Convincing evidence supports that foods rich in phytochemicals and dietary fibre decrease risk of colorectal cancer (CRC)( Reference Perera, Thompson and Wiseman 1 , Reference Li, Niu and Sun 2 , Reference Liu 3 ). With CRC being the third most common cancer in the world with approximately 1·4 million new cases( Reference Ferlay, Soerjomataram and Ervik 4 ), a focus on lifestyle behaviours, especially diet, may be beneficial for cancer prevention. Dietary fibre is comprised of NSP that are found in fruits, vegetables, legumes and whole grains( Reference Perera, Thompson and Wiseman 1 ). Phytochemicals are diverse classes of compounds including, but not limited to antioxidants, polysaccharides and saponins( Reference Li, Niu and Sun 2 ). Whole grains are a rich source of dietary fibre and phytochemicals that have been studied for cancer chemopreventive properties( Reference Slavin 5 , Reference Windey, De Preter and Huys 6 ). Whole grain brown rice and rice bran (RB) contain multiple unique plant bioactive components (i.e. β-sitosterol, γ-oryzanol, vitamin E isoforms, prebiotics and dietary fibres) that have been shown to inhibit CRC growth( Reference Katyama, Yoshimi and Yamada 7 – Reference Henderson, Ollila and Kumar 10 ). For example, consuming 30 % fibre from bran had a chemopreventive effect of reduced adenoma burden in mice, equating to approximately 1·2 g/d per kg RB fibre for human consumption( Reference Norris, Malkar and Horner-Glister 11 ).
Dietary RB elicits chemopreventive actions via multiple mechanisms, such as inhibition of cyclo-oxygenase 2, induced nitric oxide synthase expression, NF-κB signalling, chromatin condensation and caspase 3 activation, among others( Reference Phutthaphadoong, Yamada and Hirata 12 – Reference Sun, Xu and Liu 14 ). RB metabolism by gut microflora may decrease the intestinal pH and increase intestinal SCFA production( Reference Sheflin, Borresen and Wdowik 15 – Reference Komiyama, Andoh and Fujiwara 17 ). During colonic fermentation, there are distinct interactions between host and microbial metabolism, resulting in production of secondary metabolites that originated from RB intake( Reference Schmidt, Goncalves and Prietto 18 ). Modulation of substrates available to gut microbiota from the diet can influence the type and amount of microbial metabolites produced, and was correlated with various health outcomes, including cancer( Reference Hullar, Burnett-Hartman and Lampe 19 ). Despite the established dietary chemopreventive actions of RB, a knowledge gap remains of the profile of RB small molecules that reach the colon and that work in an additive or synergistic manner to promote healthy metabolism and gastrointestinal functions( Reference Ryan, Heuberger and Weir 20 ).
Clinical trials established the feasibility of increased consumption of heat-stabilised RB in people and showed that RB intake can modulate the human stool microbiome( Reference Sheflin, Borresen and Wdowik 15 , Reference Borresen, Brown and Harbison 21 , Reference Sheflin, Borresen and Kirkwood 22 ). These studies helped determine what phytochemicals from RB interact for enhanced microbial metabolism in the colon and excretion in stool. The objective of this study was to integrate gut microbial metabolite profiles and identify metabolic pathway networks for CRC chemoprevention using non-targeted metabolomics. We hypothesised that stool metabolites resulting from both host and microbial metabolism following dietary RB intake represent viable, chemopreventive agents for inhibition of colonic inflammation and CRC recurrence. Results from this study have important dietary implications for prevention of primary and secondary CRC recurrence in adults.
Methods
Study population and stool sample collection
CRC survivors were recruited and enrolled in a parallel randomised controlled dietary intervention trial (NCT01929122) as previously described( Reference Borresen, Brown and Harbison 21 ). In brief, CRC survivors were deemed eligible if they were a healthy adult with a history of CRC, a minimum of 4 months post-cancer treatment, able to follow a dietary intervention for 28 d, had no history of food allergies or major dietary restrictions, and had a BMI between 25 and 35 kg/m2. Participants consumed study-provided foods which included one meal and one snack each day. All study meals and snacks were calorically matched across study groups and previously described as including either 30 g/d of heat-stabilised RB or no additional ingredient (control)( Reference Borresen, Brown and Harbison 21 ). The addition of 30 g RB to the meal and snack provided 397 kJ (95 kcal) and 6 g of total dietary fibre daily. The consumption of one study meal and snack each day ranged between approximately 2636 and 2929 kJ/d (630 and 700 kcal/d), about one-third of total energy content consumed daily. The remaining two-thirds of daily energy content, approximately 5858 kJ (1400 kcal), was not controlled and was analysed accordingly with the participant’s typical dietary intake. To track all foods consumed, participants recorded dietary intakes using 3-d food logs, which included records of the study-provided foods. These food logs were analysed using Nutrition ProTM (Axxya Systems) and nutrition results have been previously reported( Reference Borresen, Brown and Harbison 21 ).
This study was conducted according to the guidelines laid down in the Declaration of Helsinki and all procedures involving human participants were approved by Poudre Valley Hospital/University of Colorado Health – North Institutional Review Board (Protocol no. 10-1038) and Colorado State University Institutional Review Board (Protocol no. 09-1520H). Written informed consent was obtained from all participants. Stratified randomisation was completed based on sex, BMI and energy intake and participants were randomised into either a control diet group or RB diet group. Study-provided meals and snacks were coded to ensure participants remained blinded to their diet group. The study took place at the University of Colorado Health – North Cancer Center in Fort Collins, CO, between August 2010 and December 2014. Participants self-collected stool samples in specimen containers at two time points: baseline and week 4. Stool samples were lyophilised, aliquoted and stored at −80°C before metabolite extraction. The online Supplementary Fig. S1 illustrates the study flow.
Heat-stabilised rice bran for dietary intervention and metabolomics
The US Department of Agriculture-Rice Research Center provided the heat-stabilised RB used in this study. Rice paddy was milled using standard milling processes to isolate bran. Following milling, RB was heat-stabilised at 110°C for 3 min and stored at −20°C. For this study, Neptune and Dixiebell varieties were chosen due to agronomic similarities and incorporated into study meals and snacks and their nutritional composition has been previously described( Reference Borresen, Brown and Harbison 21 , Reference Sheflin, Borresen and Kirkwood 22 ). For RB food metabolome determination, 100 mg of RB was subjected to non-targeted metabolite profiling at Metabolon. RB was extracted with 80 % methanol as previously described, and analysed by GC-MS and ultra-performance liquid chromatography-MS (UPLC-MS/MS) in the positive and negative ionisation mode platforms( Reference Heuberger, Lewis and Chen 23 ).
Stool extractions for metabolomics
Stool samples were extracted for metabolite profile evaluation by Metabolon Inc. as previously described( Reference Brown, Rao and Weir 24 ). Stool samples were extracted using 80 % methanol. The final extract was divided into four aliquots for: UPLC-MS/MS Positive Ion Mode, UPLC-MS/MS Negative Ion Mode, GC-MS and backup.
GC-MS
Samples were derivatised under N2 using bistrimethyl-silyltrifluoroacetamide and separated on a 5 % diphenyl/95 % dimethyl polysiloxane fused silica column (20 m×0·18 mm ID; 0·18 µm film thickness) with He as carrier gas and a temperature ramp from 60 to 340°C in a 17·5 min period. Internal standards amylbenzene, 1-phenylhexane, 1- phenyloctane, 1-phenyldecane, 1-phenyldodecane, hexadecylbenzene, octadecylbenzene, tetradecylbenzene and 2,6-di-tert-butyl-4-methylphenol were added to each sample (250 ng of each standard per sample). Samples were analysed on a Thermo-Finnigan Trace DSQ fastscanning single-quadrupole mass spectrometer using electro impact ionisation and operated at unit mass resolving power. The scan range was from 50 to 750 m/z.
Ultra-performance liquid chromatography-MS
The UPLC-MS/MS portion of the platform was based on a Waters ACQUITY UPLC and a Thermo-Finnigan LTQ MS operated at nominal mass resolution, which consisted of an electrospray ionisation source and linear ion-trap mass analyser. The dried sample extract was reconstituted in acidic or basic UPLC-compatible solvents, each of which contained eleven to thirteen injection standards at fixed concentrations( Reference Evans, DeHaven and Barrett 25 ). One aliquot was analysed using acidic ten positive ion-optimised conditions and the other using basic, negative ion-optimised conditions in two independent injections using separate dedicated columns (UPLC BEH C18-2·1×100 mm, 1·7 µm; Waters). Extracts reconstituted in acidic conditions were gradient eluted using water and methanol containing 0·1 % formic acid, whereas the basic extracts, which also used water/methanol, contained 6·5 mm ammonium bicarbonate. The MS analysis alternated between MS and data-dependent MS/MS scans using dynamic exclusion and the scan range was from 80 to 1000 m/z. Raw data were extracted, peak-identified and quality control processed as previously described( Reference Brown, Rao and Weir 24 ).
Metabolic pathway networks and analysis
To visualise relevant networks of metabolic pathways from stool metabolites, the relative abundances were uploaded into a pathway analysis software and metabolite classification system called Metabolync (portal.metabolon.com). Pathway impact scores were calculated as previously described, and takes into account the number of significantly different compounds in a pathway with all metabolites identified( Reference Brown, Rao and Weir 24 ). A score ≥2 indicates significant alterations of a metabolic pathway when compared with other pathways. Metabolites were visualised via Cytoscape version 2.8.3. Each metabolite was represented by a closed, coloured node extending from a sub-metabolic pathway node (via pathway enrichment score), which connect to a super-metabolic pathway hexagonal node. A node coloured red represents metabolites with significantly higher expression at 4 weeks compared with baseline (P≤0·05). A node coloured dark green represents metabolites with significantly lower expression at 4 weeks compared with baseline. Nodes coloured pink or light green represent metabolites trending towards significance with higher expression at 4 weeks and lower expression at 4 weeks, respectively (0·05<P≤0·1). Node size is proportional to fold change magnitude over the 4-week time period for RB diet group.
Statistical analyses
RB and stool metabolite profiles from CRC survivors were quantified in terms of relative abundance and median scaled to 1 as previously described( Reference Brown, Rao and Weir 24 ). Metabolomics as an endpoint can be difficult to calculate power due to the complexity of metabolomics studies, including the constant expansion of metabolite libraries and the large number of metabolites identified. There are currently no accepted methods for sample size calculations( Reference Nyamundanda, Gormley and Fan 26 , Reference Blaise 27 ). However, a previous trial in seven healthy adults with no history of CRC, with four participants consuming RB found detectable stool metabolome changes( Reference Sheflin, Borresen and Wdowik 15 ). Thus, a sample group of approximately ten is a reasonable sample size to detect significant changes.
Statistics were performed in ArrayStudio (Omicsoft), GraphPad Prism 6.06 and SAS version 9.4. P values of ≤0·05 were statistically significant. A matched-pairs two-way ANOVA with post-test contrasts was completed using the scaled relative abundance of each metabolite to evaluate the impact of dietary RB supplementation over 4 weeks’ time and with regards to changes in the stool metabolome. An estimate of false discovery rate (q value) was calculated to determine false discoveries common to multiple-comparison-based metabolomics studies. Metabolites with q values ≥0·1 were excluded from further analysis( Reference Brown, Rao and Weir 24 ).
Results
A total of nineteen CRC survivors were randomised to a diet intervention group that was either 30 g/d of RB or none (control). In all, twenty-five participants were initially randomised to control (n 10) or RB (n 15). Six participants withdrew from the trial due to non-compliance (n 5) and family emergency (n 1). No adverse events were reported. Participant characteristics at baseline are presented in the online Supplementary Table S1.
Rice bran modulation of stool metabolite profiles and metabolic pathway networks
The principal component analysis of baseline and 4-week time points revealed inter-individual stool metabolome differences and minor clustering of individuals after 4 weeks of RB intake (online Supplementary Fig. S2). Table 1 lists ninety-three significant stool metabolites from CRC survivors consuming RB at 4 weeks compared with baseline. Stool metabolites are presented as the mean fold change or mean fold difference with respect to the scaled relative abundance of each metabolite for all participants. Of the ninety-three stool metabolites changed with RB consumption, thirty-three metabolites increased and sixty metabolites decreased after 4 weeks (Fig. 1(a) and (b)). β-Hydroxyisovaleroylcarnitine and hydroxybutyrlcarnitine showed a 265-fold and 66-fold increase over time, respectively. In addition, hesperidin (37-fold), apigenin (15-fold), narirutin (15-fold), hesperitin (12-fold), enterodiol (11-fold), diosmetin (5-fold), enterolactone (4-fold), eriocitrin (3-fold) and homostachydrine (1·14-fold) were reflective of increased phytochemical consumption and differentially expressed in the stool metabolome.
Stool metabolite profile differences in the rice bran group over time. (a) Stool metabolites that were significantly different between 4-week and baseline time points with fold change values ≥5·00 and (b) stool metabolites that were significantly different between 4-week and baseline time points and had fold change values <5·00. 
, Metabolites with significance and were identified in the rice bran food metabolome (see online Supplementary Table S3).
Stool metabolites distinguishing rice bran consumption at 4 weeks compared to baseline
↑, Metabolites with significantly (P≤0·05) higher expression in rice bran consumers at 4 weeks when compared with rice bran consumers at baseline (metabolite ratio of ≥1·00); ↓, metabolites with significantly (P≤0·05) lower expression in rice bran consumers at 4 weeks when compared with rice bran consumers at baseline (metabolite ratio of <1·00).
* Compounds that have not been officially confirmed based on a standard, but we are confident in its identity.
Metabolites were classified into metabolic pathway networks and analysed by diet group and time. At 4 weeks, a total of eight metabolic pathways were affected by RB consumption (Fig. 2(a)). Pathways included advanced glycation end product (AGE) (pathway impact score of 8·69), steroid metabolism (5·04), primary bile acid metabolism (2·9), leucine, isoleucine and valine metabolism (2·79), methionine, cysteine, S-adenosyl methionine and taurine metabolism (2·29), inositol metabolism (2·17), vitamin B6 metabolism (2·17) and benzoate metabolism (2·0) (Fig. 2(a)).
Pathway set enrichment scores for stool metabolome. (a) Metabolic pathways in rice bran over time that were significantly different between 4-week and baseline time points and (b) rice bran diet compared with control diet at 4 weeks. Pathway set enrichment scores were determined as defined in the ‘Methods’ section, and are presented for pathways with scores≥2. SAM, S-adenosyl methionine.
Visual metabolite pathway classification and nodal networks for rice bran consumers
Fig. 3 and 4 display Cytoscape pathway classification nodal network views of the stool metabolic pathways modulated by RB consumption at 4 weeks compared with baseline levels. This visual representation includes metabolites from lipid, cofactor and vitamins metabolic pathways in RB consumers (Fig. 3(a)–(d)). Carbohydrates and amino acid pathways are shown in Fig. 4(a)–(d). Network views confirmed the ninety-three metabolites significantly modulated by RB intake and the suite of metabolites trending towards significance. Fig. 3(a) shows the steroid metabolic pathway which was significantly affected after RB consumption with sixteen metabolites significantly decreased. An increase in 4-andorsten-3α, 17α-diol monosulfate and 4-andorsten-3β, 17β-diol monosulfate were observed. Fig. 3(b) shows primary bile acid metabolism, whereby glycochenodeoxycholate and taurocholate were significantly decreased. Fig. 3(c) displays inositol metabolism whereby chiro-inositol had a 26-fold increase after RB consumption. Fig. 3(d) showed significant increases in γ-carboxyethyl-hydroxychroman and pyridoxine (vitamin B6), which are metabolites in the tocopherol and vitamin B6 metabolic pathways, respectively.
Stool metabolomic pathway network analysis of lipid, cofactor and vitamin metabolites significantly different in rice bran consumers at 4 weeks compared with baseline. Pathway specific network views (Cytoscape) are presented for (a) steroid metabolism, (b) primary bile acid metabolism, (c) inositol metabolism, (d) cofactors and vitamin metabolism. Each metabolite is represented as a node, extending from a central metabolic pathway (with pathway impact score). A pathway set enrichment score ≥2 implies that pathway contains more metabolites of statistical significance relative to the entire study. The size of a node is proportional to the fold change between baseline and 4-week time points. A node coloured red represents metabolites with significantly (P≤0·05) higher expression at 4 weeks compared with baseline. A node coloured dark green represents metabolites with significantly lower expression at 4 weeks compared with baseline. Nodes coloured pink or light green represent metabolites trending towards significance (0·05<P≤0·1) with higher expression at 4 weeks and lower expression at 4 weeks, respectively. CEHC, carboxyethyl-hydroxychroman.
Stool metabolomic pathway network analysis of carbohydrate and amino acid metabolites significantly different in rice bran consumers at 4 weeks compared with baseline. Pathway specific network views (Cytoscape) are presented for (a) advanced glycation end product metabolism, (b) leucine, isoleucine and valine metabolism, (c) methionine, cysteine, S-adenosyl methionine (SAM) and taurine metabolism, (d) urea cycle; arginine and proline metabolism. Each metabolite is represented as a node, extending from a central metabolic pathway (with pathway impact score). A pathway set enrichment score ≥2 implies that pathway contains more metabolites of statistical significance relative to the entire study. The size of a node is proportional to the fold change between baseline and 4-week time points. A node coloured red represents metabolites with significantly (P≤0·05) higher expression at 4 weeks compared with baseline. A node coloured dark green represents metabolites with significantly lower expression at 4 weeks compared with baseline. Nodes coloured pink or light green represent metabolites trending towards significance (0·05<P≤0·1) with higher expression at 4 weeks and lower expression at 4 weeks, respectively.
Fig. 4(a) showed significantly perturbed AGE metabolism with N6-caboxymethyllysine decreasing 2-fold compared with baseline for the RB group. Fig. 4(b) displays the nine metabolites from the leucine, isoleucine and valine metabolic pathway with significant change in RB consumers compared with baseline (β-hydroxyisovaleroylcarnitine, β-hydroxyisovalerate, 4-methyl-2-oxopentanoate, 3-methyl-2-oxobutyrate, 3-methyl-2-oxovalerate, allo-isoleucine, ethylmalonate, isovalerylglycine and N-acetylleucine). Using this pathway visualisation tool, N-acetylvaline (0·64–fold) and N-acetylisoleucine (0·54-fold) had decreased expression and tiglyl-carnitine (C5:1-DC) (41-fold) increased at 4 weeks, and was trending towards significance compared with baseline in RB consumers. Fig. 4(c) shows the methionine, cysteine, S-adenosyl methionine and taurine metabolic pathways with 4-methylthio-2-oxobutanoate, N-acetylmethionine sulfoxide and N-acetylmethionine significantly decreased and cystathionine and N-formylmethionine increased. Fig. 4(d) displays the urea cycle; arginine and proline metabolism pathways and significantly decreased homocitrulline, N-methylproline and citrulline metabolites.
Metabolite profile and pathway distinctions between control and rice bran consumers
Compared with control diet, thirty-nine stool metabolites (sixteen of which had higher expression in RB consumers) were significantly different in the RB group after 4 weeks of consumption. This includes 28-fold and 14·5-fold increased expression of the phytochemicals hesperidin and narirutin, respectively, in the stool of RB consumers compared with control diet. Table 2 features nineteen stool metabolites that were also increased in the RB 4 weeks/RB baseline statistical comparison (Table 1). These stool metabolites included lower expressed amino acid metabolites S-methylglutathione (3-fold), formimoglutamate (5-fold), 4-methyl-2-oxopentanoate (3-fold), 3-methyl-2-oxobutyrate (3-fold), 3-methyl-2-oxovalerate (3-fold), allo-isoleucine (3-fold), N-acetylleucine (2-fold), 4-methylthio-2-oxobutanoate (5-fold), 4-hydroxyphenylpyruvate (3-fold) and phenylpyruvate (3-fold), and the phytochemicals hesperidin (37-fold), narirutin (15-fold), diosmetin (5-fold), eriocitrin (2-fold) and solanidine (33-fold) than was observed compared with control diet consumers after 4 weeks of RB dietary intervention and over time. In addition, we observed higher amino acid metabolites expressed compared with control diet after 4 weeks and over time in RB consumers, such as xanthurenate (4-fold), the vitamin metabolite pyridoxine (vitamin B6; 3-fold), 2-hydroxyhippurate (salicylurate; 9-fold) and diosmin (7-fold).
Stool metabolites distinguishing rice bran (n 9) and control diet (n 10) groups at 4 weeks
↑, Metabolites with significantly (P≤0·05) higher expression in rice bran consumers at 4 weeks when compared with control at 4 weeks (metabolite ratio of ≥1·00); ↓, metabolites with significantly (P≤0·05) lower expression in rice bran consumers at 4 weeks when compared with control at 4 weeks (metabolite ratio of <1·00).
There was modulation of eight metabolic pathways between RB and control diet consumers at 4 weeks. Fig. 2(b) shows fatty acid metabolism (pathway impact score of 10·36), leucine, isoleucine and valine metabolism (5·18), vitamin B6 metabolism (5·18), aminosugar metabolism (4·6), drug metabolism (3·55), medium-chain fatty acid (2·3) and glycine, serine and threonine metabolism (2·07).
There were forty-one metabolites with significant stool metabolome differences between CRC survivors consuming a control diet at 4 weeks compared with baseline. These control diet modifications in stool are listed in the online Supplementary Table S2. Increased 3-(2-hyroxyphenyl)propionate, salicylate, apigenin and enterolactone and decreased abscisate were observed among both diet groups at 4 weeks. N6-carboxymethyllysine showed increased relative abundance after 4 weeks of consumption in control diet consumers, and a decreased relative abundance in RB consumers (Table 1 and online Supplementary Table S2).
Rice bran food metabolome
The RB food metabolome demonstrates extensive phytochemical diversity and provides an opportunity to look for cross-over with the stool metabolome, following consumption. The RB food metabolome was analysed for comparison with stool metabolite profiles. There were 453 metabolites identified from RB that were classified into seventy-four metabolic pathways. The online Supplementary Table S3 lists the RB metabolites and the metabolic pathway classifications identified in the stool of the RB group. Additional details on the RB metabolome can be found in Zarei et al.( Reference Zarei, Brown and Nealon 28 ) .
Overlap between the stool and rice bran food metabolomes
The thirty-nine significantly changed stool metabolites from RB consumers that were also present in the RB food metabolome are shown in Fig. 1(a) and (b). These included N6-carboxymethyllysine, an AGE, which decreased by 2-fold in RB consumers’ stool over 4 weeks. Apigenin (15-fold increased) and abscisate (2-fold decreased) were also present in the RB food metabolome, as well as the inositol metabolite chiro-inositol, which increased 26-fold over time in the RB stool metabolome. Xanthurenate, a tryptophan metabolite, increased 4-fold in stool after RB consumption and was present in the RB. Pyridoxine was another metabolite involved in vitamin B6 metabolism that increased in stool and was present in RB. Fatty acids azelate (nonanedioate) (2-fold decreased), hexadecanedioate (2-fold increased), sebacate (decanedioate) (2-fold decreased), 2-hydroxyoctanoate (4-fold decreased) and α-hydroxycaproate (6-fold decreased) also had dual significance for presence in the RB food metabolome (Fig. 1). Furthermore, 2-hydroxyhippurate (salicylurate) was expressed in the RB food metabolome, as well as after 4 weeks of RB consumption and when compared with the control diet.
Discussion
The primary objective of this study was to understand the stool metabolite profile differences between CRC survivors consuming a diet supplemented with RB for 4 weeks compared with a control group. These findings revealed that the consumption of RB leads to significant up-regulation of molecules from the carbohydrate, lipid, amino acid and cofactors and vitamin metabolic pathways in CRC survivors (Fig. 2–4). The ninety-three stool metabolites and eight metabolic pathways analysed within the context of the RB food metabolome may have functional potential for dietary chemoprevention and inform nutritional intakes in CRC survivors. Furthermore, multiple stool metabolites that originate from consumption of RB phytochemicals were identified.
Two stool metabolites with the highest degree of fold change between baseline and 4-week time points in RB consumers were β-hydroxyisovaleroylcarnitine and β-hydroxyisovalerate (a 265-fold and 18-fold, increase after consumption, respectively). Currently, no data have been published on β-hydroxyisovaleroylcarnitine and cancer, including CRC and limited data exists for β-hydroxyisovalerate. β-Hydroxyisovalerate has been used as a biomarker for lung cancer in a urinary metabolomics study( Reference Silva, Cavaco and Perestrelo 29 ). This compound is closely related to biotin-mediated metabolism and has been detected in plasma metabolome of CRC patients( Reference Fukui and Itoh 30 , Reference Mock, Malik and Stumbo 31 ). Dual positive detection herein from the stool and RB food metabolome (online Supplementary Table S3) merits continued investigation. The large differences further suggest the need to evaluate the functional potential of RB modifications to leucine, isoleucine and valine in metabolism for enhanced colonic health.
Pyridoxine (vitamin B6) increased 3-fold in the stool at 4 weeks in RB consumers compared with baseline (Table 1), as well as compared with control at end of study (Table 2) and was present in the RB food metabolome (online Supplementary Table S3). Dietary nutritional intake analysis via weekly food logs also showed an increase in vitamin B6 before analysing the stool metabolome( Reference Borresen, Brown and Harbison 21 ). Vitamin B6 is a water-soluble vitamin present in the human body as pyridoxine, pyridoxal and pyridoxamine and serves as a coenzyme in folate metabolism, which has associations with CRC carcinogenesis( Reference Clayton 32 ). Vitamin B6 has potential to influence CRC risk in various ways, including DNA synthesis and methylation and reduction of inflammation, cell proliferation and oxidative stress( Reference Zhang, Ma and Smith-Warner 33 ). Recommended intake is 1·7 mg/d for men and 1·5 mg/d for women, with 30 g of RB providing 1·2 mg/d( Reference Borresen, Brown and Harbison 21 ). Furthermore, dose-dependent associations were found between CRC risk and dietary vitamin B6 intake( Reference Theodoratou, Farrington and Tenesa 34 ). Moreover, xanthurenate, a tryptophan metabolite tied to vitamin B6 maintenance( 35 ), was significantly increased over time in RB consumers and when compared with control diet group. Based on the results and what is known about pyridoxine in CRC literature, we postulate that vitamin B6 represents a diet modifiable biomarker of RB intake in CRC survivors and should be confirmed with a targeted investigation in a larger cohort.
Phytochemicals related to citrus intake include hersperidin, eriocitrin and narirutin were significantly increased over time with RB consumption. Hesperidin has shown to alleviate oxidative stress and down-regulate proliferative inflammatory markers in azoxymethane-induced colon carcinogenesis in mice( Reference Saiprasad, Chitra and Manikandan 36 ). Eriocitrin has shown significance in multiple analyses and has anti-oxidative properties in vitro ( Reference Miyake, Yamamoto and Morimitsu 37 ). Narirutin, has evidence for anti-oxidative( Reference Tripoli, La Guardia and Giammanco 38 ), anti-inflammatory( Reference Funaguchi, Ohno and La 39 ) and anti-allergic activities( Reference Park, Ha and Eom 40 ). These three compounds were increased with RB diet consumption in our study and may be related to RB metabolism changes by the host. Solanidine, a phytochemical of importance to CRC survivors, is a steroidal alkaloid that has anticancer activities against human colon HCT-8 and HT-29 cancer cells( Reference Jiang, Chen and Cheng 41 ). The phytochemical diosmetin and a closely related microbiota metabolism-driven derivative, diosmin, have shown oxidative activity against cancer cells (Colon205), inducing apoptosis in Colo320 CRC cells, as well as reducing the multiplicity of CRC neoplasms in azoxymethane-treated mice( Reference Xie, Yuan and Yang 42 , Reference Zheng, Hirose and Yoshimi 43 ). Phytochemical modulation and host metabolites in stool following RB consumption represents a novel chemopreventive mechanism that should be further explored.
N6-carboxymethyllysine is an AGE which are a group of heterogeneous macromolecules that are formed by the non-enzymatic glycation of proteins, lipids and nucleic acids( Reference Fu, Requena and Jenkins 44 ). There are two main sources of AGE in humans: food and endogenously produced exposures( Reference Uribarri, Woodruff and Goodman 45 ). AGE are typically generated in higher rates in diabetics due to altered glucose metabolism( Reference Ahmed 46 ). In our study, N6-carboxymethlysine was decreased with RB consumption over time and increased in control diet consumers at 4 weeks, which is consistent with rice bran and whole grain brown rice having the ability to lower glucose and serum lipids in Type I and II diabetics( Reference Kim, Kim and Lee 47 – Reference C Chou, Ma and Cheng 49 ). Another stool metabolite of importance to diabetes control and CRC prevention is 2-hydroxyhippurate (salicylurate). This compound increased by 9-fold following RB consumption after 4 weeks, and had an increased fold change of 3·75 compared with control at 4 weeks (Tables 1 and 2). 2-hydroxyhippurate (salicylurate) is a metabolite of aspirin, and has not only shown the ability to prevent CVD complications in diabetes mellitus patients, but also has demonstrated high efficacy in reducing incidence of colorectal neoplasias( Reference Levy 50 – Reference Pignone, Alberts and Colwell 52 ). Moreover, faecal 2-hydroxyhippurate (salicylurate) levels were associated with a decreased risk of CRC in a prior faecal metabolomics investigation( Reference Goedert, Sampson and Moore 53 ). Therefore, we postulate that decreased AGE and increased 2-hydroxyhippurate (salicylurate) levels in stool due to RB consumption represent potential mechanisms for chemoprevention. This dietary compound has relevance as candidate markers of glucose, lipid and xenobiotic metabolism changes.
The individual metabolites, as well as the network of metabolic pathways that were modulated by RB consumption, identified primary bile acid metabolism with a pathway impact score of 2·9. Primary bile acids included glycochenodeoxycholate and taurocholate that were decreased nearly 4-fold in stool after 4 weeks of RB consumption. In addition, there was a 5-fold decreased expression of glycodeoxycholate, a secondary bile acid. Primary bile acids are synthesised by the human liver and undergo transformation to secondary bile acids by the microbiota( Reference Sayin, Wahlstrom and Felin 54 ). Secondary bile acids have associations with reactive O2/N species generation and oxidative DNA damage( Reference Payne, Bernstein and Dvorak 55 ). The reduction of bile acid metabolites (primary and secondary) in the stool metabolome of CRC survivors is consistent with healthy adults( Reference Sheflin, Borresen and Wdowik 15 ) and was identified as a novel mechanism for RB chemoprevention that should be further explored.
Nutrition-based therapies have proven effective for improving long-term prognosis in CRC patients( Reference Ravasco, Monteiro-Grillo and Camila 56 , Reference Ravasco, Monteiro-Grillo and Vidal 57 ). Increasing dietary fibre has been inversely correlated with CRC risk( Reference Dahm, Keogh and Lentjes 58 ), and RB represents a rich source of dietary fibre that can be successfully incorporated into the diet( Reference Sheflin, Borresen and Wdowik 15 , Reference Borresen, Brown and Harbison 21 , Reference Sheflin, Borresen and Kirkwood 22 ). This unique metabolome analytical approach was to dually investigate the stool metabolite profiles and RB food metabolome, where the online Supplementary Table S3 illustrates the thirty-nine metabolites identified in both the RB metabolome and stool metabolome of RB consumers. β-Hydroxyisovalerate and pyridoxine represent two metabolites of potential dietary biomarkers of RB intake and identified in stool. Foodomics is a branch of metabolomics aimed at quantifying metabolite intakes and to correlate these exposures with biomarkers and health benefits( Reference Ibanez, Valdes and Garcia-Canas 59 ). The combined foodomics and stool metabolomics analytical approach with CRC survivors led to the identification metabolic pathway networks associated with dietary chemoprevention. Major strengths of this study were the confident metabolite identifications from a non-targeted stool and RB food metabolomics analysis. Study limitations included low sample size and high inter-group variation across individuals. Based on these results, we postulate that urine and blood matrices merit a similar analytical integration to determine specific, sensitive biomarkers of food intake and biological accessibility. The compelling findings from this subset of individuals included in this investigation provide rationale for RB consumption in larger studies and over longer periods of time to further analyse dietary biomarkers of intake, as well as to identify biomarkers of CRC recurrence and prevention, particularly in the colon tissue. Our findings demonstrate that the integration of food and stool metabolomes is useful for understanding dietary chemopreventive mechanisms for CRC prevention and control.
Acknowledgements
The authors would like to acknowledge the University of Colorado Health – North Oncology Research group, specifically Joann Lovins, Erica Dickson, Joanne O’Malia and Marlon Bazan, for helping with regulatory approvals, assistance with study recruitment, implementation, consents and coordination of sample collection. The authors thank Dr Anna McClung from the US Department of Agriculture for supplying the study intervention heat-stabilised rice bran; Dr Tiffany Weir for assistance in study design, food preparation and stool collection; and Iman Zarei for helping with proofreading. Katie Schmitz, Genevieve Forster, Allie Reava, Kerry Gundlach and Brianna Nervig provided technical support with study food preparations and nutrition analysis. Finally, the authors wish to thank all the study participants for their contribution to the trial.
This work was supported by the National Institutes Health – National Cancer Institute (1R21CA161472), National Institute of Food and Agriculture (2016-67001-24538) and the University of Colorado Cancer Center – Division of Cancer Prevention and Control pilot award programme.
The authors’ responsibilities were as follows – E. P. R. and R. J. B.: designed research and maintained study oversight; E. C. B.: conducted research; D. G. B.: analysed stool and rice bran for metabolites; D. G. B., E. C. B. and E. P. R. wrote paper; E. P. R. had primary responsibility for the final product. All authors read and approved the final manuscript.
None of the authors has any conflicts of interest to declare.
Supplementary material
For supplementary material/s referred to in this article, please visit https://doi.org/10.1017/S0007114517001106
Open access
