Non‐digestible carbohydrates supplementation increases miR‐32 expression in the healthy human colorectal epithelium: A randomized controlled trial

Colorectal cancer (CRC) risk is modulated by diet and there is convincing evidence of reduced risk with higher non‐digestible carbohydrates (NDCs) consumption. Resistant starch (RS), a NDC, positively modulates the expression of oncogenic microRNAs, suggesting that this could be a mechanism through which NDCs protect against CRC. The present study aimed to investigate the effects of supplementation with two NDCs, RS, and polydextrose (PD), on microRNA expression in the macroscopically‐normal human rectal epithelium using samples from the DISC Study, a randomized, double‐blind, placebo‐controlled dietary intervention. We screened 1008 miRNAs in pooled post‐intervention rectal mucosal samples from participants allocated to the double placebo group and those supplemented with both RS and PD. A total of 111 miRNAs were up‐ or down‐regulated by at least twofold in the RS + PD group compared with the control group. From these, eight were selected for quantification in individual participant samples by qPCR, and fold‐change direction was consistent with the array for seven miRNAs. The inconsistency for miR‐133b and the lower fold‐change values observed for the seven miRNAs is probably because qPCR of individual participant samples is a more robust and sensitive method of quantification than the array. miR‐32 expression was increased by approximately threefold (P = 0.033) in the rectal mucosa of participants supplemented with RS + PD compared with placebo. miR‐32 is involved in the regulation of processes such as cell proliferation that are dysregulated in CRC. Furthermore, miR‐32 may affect non‐canonical NF‐κB signaling via regulation of TRAF3 expression and consequently NIK stabilization.

the large bowel to produce the short-chain fatty acids (SCFAs) butyrate (the preferred energy source for colonocytes), propionate and acetate. Butyrate, produced primarily by bacterial species of the Clostridia, Roseburia, and Eubacteria genera, 3 has chemoprotective anti-inflammatory and anti-proliferative properties and regulates apoptosis in both normal and cancerous cells. 4,5 Butyrate has effects on epigenetic mechanisms that regulate gene expression and are important in colorectal tumorigenesis, including its role as a histone deacetylase inhibitor (HDACi) 6 and effects on microRNA (miRNA) expression. 7,8 miRNAs are small, non-coding RNAs that regulate the expression of their target genes at a post-transcriptional level by degrading messenger RNA or blocking translation. Over 1500 miRNAs have been identified in humans and have been predicted to regulate the expression of up to 60% of genes, 9,10 consequently regulating multiple biological processes including cell proliferation, differentiation, and apoptosis. Abnormal miRNA expression is observed in colorectal tissue as well as plasma, urine, and stool samples from CRC patients. 11,12 miRNA expression is modulated by multiple dietary factors; for example the expression of 250, 198, and 99 miRNAs was associated with intakes of carbohydrate, sucrose, and wholegrains, respectively, and some of these miRNAs showed differential expression between normal and tumor tissue. 13 7 Humphreys et al were the first to investigate the effects of RS and butyrate on expression of miR-17-92 cluster members in a human intervention study. 15

| Intervention methodology
The samples utilized for the present study were collected as part of the DISC Study, a double-blind, randomized, placebo-controlled dietary intervention that supplemented healthy participants with RS and/or PD for 50 d in a 2 × 2 factorial design. 16 For the purpose of this study, the following two groups of participants were used: Anthropometric measurements were taken at baseline and postintervention. Participants also completed a food frequency questionnaire and a lifestyle questionnaire that collected information on physical activity and smoking status.

| Study design
We designed a two-phase study to identify miRNAs modulated by NDCs (see Fig. 1). In the screening phase, we pooled equivalent 2.6 | Synthesis of cDNA cDNA was synthesized from 0.8 μg of RNA using the miScript II RT Kit (Qiagen) as described by the manufacturer. The miScript HiSpec Buffer (5×) was used during reverse transcription to enable subsequent quantification of mature miRNAs. The samples were incubated for 60 m at 37°C followed by 5 m at 95°C using the Sensoquest lab cycler (Göttingen, Germany).

| miRNome array
Prior to analyses, cDNA samples were diluted ten times. Expression of 1008 miRNAs plus six quality controls and six reference RNAs were quantified by qPCR using the 96 well plate Human miRNome miScript® miRNA PCR array (Qiagen) as described by the manufacturer (see Fig. 1). miRNAs with Ct values over 35 were classed as undetected. Raw data were normalized against the six reference RNAs and, as subsequent validation of miRNA expression by qPCR was normalized against only two reference RNAs, also against the selected two reference RNAs (RNU-6 and SNORD68) to check that these data were comparable with both normalization methods. Technical performance of the plates was checked using the integrated quality controls included on all the array plates.

| Validation of miRNA expression by quantitative PCR
Quantification of the eight selected miRNAs for validation and two reference RNAs (SNORD68 and RNU-6) in individual samples was performed by qPCR using the Applied Biosystems® StepOnePlus™ system, the miScript SYBR Green PCR Kit and custom miScript primer assays (Qiagen, UK) as described by the manufacturer.
Prior to analysis of the qPCR data, the melt curves were checked for a single peak and a constant threshold value was set for each miRNA separately for all of the samples by taking the average of the set thresholds for each miRNA for every plate. Ct values greater than 35 were classed as undetected and raw data were normalized relative to the geometric mean of RNU-6 and SNORD68.

| Statistical analyses
The data, expressed as adjusted relative copies (2 −ΔCt × 1000, relative to the geometric mean of SNORD68 and RNU-6), were analyzed using Minitab® v.17. Prior to analyses, data were checked for a normal distribution using the Kolmogorov-Smirnov test. Data that were not normally-distributed were transformed appropriately, using log 10 or by taking the square root.
Differences between the two treatment groups were analyzed using the ANOVA General Linear Model (GLM), adjusting for preintervention expression, age, gender, endoscopy procedure, smoking status, and BMI as covariates.

| Participant characteristics
Samples from a total of 29 participants were utilized for this study, comprising 15 Control participants and 14 Double Intervention participants. Participant characteristics are summarized in Table 1.
Mean age of participants in both groups was similar (49 and 51 years for Control and RS + PD, respectively), with the youngest participant being aged 30 and oldest aged 74. With the exception of one Control participant, all participants were Caucasian. In both groups, colonoscopy was the main endoscopy procedure at baseline.
The participants were well-matched for BMI (most were overweight or obese) and smoking behavior (73% and 71% were never or former smokers for Control and RS + PD groups, respectively). The observed similarity in participant characteristics is evidence of the success of the randomization process.

| Human miRNome miScript miRNA PCR array
The expression of 1008 miRNAs was quantified in pooled postintervention samples from participants in the Control and from the RS + PD group using Qiagen's Human miRNome array. All of the array plates passed the quality control checks. A total of 217 miRNAs (Ct values ≥35), equating to 22% of the 1008 quantified miRNAs, were undetected in one or both of the Control and RS + PD samples. The proportions of miRNAs in each Ct value range were comparable for both the Control and Intervention samples ( Fig. 2A).
A cut-off value of twofold difference (up-or down-regulated) in expression, relative to all six reference RNAs, for the miRNAs between the RS + PD group and the Control group was used as an index of potentially differentially-expressed miRNAs. A total of 33 miRNAs were up-regulated ≥2-fold and 78 miRNAs were down-regulated ≥2-fold in the Intervention compared with Control sample (Additional File 2).

| Normalization of PCR array data
Raw data were normalized relative to all six reference RNAs included on the array plates and to the two selected reference RNAs (RNU-6 and SNORD68) used during the validation stage of the study. We checked that both normalization methods were comparable for all 111 miRNAs showing a fold-change ≥ 2 (Fig. 2B). From the 1008 miRNAs for which there were probes on the array, 69 miRNAs showed ≥2-fold difference in expression when normalized relative to only RNU-6 and SNORD-68 compared with 111 for normalization to all six reference RNAs. In all cases where the normalization methods resulted in different fold-change results, fold-changes were very small (between −1.13 and 1.12) and the difference in foldchange between the two normalization methods did not exceed 0.14. These were: miR-135a-5p, miR-517-5p, and miR-26a-1-3p (upregulated genes) and miR-133b, miR-640, miR-1287-5p, and miR-127-3p

| Selection of miRNAs for individual participant analysis
(down-regulated genes). In addition, miR-32 was selected due to its role in colorectal carcinogenesis. 17 3.5 | Quantification of miRNA expression in individual participants by qPCR qPCR was run using samples from each individual participant in the Control group (n = 15) and in the RS + PD group (n = 14) for the   Table 2). There were no statistically significant differences in the expression of miR-26a, miR-127, miR-133b, miR-135a, miR-517, miR-640, or miR-1287 between the Control and RS + PD groups.

| Comparison of PCR array and qPCR of individual participant data
Quantification of the eight selected miRNAs in individual samples allowed for comparison with the originally obtained PCR array data.
For the PCR array data, fold-change was calculated for each miRNA by dividing the relative copies (2 −ΔCt ) for the pooled RS + PD group sample by that for the pooled Control group sample. For the individual  participant analysis stage, the means of the relative copies for each treatment group were calculated and from this the fold-change was estimated. The resulting data are summarized in Table 3. The direction of fold-change were the same for seven out of eight miRNAs, the exception being miR-133b, and we observed some differences in the degree of change, for example the magnitude of fold-change in miR-135a for individual participant data was half that observed in the screening array (1.49 vs 3.19).

| DISCUSSION
miRNAs regulate the expression of thousands of genes by inhibiting their expression and, consequently, have effects on many biological processes including those involved in tumorigenesis. 7,15,18 Abnormal expression of miRNAs is implicated in the pathogenesis of CRC, where it is associated with disease progression, treatment response, and prognosis. 19,20 miRNAs have potential to be used as cancer biomarkers since altered expression is observed not only in colorectal tissue but also in the plasma, stool, and urine from CRC patients. 11,12 However, little is known about the potential of miRNAs as early markers of CRC risk and how dietary factors may modulate these markers.
In the current study, we used a multi-step approach to identify miRNAs in the colorectal mucosa of healthy individuals that are potentially regulated by two NDCs (RS and PD). Using a whole miRNome PCR array screen, we first compared the expression of 1008 miRNAs in pooled samples from rectal mucosal biopsies from participants who were randomized to RS + PD or to Control for 50 d.
The use of pooling techniques has been applied widely in "omics" studies including those analyzing miRNA expression [21][22][23]   Although, we did not observe significant effects of NDCs on expression of the additional seven miRNAs quantified, further investigation of possible effects of NDCs and other dietary factors on these miRNA in larger studies is warranted. We observed considerable inter-individual variation in response to supplementation (evidenced by relatively large SEM values- Table 3) so that despite mean differences of up to 50%, these were not statistically significant.
Furthermore, some of these miRNAs may have potential for development as biomarkers in investigations of the protective or detrimental effects of dietary factors, such as red meat and dietary fibre, on CRC risk. For example, miR-1287 41 and miR-517 42 are upregulated in CRC tissue and miR-133b (which has been proposed to be a tumor suppressor) is downregulated in SW-620 and HT-29 colon cells. 43 In addition, miR-135a has been identified as a biomarker of CRC in serum 44 and was downregulated by butyrate and trichostatin A, another HDACi, in LT97 colon adenoma cells. This effect was associated with reduced cell proliferation, suggesting that miR-135a may be involved in butyrate-mediated inhibition of cell proliferation in CRC cells. 8 This study is the first to use a multi-step approach to identify miRNAs in the colorectal mucosa of healthy individuals that are regulated by RS and PD. Participants in the study had been shown at endoscopy to have a normal colon and rectum and supplementation using a randomized double-blind design ensured that observed effects are likely to be causal. This study has shown that miR-32 expression in the colorectal mucosa of healthy human participants increased substantially after supplementation with RS + PD for 50 d. Given the potential roles of this miRNA in regulation of the cell cycle, inflammation, and interactions with the gut microbiota, further studies are warranted to confirm our findings and to investigate the utility of miR-32 as a diet-responsive biomarker of gut health. The findings from this RCT study, although relatively small, provide valuable evidence of causality which is free from confounding issues associated with observational studies, such as the subjectivity of self-reported dietary data.