogeneous assemblages in the mucus layer and that some of the ileal biopsies harboured large numbers of helical bacteria.

The large intestine consists of the colon and caecum. In humans the caecum is rudimentary whereas in many animals, like pigs and rodents, it is more prominent. The transit time in the large intestine can vary from 10 h to several days in humans. The large intestine is the most heavily colonized area of the GI-tract and has been reported to hold 10¹¹−10¹² bacteria per ml of intestinal content (Laux et al., 2005). Dietary compounds that are not degraded in the upper GI-tract reach the large bowel where they support the microbiota with nutrients and energy. The microbiota ferment carbohydrates into CO₂, H₂, CH₄, and short-chain fatty acids (SCFA), primarily acetate, propionate and butyrate. Most of the SCFA produced in the large intestine are absorbed by the host and provide an energy source. In humans, the amount of energy derived from SCFA accounts for 6–9% of the total energy requirement (McNeil, 1984). However, animals with a (relatively) large intestine and well-developed caecum may acquire as much as 44% of their energy intake from microbial SCFA (Hume, 1997). Butyrate is the preferred energy source of colonocytes and is readily absorbed and metabolized in the epithelium (Fitch and Fleming, 1999). Butyrate in concentrations comparable with those occurring in the large bowel regulate cell proliferation and apoptosis (Avivi-Green et al., 2002), possibly by controlling gene expression through hyperacetylation of specific nuclear histones (Della Ragione et al., 2001). Butyrate is mainly produced by Roseburia spp. and Eubacterium rectale, both members of the clostridial cluster XIVa, and to a lesser extent by Faecalibacterium prausnitzii-like organisms. Populations of these bacterial groups are susceptible to changes in dietary intake of carbohydrates which influences colonic butyrate concentrations (Duncan et al., 2007).

Bacteria colonize both the intestinal content on the surfaces of digestive residues and the outer mucus layer (Fig. 2C). The loosely adherent mucus layer facing the intestinal lumen of the large intestine is thick (700–900 μm) and densely populated by bacteria (Atuma et al., 2001), while there is no contact between bacteria and the colonic mucosa in healthy individuals (van der Waaij et al., 2005; Swidsinski et al., 2007). The mucus-associated bacterial community differs from that of the lumen and appears to be host specific and controlled by host-related factors (Zoetendal et al., 2002; Eckburg et al., 2005). The spatial organization of the microbial colonization in the outer mucus layer and on digestive residues is relatively unknown. Bacteria can be found diffusely spread, in microcolonies, and perhaps even in biofilms (Macfarlane et al., 2004; Swidsinski et al., 2005; Macfarlane and Macfarlane, 2006).

Despite the continuous flow of colonic contents, the bacterial density and diversity of the colon is high and remains relatively stable. This reflects a system maintaining a firm degree of homeostasis. It is not understood how the constancy of this complex system is ensured, but contributing factors may be a structured bacterial organization regulated by metabolic cross-talk and cell-to-cell communications among bacteria and between bacteria and the host epithelium.

The microbiota affects the gut epithelium

The dense population of microorganisms in the GI-tract has a marked impact on the host epithelium. Extensive cross-talk between the microbiota and the gut epithelium takes place, and the gut microbiota influences the GI-tract and the host as such. However, life without a microbiota is possible and germ-free (GF) animals have been used to study how the microbiota affects the host and how the host adapts to the microbiota. The GI-tract of animals raised under GF conditions exhibits morphological and physiological characteristics that are different from conventional (CONV) animals. For example, the epithelial cell turnover is twice as fast in CONV as in GF animals (Savage et al., 1981), and the number of secretory goblet and enteroendocrine cells is increased in CONV animals (Bates et al., 2006). Other differences include changes in the immune system, body metabolism, electrolyte and fluid handling, the vasculature, the liver, the endocrine system and behaviour (data compiled in Smith et al., 2007).

The postnatal gut is partly maturated by the microbiota. Although intrinsic factors like glucocorticoid hormones drive a developmental scheme, the gut microbiota influences intestinal epithelial ontogony. Activities of brush-border enzymes, such as disaccharide hydrolases, peptidases and alkaline phophatase, which are markers of enterocyte differentiation, are modified by intestinal colonization (Whitt and Savage, 1988; Kozakova et al., 2006; Siggers et al., 2008). Postnatal gene expression of β1,4-galactosyltransferase (βGT), a marker of glycosyltransferase development, increases rapidly in CONV mice, but stays at low levels in GF mice. βGT expression is rapidly induced upon conventionalization of the GF mice with an adult microbiota. When colonized by a microbiota derived from suckling mice, βGT gene expression remains low, suggesting that specific components of the microbiota are responsible for the induction of this gene (Nanthakumar et al., 2005).

Germ-free animals can be colonized with one or a few bacterial species, becoming gnotobiotic. In the gnotobiotic set-up, the effects of individual bacterial strains or defined groups of strains on the host epithelium can be investigated. Gnotobiotic studies have demonstrated a bacteria-driven modification of the host's glycosylation pattern. Glycoconjugates cover the epithelial cell surface and serve as receptors for bacterial adhesins and as a source of nutrients to the bacteria. When GF mice are colonized with segmented filamentous bacteria (SFB), the small intestinal epithelial cells (IECs) induce α1,2 fucosyltransferase activity that mediates synthesis of the fucosyl asialoGM1 glycolipid (Umesaki et al., 1995). Segmented filamentous bacteria are indigenous intestinal bacteria in mice, which attach to the upper parts of the villi. Bry and colleagues (1996) showed a similar fucosyltransferase induction and production of fucosylated glycoconjugates in GF mice colonized by the human strain Bacteroides thetaiotaomicron VPI-5482. Unlike the situation with SFB, signalling occurs without direct bacterial attachment to the epithelium, but depends on the ability of B. thetaiotaomicron to use fucose as a carbon source, and is possibly mediated via a soluble compound (Bry et al., 1996). When mice were colonized by other bacterial species, including Peptostreptococcus micros or Bifidobacterium infantis, no changes in expression of fucosylated glycoconjugates were found indicating that the induction of the fucosylation programme is a consequence of specific bacterial–host interactions.

With the advent of microarray technology, it became possible to screen cell or tissue responses to microbial colonization by transcriptome profiling. In a pioneering work, Hooper and colleagues (2001) showed how colonization of a single gut commensal (Bacteroides thetaiotaomicron) modulates host gene expression in the mouse ileal epithelium. Microarray analysis of intestinal transcriptional responses established that colonization influences diverse and fundamental physiological functions (Table 1). The range and kind of functions regulated by colonization were surprising, e.g. the upregulation of the ileal Na⁺/glucose co-transporter, and four genes involved in lipid absorption and transport, suggesting microbial effects on host metabolism. These findings may partly explain why GF rodents require a higher caloric intake to maintain their body weight than CONV animals (Gordon, 1968). The finding of four genes associated with steroid function highlighted the maturational effects of microbial colonization of the gut. The upregulation of seven genes annotated to mucosal barrier function, and the upregulation of several genes involved in the cytoskeleton and extracellular matrix indicated a strong stimulation of epithelial barrier integrity. A discernible effect on immune response genes was absent suggesting that colonization by B. thetaiotaomicron produces no inflammatory response. A lack of immune response was also found when GF mice were mono-associated with different lactobacilli while CONV mice mounted an inflammatory immune response upon colonization with the same strains (Nerstedt et al., 2007). This could be due to an incompetent immune system in ex-GF animals. Thus, these results and other studies clearly demonstrate that the microbiota affects a multitude of epithelial functions and that these extend far beyond the immune response.

The host may respond differently to subsets of the gut microbiota. Thus, B. thetaiotaomicron colonization upregulated expression of the small proline-rich protein 2a 205-fold in gnotobiotic mice, while colonization with Escherichia coli or B. infantis only slightly upregulated the gene (Hooper et al., 2001). Colonization of CONV mice with Lactobacillus paracasei spp. paracasei F19, or L. acidophilus NCFB 1748, induced significant regulation of 22 and 55 genes in the distal ileum respectively. Twenty genes were regulated by both strains (Nerstedt et al., 2007). Seemingly, host responses are specific to different members of the microbiota.

Throughout the GI-tract the most significant molecular function affected by the microbiota is ‘defence/immunity protein activity’, and all genes with this annotation are upregulated in the small and large intestines of CONV mice (Mutch et al., 2004). When GF piglets were conventionalized, epithelial inflammatory responses were detected in the ileum using high-density porcine oligonucleotide microarrays. Genes involved in epithelial cell turnover, mucus biosynthesis and priming of the immune system were induced. Receptors and transcription factors related to INF-inducible genes were upregulated, yet at the same time, the inflammatory response seemed to be controlled through activation of genes in pathways that prevent excessive inflammation, supporting the concept of a homeostatic epithelium that maintains a tight intestinal barrier without producing excessive inflammatory responses, which would compromise barrier function (Chowdhury et al., 2007). It is a characteristic feature of a healthy mucosal defence system that protective immune responses against pathogenic organisms are allowed to proceed while the indigenous microbiota is tolerated. Disturbance of this homeostasis between microbiota and host is associated with chronic intestinal inflammatory diseases. Even in mice with an established microbiota, colonization by commensal bacteria not previously encountered by the animals activates IECs. However, the activation is modest, controlled and transient (Hoffmann et al., 2008). Intestinal epithelial cells are closely linked to the regulation of innate and adaptive defence mechanisms through their contact with luminal antigens and with immune cells (dendritic cells, macrophages) of the lamina propria (Clavel and Haller, 2007; Zaph et al., 2007). Thus, activation of Toll-like receptors (TLRs) on the apical surface of IECs, typical of commensal, non-invasive microbes, induces a partial and protective activation of NF-κB, while basolateral activation of TLRs by invasive pathogens leads to a robust inflammatory response (Rescigno et al., 2008). Pathogenic bacteria are able to invade spaces that are usually devoid of bacteria, such as epithelial crypts, the epithelial surface, epithelial cells and the lamina propria where they elicit defence responses. Other factors that contribute to control intestinal TLR activation are known (Shibolet and Podolsky, 2007; Zaph et al., 2007); however, the mechanisms by which the microbiota is discerned from pathogens and is tolerated by the host is far from understood.

The gut microbiota assists in ensuring the integrity of the mechanical mucosal barrier. The mucus gel overlying the GI-tract epithelium consists of heavily glycosylated proteins, and the large gel forming mucin, Muc2, is the major structural component. The microbiota stimulates the mucus layer, thus, in CONV rats the colonic mucus layer is twice as thick as in GF rats (Szentkuti et al., 1990), and the mucin chemical composition is altered (Sharma et al., 1995; Meslin et al., 1999). Orally ingested probiotic lactobacilli, but not bifidobacteria, have been shown to stimulate MUC2 gene expression and mucin secretion in the colon of CONV rats (Caballero-Franco et al., 2007). Stimulation of MUC2 and MUC3 expression in HT-29 cells by lactobacilli inhibits adherence of enteropathogenic E. coli to the cells possibly because E. coli epithelial attachment sites are mimicked by the mucus (Mack et al., 1999).

Further protection of the host is provided by epithelial production of antimicrobial compounds. The microbiota induces the host mucosal immune system to produce immunoglobulin A (IgA), which is released into the intestinal lumen in large amounts rendering the intestinal mucosa quantitatively the most important effector organ of antibody-mediated immunity (Brandtzaeg, 2007). According to the model proposed by Macpherson and colleagues (2005), IgA limits local epithelial bacterial colonization and prevents penetration of bacteria through the epithelial layer. Commensals reside almost entirely in the gut lumen and in the outer layer of the mucus coat, but are constantly sampled by dendritic cells in Peyer's patches leading to induction of local IgA responses (McCracken and Lorenz, 2001; Macpherson and Uhr, 2004). Unlike pathogens, which have evolved mechanisms of avoiding phagocytosis (Sansonetti, 2001), commensals breaching the mucus/epithelium barrier are readily destroyed (Macpherson et al., 2005). A multitude of bactericidal peptides and proteins are produced by the gut epithelium. These molecules commonly have a broad specificity against microbes. Paneth cells which are specialized secretory cells located at the base of the crypts in the small intestine produce α-defensins, phospholipase A2 and lysozyme but also IECs produce antimicrobial compounds including β-defensins, cathelicidins, bactericidal/permeability-increasing protein (BPI) and chemokine (Müller et al., 2005; Lievin-Le Moal and Servin, 2006). A recent study has shown that secreted antibacterial activity is confined to the mucus layer, which then provides a strong mucosal physical and antibacterial barrier while allowing the presence of a luminal microbiota (Meyer-Hoffert et al., 2008).

Implications for human lifestyle diseases

Recently, the gut microbiota has been implicated in human obesity and related lifestyle diseases such as diabetes and cardiovascular diseases. When adult GF mice were conventionalized with a microbiota harvested from the caecum of conventionally raised mice the body fat content increased 60% after 14 days in spite of a 27% lower chow intake compared with GF animals. Lean body mass decreased 7%, leaving total body weight unchanged (Bäckhed et al., 2004). Compared with GF mice the conventionalized animals had a higher metabolic rate, increased leptin levels, increased insulin resistance and increased monosaccharide uptake. De novo hepatic lipogenesis was induced, and it was found that the introduction of a microbiota suppressed fasting-inducing adipocyte factor (Fiaf) expression in the intestinal epithelium. Fiaf is a circulating inhibitor of lipoprotein lipase (Lpl), and it was suggested that the microbiota, through its suppression of Fiaf, causes an increased Lpl activity leading to adiposity (Bäckhed et al., 2004; 2007).

The mouse caecum is dominated by Firmicutes (60–80% of phylotypes) and Bacteroidetes (20–40% of phylotypes), but Ley and colleagues (2005) found a statistically significant 50% reduction in the abundance of Bacteroidetes in obese ob/ob mice relative to lean mice and a proportional increase in Firmicutes. It was found that the obese microbiota is enriched for genes encoding enzymes that break down otherwise indigestible dietary polysaccharides and may have an increased capacity to harvest energy from the diet (Turnbaugh et al., 2006). When GF mice were colonized with an ob/ob microbiota they gained significantly more body fat than mice receiving microbiota from lean mice, suggesting a contributory effect of the microbiota composition on development of obesity (Turnbaugh et al., 2006). An increased proportion of the Mollicute lineage within the Firmicutes was found in diet induced obese mice together with a suppression of Bacteroidetes. Mollicutes appear to have a high capacity to import the type of carbohydrates found in a Western diet and to metabolize these into SCFA that are readily absorbed by the host (Turnbaugh et al., 2008). In humans, a reduced proportion of Bacteroidetes was found in obese individuals. Over a 1-year time-course, the abundance of Bacteroidetes increased and the abundance of Firmicutes decreased in obese individuals who were assigned to a low-calorie diet. The change was division-wide and bacterial diversity remained constant over time (Ley et al., 2006). The division-wide difference associated with obesity has recently been questioned. When quantitative FISH or PCR was applied to analyse the microbiota composition, no significant relationship was found in human stool between body mass index and the relative proportion of Bacteroidetes and Firmicutes and the percentage of Bacteroidetes was unchanged in volunteers undergoing weight loss on carbohydrate restricted diets. However, diet composition did cause a significant reduction in the number of butyrate producing Firmicutes, while total numbers of Firmicutes was unaffected (Duncan et al., 2008).

A different hypothesis of the involvement of the microbiota in obesity and associated lifestyle diseases has been proposed by Cani and co-workers, who introduced the concept of metabolic endotoxaemia (ME). Metabolic endotoxaemia is a condition of slightly elevated plasma concentrations (two to three times normal) of lipopolysaccharides (LPS). Lipopolysaccharide is found in the outer membrane of Gram-negative bacteria, and is released in the GI-tract upon bacterial lysis. Obesity is associated with a state of chronic low-level systemic inflammation increasing the risk of insulin resistance and cardiovascular diseases (Wellen and Hotamisligil, 2005), and Cani hypothesized that LPS is the triggering factor of inflammation during a sustained intake of a high-fat diet (Cani et al., 2007a). Mice fed a high-fat diet became obese and, indeed, had elevated plasma LPS levels and a metabolic response similar to that of high-fat feeding was induced by LPS infusion in normal diet-fed mice. The high-fat diet decreased abundance of some groups of both Firmicutes and Bacteroidetes, as well as Bifidobacterium spp. Introduction of oligofructose to the high-fat diet restored the abundance of bifidobacteria as well as circulating LPS concentrations to normal conditions. Endotoxemia significantly and negatively correlated with bifidobacteria, but no correlation was seen for any other bacterial group leading the authors to conclude that modulating the microbiota in favour of bifidobacteria could reduce the risk of metabolic diseases caused by a high-fat diet (Cani et al., 2007b).

These studies suggest a role for the microbiota in regulating body fat accumulation; however, the causative mechanisms remain unknown. Research should be directed towards linking these mechanisms to specific groups of bacteria and to determining if the microbiota shifts are causing the effects on host physiology, or they are the result of changes in the host.

Manipulating the gut microbiota

In the beginning of the 20th century, Metchnikoff speculated that the ageing process was caused by toxic compounds produced by proteolytic clostridia in the colon, and that beneficial or harmless lactic acid bacteria (LAB) could replace the proteolytic species for health benefits. Metchnikoff's ideas started a consumer and industry interest in using beneficial microorganisms as dietary supplements (probiotics), while scientific interest, until recently, has been scarce. Most probiotic bacterial strains are LAB belonging to the Lactobacillus and Bifidobacterium genera which are considered health-promoting. Yet, taking into consideration the vast diversity of the gut microbiota and its functions, health-promoting strains unrelated to LAB are likely to exist. Initially the concept of probiotics implied an improvement of the gut microbiota balance, but today, specific health effects are being investigated and documented such as alleviation of chronic intestinal inflammatory diseases (Mach, 2006), prevention and treatment of pathogen-induced diarrhoea (Yan and Polk, 2006), or treating atopic diseases (Vanderhoof, 2008). Ideally, probiotics should survive passage through the harsh conditions in the proximal GI-tract to reach their site of action where they should proliferate. Accordingly, probiotic bacterial strains have been selected on their ability to endure acid and bile stress and their capability to adhere to epithelial cells in vitro. Nevertheless, probiotic strains can usually only be detected in human and animal faeces while the probiotic is being consumed, and their beneficial effects are limited to that period (Tannock et al., 2000; Cronin et al., 2008; Leser et al., 2008). Stable probiotics colonization is possible in gnotobiotic animals (Sonnenburg et al., 2006; Ménard et al., 2008), emphasizing the importance of the microbiota in preventing invading bacterial species from establishing in the gut. Recently analysis of genome sequences has revealed specific traits of autochthonous intestinal bacteria that contribute to their adaptation to the ecological niches in the GI-tract (Azcarate-Peril et al., 2008), and such characteristics may be used for improving physiological robustness of probiotic cultures (Sheehan et al., 2007).

Prebiotics are non-digestible food ingredients which stimulate beneficial bacteria already resident in the colon (Gibson and Roberfroid, 1995). Dietary intake of inulin, fructo-oligosaccharides or galacto-oligosaccharides stimulates the growth of bifidobacteria and lactobacilli, and most likely other bacterial genera (Macfarlane et al., 2006). Accordingly, prebiotics are another way to obtain the health effects of LAB.

The ‘omics’ era

It is evident that the GI-tract and its microbiota represent a very complex ecosystem to study. But with the emerging ‘omics’ technologies, such as metagenomics, metatranscriptomics, metaproteomics and metabonomics, it will be possible to understand the activities, functions and interactions of the gut microbiota in a systems biology perspective. The prospects and technical approaches of this type of research have been reviewed in several recent papers (Kowalchuk et al., 2007; Medini et al., 2008; Nicholson and Lindon, 2008; Zoetendal et al., 2008; Hattori and Taylor, 2009), and will not be discussed here. Results are starting to accumulate from these technologies, thus, large-scale sequencing has been used for whole genome sequencing (Manichanh et al., 2006), deep 16S rRNA gene sequencing with thousands of reads from each sample (Huse et al., 2008), and for metagenomic analysis of the gastrointestinal ecosystem (Gill et al., 2006). Next-generation sequencing platforms have increased throughput and lowered cost: where the Humane Genome Project lasted 13 years to complete; a human genome was recently sequenced in 2 months with a 7.4-fold redundancy (Wheeler et al., 2008). The sequencing boost has also impacted the number of available microbial genomes; in the beginning of 2009, 782 whole genomes were sequenced and publicly available. Of these, 419 were Proteobacteria, 169 Firmicutes, 58 Actinobacteria, and 26 belonged to the Bacteroidetes/Chlorobi group (http://www.cbs.dtu.dk/services/GenomeAtlas/). Whole genome sequences provide not only essential insights into the functional potential of the sequenced bacterial isolates, but also make possible the comparison of different isolates of the same species (pan-genome) and/or other species. Thus, by comparing genome sequences of Bacteroides vulgatus and Bacteroides distasonis with sequences of other gut and non-gut Bacteroidetes, Xu and colleagues (2007) found that Bacteroidetes have adapted to the gastrointestinal habitat by varying their cell surface, environment sensing, and nutrient harvest through evolutionary mechanisms. Lateral gene transfer, mobile elements and gene amplification have shaped the genomes to adapt to niches within the gut habitat allowing for a non-competitive mutual existence, yet at the same time providing sufficient functional redundancy to ensure system stability towards perturbation.

The lack of adequate culturing methods for the majority (> 80%) of gastrointestinal bacteria and the large variation of the pan-genomic gene pools limits the ecological information that can be obtained by whole genome sequencing. Metagenomics, which targets microbial communities in the environment, represents one solution to this problem. Through direct extraction, cloning and analysis of all of the genomes as a single entity in microbial assemblages, phylogenetic and functional information is provided, from culturable as well as not yet culturable microbes (Stein et al., 1996; Handelsman et al., 1998; Handelsman, 2004). Metagenomics has showed that the human microbiome is enriched of microbial genes involved in the metabolism of glycans, amino acids and xenobiotics, methanogenesis and biosynthesis of vitamins (Gill et al., 2006), and that bile salt hydrolases are present in all major bacterial divisions and archaeal species in the gut (Jones et al., 2008). In a large-scale comparative metagenomic analysis of faecal samples from 13 healthy individuals of various ages, Kurokawa and colleagues (2007) found a fundamentally different gut microbiota in unweaned infants from that in adults and weaned children. The infant microbiota was structurally simple with a high level of interindividual variation in taxonomic and gene composition, while the adult and weaned children microbiota was complex with a high functional uniformity regardless of age or sex. By analysing 662 548 predicted genes it was found that the genes in the category of carbohydrate transport and metabolism were over-represented and those for lipid transport and metabolism were under-represented in the faecal microbiota compared with microorganisms from other environments such as sea water and soil. The groups of enriched genes varied remarkably between infants and adults. Although the number of individuals sampled for metagenomic studies has been small, the available data sets already assist in creating knowledge about gene sets encoding core functions of the gut microbiota and how these are affected by intrinsic and environmental factors like age, diet and host genotype. While metagenomic analyses are costly, their biggest challenge is to bring together the expertise of microbiologists, system biologists, mathematicians and bioinformatics in order to make the most of the large amounts of data generated and to interrogate metagenomes for important microbial functions. Two large consortia have been initiated: the Human Microbiome Project and MetaHIT, aiming at characterizing the composition, diversity and distribution of human-associated microbial communities (Turnbaugh et al., 2007; Karow, 2008).

DNA-targeting metagenomics can be combined with other approaches like transcriptomics, proteomics and metabonomics to gain insight into the microbial genes, proteins and metabolites that are actually active and present in the gut. Metabonomics aim at measuring the global, dynamic metabolic response of living systems to biological stimuli or genetic manipulation (Nicholson and Lindon, 2008). Typically, high-resolution ¹H nuclear magnetic resonance (NMR) spectroscopy and mass spectrometry is used to detect metabolites in biological fluid samples. Top-down multivariate analyses of metabolic profiles have demonstrated that host metabolism is sensitive to gut-microbial metabolism which provides complimentary pathways of diet and drugs. Effects are not limited to the gut epithelium, but include organs like liver and kidneys eventually affecting the global metabolic phenotype (Claus et al., 2008). Through combined analyses of faecal microbiota composition and urinary metabonomic profiles, Li and colleagues (2008) showed that specific metabotypes can be associated with the structure of the gastrointestinal microbiota. Thus, in humans, urinary metabolic markers could be linked to specific members of the gut microbiota, e.g. the abundance of F. prausnitzii associated with modulation of eight urinary metabolites. In blood plasma a significant effect of the microbiota on metabolite profiles was found in CONV versus GF mice. Fifty-two metabolites were present only in GF mice, while 145 metabolites were found exclusively in CONV animals. Approximately 400 of the metabolites in common differed significantly in concentration between GF and CONV mice. Amino acid metabolites were affected by the microbiota and several of the microbiota-affected metabolites were potentially harmful (uraemic toxins) or beneficial (antioxidant) to the host. Substantial numbers of metabolites derived from detoxification of bacterial substances were found in CONV mice (Wikoff et al., 2009). Host lipid metabolism is strongly influenced by microbial modulation of bile acids. Bacterial de-conjugation of bile acids in the intestine subsequently affects host lipid absorption, cholesterol metabolism, liver fatty acids storage and lipoperoxidation (Martin et al., 2007; Martin et al., 2008). Metagenomics and metabolic profiling have demonstrated how diet can influence the structure of the gut microbiota causing changes in gut-microbial metabolism that eventually alter host metabolism and contribute to insulin resistance and obesity (Dumas et al., 2006; Turnbaugh et al., 2008).

While these new methods undoubtedly will increase our understanding of the structure and function of gastrointestinal microbial ecosystems and eventually may link the microbiota to certain diseases, the ‘omics’ technologies cannot stand alone. Hypotheses generated by holistic, systems biology approaches must be tested in simple and controlled experimental settings. One key constraint is our current inability to culture the majority of gastrointestinal microbial species, highlighting the importance of expanding the culturable range of microorganisms from the GI-tract. Animal models, e.g. disease models, are required for experimental validation of hypotheses, and to provide easy access to samples from those GI-tract regions that cannot be represented by faecal samples.

Concluding remarks

In Introduction, we proposed that a favourable gut microbiota ensures optimal epithelial functioning and that optimal epithelial functioning implies:

Today, experimental data support the view that these functions are all affected by the microbiota through diverse processes (Fig. 3), and that the mammalian host and its microbiota have evolved into a homeostatic, symbiotic relationship. Although the microbiota acts locally on the epithelium this eventually results in secondary, systemic effects.