Identification of Unique miRNA Biomarkers in Colorectal Adenoma and Carcinoma Using Microarray : Evaluation of Their Putative Role in Disease Progression

1 Center of Excellence in Genomic Medicine Research, Faculty of Applied Medical Sciences, King Abdulaziz University, P.O. Box 80216, Jeddah 21589, Saudi Arabia 2 King Fahd Medical Research Centre, Faculty of Applied Medical Sciences, King Abdulaziz University, P.O. Box 80216, Jeddah 21589, Saudi Arabia 3 Faculty of Life and Physical Sciences, The University of Western Australia (M011), 35 Stirling Highway, Crawley, WA 6009, Australia


Introduction
Colorectal cancer is one of the most common gastrointestinal cancers that shows a rising trend due to the dietary habits and lifestyle modifications.Accumulation of genetic and epigenetic aberrations predisposes the colonic epithelium to undergo gradual transformation with loss of cellular architecture and initiation of a benign adenoma, which subsequently develops into a malignant adenocarcinoma [1].Successful screening methods have enabled early detection and hence reduction in the mortality rate associated with colorectal adenocarcinoma [2].In addition to the oncogenes and the tumor suppressor genes involved in cancer initiation and progression, many other small molecules such as the oncoproteins, antisense RNAs, and microRNAs (miRNAs) are also implicated in cancer and its signaling mechanisms [3].
Among the small molecules involved in cancer progression and metastasis, the microRNAs (miRNAs) which are The above evidence table depicts the authors of the study, high-throughput platform used, data repository accession number, species, source material, and the number of arrays used in each study.[21][22][23][24][25] nucleotides in length belong to the class of noncoding small RNA molecules that act as posttranslational regulators and are known to exert significant influence on most cancers [4][5][6].Although the miRNAs constitute only 1-3% of the human genome, the fact that about 50% of the miRNA genes are located in the cancer associated loci indicates that these molecules are involved in either cancer progression or inhibition [5,7].Many different miRNAs such as let-7, miR-24, miR-143, and miR-192 are involved in direct or indirect regulations of the KRAS or dihydrofolate signaling pathways of colorectal cancer cell proliferation [8][9][10].Additionally, downregulations of miR-143 and miR-195 are reported to target the antiapoptotic Bcl-2 and hence induce apoptosis in colorectal cancer cell lines [11,12].As such the miRNAs are considered to hold great potential as diagnostic and prognostic markers of colorectal cancer [13].
Currently, there are about 2019 unique mature human miRNA sequences as revealed by the miRBase 19 (miRNA database of published miRNA sequences and annotation).Given the increasing numbers of miRNAs and the fact that they have diverse expression patterns and regulations ranging from developmental biology to cancer pathology [14], it becomes a practical difficulty to validate each miRNA in order to understand their biological targets.Though highthroughput technologies such as gene chip facilitate genomewide expression analysis of genes either in normal or diseased states, this leads to generation of enormous data which again pose great difficulty in analysis.As such bioinformatics algorithms based analysis becomes inevitable to select lead candidates and targets which could be then taken up for further validation.Recently, Li et al., 2011 [15], have used partial least square regression analysis approach for predicting miRNA targets in human colon cancer and reported that they were able to detect more miRNA-mRNA targets than a simple correlation based association in colorectal cancer.However, comparisons between numerous studies on miRNA-mRNA associations are necessary for the validation of such technology based predictions before it can be considered as an effective diagnostic/prognostic indicator.
Given the scope of high-throughput technologies and the power of bioinformatics analysis, we have carried out an in-depth in silico analysis of miRNAs associated with colorectal adenoma and carcinoma.Using high-throughput gene expression as well as integrated network and pathway analyses strategies, we have critically dissected the differentially expressed miRNAs and associated gene networks in colorectal adenoma and carcinoma, respectively.We have also identified some of the predictive and therapeutic markers associated with adenoma and carcinoma and their potential to activate the adenoma-carcinoma sequel.

Methodology
Five different microarray data sets with genome-wide and miRNA expression associated with colorectal adenoma and carcinoma were used in this study.All data were obtained from recent studies on homosapiens, and the details are provided in Table 1.

Genome-Wide Transcriptomic and miRNA Analysis
Raw CEL files corresponding to gene expression data GSE15960 [16] were transferred to PARTEK Genomics Suite version 6.5 0 (Partek Inc., MO, USA) and normalized using GCRMA with quantile normalization to correct for variances in distribution patterns and GC nucleotide content.variance (ANOVA) was applied on the whole data set, and differentially expressed gene list was then generated using an FDR (Benjamini Hochberg) of 0.05 with 2-fold change cutoff.

Functional and Pathway Analyses
To define biological networks among the differentially regulated genes in adenoma and carcinoma, pathway analyses were performed using Ingenuity Pathways Analysis software (IPA) (Ingenuity Systems, Redwood City, CA, USA).Differentially expressed mRNAs and miRNAs that were specific for colorectal adenoma and carcinoma was imported into IPA.The association analysis between miRNAs and gene signatures from the list of imported molecules were deduced using IPA knowledgebase.Expression values were plotted against the respective miRNAs and genes to identify the expression patterns, and only those miRNAs and genes that were differentially expressed between adenoma and carcinoma were considered for further IPA analysis.The functional analysis of IPA identified the biological functions and/or diseases that were most significantly altered for the differentially expressed gene set.The canonical pathway analysis identified the pathways that were most significantly induced in adenoma and carcinoma of the colon.The significance of the association between the data set and the canonical pathways was calculated by ratio and/or Fisher's exact test.The genes or gene products are represented as nodes, and the biological relationship between two nodes is represented by a line.Those molecules that are not linked were removed from the IPA analysis.

Results
The expression analysis of miRNAs associated with colorectal adenoma and carcinoma in the present study led to the identification of several miRNAs that were differentially expressed in colorectal adenoma (see Table S1A in Supplementary Material available online at http://dx.doi.org/10.1155/2014/526075)and carcinoma (Table S1B).These miRNAs consisted of both tumor suppressors and oncogenes.The following oncomirs, namely, miR-21, miR-20a, miR-224, and miR-18a, were upregulated in colorectal carcinoma, whereas the following tumor suppressors, namely, miR-29a and miR-29c, miR-145, miR-35, and miR-26a, were downregulated.In contrast, only one of the oncomirs, the oncogenic miR-18a, was upregulated, and there were no known tumor suppressors that were downregulated in colorectal adenoma.Ten unique miRNAs were commonly expressed in both adenoma and carcinoma (Figure 1(a)).Amongst these, both miR-135b and miR-147 showed differential expression patterns, being downregulated in colorectal adenoma and upregulated in carcinoma.The following miRNAs, namely, miR-182, miR-183, miR-196b, miR-224, miR-503, and miR-96, were upregulated while miR-137 and miR-9 were downregulated in both adenoma and carcinoma, respectively.However, both upregulation and downregulation of the individual miRNAs were relatively higher in colorectal carcinoma compared to adenoma (Figure 1(b)).
Among the genes regulated by these 10 commonly expressed miRNAs, the zinc-finger BED domain containing 3 (ZBED3), and solute carrier family 10 (sodium/bile acid cotransporter family), member 3 (SLC10A3) genes showed differential expression patterns, being downregulated in colorectal adenoma and upregulated in carcinoma (Figure 2(a)).In addition, FOXQ1 gene expression was 10-fold higher in colorectal carcinoma compared to colorectal adenoma (Figure 2(b)).
Comparison of 2406 genes in adenoma and 1403 genes in carcinoma revealed unique predictive markers as well as the miRNAs and their regulatory targets associated with adenoma-carcinoma sequel (Figure 3(a)).Screening for the regulatory genes in both adenoma and carcinoma that are controlled by the two differentially expressed miRNAs (miR-135b and miR-147) led to the identification of MiR-33 being uniquely expressed in colon adenoma (Figure 3(A)).
In addition, the Myoglobin (MB) gene was commonly expressed in both colorectal adenoma and carcinoma that also showed differential expression pattern being upregulated in colorectal adenoma and downregulated in carcinoma (Figure 4(a)).Furthermore, the various signalling mechanisms/interaction networks that are associated with the MB gene based on mRNA-miRNA integration analysis were identified (Figure 4(b)).These include the tumor suppressors (TP53), the angiogenic factors (VEGFA), cell proliferation, adhesion, and inflammatory molecules (EGF, VACM1, and IFN).
The network and pathway analyses led to the identification of the cellular and molecular functions associated with miRNAs in both adenoma and carcinoma.The identified miRNAs were primarily associated with various cellular functions including cell proliferation, cell motility, cell cycle, and cell death.The miRNAs that were specific to cellular proliferation were found to be more associated with colon carcinoma (Table S2A), while those miRNAs associated with cell morphology, cell signaling, and interaction were unique to colon adenoma (Table S2B).In general, most of the miRNAs identified in the current study were associated with diseases and disorders that have been previously reported (Table S2C).
The IPA network analysis identified genes such as insulin 1 (Ins1), miR-375, pyruvate dehydrogenase kinase-1(PDK1), and 3-phosphoinositide dependent protein kinase-1(PDPK1) were more represented in colon adenoma (Table S3A), whereas bruton agammaglobulinemia tyrosine kinase (BTK), CASP8 and FADD-like apoptosis regulator (CFLAR), EGF containing fibulin-like extracellular matrix protein 2 (EFEMP2), interleukin 18 (IL18), and miR-346 were the more represented gene networks in colon carcinoma (Table S3B).As Figure 3: (a) We compared 2406 genes in adenoma and 1403 genes in carcinoma to find the regulatory genes controlled by miRNAs that potentially mediate the carcinogenesis in adenoma-carcinoma sequel.(A) miR-33 is unique to colon adenoma stage and is responsible for the suppression of gamma-aminobutyric acid receptor beta-subunit gene (GABRB2).(B) The other miRNA which was differentially expressed was miR-147.We identified the matching partners for miR-147 by comparing the same gene sets used for miR-135b.Interestingly, both miR-147 and miR-135a/miR-135b were not expressed in colon adenoma; however, they are both upregulated in colon carcinoma.(b) Graphical representation of adenoma-carcinoma sequel starting from normal colon and specific miRNAs and gene (oncogenes and tumor suppressors) associated with the sequel.Ω Tumor suppressor, # oncogene, † oncomirs, and * mature miRNA sequence from the opposite arm of the precursor.
expected, the 5 gene networks identified in colon adenoma and the genes that interact with each network were primarily associated with cancer, gastrointestinal disease, carbohydrate metabolism, growth, and cellular proliferation.
In addition, the miRNA-mRNAs association studies revealed distinct gene regulatory signatures in both adenoma and carcinoma.The following tumor suppressors genes, namely, phosphoinositide-3-kinase regulatory subunit 1-alpha (PIK3R1), nibrin (NBN), and neurofibromin 2 (NF2), were downregulated by miRNAs in adenomas (Table 2).Kruppel-like factor 4 (KLF4), tumor necrosis factor, alphainduced protein 3 (TNFAIP3) and sterile alpha motif, and leucine zipper containing kinase AZK (ZAK) are some of the tumor suppressors that were downregulated in colon carcinoma.The list of miRNAs associated with these tumor suppressors in colon carcinoma is given in Table 3.

Discussion
In silico based high-throughput expression analyses are being widely used to decipher the changes at the gene, mRNA, and miRNA levels, and such analyses enable us to understand their molecular interactions and networks in both normal and diseased states.Change in the miRNA expression profile is one of the common features observed in the development of cancer [3,21].In the present study, genome-wide transcriptomic and miRNA analysis led to the identification of ten unique mRNAs that were commonly expressed in colorectal adenoma and carcinoma.In contrast to the other miRNAs that were expressed and mostly upregulated in both adenoma and carcinoma, two miRNAs, namely, miR9 and miR137, were found to be downregulated in both colorectal adenoma and carcinoma.Decreased expression of miR-137 which was primarily due to abnormal hypermethylation was identified to be one of the early events in colon carcinogenesis, and transfection of colon cancer cell lines with miR-137 resulted in inhibition of cell proliferation [22].Similarly, decreased expression of miR-137 was also reported with oral cancer and glioblastoma [23] indicating that miR-137 has tumor suppressor properties.

ISRN Cell Biology
Two of the ten miRNAs, namely, miRNA135b and miRNA147, were differentially expressed, being downregulated in colorectal adenoma and upregulated in carcinoma.Our findings are in line with earlier studies, where miR-147 was downregulated in colon adenoma [24] and miR-135b was upregulated in colon carcinoma [25].Similarly, downregulation of miR-135b was found to be associated with the progression of oral carcinoma [23].Moreover, miR-135 is also reported to target the 39 untranslated regions of adenomatous polyposis coli (APC) gene, a key multifunctional tumor suppressor in sporadic and hereditary colon carcinoma, suppresses its expression [26,27], and induces downstream Wnt signaling [25].Moreover, miR-135b has been identified as a novel biomarker for pancreatic ductal adenocarcinoma by global microRNA expression profiling of microdissected tissues [28].
Our study showed that upregulation of miR-147 and miR-135b in colon carcinoma was involved in the specific downregulation of gamma-aminobutyric acid receptor betasubunit gene (GABRB2).MiR-33 is unique to colon adenoma stage and is responsible for the suppression of GABRB2 gene.GABRB2 which is well known to be involved in transport processes of chloride channel has been previously reported as a colorectal cancer subtype classificatory [29].GABRB2 was also identified as a discriminatory transcript involved in the CRC-Benign versus CRC-Crohn's disease [29].The miR-135b, present in the chromosome position 1q32.1 [24], could also play a role in the downregulation of GABRB2 gene.
ZBED3 gene is a novel axin-binding protein that was shown to be involved in Wnt/beta-catenin signaling modulation [30].Stage specific expression of both ZBED3 and SLC10A3 was observed in our current study and has the potential to be used as a specific biomarker to differentiate adenoma and carcinoma of the colon [31,32].Forkhead box Q1 (FOXQ1) transcription factor has recently been reported to play an important role in the promotion of cancer through the upregulation of several genes that promote tumor growth through angiogenesis, antiapoptotic effects [33], and epithelial-mesenchymal transition (EMT) [34].One of the reasons for the downregulation of ZBED3 and SLC10A3 could be due to the upregulation of miR-224 and miR-96 in adenoma; however, this needs experimental validation (Figure 2(a)).Probably the mechanism of upregulation of these two genes could either be a mutation or polymorphism in the mRNA coding region that could inhibit the binding of the miRNA to these genes [35].In the presence of miR-33, miR-135a/miR-135b becomes inactive.MiRNA-miRNA synergistic network is least investigated though it has been postulated to be a key factor associated with complex diseases [36].Our results indicate a potential involvement of both positive and negative synergetic roles for miRNAs in the progression sequel associated with colon adenoma to carcinoma.
We identified MB gene as the differentially expressed target in colon adenoma and carcinoma under the control of miR-135b.The gene was upregulated in adenoma and downregulated in carcinoma.The expression was correlated with the inverse expression pattern of miR-135b in adenoma and carcinoma.Although MB is expressed chiefly in cardiomyocytes and oxidative skeletal muscle fibres, recent studies identified low level of MB being expressed in various nonmuscle tissues [37].Interestingly, MB gene has been widely implicated in epithelial cancers and given renewed importance in solid tumors [38,39].It was shown in vitro that MB was expressed in hypoxic and oxidative stress conditions associated with epithelial tumors [26].MB is important in both oxygen transport and free radical scavenging, and its expression in human tumor cells promotes differentiation and inhibits metastasis [27].Solid epithelial tumors such as colon carcinoma could take advantage of proteins such as MB to cope with hypoxic conditions and to control the metabolism of reactive oxygen and nitrogen species.Furthermore, our study showed for the first time, based on mRNA-miRNA integration analysis, the enhanced expression of MB in adenoma and drastic downregulation in carcinoma of the colon.
The downregulation of Kruppel-like family of transcription factors (KLF4) in colon cancer, and not in adenoma, was associated with the specific upregulation of known oncomirs such as hsa-miR-107 as well as novel oncomirs identified in our study such as hsa-miR-133b, hsa-miR135b, hsa-miR152, hsa-miR-7, and hsa-miR-25.KLF4 functions as a tumor suppressor in several tissues, including the colon, and its specific knockdown induces epithelial-mesenchymal transition which predisposes to the development of a subset of colorectal cancers involving Wnt/beta-catenin signalling mechanisms.[40].KLF4 also directly inhibits the expression of Bmi1 in colon cancer cells [41].
Phosphatidylinositol 3  -kinase p85alpha regulatory subunit 1 gene (PIK3R1) identified in the current study is also a tumor suppressor, which could potentially be downregulated through miRNA targets and requires validation.Studies on human tumor samples showed increased expression of a coordinately regulated module consisting of PIK3R1 in advanced malignancy [42].The PI3KR1 gene was also found to be an oncogene in human ovarian and colon tumors [43].PIK3R1 gene was also one of the genes upregulated by leptin, and this could be one of the causative factors in changing the response of colon epithelial cells possessing an APC mutation but not normal cells.Furthermore, the genes regulating the Wnt/beta-catenin-mediated pathway including PIK3R1 were upregulated by leptin [44], which was consistent with the progression of colon carcinogenesis.
In addition, Nibrin (NBN) was found to be under the regulation of miRNAs and is associated with immortalization of colon carcinoma [45].NBN was one of the nine genes showing altered expression in both low and high clinical stage colon carcinoma [45].On the other hand, Neurofibromatosis 2 gene (NF2) is a candidate tumor suppressor gene in chromosome 22.q locus [46], and the colon cancers commonly have allelic losses of chromosome 22q.Consequently, NF2 gene was found to be the target of potential miRNAs in colon carcinoma in our study.

Conclusions and Future Directions
In the present study, we have exploited the high-throughput expression analyses strategies to critically delineate the mRNA as well as miRNA expression profiles of adenoma and carcinoma of the colon.Besides, we have identified novel miRNA regulatory networks that regulate the transcription of mRNAs required for the adenoma and carcinoma sequel.It was significant to note that none of the miRNAs and their gene targets identified either in adenoma or carcinoma had overlapped in their expression patterns.This shows that specific miRNAs are expressed in a stage specific manner and regulate their candidate genes in colon adenoma and carcinoma, respectively.These transcriptional networks that regulate genes involved in the molecular functions associated with the differentiation of adenoma into carcinoma provide the predictive markers of colon adenoma-carcinoma sequel and together with RNAa/RNAi strategies to increase the expression of the tumor suppressor genes as well as silencing oncogenes would add tremendously to early detection and management of colon carcinoma.
was supported by the NSTIP strategic technologies program in the Kingdom of Saudi Arabia, Award no.11-MED1550-03, and the authors, therefore, acknowledge with thanks STU, KAU, for technical financial support.

Figure 2 :
Figure 2: (a) Zinc-finger BED domain-containing 3 (ZBED3) and the solute carrier family 10 (sodium/bile acid cotransporter family) and meMBer 3 (SLC10A3) genes that were regulated by the 10 common miRNAs showed differential expression patterns between adenoma and carcinoma, respectively.(b) Similarly, FOXQ1 gene expression was 10-fold higher in colon carcinoma when compared to the adenoma of the colon.(c) The upregulation of miR-224 and miR-96 in colon adenoma could cause the downregulation of both ZBED3 and SLC10A3.(d) Though both miRNAs were upregulated in colon carcinoma, the expression of ZBED3 and SLC10A3 was not influenced by these miRNAs.

Figure 4 :
Figure 4: (a)-(A) We identified MB gene as the differentially expressed target in colon adenoma and carcinoma that was under the control of miR-135b.(B) MB was upregulated in adenoma and downregulated in carcinoma, respectively.(C) The expression of MB was correlated with the inverse expression pattern of miR-135b in adenoma and carcinoma.(b) Myoglobin interaction network in colon carcinoma.The interaction partner of Myoglobin is represented along with the functional role they represent in colorectal cancer.

Table 1 :
Overview of genome-wide transcriptomic and microRNA array data used for the study.

Table 2 :
Differential expression of key miRNAs and their specific gene targets in colon adenoma.Tumor suppressor, # oncogene, † oncomirs, and * mature miRNA sequence from the opposite arm of the precursor. Ω

Table 3 :
Differential expression of key miRNAs and their specific gene targets in colon carcinoma.