Prebiotic stimulation of human colonic butyrate-producing bacteria and bifidobacteria, in vitro

Authors

  • Karen P. Scott,

    Corresponding author
    1. Microbial Ecology Group, Rowett Institute of Nutrition and Health, University of Aberdeen, Bucksburn, Aberdeen, UK
    • Correspondence: Karen P. Scott, Microbial Ecology Group, Rowett Institute of Nutrition and Health, University of Aberdeen, Greenburn Road, Bucksburn, Aberdeen, AB21 9SB, UK. Tel.: 01224 438730; fax: 01224 438699; e-mail: k.scott@abdn.ac.uk

    Search for more papers by this author
  • Jennifer C. Martin,

    1. Microbial Ecology Group, Rowett Institute of Nutrition and Health, University of Aberdeen, Bucksburn, Aberdeen, UK
    Search for more papers by this author
  • Sylvia H. Duncan,

    1. Microbial Ecology Group, Rowett Institute of Nutrition and Health, University of Aberdeen, Bucksburn, Aberdeen, UK
    Search for more papers by this author
  • Harry J. Flint

    1. Microbial Ecology Group, Rowett Institute of Nutrition and Health, University of Aberdeen, Bucksburn, Aberdeen, UK
    Search for more papers by this author

Abstract

Dietary macronutrients affect the composition of the gut microbiota, and prebiotics are used to improve and maintain a healthy gut. The impact of prebiotics on dominant gut bacteria other than bifidobacteria, however, is under-researched. Here, we report carbohydrate utilisation patterns for representative butyrate-producing anaerobes, belonging to the Gram-positive Firmicutes families Lachnospiraceae and Ruminococcaceae, by comparison with selected Bacteroides and Bifidobacterium species. Growth assessments using anaerobic Hungate tubes and a new rapid microtitre plate assay were generally in good agreement. The Bacteroides strains tested showed some growth on basal medium with no added carbohydrates, utilising peptides in the growth medium. The butyrate-producing strains exhibited different growth profiles on the substrates, which included starch, inulin, fructooligosaccharides (FOS), galactooligosaccharides (GOS) and xylooligosaccharides (XOS). Eleven were able to grow on short-chain FOS, but this number decreased as the chain length of the fructan substrates increased. Long-chain inulin was utilised by Roseburia inulinivorans, but by none of the Bifidobacterium species examined here. XOS was a more selective growth substrate than FOS, with only six of the 11 Firmicutes strains able to use XOS for growth. These results illustrate the selectivity of different prebiotics and help to explain why some are butyrogenic.

Introduction

The human large intestine is inhabited by a highly complex bacterial community dominated by hundreds of different species of obligate anaerobes (Flint et al., 2007). Culture-independent 16S rRNA gene analysis has indicated that the two most abundant bacterial phyla in adults are the Bacteroidetes (normally between 10% and 50%) and the Firmicutes (up to around 75%) (Eckburg et al., 2005; Duncan et al., 2008; Tap et al., 2009; Qin et al., 2010; Walker et al., 2011). The dominant species of Firmicutes mainly belong to the families Lachnospiraceae and Ruminococcaceae (Duncan et al., 2007a). Members of the phylum Actinobacteria, especially Bifidobacterium spp., can also be abundant in the adult human colon (normally up to 10%) (Duncan et al., 2007a; Walker et al., 2011), but have often been underestimated by 16S rRNA gene sequence analysis if the correct primers are not used (Sim et al., 2012). Although at least two-thirds of bacterial phylotypes detected in the human intestinal microbiota by molecular approaches are not represented in culture collections, a recent study estimated that among the most abundant phylotypes, each representing more than 0.5% of the total microbiota, more than 60% were represented by cultured species (Walker et al., 2011; Flint et al., 2012ab). These cultured bacterial representatives are an extremely valuable resource and can be used to test, for example, substrate utilisation and cross-feeding interactions. Moreover, complete genome sequences of cultured bacteria provide a framework for metagenomic studies (Qin et al., 2010; Lepage et al., 2013).

There is considerable interest in using dietary approaches, including prebiotics, to modulate the composition of the gut microbiota. The most widely used prebiotics are fructans, both short-chain fructooligosaccharides [scFOS, degrees of polymerisation (DP) 2 – 9; β,1-2-linked fructose residues] and long-chain (lc) inulin (DP between 10 and 60), galactooligosaccharides (GOS) and xylooligosaccharides (XOS). Most of these prebiotics have been reported to increase numbers of bifidobacteria detected in faeces (Roberfroid et al., 1998; Macfarlane et al., 2008; Ramirez-Farias et al., 2009; Lecerf et al., 2012), and many previous studies have focused on changes in bifidobacteria abundance as a measure of prebiotic effects. Interestingly, lc-inulin has also been shown to enhance the production of butyrate by the faecal microbiota (Kleessen et al., 2001), and butyrate has the potential to benefit colonic health (Pryde et al., 2002; Hamer et al., 2008). Bifidobacteria, however, do not produce butyrate, leaving the mechanism of enhanced butyrate production unclear, although bacterial cross-feeding is likely to play a role (Belenguer et al., 2006). However, mixed lc-inulin/scFOS and GOS supplementation have been shown to increase the abundance of Faecalibacterium prausnitzii in human volunteers (Ramirez-Farias et al., 2009; Davis et al., 2010; Dewulf et al., 2013), suggesting that certain butyrate-producing Firmicutes species might be directly stimulated by these prebiotics.

In the current study, our aim was to investigate the ability of a range of dominant human colonic butyrate producers and Bifidobacterium species, to utilise selected nondigestible dietary carbohydrates, including prebiotics, for growth. The most abundant butyrate-producing species detected in recent 16S rRNA gene studies of adults are Faecalibacterium prausnitzii, Eubacterium rectale, Eubacterium hallii and Anaerostipes hadrus (Tap et al., 2009; Walker et al., 2011). These species were included, together with additional species of human colonic butyrate producers (Duncan et al., 2004a, 2006; Louis & Flint, 2009).

Materials and methods

Substrates

The substrates studied included fructans of increasing chain length ranging from P95 short-chain (sc) FOS (DP 2-8), Synergy1 (50 : 50 mixture of P95 scFOS and HP) and HP inulin (degree of polymerisation (DP) > 5; all gifted by BENEO-Orafti, Belgium); dahlia long-chain (lc) inulin (DP c. 25, purified from Dahlia tubers at RINH, A. Gordon); xylooligosaccharides (XOS; Suntory Ltd, Tokyo, Japan); and Vivinal GOS (FrieslandCampina Domo, the Netherlands) which has a dry matter content of 75% consisting of 59% GOS, 21% lactose, 19% glucose and 1% galactose. Amylopectin potato starch was purchased from Sigma-Aldrich (catalogue number A8515).

Bacterial strains and growth conditions

Bacterial strains used are listed in Table 1. Routine culturing of bacterial strains was in M2GSC medium (Miyazaki et al., 1997), using the Hungate tube method and maintaining the media under CO2. Growth analyses on single carbon sources were performed in 9.5 mL basal YCFA medium (Lopez-Siles et al., 2012) supplemented with 0.5% w/v of the specific substrate indicated. These tubes were inoculated in triplicate with 0.2 mL of a 16- to 18-h M2GSC culture and incubated anaerobically at 37 °C. Growth was determined by following changes in optical density at 650 nm (OD650 nm), taking hourly measurements. Maximum specific growth rates (μmax) were determined during exponential growth (Pirt, 1975). Growth was also monitored in microtitre plates, setting the plates up within an anaerobic cabinet and maintaining anaerobic conditions throughout. In this case, readings were obtained from six replicate samples. Overnight cultures (10 or 20 μL) were added to prereduced YCFA supplemented with the appropriate substrate (final volume 200 μL) in flat-bottomed 96-well microtitre plates (Corning, Sigma-Aldrich). Sample blanks containing uninoculated medium were used as controls. After inoculation, microtitre plates were sealed with PCR film to prevent evaporation and to maintain the anaerobic atmosphere (Cernat & Scott, 2012). Cells were incubated for 23 h at 37 °C in a Tecan Safire 2 microplate reader (Tecan Group Ltd), with optical density readings at 650 nm taken automatically every hour using the Magellan software (Tecan Group Ltd), with low speed shaking for 5 s prior to each reading. Maintenance of anaerobic conditions was verified by performing control growth experiments using either glucose or maltose as an energy source and by including the anaerobic indicator dye resazurin in the growth media.

Table 1. Bacterial strains and characteristics
Bacterial phylumRepresentative strainReference
  1. a

    Bifidobacterium animalis was isolated from a dairy product by K. Scott.

  2. b

    Alternatively ATCC Bacteroides thetaiotaomicron VPI 5482.

Firmicutes, cluster IVFaecalibacterium prausnitzii A2-165Barcenilla et al. (2000)
Firmicutes, cluster XIVaRoseburia intestinalis L1-82Barcenilla et al. (2000)
Roseburia inulinivorans A2-194Barcenilla et al. (2000)
Roseburia faecis M72/1Louis et al. (2004)
Roseburia hominis A2-183Barcenilla et al. (2000)
Eubacterium rectale A1-86Barcenilla et al. (2000)
Eubacterium hallii L2-7Barcenilla et al. (2000)
Anaerostipes caccae L1-92Barcenilla et al. (2000)
Coprococcus comes A2-232Barcenilla et al. (2000)
Coprococcus eutactus L2-50Barcenilla et al. (2000)
Anaerostipes hadrus SS2/1Allen-Vercoe et al. (2012)
Actinomycetes Bifidobacterium adolescentis 20083DSM culture collection
Bifidobacterium breve 20213DSM culture collection
Bifidobacterium longum 20219DSM culture collection
Bifidobacterium longum 8809NCIMB culture collection
Bifidobacterium animalis KS1This studya
Bifidobacterium infantis 20088DSM culture collection
Bifidobacterium bifidum 20456DSM culture collection
Bifidobacterium angulatum 20098DSM culture collection
Bifidobacterium pseudocatenulatum 20438DSM culture collection
Bacteroidetes Bacteroides vulgatus B1447Gift from A.A. Salyers, U. Illinois
Bacteroides thetaiotaomicron 2079DSM culture collectionb

Short-chain fatty acid production (SCFA)

SCFA production was assessed in culture supernatants (1 mL) after 24-hr growth at 37 °C using gas chromatography as described previously (Richardson et al., 1989). After derivatisation of duplicate samples, 1 µL was analysed using a Hewlett-Packard gas chromatograph fitted with a fused silica capillary column with helium as a carrier gas. The SCFA concentrations were calculated from the relative response factor with respect to the internal standard (2-ethylbutyrate).

Results

Comparison of growth of selected faecal anaerobes in tubes and microtitre plates

Eleven of the predominant Firmicute bacteria isolated from human faeces (Barcenilla et al., 2000) were compared with some of the most prevalent Bifidobacterium and Bacteroides species from the human colon for their ability to utilise a range of carbohydrate substrates for growth under anaerobic conditions. Butyrate-producing members of the families Ruminococcaceae (clostridial cluster IV) and Lachnospiraceae (cluster XIVa) were included in the study (Table 1). YCFA medium (Lopez-Siles et al., 2012) contains a mixture of short-chain fatty acids (SCFA) and supported good growth of all of the strains tested, in the presence of their preferred carbohydrate energy sources (glucose or maltose). Only B. vulgatus 1447 and B. thetaiotaomicron B5482 gave significant growth in the absence of any added carbohydrates due to their ability to use peptides present in the casitone and yeast extract components of the basal YCFA medium (Fig. 1). In the remaining data presented, all such basal growth has been subtracted from the OD650 nm values for the carbohydrate substrates.

Figure 1.

Comparison of the growth of representative strains on fructans (P95 scFOS, HP and dahlia inulin), amylopectin starch and XOS in Hungate tubes. Carbohydrate substrates were added at 0.5% and the OD650 measured hourly. Data plotted are the mean of triplicate maximum OD650 readings ± standard deviation. OD650 values attributable to growth on basal YCFA have been subtracted from the readings on the carbohydrate substrates. Bacterial strain codes are the following: B. theta – Bacteroides thetaiotaomicron; B. vulg – Bacteroides vulgatus; F. prau – Faecalibacterium prausnitzii; E. rect – Eubacterium rectale; R. homi – Roseburia hominis; R. inte – Roseburia intestinalis; R. faec – Roseburia faecis; R. inul – Roseburia inulinivorans; E. hall – Eubacterium hallii; A. cacc – Anaerostipes caccae; C. euta – Coprococcus eutactus; C. come – Coprococcus comes; B. long8809 – Bifidobacterium longum 8809; B. long20219 – Bifidobacterium longum 20219; B. infa – Bifidobacterium infantis.

The maximum OD650 nm achieved by the bacterial isolates during growth on selected substrates is shown in Fig. 1 (growth in tubes, detailed data in Supporting Information, Table S1) and Table 2 and Table S2 (growth in high-throughput microtitre plates). Generally, the results obtained using the two methods are comparable, although the maximum OD values attained in the microtitre plates tended to be lower, rarely exceeding an OD of 1.0 (Fig. 2). Bifidobacterium infantis 20088, for example, grew well on scFOS in tubes, but less well in the microtitre plate system. The reason for this has not been established.

Table 2. Growth of Bifidobacterium strains in microtitre platesa
Bacterial speciesStrainBYCFAStarchP95 scFOS DP 2-8Synergy1 DP > 2HP inulin DP > 5Dahlia inulin DP ~25GOSXOS
  1. ng – no growth (final OD after 24 h < 0.05); mean OD650 nm values of > 0.5 are shown in bold.

  2. a

    Data are average maximal OD650 nm readings from two independent experiments, with four to six replicates in each.

Bifidobacteria
B. adolescentis 200830.28 ± 0.030.17 ± 0.040.51 ± 0.130.53 ± 0.100.5 ± 0.110.15 ± 0.080.83 ± 0.020.40 ± 0.01
B. angulatum 200980.08 ± 0.07ng0.16 ± 0.050.14 ± 0.090.15 ± 0.050.1 ± 0.060.80 ± 0.060.17 ± 0.04
B. animalis KS10.19 ± 0.04ng0.37 ± 0.080.27 ±  0.20.3 ± 0.110.14 ± 0.05ngng
B. bifidum 204560.14 ± 0.14ng0.64 ± 0.020.41 ± 0.060.29 ± 0.010.06 ± 0.060.77 ± 0.02ng
B. breve 202130.1 ± 0.050.91 ± 0.080.73 ± 0.090.71 ± 0.050.42 ± 0.060.11 ± 0.040.78 ± 0.010.13 ± 0.04
B. infantis 200880.01 ± 0.01ng0.24 ± 0.030.35 ± 0.080.44 ± 0.03ng0.98 ± 0.050.15 ± 0.04
B. longum 202190.06 ± 0.050.12 ± 0.061.16 ± 0.080.69 ± 0.050.12 ± 0.03ng0.92 ± 0.01ng
B. pseudocatenulatum 204380.04 ± 0.01ngng0.05 ± 0.02ngng0.57 ± 0.060.05 ± 0.02
Figure 2.

Comparison of maximum OD650 achieved during bacterial growth on 0.5% scFOS or 0.5% dahlia inulin in Hungate tubes compared with microtitre plates. Data are the average of triplicate readings (tubes) or four to six replicates from two independent experiments (microtitre plates). Bacterial strain codes are the following: B. long – Bifidobacterium longum 20219; B. infa – Bifidobacterium infantis; F. prau – Faecalibacterium prausnitzii; E. rect – Eubacterium rectale; R. homi – Roseburia hominis; R. inte – Roseburia intestinalis; R. faec – Roseburia faecis; R. inul – Roseburia inulinivorans; E. hall – Eubacterium hallii; A. cacc – Anaerostipes caccae; C. euta – Coprococcus eutactus.

Virtually, all of the bacterial strains tested were able to grow well on scFOS (P95), indicating that it is not a particularly selective growth substrate (Fig. 1, Table S1). However, as the complexity of the fructan substrate increased from scFOS to long-chain inulin, fewer bacterial strains were able to grow to high ODs. Some of the bacteria were possibly able to utilise any short-chain FOS present for initial growth and were unable to grow further once these were depleted, resulting in lower maximum ODs on Synergy1 (a 1 : 1 mixture of P95 scFOS and HP inulin) compared with scFOS (Table S1, Table 2). Several Bifidobacterium isolates grew well on both scFOS and HP inulin, but generally, higher ODs were achieved on scFOS (Fig. 1, Table 2). Only four of the Firmicutes, and none of the bifidobacteria, were able to grow utilising dahlia inulin (DP~25; Fig. 1), with Roseburia inulinivorans A2-194 achieving the highest OD and the highest specific growth rate (Table S2, Table 3). Growth in microtitre plates of strain SS2/1 belonging to the newly proposed species Anaerostipes hadrus (Allen-Vercoe et al., 2012) on long-chain inulin substrates occurred only after a prolonged lag period of 10 h on HP inulin and 15 h on dahlia inulin (Table S2). This suggests that for this bacterium, a switch in gene expression is a prerequisite for growth, and also illustrates that the shorter chain fructans are more readily fermentable. The genes required for inulin utilisation by R. inulinivorans have been shown to be inducible (Scott et al., 2011).

Table 3. Maximum specific growth rates (h−1) of selected bacteria in Hungate tubes on starch and increasing chain lengths of fructan substrates (0.5%)
SubstrateStarchP 95 scFOSSynergy1HP inulinDahlia inulin
  1. Data are the average of triplicate results.

E. rectale A1-860.40 ± 0.050.34 ± 0.010.38 ± 0.020.36 ± 0.070.41 ± 0.07
R. inulinivorans A2-1940.40 ± 0.030.38 ± 0.060.40 ± 0.030.35 ± 0.060.54 ± 0.04
F. prausnitzii A2-165No growth0.28 ± 0.020.25 ± 0.030.21 ± 0.16No growth
E. hallii L2-7No growth0.54 ± 0.090.58 ± 0.040.46 ± 0.07No growth
B. longum 20219No growth0.34 ± 0.060.38 ± 0.080.11 ± 0.08No growth

Only six bacterial species grew well in tubes on amylopectin starch (achieving an OD of > 0.4, Fig. 1, Table S1). Four of these strains were Roseburia species, which corresponds to published data, indicating that the proportion of Roseburia sp. in the gut microbiota can increase as a consequence of carbohydrate, particularly starch, supplementation (Duncan et al., 2007b; Walker et al., 2011). Bifidobacterium breve 20213 was the only Bifidobacterium isolate tested able to utilise amylopectin starch for growth (Table 2). However, different Bifidobacterium species and strains, including B. adolescentis L2-32, have been shown to grow on starch (Duncan et al., 2004a; Belenguer et al., 2006).

Bacterial growth rates

Similar maximal specific growth rates (μmax h−1) were achieved on the scFOS (P95) and the other substrates (Table 3). Roseburia inulinivorans was the only Roseburia species tested to utilise all the substrates, and the growth rate on the HP inulin was similar to that on the shorter chain substrates and was actually higher on dahlia inulin, although the final OD was lower. Interestingly when the growth rates of cultures in tubes and microtitre plates were compared, similar specific growth rates were attained in the microtitre plates, even though the maximal OD values were lower (Fig. 2) and the exponential growth phases shorter.

The growth of the isolates was assessed on additional prebiotic sources, including XOS and GOS (microtitre plate method only; Table 2, Table S2). Most of the isolates tested grew efficiently on GOS in microtitre plates, including the bifidobacteria species (Fig. 3a and b). In fact, GOS was the only substrate to support good growth of B. pseudocatenulatum. The strains that grew more poorly (final OD650 nm < 0.3) may have been utilising the small amounts of glucose in the substrate for minimal growth. XOS was much more selective than FOS and GOS, only supporting reasonable growth (to OD650 nm > 0.3) of six isolates tested (Fig 3c and d).

Figure 3.

Growth of bacterial strains on 0.5% GOS (a and b) or XOS (c and d) in microtitre plates. Growth points are the average OD650 from four to six replicates in two experiments, with standard deviation bars included. Bacterial strain codes are the following: R. faec – Roseburia faecis; R. inte – Roseburia intestinalis; A. cacc – Anaerostipes caccae; R. inul – Roseburia inulinivorans; E. rect – Eubacterium rectale; E. hall – Eubacterium hallii; C. euta – Coprococcus eutactus; F. prau – Faecalibacterium prausnitzii; R. homi – Roseburia hominis; B. infa – Bifidobacterium infantis; B. long – Bifidobacterium longum 20219; B. thet – Bacteroides thetaiotaomicron; B. adol – Bifidobacterium adolescentis; B. brev – Bifidobacterium breve; B. bifi – Bifidobacterium bifidum; B. angu – Bifidobacterium angulatum.

Bacterial metabolism

The metabolic activities of the bacterial strains were compared by assessing the main SCFA fermentation products after 24 h. The fermentation products for a given bacterium were generally similar for amylopectin starch and fructan substrates (Fig. 4). The main fermentation product for most of the Firmicutes selected in this study was butyrate (10–20 mM), followed by formate (6–10 mM) with many of these bacteria consuming acetate during growth, as previously reported (Barcenilla et al., 2000). Both E. rectale and R. inulinivorans also produced detectable amounts of lactate (6 to < 20 mM) during growth in pure culture, particularly on substrates that supported good growth, and this contributed to the lower final pH in these cultures (Fig. 4). The main products for all the Bifidobacterium spp. were acetate (c. 20 mM) followed by lactate (10–15 mM), with formate (c. 3 mM) also produced by some species. In the gut ecosystem, both lactate and acetate are utilised by specialised groups of cross-feeding bacteria, often producing butyrate or propionate (Belenguer et al., 2006; Falony et al., 2009a).

Figure 4.

Short-chain fatty acid production by Eubacterium rectale and Roseburia inulinivorans following growth in Hungate tubes on selected substrates. SCFA concentrations and final culture pH were measured in triplicate tubes, after 24-h growth. Concentrations in the basal media were subtracted to give the final concentrations.

Discussion

Manipulation of the human colonic microbiota through diet to improve healthy gut function and prevent systemic disease has been a long-term goal, and indeed, it has been shown that dietary carbohydrate intake has a major impact on the composition of the gut microbiota and its metabolic output (Duncan et al., 2007b; Walker et al., 2011; Wu et al., 2011; Flint et al., 2012ab). In particular, there has been considerable interest in the concept of using fructans, especially scFOS, as prebiotics to promote the growth of specific beneficial groups of gut bacteria (Gibson et al., 2004) to enhance health. Thus far, many of these investigations have focused only on the stimulation of Bifidobacterium and Lactobacillus species by prebiotics. As many of the major groups of colonic bacteria, including those producing butyrate, were uncultured or poorly characterised until recently, it has only now become feasible to predict the exact consequences of a given prebiotic on the overall ecosystem of the gut. The data presented here indicate that distinct prebiotics have the potential to stimulate many different, abundant, Firmicute members of the gut microbiota, and it is probable that additional species (not included here) will also be able to use these substrates. scFOS in particular supported growth of all the bacterial species tested in this study. It is interesting that in spite of this, selective stimulation of bifidobacteria by FOS is regularly reported in vivo (Roberfroid et al., 1998; Ramirez-Farias et al., 2009). This presumably reflects the competitive success under prevailing conditions in the gut, which cannot be predicted from growth of pure cultures. Our data do, however, demonstrate a clear potential for multiple responses. It is also clear that the chain length of the fructan is crucial in determining its fermentability, with few bacteria able to utilise the long-chain fructans for growth. Even within the Bifidobacterium genus, it has been demonstrated that there are four subgroups of bifidobacteria with markedly different degradation abilities (Falony et al., 2009b). The differing chain length of fructans may also affect the site of degradation (Van De Wiele et al., 2007).

Many investigations have illustrated the beneficial role of butyrate in gut health (McIntyre et al., 1993; Archer et al., 1998; Inan et al., 2000; Lührs et al., 2001; Hamer et al., 2008); consequently, a substrate that promotes growth of butyrate-producing bacteria, as shown here, would fulfil the ‘prebiotic concept'. Moreover, dietary supplementation of a mixed scFOS/inulin prebiotic stimulated numbers of both the Bifidobacterium genus and F. prausnitzii (Ramirez-Farias et al., 2009; Dewulf et al., 2013). Such stimulation of additional bacterial species also helps to explain the ‘prebiotic conundrum', whereby prebiotics have been reported to enhance faecal butyrate levels and numbers of bifidobacteria (Kleessen et al., 2001), even though these bacteria do not produce butyrate. In a human-flora-associated (HFA) animal study, only longer-chain inulin and not scFOS resulted in increased butyrate detection, and this was accompanied by increases in the Clostridium coccoides/Eubacterium rectale population (Kleessen et al., 2001), which includes the key butyrate-producing bacteria included in the growth study presented here. The ability of the Bacteroides species (B. vulgatus and B. thetaiotaomicron) to utilise both scFOS and XOS for growth in pure culture may be beneficial in terms of enhanced production of propionate, which is often one of the fermentation products of Bacteroides species (Macfarlane & Gibson, 1997).

Amylopectin starch promoted good growth of five of the butyrate producers included in the study, and diets high in resistant starch content have previously been reported to shift the proportions of the major SFCA towards enhanced butyrate production (Englyst et al., 1992; Topping & Clifton, 2001; Duncan et al., 2007b). In contrast, reducing the carbohydrate (specifically starch) content of the diet significantly reduced numbers of the butyrate-producing Roseburia species, with a concomitant reduction in butyrate formation (Duncan et al., 2007b).

Increased butyrate levels may also result from bacterial cross-feeding. Lactate produced by the bifidobacteria can be metabolised by a group of gut bacteria known as ‘lactate utilisers’ to produce butyrate. This bacterial cross-feeding has been demonstrated for a combination of E. hallii and bifidobacteria grown on starch (Duncan et al., 2004b; Belenguer et al., 2006; Falony et al., 2009a). Alternative cross-feeding mechanisms whereby bacteria can scavenge for sugars released by other bacteria during degradation of complex oligo- or polysaccharides have also been demonstrated (Duncan et al., 2003; Falony et al., 2006). For instance, the keystone starch degrader Ruminococcus bromii may promote growth of A. hadrus, a butyrate-producing, nonstarch utiliser, through the provision of starch breakdown products (Ze et al., 2013).

GOS metabolism was prevalent among the strains tested, with a phylogenetically diverse subset of bacteria utilising it efficiently. A large collection of lactobacilli and bifidobacteria have been shown to utilise GOS effectively for growth (Watson et al., 2013). The effect of GOS (an emerging prebiotic) supplementation on the gut microbiota has now been investigated across all age groups. In a study of healthy adults, Davis and co-workers (Davis et al., 2010) found a 100-fold increase in the abundance of bifidobacteria following GOS supplementation, with a smaller increase in numbers of the butyrate producer F. prausnitzii (Davis et al., 2011). Interestingly, the increase in bifidobacteria was only observed in half of the adult subjects, revealing a considerable degree of interindividual variation (Davis et al., 2010). The authors also showed that different Bifidobacterium species/strains were enriched on GOS compared with resistant starch, which enhanced detection of B. adolescentis (Davis et al., 2011). GOS supplementation increased numbers of bifidobacteria and improved calcium absorption in adolescent girls (Whisner et al., 2013). GOS is one of the most abundant oligosaccharides contained in human breast milk and is therefore an important energy source for the intestinal microbiota of babies (German et al., 2008; Sela & Mills, 2010). In our study, the Bifidobacterium isolates tested were able to utilise GOS better than the other substrates.

XOS was found to be the most selective oligosaccharide tested, with many of the bacteria unable to grow on this substrate. Xylooligosaccharides can be extracted from solid lignocellulosic industrial waste (e.g. from malting industries; (Gullón et al., 2011)) and are considered as emerging prebiotics (Broekaert et al., 2011; Carvalho et al., 2013). Various studies have shown that increased SCFA production in mixed incubations and increased numbers of bifidobacteria were observed in a human study (Lecerf et al., 2012). In the latter study, a combination of XOS and mixed chain length inulin seemed to reduce the inflammatory effects of a high-fat diet (Lecerf et al., 2012), which may be due to down-regulation of the LPS-induced inflammatory response (Chen et al., 2012). Similarly to fructans, the chain length and composition of XOS-based prebiotics have been shown to influence bacterial fermentation and the actual site of degradation in the human colon (Sanchez et al., 2009).

Conclusions

Strategies to enrich specific bacterial populations in the large intestine have so far almost exclusively targeted lactic acid-producing bacteria, specifically bifidobacteria and lactobacilli. The work presented here shows that populations of certain species of Ruminococcaceae and Lachnospiraceae that include potentially important beneficial strains could become enhanced through prebiotic strategies. In particular, certain fructan-based prebiotics might promote butyrate-producing species such as R. inulinivorans A2-194 and F. prausnitzii. Indeed, studies in a human colonic fermentor system (Duncan et al., 2003) illustrated that numbers of supplemented R. inulinivorans A2-194 cells increased against a background of the total faecal microbiota only when inulin was supplied as the major carbohydrate substrate. The response of F. prausnitzii to FOS stimulation (shown in vitro here and in vivo by Ramirez-Farias et al. (2009)) is potentially important, given the ubiquity of this strain in healthy individuals and its reduced abundance in ileal Crohn's disease patients (Sokol et al., 2009; Willing et al., 2009). The information described here is also relevant to the development of synbiotics (Drakoularakou et al., 2004; Macfarlane et al., 2008; Saulnier et al., 2008), whereby both the growth substrate and the relevant bacterial cells are supplied in combination. This study also indicated that future investigations on the effects of prebiotics on gut health should assess the complete microbial ecosystem, and not just selected bacterial groups. Only by monitoring total population shifts will we improve our understanding of the mode of action of prebiotics and our ability to determine their true role in promoting health.

Acknowledgements

We wish to thank Cecile Crost, Cyril Nourrisat, Virgile Duclos, Marlene Clerget, Melissa Richardeau, Clemence Defois and Kenneth Young for help with the growth experiments. This work was funded by the Scottish Government (SG-RESAS).

Ancillary