The intestinal microbiota are a complex ecosystem of several hundred microbial species(Reference Dethlefsen, Eckburg and Bik1). The microbiota have numerous functions in the maintenance of health, including the ability to decrease colonisation and infection with pathogens, a mechanism referred to as colonisation resistance(Reference van der Waaij, Berghuis-de Vries and Lekkerkerk-van der Wees2). The intestinal microbiota also produce SCFA through metabolism of fermentable substrates such as dietary fibre. SCFA stimulate colonic water absorption(Reference Bowling, Raimundo and Grimble3) and, in particular butyrate, are the primary energy source for the colonocytes that constitute the epithelial lining of the large intestine(Reference Pryde, Duncan and Hold4). Butyrate promotes the growth of colonocytes in animal models, preventing mucosal atrophy, and appears to lower the risk of malignant transformation in the colon(Reference Koruda, Rolandelli and Bliss5). Therefore, butyrate in the intestine is beneficial for the integrity of the mucosal barrier function of the colon. However, not all species of the intestinal microbiota produce butyrate, with the predominant butyrate producers belonging to the genus Faecalibacterium prausnitzii (Reference Duncan, Hold and Harmsen6) and Roseburia spp.(Reference Pryde, Duncan and Hold4).
Colonisation resistance and SCFA production are therefore important functions of the intestinal microbiota(Reference Whelan, Judd and Tuohy7). However, when the intestinal microbiota are altered, for example through the use of antimicrobials(Reference Sullivan, Edlund and Nord8) or by a reduction in fermentable biomass(Reference Scott, Duncan and Flint9), their capacity to exert these important functions is reduced. The major source of fermentable biomass is NSP (i.e. dietary fibre). Both the US Institute of Medicine(10) and the American Dietetic Association(Reference Slavin11) have recommended increased dietary fibre consumption as evidence accumulates that higher intakes are related to a lower risk of CVD, and potentially colon cancer, diabetes and obesity. These beneficial effects of dietary fibre may be mediated in part by the augmentation of colonic fermentation and SCFA production.
Research on the impact of dietary fibre on the intestinal microbiota and SCFA production in human subjects is impeded by the ubiquitous presence of fibre in the normal diet, making it difficult to achieve total dietary exclusion. However, one approach to resolving this is through the use of liquid enteral formulas that are used to provide artificial nutrition support. To date, many of the most commonly used standard formulas have been completely devoid of fibre, but formulas supplemented with various types of dietary fibre are now available.
Use of such formulas provides a way of investigating the complete removal of dietary fibre from the diet as well as the effect of addition of specific types and amounts of dietary fibre on the intestinal microbiota and SCFA in healthy human subjects. In addition, they may also enable investigation of two clinically relevant situations. First, patients receiving enteral nutrition in the intensive care unit, and other settings, can experience impairments in gastrointestinal function, including diarrhoea(Reference Stroud, Duncan and Nightingale12). Second, enteral formula nutrition is effective in inducing remission in active Crohn's disease(Reference Zachos, Tondeur and Griffiths13). Alterations of the intestinal microbiota implicated in disease activity can also be effective in inducing remission(Reference Hedin, Whelan and Lindsay14). Furthermore, faecal concentrations of F. prausnitzii are lower in patients with active inflammatory bowel disease(Reference Sokol, Seksik and Furet15), whilst butyrate may be a potential treatment for Crohn's disease(Reference Di Sabatino, Morera and Ciccocioppo16) through its immunoregulatory effects(Reference Cavaglieri, Nishiyama and Fernandes17).
Our hypothesis is that the exclusion of dietary fibre will cause a significant change in the numbers of the butyrate producers F. prausnitzii and Roseburia spp., and thereby reduce SCFA concentrations; in contrast, the addition of dietary fibres will prevent such effects on the intestinal microbiota and SCFA.
We have previously described the effects of consumption of enteral formulas devoid of fibre on the total microbiota and SCFA. The reduction of both total bacterial count and SCFA production that occurred was partially prevented by the addition of fibre and fructo-oligosaccharides to the diet(Reference Whelan, Judd and Preedy18). Interestingly, butyrate concentrations were reduced following both fibre-free and fibre-supplemented formulas. However, in that study there was no extensive analysis of bacterial groups, including the butyrate producers F. prausnitzii and Roseburia spp., nor was there analysis of the association between specific bacterial groups and the concentrations of SCFA.
The majority of bacteria in the colon are anaerobic, many are hard to culture and enumerating using selective media typically underestimates the numbers of bacteria in faecal samples compared with molecular techniques such as fluorescent in situ hybridisation (FISH)(Reference Harmsen, Gibson and Elfferich19). Our extended probe set covers approximately 88 % of the total intestinal microbiota in healthy volunteers, including the butyrate producers F. prausnitzii and Roseburia spp.(Reference Harmsen, Raangs and He20).
Here we describe an independent blinded reanalysis of the faecal samples from the previous study of liquid formulas devoid of, and supplemented with, fibre(Reference Whelan, Judd and Preedy18), using the extended probe set, in order to provide further insights into the relationships between fibre, the intestinal microbiota and faecal SCFA concentrations.
Methods and materials
Subjects
Samples were obtained from healthy volunteers recruited to a prospective, double-blinded, randomised, cross-over trial(Reference Whelan, Judd and Preedy18). The subjects were free of gastrointestinal diseases, or self-reported eating disorders; they were not following a special diet, had not used antibiotics in the previous 3 months and had not consumed products containing prebiotics or probiotics in the previous 1 month. The study was conducted according to the guidelines laid down in the Declaration of Helsinki and was approved by the King's College London Research Ethics Committee (reference 01-62). Written informed consent was obtained from each subject before recruitment.
Protocol
Subjects consumed their normal diet for 14 d (baseline period), then an enteral formula for 14 d (diet period), followed by their normal diet for 6 weeks as a washout phase (baseline period) and then an alternative enteral formula again for 14 d (diet period). All consumption of prebiotics and probiotics was avoided for the duration of the study. During the liquid diet periods, the formulas were the only source of nutrition except for water ad libitum and black tea and/or coffee to a maximum of 600 ml/d. Subjects were advised on the amounts of formula required to achieve their calculated total energy expenditure and provided with sufficient formula to maintain their weight. During the diet periods, subjects consumed either a fibre-free enteral formula or a formula supplemented with dietary fibre (14 g/l). The dietary fibre formula consisted of pea fibre (8·9 g/l) and fructo-oligosaccharides (5·1 g/l) and therefore contained 48 % soluble and 52 % insoluble fibre fractions. The nutritional values of both formulas were almost identical except for the difference in fibre content. The order of consumption of each formula was randomised and this randomisation was blinded for both study subjects and researchers. This blinding was maintained for the present analysis.
Dietary intake during the normal diet (baseline) was recorded in a 7 d semi-weighed food diary. The data were entered into a dietary analysis package (CompEat v4; Nutrition Systems, Banbury, Oxon, UK). Intake of the formulas during the fibre-free diet and fibre-supplemented diet was recorded using self-reported diaries(Reference Whelan, Judd and Preedy18). Faecal frequency was calculated from self-reported diaries during the last 7 d of both normal diets periods (baseline), the fibre-free diet and fibre-supplemented diet. Mean daily faecal weight was calculated from a 3 d total faecal collection during the last 3 d of each diet period.
The final faecal specimen from each period was collected and stored for analysis. A sample was frozen at − 80°C for analysis of SCFA using GLC, as published earlier(Reference Whelan, Judd and Preedy18). A sample was also processed and stored for analysis of microbiota using FISH(Reference Franks, Harmsen and Raangs21). Briefly, samples were diluted 1:10 (w/v) in PBS (NaCl (8 g/l), KCl (0·2 g/l), Na2HPO4·2H2O (1·44 g/l), KH2PO4 (0·24 g/l), pH 7·4) and fixed in 4 % paraformaldehyde in PBS for at least 4 h. Washed cells were re-suspended in PBS–ethanol solution (1:1, v/v) and stored at − 20°C before the present analysis.
Fluorescent in situ hybridisation
For quantification of the bacteria in these faecal samples, multiple slides with 1 cm2 wells were prepared for cell counting. Per well, 10 μl of diluted sample were spread. After drying, the cells were fixed to the glass surface with 96 % ethanol for 10 min. In the present study hybridisation was performed with an extended set of 16S rRNA-targeted probes (Table 1). The probe set used for bacterial groups covers approximately 88 % of the total number of bacteria which hybridise with the EUB338 probe in healthy volunteers(Reference Harmsen, Raangs and He20). The probes were manufactured by Eurogentec (Seraing, Belgium) and were labelled at the 5′ end with either fluorescein isothiocyanate (FITC) or Cy3.
Probes used for the detection of the intestinal microbiota
FITC, fluorescein isothiocyanate.
The samples were hybridised overnight at 50°C in hybridisation buffer (0·9 m-NaCl, 20 mm-2-amino-2-hydroxymethyl-propane-1,3-diol (Tris)-HCl (pH 7·2), 0·1 % SDS (w/v)) containing 9 ng labelled probe per slide. The slides were washed for 20 min in wash buffer (0·9 m-NaCl, 20 mm-Tris-HCl (pH 7·2)), rinsed briefly in Milli-Q and dried using compressed air. Total cells were enumerated after staining with 4′,6-diamidino-2-phenylindole (DAPI). Slides were mounted in Vectashield® (Vector Labs, Burlingame, CA, USA) to minimise fading of the fluorescent signal.
The fluorescent cells in the samples were counted automatically with a Leica® DMRA2 epifluorescence microscope using a modified version of the Leica® Q-win® software (Leica, Wetzlar, Germany)(Reference Jansen, Wildeboer-Veloo and Tonk22). The detection limit used with this method was set at 106 cells/g.
Concentrations are presented per g dry faeces in order to standardise comparisons between samples with different water content. Dry faecal weight was calculated following lyophilisation of duplicate faecal samples.
Statistics
All data were analysed on SPSS for Windows (version 15.0; SPSS, Inc., Chicago, IL, USA). The concentrations of faecal bacteria were log transformed in order to have a normal distribution, and the geometric mean calculated. The log-transformed concentrations and proportions of bacteria were compared between baseline and diet periods using a paired t test. To investigate associations between the microbiota and SCFA, the correlation between the proportion of each bacterial group (as percentage of the total cell count) and the proportion of each major SCFA (as percentage of the total) was calculated using a Pearson's correlation coefficient at each time point. To investigate the impact of excluding and supplementing fibre on the associations between the microbiota and SCFA, the change in the proportion of each bacterial group and the change in the proportion of each major SCFA between baseline and diet period was calculated. Then, the correlation between the change in the proportion of each of these was calculated using a Pearson's correlation coefficient. P < 0·05 was considered statistically significant.
Results
Fourteen healthy subjects, five men and nine women, aged between 21 and 34 years old were included in the original study(Reference Whelan, Judd and Preedy18). Two women withdrew because they were unable to consume the formula as their sole source of nutrition; one woman withdrew because of personal reasons unrelated to the study. One man was excluded after testing positive for Giardia lamblia during the study. In total, ten healthy subjects (six women and four men) completed the study and were included in this analysis.
Intakes of fibre (19·6 and 18·0 g/d) and NSP (14·3 and 12·8 g/d) during normal (baseline) diets are similar to values previously recorded in the UK (Table 2).
(Mean values and standard deviations)
FOS, fructo-oligosaccharides.
* Values for the energy, macronutrient and fibre content of the fibre-free diet (Nutren 1·0) and the fibre supplemented diet (Nutren fibre) were provided by Nestlé UK.
† Data not available.
There was a reduction in faecal frequency between the normal diet and the fibre-free diet (1·0 (sd 0·3) v. 0·6 (sd 0·2); P = 0·001), whilst the reduction between the normal diet and the fibre-supplemented diet (1·1 (sd 0·3) v. 0·9 (sd 0·3); P = 0·056) did not reach statistical significance. Faecal frequency was higher during the fibre-supplemented diet than the fibre-free diet (P = 0·019). There was a reduction in daily faecal weight between the normal diet and the fibre-free diet (132·4 (sd 68·5) v. 43·8 (sd 30·1) g; P = 0·005) and the normal diet and the fibre-supplemented diet (127·5 (sd 71·2) v. 73·2 (sd 37·5) g; P = 0·034). However, there were no differences in daily faecal weight between the fibre-free and fibre-supplemented diets (P = 0·149).
In general, the results show a declining trend in all bacterial species during both diet periods, except for bifidobacteria, which increased during the fibre-supplemented diet (Table 3). Interestingly, statistically significant and large reductions in both the concentration and the proportion of the F. prausnitzii group and Roseburia intestinalis group occurred during both diet periods, irrespective of fibre content, whereas the Bacteroides group underwent significant decline in concentration and proportion only during the fibre-free diet period only.
(Mean values and standard deviations)
* P value for diet v. baseline.
† P value for fibre-free diet v. fibre-supplemented diet.
There were significant reductions in concentrations of total SCFA and each of the major SCFA (acetate, propionate, butyrate) following the fibre-free diet, whilst only butyrate was reduced following the fibre-supplemented diet. There was an increase in the proportion of acetate and a reduction in the proportion of butyrate following both the fibre-free diet and the fibre-supplemented diet. Details of these results are presented elsewhere(Reference Whelan, Judd and Preedy18).
There were significant and strong positive correlations between the proportion of F. prausnitzii and the proportion of butyrate during the normal diet (Table 4) and these occurred before both the fibre-free (r 0·881; P = 0·001) and the fibre-supplemented diet (r 0·844; P = 0·002). Interestingly, these correlations were no longer found during either the fibre-free (r 0·359; P = 0·308) or the fibre-supplemented diet (r 0·090; P = 0·805).
Correlations between the proportion of intestinal microbiota (as percentage of total cells) and the proportion of acetate, propionate and butyrate (as percentage of total SCFA) at the end of each dietary period
* P < 0·05.
The dynamic effects of switching from a normal diet to a fibre-free or a fibre-supplemented diet were also investigated (Table 5). There was a significant correlation between the change in proportion of F. prausnitzii and the change in proportion of butyrate following the fibre-free diet (r 0·806; P = 0·005) and between the change in proportion of acetate (r 0·671; P = 0·034) and butyrate (r 0·749; P = 0·013) following the fibre-supplemented diet. There were no other significant correlations in these changes for any bacterial group and any SCFA.
Correlation between the change in proportion of intestinal microbiota (as percentage of total cells) and the change in the proportion of acetate, propionate and butyrate (as percentage of total SCFA) between the standard diet (baseline) and a fibre-free or a fibre-supplemented diet
* P < 0·05.
Discussion
This is the first study to demonstrate a strong correlation between F. prausnitzii and butyrate production in healthy volunteers during a normal diet, and that the reduction in F. prausnitzii when switching to a fibre-free or fibre-supplemented liquid diet also correlates with the reduction in faecal butyrate. Duncan et al. have previously shown a correlation between Roseburia and butyrate production in obese patients on a carbohydrate-restricted diet; however, in that study the correlation with F. prausnitzii was poor(Reference Duncan, Belenguer and Holtrop23).
We describe an in-depth analysis of faecal samples from a study of healthy subjects consuming defined liquid diets either devoid of fibre or fibre-supplemented. In the previous report(Reference Whelan, Judd and Preedy18), SCFA were measured in the stool samples and all major SCFA (acetate, propionate, butyrate) were reduced following a fibre-free diet, whereas following a fibre-supplemented diet acetate and propionate did not significantly decrease from baseline, whilst butyrate did. However, without the present analysis, and in particular the detection of the decrease in F. prausnitzii during both diets, the reduction in butyrate during both diets would have remained unexplained(Reference Schneider, Girard-Pipau and Anty24). Although there were also correlations between butyrate and other species, only in F. prausnitzii did the correlation occur in both baseline diets and in the dynamic effects analysis. The results for FISH analysis were remarkably reproducible, with bacterial counts similar to the original analysis following long-term storage in ethanol–PBS at − 20°C. The present analysis used automated, rather than manual, counting and this may explain the subtle differences in results found here compared with the previous analysis.
There was an increase in proportion of bifidobacteria during the fibre-containing diet, which contained prebiotic fructo-oligosaccharides. This may explain why acetate concentrations were higher compared with during the fibre-free diet, as bifidobacteria are major producers of acetate(Reference Whelan, Judd and Preedy18). Other bacteria responsible for butyrate production, such as Roseburia spp., are associated with insoluble fibres and ferment these as an energy source(Reference Walker, Duncan and Harmsen25). In addition, ruminococci also attach to dietary fibres(Reference Aminov, Walker and Duncan26) and these bacteria also show a trend towards lower numbers during both fibre-free and fibre-supplemented diets.
We speculate that the inability of the fibre-supplemented formula to fully maintain the original microbiota might be related to the composition of the insoluble fibres contained within the liquid enteral formula. The fibre-supplemented formula contained 48 % soluble and 52 % insoluble fractions from fructo-oligosaccharides and pea fibre, and it has been shown that different sources of dietary fibre lead to the formation of different SCFA profiles(Reference Salvador, Cherbut and Barry27). In pea fibres the strong linkages between uronic acid and xylose protect the xylose from fermentation, whereas in wheat-bran fibres the majority of the xylose molecules are ordered in linear xylans which are preferentially degraded. The fermentation of xylose provides a preferential substrate for the production of butyrate; therefore wheat-bran fibre leads to a higher production of butyrate than pea fibre(Reference Salvador, Cherbut and Barry27). Whilst the reduction in Roseburia spp. and F. prausnitzii during the fibre-free diet may be caused by the absence of fibre, their reduction during the fibre-supplemented diet may relate to the presence of a fibre that does not support proliferation of Roseburia and F. prausnitzii. A recent study showed that inulin increased the numbers of both bifidobacteria and F. prausnitzii (Reference Ramirez-Farias, Slezak and Fuller28). However, we cannot confirm this increase in F. prausnitzii with the data from the present study.
There was a reduction in faecal weight and frequency during both liquid diet regimens. This, together with the lower concentrations of bacteria and SCFA, reflects an even larger fall in total amounts of bacteria and SCFA. As none of the volunteers reported diarrhoea, a common complication of enteral nutrition, this is not a cause of the reduction in F. prausnitzii in these samples(Reference Swidsinski, Loening-Baucke and Verstraelen29).
Reduction of the butyrate-producing capacity of the microbiota is likely to be important in both health and in specific disease states. Low amounts of this primary energy source for the colonocytes may result in decreased intestinal barrier integrity. Furthermore, abnormal water excretion in the colon has been observed during enteral feeding which probably contributes to the pathogenesis of diarrhoea(Reference Bowling, Raimundo and Grimble30). This effect has been shown to be reversed when supplementing SCFA, including butyrate, to the colon(Reference Bowling, Raimundo and Grimble3). A case study on antibiotic-induced diarrhoea supports the concept that lowered levels of SCFA coincide with the occurrence of diarrhoea(Reference Young and Schmidt31).
In a recent study on the microbiota of patients with Crohn's disease, those who were still in remission 6 months after surgery were found to have higher mucosal concentrations of F. prausnitzii compared with those who had relapsed(Reference Sokol, Pigneur and Watterlot32). Meanwhile, patients with active inflammatory bowel disease have lower faecal concentrations of F. prausnitzii than healthy controls(Reference Sokol, Seksik and Furet15). Studies have demonstrated that this bacterial strain has pronounced immunoregulatory effects. These include decreased IL-12 and interferon-γ and increased IL-10 production in peripheral blood mononuclear cells, and marked improvement in inflammation in a murine model of colitis(Reference Sokol, Pigneur and Watterlot32). A significant reduction in F. prausnitzii, as shown in the present study, might therefore potentially lead to an increased inflammation and disease activity in Crohn's disease.
Whilst the present analysis relates to luminal, not mucosal, microbiota the present results seem somewhat paradoxical, as we have shown that enteral formulas cause a reduction in F. prausnitzii, despite the fact that they are used as primary treatment in active Crohn's disease. The exact mechanism of effect of enteral formulas in inducing remission in Crohn's disease is not clear, and the role of the mucosal and luminal microbiota requires further research. The present study suggests that this beneficial effect of enteral formula use is probably not mediated by F. prausnitzii.
Patients with isolated colonic Crohn's disease, who have no ileal disease, respond poorly to treatment with enteral formulas alone(Reference Afzal, Davies and Paintin33). As F. prausnitzii has been found in lower numbers in patients with Crohn's colitis(Reference Sokol, Seksik and Furet15), our present results might provide an explanation why patients with isolated colonic Crohn's disease do not respond well to treatment with only enteral formulas, in contrast to patients with isolated ileal Crohn's disease. The mechanisms necessary for inducing remission probably differ for the different disease states and localisation of Crohn's disease. Whereas microbial debulking might prove beneficial in ileal Crohn's disease, the coinciding depletion of colonic F. prausnitzii might not be beneficial in isolated colonic Crohn's disease. What is of striking relevance is the positive association between F. prausnitzii and butyrate production in the gastrointestinal tract, especially since butyrate may be therapeutic for active ulcerative colitis and Crohn's disease(Reference Segain, Raingeard de la Bletiere and Bourreille34). More studies of the effect of enteral formulas on the microbiota and SCFA, in particular F. prausnitzii and butyrate, are required in Crohn's disease.
For future research it is important to include an extensive probe set when investigating effects of the intestinal microbiota on SFCA production with FISH. Furthermore, alternative fibre supplements should be designed to restore the numbers of butyrate-producing bacteria and concentrations of butyrate in the faecal samples; these should be investigated for use in enteral formulas.
Conclusion
We have shown that a diet either devoid of fibre, or a diet supplemented with dietary fibre wherein pea fibre constitutes the main insoluble fraction, is accompanied by lower numbers of bacteria from the F. prausnitzii and Roseburia groups. The decrease of bacteria from the F. prausnitzii group is correlated with lower concentrations of faecal butyrate.
Acknowledgements
The present study was jointly supported by King's College London and Nestlé UK.
R. F. J. B. has received a grant from the University Medical Centre Groningen (‘UMCG Stimuleringsgelden’).
R. F. J. B. conducted the microbiological analysis and wrote the paper; T. S. v. d. W. reviewed and edited the paper; G. W. W. reviewed and edited the paper and provided essential reagents; P. A. J. designed and supervised the trial, and reviewed and edited the paper; M. A. T. designed and supervised the trial, and reviewed and edited the paper; H. J. M. H. reviewed and edited the paper and provided essential reagents; K. W. designed the trial, recruited subjects, performed statistical analysis and had primary responsibility for the final content. All authors read and approved the final manuscript.
All authors declare that they have no conflicts of interest.
