Understanding the effects of diet on bacterial metabolism in the large intestine


Harry James Flint, Microbial Ecology Group, Rowett Research Institute, Greenburn Road, Bucksburn, Aberdeen AB21 9SB, United Kingdom. E-mail: H.Flint@rowett.ac.uk


Recent analyses of ribosomal RNA sequence diversity have demonstrated the extent of bacterial diversity in the human colon, and have provided new tools for monitoring changes in the composition of the gut microbial community. There is now an excellent opportunity to correlate ecological niches and metabolic activities with particular phylogenetic groups among the microbiota of the human gut. Bacteria that associate closely with particulate material and surfaces in the gut include specialized primary degraders of insoluble substrates, including resistant starch, plant structural polysaccharides and mucin. Butyrate-producing bacteria found in human faeces belong mainly to the clostridial clusters IV and XIVa. In vitro and in vivo evidence indicates that a group related to Roseburia and Eubacterium rectale plays a major role in mediating the butyrogenic effect of fermentable dietary carbohydrates. Additional cluster XIVa species can convert lactate to butyrate, while some members of the clostridial cluster IX convert lactate to propionate. The metabolic outputs of the gut microbial community depend not only on available substrate, but also on the gut environment, with pH playing a major role. Better understanding of the colonic microbial ecosystem will help to explain and predict the effects of dietary additives, including nondigestible carbohydrates, probiotics and prebiotics.


The activities of the microbial community that inhabits the large intestine are considered to play an important role in the maintenance of gut health and in the aetiology of gut disease in humans. It is estimated that less than 25% of colorectal disease has an obvious genetic basis, suggesting that nutrition plays a major role, and this is thought to be mediated to a large extent via its effects on the colonic microbiota (Gill and Rowland 2002). Dietary components that escape digestion by endogenous enzymes in the upper gastrointestinal tract become available as substrates in the large intestine. These ‘nondigestible’ (ND) dietary carbohydrate substrates include resistant starch, plant cell wall material and oligosaccharides (Cummings and Macfarlane 1991). In addition, some dietary protein may reach the colon, while endogenous secretions, including mucin, provide diet-independent substrates. Many secondary plant metabolites ingested with the diet, such as polyphenolic substances, may also reach the large intestine and are subject to bacterial transformations. The purpose of this review is to discuss how recent advances in the molecular ecology of gut microbial communities have helped our understanding of responses to dietary change.

Alterations in diet composition result in both quantitative and qualitative changes in the supply of substrates to the large intestinal microbiota. The impact that dietary changes have upon microbial metabolism occurs through several inter-related mechanisms (Fig. 1). First, metabolism is regulated within each individual species of gut bacterium. Alternative substrates can give rise to different products as a result of fermentation via different metabolic routes, while the same substrates can be processed via different routes depending on their rate of supply, or the physiology and environment of the bacterial cell (Macfarlane and Macfarlane 2003; Scott et al. 2006). Second, there is evidence from in vivo studies with prebiotics (mainly oligofructose) that sustained changes in the supply of ND dietary carbohydrate can lead to shifts in the species composition of the colonic bacterial community, as monitored by faecal sampling (e.g. Gibson et al. 1995). If the groups that are selected for and against differ in their metabolic characteristics, this is expected to alter the metabolic outputs from the community. In addition, both bacterial metabolism and bacterial competition are strongly influenced by the gut environment, i.e. local conditions of pH, oxygen and hydrogen, metabolite concentration and gut transit time (Fig. 1). The gut environment is also determined by host secretions and secondary bacterial metabolites, such as antimicrobials and quorum-sensing molecules.

Figure 1.

 Schematic diagram of the gut microbial ecosystem. Metabolic flows are shown with solid arrows and other influences are shown with dashed arrows. Further details are given in the main text.

Microbial diversity in the human large intestine

The availability of a range of molecular methods, including fluorescent in situ hybridization (FISH) and 16S rRNA clone libraries has provided new insights into the composition of the human gastrointestinal microbiota (e.g. Eckburg et al. 2005; Lay et al. 2005; Flint 2006) (Table 1). It was recently estimated that at least 90% of the bacterial phylotypes detected by 16S rRNA sequencing share less than 99% identity with any from known cultured species (Eckburg et al. 2005). The two most abundant bacterial phyla that colonize the large intestine are the gram-negative Bacteroidetes and the low % G+C Firmicutes. The most abundant Firmicutes are members of the clostridial clusters IV and XIVa with lower abundance of the cluster IX group. In total, these groups may comprise up to 60% of the colonic microbiota (Table 1). Species belonging to the Bacteroidetes, along with many of their metabolic activities, have been relatively well described (Salyers et al. 1977). By contrast, the gram-positive Firmicutes are estimated to include some 78% of the unidentified phylotypes from the human gut (Eckburg et al. 2005). A far better understanding of the functional roles of this dominant group of bacteria is needed to understand the impact of diet on microbial activity and health. In addition to diet, host genetic factors, and microbial inocula received from the mother and from the environment have a role in determining the composition of the microbiota for each individual (Zoetendal et al. 1998).

Table 1.   Abundance of main colonic bacterial groups in adult human faecal samples and fermentation profiles
Bacterial groupAbundance*Main acidic fermentation products from hexoses
  1. *Percent of total bacteria (mean values of several volunteers) based on fluorescent in situ hybridization (FISH) data from several publications as reported by Flint (2006).

  2. Some groups are heterogenous and/or under-researched; therefore reported acids are just indicative of cultured representatives.

Low G+C-content gram positives
 Clostridial clusters XIVa+b10·8–29 
  Roseburia/Eubacterium rectale group 2·3–8·8Butyrate, formate, l-lactate
  Eubacterium hallii 0·6–3·8Butyrate, formate, acetate
  Ruminococcus obeum 2·5Acetate
  Lachnospira spp. 3·6Formate, acetate, lactate, succinate
 Clostridial cluster IV25·2 
  Faecalibacterium prausnitzii 3·8–15·4Butyrate, formate, d-lactate
  Ruminococcusbromii, Ruminococcus flavefaciens 1·7–10·3Acetate, formate, lactate, succinate
 Clostridial cluster IX  7·1Propionate, various minor acids
 Clostridial cluster XVI
  Eubacterium cylindroides 0·4–1·4Butyrate, acetate, lactate, succinate, formate
  Lactobacillus/Enterococcus 0·01–1·8Lactate, acetate
High G+C-content gram positives
 Bifidobacterium spp. 2·5–4·9l-Lactate, acetate, formate
 Atopobium spp. 2·1–11·9Acetate, formate, lactate
Gram negatives
 Bacteroides-Prevotella group 8·5–27·7Acetate, propionate, succinate
 Proteobacteria 0·1–0·2Lactate, acetate, succinate, formate

Microbial breakdown of carbohydrate substrates

Members of the gut microbiota are in competition for most types of ND plant polysaccharides and oligosaccharides present in the colon. Analyses of the genome sequences for some commensal gut bacteria illustrate their dependence on complex carbohydrates for growth. Both Bifidobacterium longum (Schell et al. 2002) and Bacteroides thetaiotaomicron (Xu et al. 2003) dedicate at least 8% of their genomes to carbohydrate transport and metabolism functions. In fact, B. longum, with a genome size of only 2·26 Mb, encodes over twice as many, and B. thetaiotaomicron (genome size 6·26 Mb) over nine times as many such enzymes compared with Escherichia coli (genome size 4·64 Mb) (Xu et al. 2003). Complete genome sequences are unfortunately not yet available for the specialized bacteria that degrade plant cell wall polymers.

Plant cell wall polysaccharides

Insoluble plant cell wall material reaching the human colon consists of cellulose, arabinoxylan, xyloglucan, β-glucan, mannan, pectins and lignin. These complex polymers are intimately associated in the plant cell wall, and are degraded by a battery of microbial hydrolases, esterases and lyases. Microbial degradation in the large intestine, particularly of lignin and cellulose, is incomplete resulting in particles of plant fibre that persist through to the distal bowel. To date, the systems responsible for plant cell wall degradation by gut bacteria have been investigated in detail only for a few rumen anaerobes. The cellulolytic species Ruminococcus flavefaciens exhibits a cell surface enzyme complex of the cellulosome type (Rincon et al. 2005; Flint and Rincon 2006). Ruminococcal enzymes associated with cellulosomes have characteristic dockerin sequences that interact specifically with cohesins present in structural components; the resulting cellulosome complex is anchored to the bacterial cell wall, while a variety of substrate-binding modules and proteins mediate adhesion to the substrate. Ruminococcus species have also been isolated from human faeces (Robert and Bernalier-Donadille 2003) and R. flavefaciens-related 16S rRNA sequences can be demonstrated among bacteria that are tightly adherent to fibrous material recovered from human faecal samples (Walker 2006).

Hemicelluloses and pectins are generally subject to more complete breakdown by bacteria in the large intestine. Although initially present in cell wall matrices, these will be partially released in soluble form by the activities of primary cell wall-degrading bacteria, and become available to other gut bacteria. Such soluble oligosaccharides, including xylo-oligosaccharides, galacto-oligosaccharides and manno-oligosaccharides, which can be readily produced commercially, form the basis of many potential prebiotic preparations. Only certain species of bacteria from the human colon are known to utilize xylans, heterogeneous polymers of β-1,4-linked xylose residues substituted with acetyl, arabinosyl and glucuronic acid residues. These include Bacteroides ovatus, which possesses a xylan utilization operon similar to the one found in the rumen relative Prevotella bryantii (Weaver et al. 1992; Miyazaki et al. 2003) but that is not found in B. thetaiotaomicron (Xu et al. 2003). Among gram-positive species, Roseburia intestinalis is known to utilize xylans (Chassard and Bernalier-Donadille 2006), while human gut relatives of R. flavefaciens are also likely to degrade and utilize these molecules.

Pectin comprises long linear chains of α-1,4-linked galacturonic acid residues (smooth pectin) that may be partially esterified with methanol, or may be branched forming rhamnogalacturonans (hairy pectin). The side chains can contain residues such as d-xylose, l-fucose, d-glucuronic acid, and others. In common with other complex dietary substrates, the rate of bacterial fermentation is dependent on the degree of methylation and branching of the basic polymer. Pectin was completely degraded during 24-h batch culture incubations with a total human faecal inoculum (Dongowski et al. 2000).


Another important source of fermentable substrate to reach the colon is resistant starch, which is that fraction of starch not degraded by host amylases. Dietary starch comprises a mixture of amylose (α-1,4-linked glucose residues) and amylopectin, a branched polymer composed of amylose chains linked to an amylose backbone by α-1,6-linkages. Starches with a higher amylopectin content generally are more readily degraded. Starch-degrading enzymes are grouped into different families depending on their structure (MacGregor et al. 2001, http://afmb.cnrs-mrs.fr/CAZY/) and include α-amylases that hydrolyse α-1,4- linkages, type I pullulanases that specifically cleave α-1,6- bonds and amylopullulanases that possess both α-1,4- and α-1,6- activities (MacGregor et al. 2001).

The starch degrading system of the gram-negative bacterium B. thetaiotaomicron has been studied in some detail. The genome encodes multiple gene products that are involved in attaching to, hydrolysing and translocating starch molecules (Xu et al. 2003). Starch molecules are first bound tightly to the cell surface by several outer membrane proteins (Cho and Salyers 2001). While most of the hydrolysis occurs in the periplasm, one of the outer membrane proteins, SusG, is able to degrade starch (Shipman et al. 1999), presumably releasing smaller breakdown products that can be transported into the periplasm for complete hydrolysis. By binding the initial starch substrate tightly prior to degradation, B. thetaiotaomicron ensures that it becomes the beneficiary of its own hydrolytic activity.

Certain high G+C gram-positive Bifidobacterium sp. are also actively amylolytic and adhere to starch granules, via an interaction with a specific cell surface protein (Crittenden et al. 2001). Adhesion was not a prerequisite for starch utilization, although all highly adherent strains were starch utilizers. It is also likely that low % G+C gram-positive anaerobes contribute significantly to starch breakdown in the colon, but their contribution has received little study. Two representatives of this group from the human colon, Butyrivibrio fibrisolvens 16/4 and Roseburia inulinivorans A2-194, encode very large α-amylase enzymes of 1333 and 1674 amino acids, respectively (Ramsay et al. 2006). These multidomain enzymes have family 13 α-amylase catalytic domains but additionally carry putative carbohydrate binding domains and C-terminal cell wall anchoring domains, similar to the sortase mediated cell-wall attachment systems described in other gram-positive bacteria (Comfort and Clubb 2004). While an N-terminal signal peptide dictates export of the enzyme, the C-terminal region ensures its anchorage to the cell surface (Ramsay et al. 2006); again this positioning is presumed to favour the uptake of the hydrolysis products by the cell.

Inulin and fructo-oligosaccharides

Inulin and fructo-oligosaccharides (FOS) are linear polymers [degree of polymerization (DP), 3–60 for inulin, up to 3–9 for FOS] of β-2,1-linked fructose monomers with a terminal glucose residue. Both inulin and FOS have been shown to selectively stimulate the growth of Bifidobacteria and Lactobacilli (reviewed in Van Loo 2004), and are consequently classed as prebiotics (Gibson and Roberfroid 1995). Many of the original studies investigating the effects of inulin and FOS have focussed only on the response of targeted bacterial groups, specifically Bifidobacteria and Lactobacilli. It is clear however that other groups of bacteria are also stimulated. One such bacterium is likely to be R. inulinivorans, a butyrate-producing inulin degrader belonging to clostridial cluster XIVa that can account for up to 2% of faecal bacteria (Eckburg et al. 2005). Inulin enhanced the survival of one R. inulinivorans strain against a background of total faecal bacteria following the introduction into a fermentor system designed to simulate the proximal colon (Duncan et al. 2003). Manderson et al. (2005) reported an increase in the numbers of clostridial cluster XIVa bacteria upon in vitro fermentation of both pectic oligosaccharides and fructo-oligosaccharides, determined by FISH. Increased production of butyric acid from FOS noted in other studies (Manderson et al. 2005; Rossi et al. 2005) may therefore be attributed in part to direct stimulation of butyrate-producing species.

Kleessen et al. (2001) investigated changes in bacterial species in human flora associated rats fed on diets containing various mixtures of short- and long-chain fructose polymers. Bacteria were enumerated using FISH with group-specific probes. They showed that a mix of FOS and long-chain inulin or inulin alone enhanced the numbers of the clostridial cluster XIVa group, which was unaffected by FOS alone. In contrast, the numbers of Bifidobacteria were stimulated by FOS but not by inulin (Kleessen et al. 2001). It seems likely that the DP of these substrates is critical in determining their impact on the microbial community, and hence their consequences for health. It has also been reported that FOS increased the colonization and translocation of Salmonella in an animal model (Ten Bruggencate et al. 2004) and in human volunteers on a low calcium diet led to increased mucin excretion, which might indicate mucosal irritation (Ten Bruggencate et al. 2006). This was not observed, however, in volunteers on a regular diet (Scholtens et al. 2006).

Different bacterial β-fructofuranosidases vary in their ability to cleave β-2,1 bonds in sucrose, FOS and inulin, with some enzymes capable of hydrolysing FOS more rapidly (Warchol et al. 2002). A single β-fructofuranosidase is encoded by the genome of B. longum (Schell et al. 2002). More recently, a β-fructofuranosidase with greatest activity against β-2,1 glucose–fructose links was sequenced in Bifidobacterium breve (Ryan et al. 2005), while the β-fructofuranosidase isolated and sequenced from Bifidobacterium lactis was most active against β,2-1 fructose–fructose links in a study by Janer et al. (2004). The two Bifidobacterial β-fructofuranosidases had only low activities against long-chain inulin molecules (Janer et al. 2004; Ryan et al. 2005). A recent study investigating the growth of 11 species of Bifidobacteria (55 strains) found that only eight strains, belonging to five different species, were able to grow on inulin, while they all grew well on FOS (Rossi et al. 2005). The same authors also illustrated the bifidogenic effect of inulin in batch cultures containing faecal slurries, but theorized that this effect was the result of cross-feeding by the Bifidobacteria on the monosaccharide products released by the primary degraders (Rossi et al. 2005).

Simple sugars and sugar alcohols

Unabsorbed sugars and sugar derivatives also provide substrates for the colonic microbiota. Vogt et al. (2004) investigated the effect of the synthetic disaccharide lactulose and the sugar l-rhamnose on acetate and propionate serum levels in healthy volunteers. Lactulose ingestion increased serum acetate and l-rhamnose ingestion increased serum propionate, suggesting a modulation of the gut microbiota. A similar effect was also found in vitro, with a concomitant increase in butyrate in the case of lactulose fermentation (Fernandes et al. 2000). The consumption of unphysiologically high amounts of unabsorbable mono- and oligosaccharides, while showing potential to modulate the gut microbiota, has also been reported to lead to undesirable symptoms, such as bloating and abdominal pain in some volunteers (Vogt et al. 2004; Scholtens et al. 2006; Ten Bruggencate et al. 2006).

Formation of butyrate and propionate in the human colon

The total concentration of short-chain fatty acids (SCFA) in human faecal samples, which are largely derived from microbial fermentation, may exceed 100 mmol l−1, and the three major short-chain fatty acids detected are acetate, propionate and butyrate (Cummings et al. 1987). The molar proportion of these products is likely to be dependent on a number of factors, including diet. The fermentation pathways employed by the gut microbiota to generate fermentation products have been extensively reviewed (Marfarlane and Gibson 1997; Bernalier et al. 1999). Acetate makes up around 60–75% of the total SFCA detected in faeces and is formed by many of the bacterial groups that inhabit the colon, with approximately one-third coming from reductive acetogenesis (Miller and Wolin 1996). Substantial utilization of acetate also occurs during butyrate formation (discussed later). The bacterial groups that form propionate and butyrate are more restricted (Table 1), and are of particular interest in view of their potentially beneficial effects on health. Propionate is largely metabolized in the liver, is gluconeogenic, and may inhibit lipogenesis (Vogt et al. 2004). Butyrate, on the other hand, is the major energy source for the colonocytes, and has been implicated in the prevention of colitis and colorectal cancer (Pryde et al. 2002).

Propionate-forming bacteria

Several different pathways for propionate formation are known (Marfarlane and Gibson 1997; Bernalier et al. 1999) (Fig. 2). Bacteroides species generally employ the succinate route for propionate formation, while the acrylate route from lactate is found in bacteria belonging to the clostridial cluster IX group. In addition, a third pathway employed by the butyrate-producing bacterium R. inulinivorans with fucose as substrate has recently been described (Scott et al. 2006) (Fig. 2). While it has been reported that the acrylate pathway is unlikely to play an important role in propionate formation in the human gut (Marfarlane and Gibson 1997), the picture of the main propionate-producing bacteria in the human colon is still emerging. A recent in vitro study investigating the conversion of isotopically labelled lactate by human faecal inocula found that the main metabolic pathways utilized were significantly different between the three volunteers studied, with the acrylate pathway being utilized by the microbiota of one of the three volunteers (Bourriaud et al. 2005). Thus interindividual variation with respect to the predominant metabolic activities seems to be high, which is in accordance with phylogenetic studies investigating the interindividual variability of the gut microbiota (Zoetendal et al. 1998; Lay et al. 2005). The bacterial species that form propionate in other gut ecosystems, such as the rumen, include Megasphaera and Veillonella species which belong to clostridial cluster IX. In humans, FISH analysis has indicated that this group is numerically significant, making up around 7–9% of the total faecal microbiota (Flint 2006), but the species composition of this cluster in humans remains poorly understood.

Figure 2.

 Alternative fermentation pathways leading to butyrate (1) and propionate (2–4) formation. 1a, butyrate kinase pathway; 1b, butyryl-CoA CoA-transferase pathway; 2, acrylate pathway; 3, succinate pathway; 4, propanediol pathway. Dotted arrows indicate several intermediates. DHAP, dihydroxyacetone phosphate; P, phosphate; PEP, phophoenolpyruvate.

Butyrate-forming bacteria

Colonic bacteria that produce butyrate belong to the clostridial clusters I, III, IV, XI, XIVa, XV and XVI. Two particularly abundant groups that are together estimated to constitute 7–24% of the total gut bacteria in healthy subjects are cluster IV bacteria related to Faecalibacterium prausnitzii, and cluster XIVa bacteria related to Eubacterium rectale and to Roseburia spp. (Barcenilla et al. 2000; Hold et al. 2003; Aminov et al. 2006).

The fermentation pathway for the formation of butyrate (Fig. 2) consists of a central pathway analogous to the reactions of fatty acid β-oxidation in reverse sequence (Bernalier et al. 1999) and two alternative pathways for the formation of butyrate from butyryl-CoA. The pathway employing phosphotransbutyrylase and butyrate kinase (Fig. 2, 1a) appears to be widespread among different butyrate-producing Clostridium species, as homologues to the respective genes are present in Clostridium acetobutylicum, Clostridium perfringens, Clostridium tetani, Clostridium botulinum and Clostridium difficile (ClostriDB: http://clostri.bham.ac.uk/) and butyrate kinase activity was found in several clostridia (e.g. Zhu et al. 2005). The second pathway employing a CoA-transferase that transfers the CoA-moiety from butyryl-CoA onto acetate (Fig. 2, 1b) has been described in several bacteria, for example, in the isolates from the rumen (Diez-Gonzalez et al. 1999) and human gut (Duncan et al. 2002). The description of this route was based on enzymatic studies, as no corresponding CoA-transferase gene had been assigned to this activity. A screen of 38 butyrate-producing bacteria from the human colon for both pathways revealed that the CoA-transferase route was the predominant route, whereas butyrate kinase was only found in a few isolates (Louis et al. 2004). Although a recent metagenome study based on faecal samples from two subjects reported butyrate kinase as a highly enriched sequence (Gill et al. 2006), we were only able to confirm one butyrate kinase sequence among these sequences. A novel CoA-transferase gene encoding butyryl-CoA CoA-transferase was recently identified from the human isolate Roseburia hominis A2-183 (Charrier et al. 2006).

Many Clostridium sp. that possess butyrate kinase have been reported from soils and from aquatic environments. This suggests that the CoA-transferase route of butyrate formation might be the preferred route mainly in bacteria adapted to thrive within gut ecosystems. Acetate is a cosubstrate of this reaction, and it has been shown that external acetate stimulates growth in bacteria possessing the CoA-transferase (Duncan et al. 2002). Therefore, the high acetate concentrations normally present in the large intestine (Cummings et al. 1987) might promote CoA-transferase carrying strains. This route of butyrate formation might even help to detoxify excess acetate entering the bacterial cell under the slightly acidic conditions often prevailing in the proximal colon (Nugent et al. 2001). In bacteria that are adapted to thrive outside the gut environment, however, possessing the butyrate kinase route, which does not depend on high concentrations of acetate, might enable those strains to adapt to a wider range of environmental conditions. It remains to be seen whether the prevalence of the alternative routes for butyrate formation in different environments will be confirmed by further studies.

Metabolic cross-feeding

Metabolic cross-feeding between bacteria is likely to play an important role in substrate breakdown in this densely populated ecosystem. A number of bacterial species, including Eubacterium hallii and Anaerostipes caccae, have been isolated from human faecal samples that can convert lactate and acetate into butyrate (Duncan et al. 2004). The presence of these bacteria may help to explain the fact that lactate does not accumulate in the colon of healthy individuals, although a significant proportion of colonic bacteria can produce this acid. Propionate-producing bacteria belonging to clostridial cluster IX may also contribute to lactate utilizing activity in the large intestine, but in addition, lactate is a favoured cosubstrate for sulphate-reducing bacteria (Gibson et al. 1990). Lactate can reach very high concentrations in the colon of ulcerative colitis (UC) sufferers (Vernia et al. 1988), possibly indicating a failure in the normal utilization of lactate under these conditions.

It has recently been shown in vitro that lactate was converted mainly into butyrate by faecal inocula of two volunteers, although propionate was a major product in a third volunteer (Bourriaud et al. 2005). Carbon flow from lactate to butyrate has been confirmed by stable isotope labelling in cocultures of Bifidobacterium adolescentis and butyrate producers grown on either starch or FOS, which cannot be utilized directly by the butyrate producers (Belenguer et al. 2006). A second mechanism of cross-feeding, likely to be the result of the liberation of carbohydrate breakdown products, was also observed in the cocultures of B. adolescentis and R. hominis, a butyrate producer which is unable to utilize lactate (Belenguer et al. 2006). An investigation into the breakdown of FOS and inulin by the isolates of the genus Bifidobacterium and by faecal microbiota also points towards primary degraders of carbohydrates that produce extracellular hydrolytic enzymes and thus supply nutrients to scavengers of partially hydrolysed substrates (Rossi et al. 2005).

Metabolic regulation

Bacteroides thetaiotaomicron and R. inulinivorans are both able to grow on l-fucose (Hooper et al. 1999; Scott et al. 2006). The initial uptake and conversion of l-fucose to dihydroxyacetone phosphate (then channelled into the glycolytic pathway) and l-lactaldehyde involves similar enzymes in these two bacteria. The l-lactaldehyde is then reduced to propane-1,2-diol, which is thought to be excreted from B. thetaiotaomicron. Roseburia inulinivorans, however, employs the propanediol utilization pathway (Fig. 2), previously identified in enteric pathogens (Bobik et al. 1999), to ferment the propanediol into propionate and propanol (Scott et al. 2006). Fucose utilization and propandiol utilization genes are interspersed in an operon in R. inulinivorans, illustrating the codependence of the two pathways. In addition to being a component of pectin, fucose is one of the terminally linked sugars in glycoconjugates found on the gut epithelial surface (Finne et al. 1989); hence, the ability to use this sugar gives the bacterium an alternative potential energy source that may provide an important competitive advantage during times of dietary starvation. Roseburia inulinivorans was originally isolated as a producer of butyrate from starch or glucose (Barcenilla et al. 2000). The fucose utilization genes in R. inulinivorans are strongly induced following transfer from glucose to fucose as energy source, resulting in a dramatic metabolic switch over to propionate formation (Scott et al. 2006). This provides just one example out of many of the key role of substrate-driven metabolic regulation upon fermentation product formation.

Impact of diet on butyrate and propionate formation

The dietary intake of fermentable carbohydrates is known from many studies to influence short-chain fatty acid synthesis in the large intestine (reviewed in Cummings and Macfarlane 1991; Topping and Clifton 2001). Resistant starch has consistently been shown to have a butyrogenic effect on the colonic microbiota, both in vitro (Wang and Gibson 1993) and in vivo (Topping and Clifton 2001). The addition of starch to the diet of human flora-associated rats led to an increase in the numbers of lactobacilli and bifidobacteria, and to elevated production of butyrate (Silvi et al. 1999).

Duncan et al. (2007) conducted a study in which obese human subjects each consumed diets with normal, reduced or dramatically reduced carbohydrate content. Faecal concentrations of all three major short-chain fatty acids decreased with reduced total carbohydrate intake, but in particular, the concentration of butyrate decreased from 17·7 to 4·4 mmol l−1. In parallel, the major bacterial populations were monitored using FISH. While the overall bacterial numbers and the proportion of the dominant cluster XIVa and CFB groups changed little, the proportion of the cluster XIVa subgroup Roseburia/E. rectale decreased on average from 11·4% to 3·3% of total bacteria detected between the highest and lowest carbohydrate intakes. This tends to confirm a dominant role for this group of bacteria in the formation of butyrate in the colon, while also indicating that the Roseburia/E. rectale group is particularly dependent on residual dietary carbohydrate to maintain its competitiveness in the colon. Schwiertz et al. (2002) also observed increased butyrate during a human intervention study with resistant starch, but the bacterial species monitored did not include the Roseburia group.

Butyrate and propionate are both relatively reduced products. Their formation may thus be influenced by the activities of hydrogen-consuming bacteria, by the redox state of substrates, and by the need for the disposal of reducing equivalents resulting from substrate utilization (Macfarlane and Gibson 1997). The formation of butyrate in the human large intestine may be affected in several different ways by diet composition (Pryde et al. 2002). The evidence discussed already suggests that certain dietary carbohydrates will directly stimulate the growth of particular butyrate producing bacteria, such as E. rectale and Roseburia species, including R. inulinivorans and R. intestinalis. In addition, the stimulation of non-butyrate producing bacteria may have an indirect effect on butyrate production via metabolic cross-feeding (Belenguer et al. 2006), as discussed earlier. This may help to explain how the stimulation of lactic acid-producing bacteria, such as Bifidobacteria, by prebiotics, such as FOS, is apparently often accompanied by increases in butyrate formation.

A well-recognised general effect of increased consumption of ND fermentable carbohydrate is to lower the pH of the proximal colon as a result of increased concentrations of weak acids (SCFA). The extent of pH reduction is likely to depend on buffering from secreted bicarbonate and the equilibrium between SCFA formation and absorption. Thus, the pH of the proximal colon can become mildly acidic, increasing towards neutral pH in the distal region (Nugent et al. 2001). The pH in the colon can have a marked effect on the composition of the colonic microbiota. Walker et al. (2005) reported in fermentor systems that switching the pH from 5·5 to 6·5 resulted in a much less butyrogenic but more propiogenic fermentation. This was shown to correlate with a shift in the composition of the microbiota from pH 5·5, where the butyrate-forming Roseburia and E. rectale group comprised 20% of total bacteria, to pH 6·5, where this group became undetectable. At pH 6·5, the fermentor community became dominated by Bacteroides, indicating that Bacteroides species were able to outcompete most other bacteria for the soluble carbohydrates supplied at pH 6·5, whereas at the lower pH, other bacterial groups were able to compete for these substrates (Flint 2006). The inhibition of another group of gram negatives, the enterobacteria, at acidic pH is already recognized as an important factor tending to limit the populations of certain pathogens in the gut.

Under circumstances where the pH decreases to values below pH 5·5, however, the composition of the microbiota is likely to change again along with the overall fermentation processes and could potentially result in acidosis with elevated lactate in the colon. High lactate concentrations have been reported in faecal samples from UC sufferers (Vernia et al. 1988). The effects of diet upon colonic pH may therefore play a critical role in mediating the influence of diet upon gut health.


Molecular methods for the quantification of bacteria have led to a major revision of the description of the human gut microbiota in recent years. We now have a much clearer view of the types of bacteria present in the large intestine, but many species remain uncultured and some numerically less prevalent ones remain undetected with the techniques currently used. This microbial ecosystem is highly complex, being influenced by numerous factors as outlined in this review, and we are only starting to get an insight into the effect of diet on the composition and activity – and in turn, the effect on human health – of the gut microbiota. To further our understanding, a combination of multiple approaches ranging from the investigation of pure cultures and in vitro mixed systems to animal models and human intervention studies should be employed. While in vitro incubations are valuable to understand the physiology of single bacteria and identify metabolic links between different types of bacteria, they are likely to amplify population changes compared with the in vivo situation, which is affected by frequent fluctuations. Here, adaptations to varying substrates and environmental conditions might result in more prominent changes of activity rather than bacterial populations, and molecular techniques targeting RNA rather than DNA (e.g. Tannock et al. 2004) are now emerging to address this.


The Rowett Research Institute receives funding from the Scottish Executive Environment and Rural Affairs Department.