The colonic microbiota forms a complex interface between dietary components and colonic health. Our understanding of colonic bacterial population dynamics and their consequent influence on gut disease risks is poor. The human colon typically contains a large and diverse population of bacteria (>1011 bacteria/g contents) which play an important role in the normal development of the intestine and provide an important energy source via fermentation of ‘resistant’ starches(Reference Hill1).
Strictly anaerobic bacteria such as Bacteroides, Bifidobacterium, Fusobacterium, Clostridium and Peptostreptococcus species are thought to predominate, as measured by both cultural and molecular methodologies(Reference Finegold, Sutter, Mathisen and Hentges2–Reference Wang, Cao and Cerniglia4). Certain genera, such as Lactobacillus and Bifidobacterium, may indicate a ‘healthy’ bowel(Reference McFarland5, Reference Salminen6), whereas others have been linked with an increased colon cancer risk(Reference Moore and Moore7). Considerable variation in the distribution of bacterial species between individuals has been clearly identified(Reference Wang, Cao and Cerniglia4, Reference Zoetendal, Akkermans and De Vos8, Reference Maukonen, Satokari, Mättö, Söderlund, Mattila-Sandholm and Saarela9) but few of these studies had detailed dietary information. There is little information on how an individual's bacterial population may change over time. Importantly, many enteric bacterial species cannot be readily cultured and may have been omitted from some previous studies.
An important metabolic process of colonic bacteria is their production of the SCFA: acetate, butyrate and proprionate. These SCFA, particularly butyrate, are often absorbed and may be utilised directly as an energy source by colonic epithelial cells(Reference McNeil, Cummings and James10–Reference Livesey, Elia, Cummings, Rombeau and Sakata12). Several lines of research have indicated that butyrate may be protective for colonic cell damage. Cell culture studies have indicated that the presence of butyrate at physiological concentrations enhances the growth of normal enterocytes and inhibits that of malignant ones(Reference Hague, Elder, Hicks and Paraskeva13, Reference Hinnebusch, Meng, Wu, Archer and Hodin14). A rising butyrate concentration within the large bowel in a rodent colon cancer model was correlated with increased apoptosis in the aberrant crypt foci in affected colons(Reference Caderni, Luceri, Lancioni, Tessitore and Dolara15). Although the causes of the various forms of inflammatory bowel disease complex may be generally related to dysregulated immune responses to commensal bacteria(Reference Hendrickson, Gokhale and Cho16, Reference Swidsinski, Ladhoff and Pernthaler17) distal ulcerative colitis has also been related to SCFA metabolism and/or availability in the colon(Reference Roediger, Cummings, Rombeau and Sakata18). Colon cells isolated from distal ulcerative colitis patients had an impaired capacity to oxidise butyrate and faecal butyrate levels were higher in adult and infant colitis patients than in controls(Reference Jorgensen and Mortensen19). It is therefore possible that a defect in butyrate metabolism occurs in distal ulcerative colitis, which could be overcome through provision of more substrate. SCFA enemas (including butyrate) have been reported to induce remission in colitis-affected patients(Reference Scheppach, Sommer, Kirchner, Paganelli, Bartram, Christl, Richter, Dusel and Kasper20). Identification of individuals with low faecal SCFA concentrations could also provide a target population for dietary or microbiota modulation or therapeutics. The SCFA are produced solely via bacterial anaerobic fermentation pathways in the colon, particularly the proximal portion. However, the number and extent of SCFA and bacterial studies on colon samples from cadavers or colostomy patients have been limited(Reference Mitchell, Lawson, Davies, Kerr-Grant, Roediger, Illman and Topping21, Reference Cummings, Pomare, Branch, Naylor and MacFarlane22), and the biology of SCFA production along the length of the colon largely remains an interpolation of faecal studies in man.
This preliminary study explores bacterial population dynamics and faecal SCFA concentrations in eight healthy human volunteers consuming their regular diet. The application of new molecular techniques to biological samples enables a more comprehensive analysis of colonic bacterial species and their relationship to diet and health.
Experimental methods
Participants and sample collection
A 12-week bowel health study was designed to examine enteric bacterial microbiota and SCFA concentrations in eight healthy participants consuming their regular diets. Six unrelated free-living volunteers and one married free-living couple (coded as PH01 to PH 08) were recruited to the trial from one city of one million residents. Approval was obtained from the CSIRO Health Sciences and Nutrition Human Ethics Committee and informed, written consent obtained from each volunteer. Subjects with a history of bowel disorder, such as irritable bowel syndrome, Crohn's disease or colorectal cancer, were excluded from the study. Subjects were required to be antibiotic-free for at least 3 months prior to commencement of the trial. Volunteers were required to complete a food diary, recording all food, drink or medication consumed, over 2 separate weeks during the 12-week study. The macronutrient composition of each volunteer's diet was calculated using daily food diaries and average daily nutrient intakes based on the nutrient composition of each food or beverage item per unit weight. These figures were calculated by reference to Australian food tables using the CSIRO dietary analysis system(Reference Baghurst and Record23). The Australian food nutrient data were supplemented by British, US or manufacturer's data where Australian data are missing(Reference Holland, Welch, Unwin, Buss, Paul, Southgate, McCance and Widdowson24).
Fresh faecal specimens were collected over a 48 h period once a week for 12 weeks. In addition, for 1 week of the study bowel actions were collected daily. Each bowel movement was packaged separately and frozen directly after collection, and stored at − 20°C. Specimens were transported frozen to the central laboratory. Faecal samples were taken to be an accurate reflection of colonic organisms in number and identity(Reference Macfarlane, Gibson and Cummings25).
Specimens were defrosted at room temperature and all processing was performed under anaerobic conditions. Each stool mass was recorded prior to combining the 48 h collection from each volunteer (henceforth referred to as the specimen). The specimen was thoroughly mixed by hand homogenisation and the pH measured. Faecal aliquots (samples) were taken for SCFA analysis, DNA extraction and faecal moisture calculation.
SCFA measurement
SCFA concentrations in faecal specimens were measured according to a modified method of Patten et al. (Reference Patten, Bird, Topping and Abeywardena26). Briefly, duplicate 1 g faecal specimens were prepared for SCFA distillation by addition of 3 × volume for weight of 1·68 mm-heptanoic acid, pH 7, mixed and centrifuged at 3000 rpm for 10 min at 5°C. Supernatant (150 μl) was distilled under vacuum and 60 μl of the distillate were transferred to a glass GC vial prior to loading on the Agilent 6890N Network gas chromatograph system with automatic loader/injector. The GC column was a Zebron ZB-FFAP (Phenomenex, Lane Cove, NSW, Australia), length 30 m, internal diameter 0·53 mm, film thickness 1 μm. The GC was programmed to achieve the following run parameters: initial temperature 90°C, hold 0·5 min, ramp 20°C/min, final temperature 190°C, total run time 8·0 min. Gas flow 7·7 ml/min splitless to maintain 3·26 psi column head pressure, septum purge 2·0 ml/min.
Calibration standards were prepared to give a mixture of the following concentrations of acids (mm): 26·22 acetic, 19·86 propionic, 3·24 isobutyric, 16·32 butyric, 5·40 isovaleric, 5·46 valeric, 4·74 caproic and 5·04 heptanoic. This standard mix (0·2 μl) was used to calculate retention times and create a standard plot. Additional standards were included in each GC run of samples at five sample intervals to maintain calibration; 0·2 μl of each subsample distillate was analysed by GC and an integrator was used to plot the curve of the standards and provide the concentration of acids present in mm.
The total SCFA (the sum of acetate, propionate and butyrate values) excreted in a 48 h stool collection (pooled values or total output) were calculated using total faecal output (in g) and faecal moisture values.
DNA extraction and purification
Extraction of DNA from duplicate 0·1 g faecal samples was performed using the QIAmp® DNA Stool Mini Kit (Qiagen, Hilden GmbH, Germany), following the supplied protocol for isolation of bacterial DNA from stool. Quality, high molecular weight, DNA was determined on 1 % (w/v) agarose gels containing 10 μg/ml ethidium bromide and visually assessed under UV transillumination.
Denaturing gradient gel electrophoresis–PCR
Approximately 100–200 ng of genomic DNA were used as template for denaturing gradient gel electrophoresis (DGGE)–PCR. Universal bacterial 16S rRNA gene primer sets were used to amplify regions of the 16S rRNA gene. They were comprised of forward primer 907f (5′-AAACTCAAA GGAATTGACGG-3′(Reference Santegoeds, Ferdelman, Muyzer and de Beer27)) and GC-clamped reverse primer 1392rc (5′-CGCCCGCCGCGCCCCCGCCCGGCCCGCCGCCCCCGCCCCACGGGCGGT GTGTRC-3′(Reference Ferris, Muyzer and Ward28)) resulting in a PCR product size of about 500 bp. A mixture of diluted reamplified DGGE bands was used as a positive control template for PCR–DGGE.
Amplification of isolated bacterial 16S rRNA gene was performed using the primers 10f and 1492r(Reference Lane, Stackebrandt and Goodfellow29) resulting in a product size of about 1500 bp. Each PCR contained one unit of Taq DNA polymerase (Promega, Madison, WI, USA), 2·5 μl 10 × reaction buffer (Promega), 1·5 μl 25 mm-MgCl2, and 10 mm-deoxynucleotide triphosphates (dNTP), genomic DNA and 12·5 pmol of each primer, in a total volume of 25 μl. PCR cycling parameters followed a touch down protocol of twenty-nine cycles on a Hybaid PCR Express thermal cycler (Hybaid, UK), and incorporated 0·5°C steps from 65 to 55°C with 2 min 72°C extension and 1 min 95°C denaturation steps, with a final 4 min 72°C extension step.
Denaturing gradient gel electrophoresis analysis
DGGE was performed using the D-code Universal Mutation Detection System (Bio-Rad Laboratories, Hercules, CA, USA) according to the manufacturer's method for perpendicular gels. PCR product (15 μl) was loaded per lane and electrophoresed on a 16 cm × 15 cm 6 % acrylamide gel containing a 35–80 % denaturing gradient for 16 h at 60°C. Denaturation solutions were made by combining 0 % (6 % v/v acrylamide in 1 × TAE buffer (40 mm-Tris-acetate,1 mm-disodium EDTA (pH 8·0))) and 100 % (6 % v/v acrylamide, 40 % v/v deionised formamide, 42 % w/v urea in 1 × TAE buffer) denaturation solutions. Gels were stained for 30 min in 100 ml 1 × Sybr-gold (Molecular Probes, Eugene, OR, USA) in 1 × TAE and destained for 5 min in 100 ml Milli-Q H2O. Images were digitally captured under 302 nm UV transilluminatation using the DigiDoc apparatus and Software (Bio-Rad Laboratories). For comparison between gels the positive control PCR reaction was run in the outside lanes and was used to standardise fragment migration when comparing between gels.
Data analysis and statistics
Elementary statistical analysis (mean, standard deviation and CV) was applied to the data. Correlations were calculated using Pearsons linear regression model (R) and coefficient of determination (R 2). Multivariate analysis (multidimensional scaling and analysis of similarities)(Reference Schiffman and Reynolds30, Reference Powell, Bowman, Snape and Stark31) of DGGE banding patterns was performed using the Primer 6 package (PRIMER-E Ltd, Plymouth, UK). DGGE bands migrating to the same position in separate lanes were considered for the purpose of analysis to represent the same sequence.
Results
Study subjects
The ages of the subjects (three men, five women) ranged from 31 to 59 years. BMI ranged from 23·1 to 28 kg/m2; none of the volunteers were obese. All subjects consumed their regular diets for the duration of the study. None of the volunteers were smokers. One subject (PH01) withdrew at week 8 of the trial due to travel commitments. Occasional samples were not collected at week 1 (PH08), week 6 (PH01), week 10 (PH02), week 12 (PH02, PH04, PH05) due to personal or travel commitments.
Dietary analysis
Food diaries were completed during weeks 4 and 9 (PH01–PH05) and weeks 7 and 10 (PH06–PH08). The daily macronutrient composition, comprising forty dietary components, consumed by each volunteer during two separate weeks was calculated using CSIRO Dietary Analysis Software(Reference Baghurst and Record23) and data on a subset of ten major dietary constituents, energy, water, protein, fat, carbohydrate, complex carbohydrate, sugar, fibre, total NSP and alcohol, are presented in Table 1, and the average daily values reflect the inconsistency and variety of volunteers' regular diets. Average energy varied from 6970 to 8441 KJ/d in females and 6958 to 9472 KJ/d in males. Alcohol intake had the greatest variation between weeks for most volunteers and the CV ranged from 0 (non-consumer) to 68·2 %. Daily patterns of nutrient intake were variable, and different patterns of eating could be observed (e.g. increased intake of dietary components associated with weekends in some volunteers) (data not shown).
CHO, carbohydrates; CV, percentage CV between the 2 weeks.
Faecal specimens
Total faecal output over 48 h was calculated for each volunteer in the 12-week study period (Table 2), and collections varied from 0 to 579 g. Daily faecal output (24 h) ranged from 0 to 377 g (Table 3). Faecal output was greater in males than females for both average weekly and average daily collections but did not reach significance (ANOVA F 2·33, P = 0·178; F 4·94, P = 0·068, respectively). An irregular daily pattern of faecal output was observed in women, and for each female participant bowel movements were absent on 2 or more days of the week. Frequency of bowel movements ranged from 0 to 5 over 48 h, and per day averaged 1·02 for males and 0·90 for females (P ≥ 0·05). Faecal pH ranged from a minimum pH 5·7 to a maximum of pH 8, averaging 7·2 for females and 6·9 for males (ANOVA P = 0·144; data not shown). Faecal moisture ranged from 61 to 85 %, averaging 76 % for males and 69 % for females and variance between repeat specimens for each volunteer was small (CV < 8·3 %; data not shown). Moisture differences between genders approached significance, ANOVA P = 0·053, with male samples more moist than female samples. Faecal moisture was positively correlated with total faecal output (48 h collections; Pearson's correlation R 0·557, P = 0·00). Faecal pH values were negatively correlated with both faecal output (R − 0·222, P = 0·04) and faecal moisture (R − 0·507, P = 0·00).
NA, not available.
* Male volunteer.
* Male volunteer.
SCFA measurement
SCFA were measured in a total of eighty-six weekly faecal specimens. Faecal concentrations of SCFA in 48 h collections ranged from 21·2 to 186·6 mmol/kg and average values for each volunteer over 12 weeks varied from 36·9 to 144·4 mmol/kg, demonstrating inter-individual variation (Table 4). Intra-individual variation in total SCFA concentration was observed over 12 weeks and CV ranged from 11·8 to 39·1 % (Table 4). Total SCFA output over 48 h was more variable than SCFA concentration (Table 4), and reflects differences in faecal output (and moistures) in both inter- and intra-individual specimens. Faecal butyrate concentrations showed similar patterns but were more variable than total SCFA (Table 5). Concentrations ranged from 1·8 to 48·5 mmol/kg, and total butyrate output over 48 h ranged from 0·2 to 12·9 mmol. Scrutiny of standard deviation reveals greater variation in SCFA values between individuals than within individuals, indicating a consistent difference between individuals.
NA, not available; N/S, no sample.
NA, not available; N/S, no sample.
Overall, faecal output was positively correlated with SCFA output in weekly specimens (R 0·848, P = 0·00; note that faecal output values are used to calculate SCFA output), however, there was only weak correlation with SCFA concentration (R 0·392, P = 0·00). These patterns of correlation held true when data sets for individuals were examined. Faecal moisture content was weakly associated with SCFA output and concentration (R 0·637, P = 0·00; R 0·529, P = 0·00, respectively). A weak negative correlation was observed between pH and SCFA output and concentration (R − 0·479, P = 0·00; R − 0·679, P = 0·00), respectively.
SCFA and dietary macro-constituents
To partially compensate for intestinal transit time dietary data was derived from 24 h prior to faecal collection and the first 24 h of the 48 h faecal collection period. Of nine dietary macro-constituents analysed, only fat, sugar and carbohydrate intake correlated with SCFA concentration, and only fat and sugar correlated with SCFA output (Table 6).
CHO, carbohydrates.
Bacterial population dynamics
DNA was extracted from a total of eighty-six weekly specimens and forty-four daily specimens, and DNA quality was confirmed by agarose gel electrophoresis. Multivariate analysis of DGGE banding patterns from individual faecal samples demonstrated stable bacterial populations specific to each individual (Figs. 1 and 2). Clear inter-individual differences were observed in DGGE profiles of faecal bacterial populations (Figs. 1 and 3). Analysis of similarities analysis of total DGGE banding patterns confirmed significance of inter-individual differences for the dominant bacterial groups detected by PCR–DGGE (P < 0·001, R ≥ 0·99).
Between sixteen and twenty-five distinct bands were detected in each weekly faecal specimen and some variation in the intensity of individual DGGE bands was observed between intra-individual specimens, as illustrated in Fig. 1.
Discussion
Examination of the faecal bacterial eco-system in eight humans over a 12-week period generated detailed information on its bacterial population dynamics and on its metabolic products. In order to achieve a realistic view of possible variability between free-living adults, we did not impose a controlled diet on volunteers, but recorded dietary intake and composition as a study variable. By combining molecular analysis of bacterial populations with physiological measures we further demonstrated the complexity of the colonic environment and highlighted the potential of a multidisciplinary approach to understanding inter-relationships which are likely to have a bearing on human colonic health.
The present studies of the volunteers' colonic bacterial populations confirm their clear variability between individuals on omnivorous Western diets, as indicated in previous studies of free-living human subjects and other omnivores(Reference Satokari, Vaughan, Akkermans, Saarela and de Vos32–Reference Mueller, Saunier and Hanisch35). The DGGE analysis visualised between sixteen and twenty-five distinct bands, out of an estimated 400–500 species, per faecal sample with only three DGGE bands common to all volunteers. Besides this expected variability between individuals, we also found that the genera comprising each individual's microbiota changed very little over the study. It is yet to be clearly established whether a consistent change in diet under controlled conditions would give rise to bacterial population shifts. The 12-week period chosen for the present study was considered to reflect a likelihood of reliable extrapolation to longer periods, possibly semi-annual. In other words each individual carried a distinct but stable microbiota. This occurred despite all the temporal variations in dietary intake (energy, fat, fibre, etc) associated with a Western diet. This microbiota stability has been suggested in occasional less detailed studies(Reference Maukonen, Satokari, Mättö, Söderlund, Mattila-Sandholm and Saarela9, Reference Vanhoutte, Huys, De Brandt and Swings33), but we consider the present study to be more wide-ranging. This stability also confirms widespread notions of the ‘colonisation resistance’ effect of the resident colonic microbiota to incoming novel bacteria(Reference van der Waaij, Berghuis-de Vries and Lekkerkerk36, Reference Hopkins and Macfarlane37).
The faecal SCFA concentrations in the volunteers were broadly similar to those noted in other Western citizens in previous single-sample studies(Reference Cummings, Pomare, Branch, Naylor and MacFarlane22). However, in contrast to the bacterial microbiota, we established that the volunteers' SCFA concentrations were widely variable over the 12-week study period. This variability was noted both within each healthy adult repeatedly sampled and between adults (including between each individual in a marriage). Despite this variability, we were able to stratify individuals within their ranges of SCFA concentrations observed. Two individuals (PH02 and PH08) consistently had low concentrations of total SCFA (average of 56·4 and 36·9 mmol/kg) and butyrate (average of 7 and 6 mmol/kg) in their faeces, as compared to their same-city cohorts. The remaining individuals had higher and more variable concentrations with averages typically over 100 (total SCFA) and 20 mmol/kg (butyrate). It is of interest that some individuals (25 % of the study group) may have consistently low concentrations of SCFA in their faeces and could represent an at-risk population for bowel health. It is possible that these individuals may have altered production and/or absorption of SCFA within their colon. Bacterial metabolic activity, including SCFA production, is thought to be more intense in the proximal colon where the environment is more fluid and substrate availability greater(Reference Macfarlane, Gibson and Cummings25, Reference MacFarlane, Gibson, Cummings, Rombeau and Sakata38). It would therefore be valuable to link SCFA faecal content to distal colon cancer or other abnormalities in future studies.
The co-analysis of self-recorded dietary composition and faecal SCFA revealed only weak correlations with dietary carbohydrate, fat and sugar intake. Carbohydrates, particularly in the form of resistant starches, provide fermentable substrate to the colon and by-products of this fermentation include SCFA(Reference Topping and Clifton39). The correlations of faecal SCFA with dietary fat or sugar content may reflect the variable western patterns of eating. Dietary fat consumption was negatively associated with faecal SCFA concentrations, however, this effect was weak and the actual influence of fats on SCFA production remains unclear. Dietary sugar consumption strongly correlated with carbohydrates (R 0·77) and may not have been an independent variable. Simple sugars are predominantly digested and absorbed in the small intestine. Alcohol intake varied between participants but was not associated with SCFA values. Further studies involving more volunteers consuming controlled diets would be needed to further explore these co-relations. Dietary complex carbohydrates, including resistant starches and fibre, such as bran, are known to affect gut transit time. Dietary carbohydrates have been linked with increased faecal SCFA concentration or output(Reference Lewis and Heaton40), but in the present study we found no clear relationship.
The influence of the composition of the colonic bacterial population on the relationship between diet and SCFA status was only partially addressed. Each of the volunteers demonstrated a distinct and stable individual DGGE profile of the various bacterial species present. Interestingly, the two individuals with the lowest SCFA concentrations, PH02 and PH08, were located adjacent to each other on multidimensional scaling plot analysis of these DGGE profiles (see Fig. 3). It is therefore possible that particular and distinct colonic bacterial populations may be linked to functional production of SCFA. This relationship may not be direct or linear, and requires considerable further characterisation with respect to individual genera and their SCFA competency or redundancy. As information on colonic bacteria proceeds, more specific molecular approaches may allow analysis of functional genes and targeting groups containing butyrate production. Targeted dietary changes may be envisaged to drive microbial composition to a more competent butyrate state or other desired outcome.
Acknowledgements
We thank Sally Record for her excellent assistance with dietary analysis and Ian Saunders (CMIS) for his help with study design and statistical analysis. The work in this paper has been carried out with financial support of the Preventative Health Flagship and CSIRO Human Nutrition, CSIRO, Australia. There are no conflicts of interests associated with this paper.