Dietary and lifestyle factors account for approximately 70 % of colorectal cancer, and an optimal dietary approach may reduce the incidence of this disease(1). Consumption of resistant starch (RS) improves the colonic environment by softening stools, increasing faecal bulk, decreasing faecal pH and increasing luminal SCFA concentrations(Reference Topping and Clifton2, Reference Young and Le Leu3). The main SCFA produced by colonic fermentation of RS are acetate, propionate and butyrate, and these SCFA have important roles in the maintenance of bowel health. Butyrate, which is the preferred energy source for colonocytes(Reference Roediger4), may be protective against large-bowel cancer(Reference Cummings5) by enhancing apoptosis, inhibiting proliferation and inducing differentiation of cancer cells in vitro (Reference Hague, Manning and Hanlon6, Reference Young, Hu and Le Leu7). Acylated starches have been shown to deliver butyrate to the large bowel of rats(Reference Annison, Illman and Topping8, Reference Bajka, Topping and Cobiac9) and humans(Reference Clarke, Bird and Topping10). The esterified butyrate is released by colonic bacteria and is absorbed or utilised by colonocytes or gut microbes.
The aim of the present study was to investigate the effects of dietary RS and butyrylated RS on the faecal microbiota of rats treated with the carcinogen azoxymethane (AOM). Rats were also fed diets containing low RS and low RS with added tributyrin to supply butyrate systemically from small-intestinal absorption. Faecal samples were collected from the rats throughout the study to explore the effects of specific diet components on measures of bowel health. Gaining insight into the behaviour and metabolic activity of these microbial populations, the influence of diet and the development of cancer may provide opportunities to modulate colorectal cancer risk.
Experimental methods
The present study used faecal samples collected from a subset of animals from a larger trial, which is described by Clarke et al. (Reference Clarke, Topping and Bird11). The study involved male Sprague–Dawley rats (198 (sem 2) g), which were randomly allocated to four groups of nine rats and fed experimental diets. These rats were treated identically to the rats in the larger study, and were fed experimental diets based on a control AIN-93G(Reference Reeves, Nielsen and Fahey12) diet which contained 530 g standard low-amylose maize starch, 200 g casein, 70 g maize oil and 50 g α-cellulose/kg of diet. Choline, methionine, minerals, vitamins and antioxidant were added as described previously(Reference Clarke, Bird and Topping10). The amount of standard maize starch in the diet was reduced to allow for the addition of experimental starches and tributyrin. The experimental diets were control (low-amylose maize starch (LAMS), low RS), LAMS with 3 % tributyrin (LAMST, low RS), cooked 10 % high-amylose maize starch (HAMS, high RS) or cooked 10 % butyrylated HAMS (HAMSB, high RS). The latter was manufactured by National Starch Food Innovation (Bridgewater, NJ, USA) and had a degree of substitution of 0·25 (i.e. 0·25 of the hydroxyl groups on each starch d-glucopyranosyl unit were derivatised or replaced by substituent acids). HAMS and HAMSB were cooked with water and spray-dried (to minimise crystallite formation), as described previously(Reference Brown, Storlien and Brown13), before addition to the diets. This process was used to mimic the condition of starches consumed in foods by humans, as cooking changes the production and delivery of butyrate to the colon by starches(Reference Clarke, Bird and Topping10). All diets were prepared regularly at the Commonwealth Scientific and Industrial Research Organisation and stored at 4°C. The rats were fed a fresh diet daily. After weeks 4 and 5 of feeding the experimental diets, rats were injected with 15 mg/kg of AOM subcutaneously (Sigma Chemical Company, St Louis, MO, USA). One rat was euthanased early due to events unrelated to the experimental procedures (from the LAMS group in week 9). All procedures involving animals were approved by the Commonwealth Scientific and Industrial Research Organisation Human Nutrition Animal Ethics Committee and complied with the Australian code of practice (NHMRC, 2004).
Faecal specimens were collected at week 0 (AIN-93G base diet), week 4 (after 4 weeks of the experimental diets, before AOM injections) and at weeks 7, 11, 15, 19, 21, 27, 29 and 31 of the experiment. Each specimen was collected from the anus of the rat and consisted of a faecal pellet of approximately 0·1 g. The samples were immediately placed in NaCl–Tris–EDTA buffer (10 mm-Tris–HCl (pH 8·0); 1 mm-EDTA; 100 mm-NaCl) and stored at − 20°C. A total of 336 rat faecal specimens were collected. Analysis of pH and liquid phase-free SCFA concentrations were undertaken in duplicate as described previously(Reference Bajka, Topping and Cobiac9). At the end of the study, the rats were anaesthetised, and tissue and digesta samples were collected for analysis(Reference Clarke, Topping and Bird11). Inadequate digesta was available for distal colonic pH and SCFA concentration determinations in some rats (n 6 for the LAMS and LAMST groups; n 7 for the HAMS and HAMSB groups).
DNA extractions from rat faecal specimens, denaturing gradient gel electrophoresis (DGGE) and sequencing of DGGE bands were performed as described previously(Reference Abell, Cooke and Bennett14). Briefly, DNA was extracted using a matrix-based method, and universal bacterial 16S rRNA gene primers were used to amplify approximately 500 bp of the 16S rRNA gene using the forward primer 907f (5′-AAACTCAAAGGAATTGACGG-3′(Reference Fuhrman, Comeau and Hagstrom15)) and the GC-clamped reverse primer 1392rc (5′-CGCCCGCCGCGCCCCCGCCCGGCCCGCCGCCCCCGCCCCACGGGCGGTGTGTRC-3′(Reference Lane, Stackebrandt and Goodfellow16)). The resulting PCR amplicons were subjected to DGGE using a 30–75 % denaturing gradient and 6 % acrylamide gel. For each sample, 15 μl of the PCR product was loaded per track and electrophoresed at 110 V for 16 h at 60°C. Gels were stained for 30 min in 1 × SYBR Gold nucleic acid stain (Molecular Probes, Eugene, OR, USA). The gels were then destained in 100 ml of MilliQ water and photographed under UV light (DigiDoc System; Bio-Rad Laboratories, Hercules, CA, USA).
GraphPad Prism version 4.0 (www.graphpad.com; GraphPad Software, Inc., San Diego, CA, USA) was used for statistical analyses on final day measures and tumour indices(Reference Clarke, Topping and Bird11). Multivariate statistical analysis of DGGE-banding patterns was performed using the Primer 6 package (PRIMER-E Limited, Plymouth, UK), with the assumption that each DGGE band represents one phylotype. Multidimensional scaling plots demonstrating the relatedness of individual samples from specific time points were calculated using the Bray–Curtis similarity calculation on square root-transformed data(Reference Kenkel and Orloci17, Reference Minchin18). Differences between diet groups on the basis of DGGE-banding patterns were calculated with a one-way analysis of similarity routine(Reference Clarke19) and tested at α = 0·01; this gives a value for the similarity of two groups, where a value of 0 indicates no difference between groups and a value of 1 indicates no similarity. Similarity percentage analysis was used to identify the DGGE bands that primarily accounted for the dissimilarity between different diet groups, identified using one-way analysis of similarity, at week 27(Reference Clarke19). Analysis of the variation between rats within a diet group, for each week, was calculated using the dispersion sequence with the MVDISP routine. This calculated the average rank dissimilarity within each of the groups, whereby the higher the value, the more dispersed the samples are within a group.
The Shannon–Wiener index (H′)(Reference Shannon and Weaver20) was calculated for each DGGE profile, also taking into account relative band intensities. The index describes how the intensity of bands is distributed across each profile. Analysis of the variation in diversity over the course of the trial was performed using a linear mixed model allowing for random variation between rats and also between weeks within rats. Variability was normally distributed, so transformation was not performed.
Following similarity percentage analysis, DGGE bands of interest were extracted and re-amplified as described previously(Reference Abell and Bowman21). Sequencing and phylogenetic analysis of DGGE bands were performed using the ARB software package(Reference Ludwig, Strunk and Westram22), as described previously(Reference Abell and McOrist23). Sequences from the present study were deposited within the Genbank database under accession numbers DQ839222–DQ839231 and EU703763.
Results
Growth rates were similar in all of the experimental groups, and there were no significant differences in final rat body weights between the dietary treatments (Table 1). Large-bowel tumour incidence (i.e. the proportion of rats/group with tumours), tumour number and surface area did not differ between the subgroups of animals used in this trial, although there were differences between dietary groups in these indices when data from the main study were analysed(Reference Clarke, Topping and Bird11). There was no effect of diet on the proportion of adenomas that progressed to adenocarcinomas. Acetate concentrations and pH were not affected by diet in the distal colon; however, propionate concentrations were higher in rats fed the high-RS diets compared with the low-RS diets (P < 0·05–0·001). Distal butyrate concentrations were higher in the rats fed HAMSB (P < 0·01) than for the other groups, and were similar in the HAMS, LAMS and LAMST groups (Table 1).
* Mean values were significantly different from HAMSB (P < 0·01).
† Mean values were significantly different from HAMS (P < 0·05).
‡ Mean values were significantly different from HAMSB (P < 0·001).
§ Mean values were significantly different from HAMS (P < 0·01).
∥ Analysed using non-parametric Kruskal–Wallis ANOVA; n 8 (LAMS) or n 9.
¶ Analysed using Pearson's χ2 test.
** Analysed using contingency comparison between groups.
DGGE analysis was performed on samples collected from six to nine individual rats on each diet at the ten sampling points. Between weeks 5 and 33, DGGE bands were detected in individual samples, and between weeks 4 and 7, the faecal DGGE profiles diverged rapidly according to the diet consumed (see Table S1 of the supplementary material available at http://www.journals.cambridge.org/bjn), with the exception of the two low-RS diets (LAMS and LAMST), which were significantly different only in the week 7 sample (see Table S1 of the supplementary material available at http://www.journals.cambridge.org/bjn). For the remainder of the trial, three distinct profiles were observed: the LAMS and LAMST samples grouped together, and the HAMS and HAMSB groups formed divergent populations that were distinctly different from each other (Fig. 1). Analysis of the inter-diet variability using dispersion sequence analysis revealed a greater degree of variability in the low-RS diets (LAMST: 1·19 (sem 0·245)) and LAMS: 1·16 (sem 0·207)) than in the high-RS diets (HAMSB: 0·73 (sem 0·312) and HAMS: 0·97 (sem 0·19)) over the course of the dietary intervention.
At week 0, there was no significant difference in microbial diversity between the diet groups as measured by the Shannon index (H′ 2·47–2·64). After 4 weeks of dietary treatment, the microbial populations showed a reduction in diversity, and the mean H′ values for all diet groups decreased significantly compared with week 0 samples (H′ 1·46–2·00, P < 0·001). By week 7, the microbial diversity increased (H′ 1·90–2·65, P < 0·0001), and was no longer significantly different from week 0 samples for any of the diets; after this time point, diversity did not change significantly.
The phylogenetic affiliation of characteristic bands revealed significant differences between the composition of the predominant phylotypes in each of the three diets, with the low-RS diets characterised by bands 58, 82, 36 and 13, comprising two members of Clostridiaceae and two members of Bacteroidetes (see Fig. S1 of the supplementary material available at http://www.journals.cambridge.org/bjn). Rats consuming HAMS were characterised by bands 13, 19 and 27 which grouped with Ruminococcus bromii (96, 95 and 96 % similarity, respectively), and band 58 which grouped within a clade of uncultivated members of Bacteroidetes (see Fig. S1 of the supplementary material available at http://www.journals.cambridge.org/bjn). Rats consuming HAMSB were characterised by six bands. One of these, band 47, contained a mixed sequence and was not included in the analysis. Bands 65, 71 and 75 were grouped within the Bacteroidetes, and were all related to Parabacteroides distasonis (98, 98 and 96 % similarity, respectively); band 54 was related to Phascolarctobacterium (98 % similarity); and band 60 was related to Lactobacillus gasseri (99 % similarity). Of these, bands 65, 71 and 75 related to P. distasonis were only detected in the HAMSB diet, whereas all the dominant bands found in the HAMSB diet, including 54 and 60, were also detected at low levels (weak bands; see Fig. S2 of the supplementary material available at http://www.journals.cambridge.org/bjn) in the HAMS diets.
Signature phylotypes for each diet group were identified using similarity percentage analysis, revealing eleven key DGGE bands that had >10 % contribution to the similarity of individual rats within different diet groups (see Fig. S1 of the supplementary material available at http://www.journals.cambridge.org/bjn). One phylotype (band 36) significantly contributed to the dissimilarity between the low- and high-RS diets. The phylogenetic affiliation of characteristic bands revealed significant differences between the composition of the predominant phylotypes in each of the three diets, with the low-RS diets characterised by bands 58, 82, 36 and 13, HAMS associated with bands 13, 19, 27 and 58 and HAMSB were characterised by bands 54, 60, 65, 71 and 75; the latter three were only found on the HAMSB diet (see Fig. S2 of the supplementary material available at http://www.journals.cambridge.org/bjn).
Discussion
The present study examined the effects of HAMSB, a butyrylated starch designed to promote large-bowel health by rapidly delivering butyrate to the colon, on the faecal microbiota of rats. HAMSB was found to be particularly effective at raising butyrate concentrations in the distal colon, which is the site of most degenerative large-bowel disease in humans(Reference Topping, Gooden and Brown24). Although the effects on tumour indices or type were not significant in the subset of rats sampled for the present study, HAMSB reduced the incidence and number of large-bowel tumours/rat in the main study(Reference Clarke, Topping and Bird11). Based on these findings, it is unlikely that factors arising from differences in tumour burden (for instance, amount of luminal serosanguineous exudate produced by ulcerated colonic tumours) were responsible for the differences in faecal microbiota seen in rats in this present study.
Dietary HAMSB sponsored the dominance of a bacterial phylotype related to L. gasseri, an organism that may have potential as a probiotic(Reference Branning, Hakansson and Ahrne25). The HAMSB microbiota was also characterised by a phylotype related to Phascolarctobacterium faecium, which is an anaerobe that produces propionic acid from succinate but is not known to ferment carbohydrates(Reference Deldot, Osawa and Stackebrandt26). The prevalence of a phylotype related to P. faecium may be a result of cross-feeding of succinic acid produced by P. distasonis-related phylotypes, which also predominated in the faeces of HAMSB rats. P. distasonis-related phylotypes were only detected in faeces from rats fed the HAMSB diet, suggesting that these bacteria may be readily adaptable and capable of deacylating and utilising the starch backbone for growth. P. distasonis has been found to predominate in the faeces of volunteers consuming HAMSB in two independent clinical trials (JM Clarke, unpublished results). However, in vitro experiments need to be conducted to verify the specific role of P. distasonis in HAMSB hydrolysis and fermentation. The increase in distal colonic digesta butyrate concentration in the HAMSB group is likely to be the result of the release of esterified butyrate, rather than fermentative in origin.
The rats consuming HAMS differed from the other diets due to the predominance of bands grouping with R. bromii, and propionogenic bacteria including a clade of uncultivated members of the Bacteroidetes. The former is an organism known to be involved in amylolysis, requiring NH4+ as a nitrogen source and involved in assimilatory sulphate reduction(Reference Herbeck and Bryant27). Indeed, R. bromii and related phylotypes have been shown to be one of the major colonisers of starch particles during in vitro fermentations(Reference Leitch, Walker and Duncan28), while phylotypes closely related to this group have been shown to increase in dominance in response to high-RS diets in healthy humans(Reference Abell, Cooke and Bennett14).
The increased propionate concentration measured in samples from the HAMS diet was unexpected as dietary RS is usually reported to increase butyrate production, although propionate can be produced in preference to butyrate(Reference Topping and Clifton2). Digesta pH may be a determining factor, as increased large-bowel digesta propionate concentrations have been reported in rats fed a diet containing 10 % HAMS. However, in rats fed 20 % HAMS, the digesta pH fell to 6·5, and butyrate was the dominant SCFA produced(Reference Herbeck and Bryant27). The distal colonic digesta pH in rats in the present study were higher than 6·5, and the present results suggest that the microbiota of the HAMS group was dominated by propionogenic organisms, typically members of the Bacteriodes/Prevotella rather than butyrogenic bacteria.
At the commencement of the trial, there were no significant differences in rat faecal microbial populations between cages and future dietary groups. The greater degree of population variability in the low-RS diets suggests that these diets have a less stringent effect on the composition of the dominant colonic microbiota than the HAMS and HAMSB diets. The lack of significant difference in community composition between the two LAMS-based diets at most sampling points during the intervention was expected, as the tributyrin in the LAMST diet would be absorbed in the small intestine(Reference Egorin, Yuan and Sentz29) and would not affect colonic flora. The difference at week 7 following AOM treatment was unexpected but may have been the result of the AOM, which could have affected the microbial population. Differences in diversity between the LAMS- and HAMS-based diets are assumed to be a result of the high levels of RS in the two HAMS-based diets. It is unclear whether the differences between HAMS and HAMSB reflect differing ability of bacteria to deacylate HAMSB and hence access the starch as a substrate, or are a direct result of the released butyrate. The diversity findings were supported by Shannon index analysis suggesting that faecal bacteria populations capable of quickly utilising the new substrates became predominant, decreasing bacterial diversity in response to a strong selection for bacteria capable of degrading the major nutrients in the new diet. However, the microbial populations were robust and diversity was restored, as complex relationships were established after 4–7 weeks on a constant diet.
In conclusion, supplementation with dietary RS with and without butyrylation leads to changes in faecal microbiota profiles that may be associated with health benefits. The potential health benefits associated with the increased butyrate level observed on the HAMSB diet are well known. However, the cholesterol-lowering benefit of increased propionate levels(Reference Wong, de Souza and Kendall30, Reference Alvaro, Solà and Rosales31) as seen on both RS diets is still being substantiated. This study supports the suggestion that dietary intervention with specific RS may be used to modify the large-bowel microbiota with the potential to improve colonic health.
Acknowledgements
The authors wish to thank National Starch Food Innovation for providing the HAMS and HAMSB. We would also like to thank Dr Ian Saunders (Commonwealth Scientific and Industrial Research Organisation Mathematical and Information Sciences, Urrbrae, SA, Australia) for providing statistical advice, Ms Kerry Nyland for her analytical assistance, Ms Julie Dallimore, Mr Ben Scherer and Ms Jessica Winkler for their assistance with the animal study and Dr Mark Morrison for his critical analysis of the manuscript. No conflict of interest is declared, and the present study was supported by the Commonwealth Scientific and Industrial Research Organisation Preventative Health Flagship. The contribution of the authors is as follows: G. C. J. A., A. L. M. and J. M. C. contributed to the experimental design; G. C. J. A., C. T. C. and J. M. C. performed the data analysis and writing of the manuscript; G. C. J. A. and C. T. C performed the molecular analysis.