Tissues resected operationally were from the First Affiliated Hospital of Guangzhou University of Traditional Chinese Medicine (TCM). All samples were sporadic CRC, and were not administered any radiotherapy or chemotherapy before operation. The diagnosis of CRC were confirmed by at least two pathologists. Samples collected were immediately frozen in liquid nitrogen. Samples obtained from three sites from the same patient were, respectively, marked as no. 1, no. 2 and no. 3, which were ≥10, 5 and ≤2 cm away from the CRC lesions, respectively (as shown in Fig. 1A).

Reagents and instruments used in this study included TRIzol reagent and glycogen without RNaseA (Invitrogen Life Technologies, Carlsbad, CA, USA), RNase inhibitor, reverse transcriptase and 10X RT buffer solution (250 mM Tris-HCl, pH 8.3, 200 mM KCl, 40 mM MgCl₂, 5 mM DTT) (all from Epicentre Biotechnologies, Madison, WI, USA), 2.5 mM dNTP mixed liquids (dATP, dGTP, dCTP and dTTP 2.5 mM, respectively) (HyTest Ltd., Turku, Finland), RT primers (Invitrogen Life Technologies); 2X PCR Master mix (SuperArray Bioscience, Frederick, MD, USA), Primer 5.0 ABI Prism 7900 system (Applied Biosystems Inc., Foster City, CA, USA). TRIzol (Invitrogen) and miRNeasy mini kit (Qiagen, Hilden, Germany), miRCURY™ LNA array (v.18.0), Axon GenePix 4000B microarray scanner, GenePix Pro 6.0 (Axon Instruments, Foster City, CA, USA).

The tissues collected were fixed with 4% paraformaldehyde, and then paraffin-embedded samples were cut into 4 µm slides. After a 10-min and a 3-min treatment of dimethylbenzene and ethanol, respectively, the slices were stained with hematoxylin for 10 min and subsequently 0.5% eosin for 3 min, and then again treated with ethanol. The slices sealed with neutral gum were observed to obtain images using an inverted phase contrast microscope.

The 4 µm thickness sections from the tissue block were selected for IHC staining. Briefly, samples which were fixed with 4% paraformaldehyde for 24 to 48 h were dehydrated, permeabilized and embedded in paraffin. Four micrometers of paraffin-embedded slides were deparaffized by dimethylbenzene for 10 min and incubated with 5% fetal bovine serum at room temperature for 10 min. Antigen retrieval was performed separately using moderate heat-induced antigen retrieval two times in microwave oven. IHC was performed using rabbit anti-human CK18, CD133, vimentin, cyclin D1 (CD1) monoclonal antibodies. Primary antibodies were diluted in antibody diluent (Beyotime, Wuhan, China) to the indicated optimal dilutions of 1:100 for CK18; 1:100 for CD1; 1:50 for vimentin, 1:100 for CD133. The sections were incubated with a biotinylated goat anti-rabbit IgG secondary antibody. The slides were stained with diaminobenzidine (DAB) staining used as the chromogen for optimal time, then the sections were counterstained with hematoxylin. Optical density (OD) values of a positive immune product for each group were obtained by pathological analysis software IPP 6.0.

Total protein from tissue samples was extracted. Seperation of proteins (30 µg) was performed by 10% sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE), and then these were transferred onto PVDF membrane (Millipore Corp., Billerica, MA, USA). After the membrane was incubated with rabbit anti-human DNMT1 antibody (bs-0678R), DNMT3A antibody (bs-0301R) and DNMT3B (bs-0301R) (all from BIOSS, Beijing, China) and β-actin antibody (ab129348; Abcam, San Francisco, CA, USA) at 4°C overnight, secondary antibodies were added to the membranes for further incubation. Detection of these proteins was performed by enhanced chemiluminescence. The quantification of images was measured using AS image analysis system (Kontron AG, Augsburg, Germany). Finally, the absolute values of the protein expression was quantified through the darkness of the DNMT1, DNMT3A, and DNMT3B bands divided by the darkness of β-actin bands.

The 7th generation of miRCURY™ LNA array (v.18.0) (Exiqon, Vedbaek, Denmark) contains 3,100 capture probes, covering all human, mouse and rat miRNAs annotated in miRBase 18.0, as well as all viral miRNAs related to these species. In addition, this array contains capture probes for 25 miRPlus™ human miRNAs. Total RNA was isolated using TRIzol (Invitrogen) and miRNeasy mini kit (Qiagen) according to manufacturer's instructions, which efficiently recovered all RNA species, including miRNAs. RNA quality and quantity was measured by using NanoDrop spectrophotometer (ND-1000; NanoDrop Technologies, Wilmington, DE, USA) and RNA integrity was determined by gel electrophoresis. After RNA isolation from the samples, the miRCURY™ Hy3™/Hy5™ Power labeling kit (Exiqon) was used according to the manufacturer's guideline for miRNA labelling. One microgram of each sample was 3′-end-labeled with Hy3™ fluorescent label, using T4 RNA ligase by the following procedure: RNA in 2.0 µl of water was combined with 1.0 µl of CIP buffer and CIP (Exiqon). The mixture was incubated for 30 min at 37°C, and was terminated by incubation for 5 min at 95°C. Then 3.0 µl of labeling buffer, 1.5 µl of fluorescent label (Hy3™), 2.0 µl of DMSO, 2.0 µl of labeling enzyme were added into the mixture. The labeling reaction was incubated for 1 h at 16°C, and terminated by incubation for 15 min at 65°C.

After stopping the labeling procedure, the Hy3™-labeled samples were hybridized on the miRCURY™ LNA array (v.18.0) (Exiqon) according to array manual. The total 25 µl mixture from Hy3™-labeled samples with 25 µl hybridization buffer were first denatured for 2 min at 95°C, incubated on ice for 2 min and then hybridized to the microarray for 16 to 20 h at 56°C in a 12-Bay Hybridization Systems (Hybridization System-NimbleGen Systems, Inc., Madison, WI, USA), which provides an active mixing action and constant incubation temperature to improve hybridization uniformity and enhance signal. Following hybridization, the slides were achieved, washed several times using Wash buffer kit (Exiqon), and finally dried by centrifugation for 5 min at 400 rpm. Then the slides were scanned using the Axon GenePix 4000B microarray scanner (Axon Instruments). Scanned images were then imported into GenePix Pro 6.0 software (Axon Instruments) for grid alignment and data extraction. Replicated miRNAs were averaged and miRNAs with intensities ≥30 in all samples were chosen for calculating the normalization factor. Expressed data were normalized using the median normalization. After normalization, for quality assessment of miRNA data, box plot, correlation matrix and scatter plot for miRNAs were used, significant differentially expressed miRNAs were identified through volcano plot filtering (fold-change ≥1.5, P-value ≤0.05). Hierarchical clustering was performed using MEV software (v4.6, TIGR). Quality control was by NanoDrop 1000 spectrophotometer and standard denaturing agarose gel electrophoresis.

Total RNA was isolated and abstracted using TRIzol. Random primer 10 µmol/l was added to total RNA (1 µg) from samples, then diethylpyrocarbonate was blended for 10 min, finally, reverse transcription reaction liquids (5X RT reaction buffer 5 µl, dNTP 1 mmol/l, RNase inhibitor 1 U/µl, M-MLV RTase 8 U/µl, diethylpyrocarbonate 4 µl) were added for 1 h to obtain reverse transcription products. SYBR-Green I was used to obtain quantitative RT-PCR according to manufacturer's instructions, relative expression levels were calculated by subtracting the average U6 (internal control) Ct from the Ct of the target gene obtained from the same cDNA and applying the formula 2^−ΔΔCt. All samples were in triplicate. The primers used by Primer 5.0 ABI Prism 7900 system are listed in Table I. A 2X qRT-PCR reaction mix 8 µl (2X Master mix 5 µl, PCR sequence specific primer F 0.5 µl, PCR sequence specific primer R 0.5 µl) was added to holes corresponding to 384-PCR plate, then corresponding 2 µl cDNA was added. Three-step program was used for the PCR reaction, after the reaction, analysis of melting curves was performed.

All data analysis was performed by SPSS 17.0 software. The RT-PCR results are presented as the mean ± SEM. One-way ANOVA analysis was used to accomplish the comparison among three groups, if the data were normal distribution, otherwise, non-parametric test was performed. Significant level α was 0.05, and P-values <0.05 were considered to be statistically significant. All experiments were independently performed a triplicate.

Normal tissue structure, an important element of microenvironment, is responsible for stabilization of internal milieu. Stable tissue microenvironment exactly mediates generation and release of growth signaling responsible for cell proliferation and differentiation to sustain architecture and function of normal tissue (16). Alteration of tissue architecture and homeostasis is an early signal initiating carcinogenesis (17,18). As for large intestine, crypts are structurally fundamental element of epithelial tissues and functionally basic unit for performing the above biological activities. In this context, we explored biological characteristics during colorectal cancerization on the basis of the crypt. To obtain histologically legible stucture, we performed HE staining. As shown in Fig. 1B, regular crypt structure is shown in no. 1, lumens of gland and monolayer cell nuclei in crypt were normal in dimension. However, crypt structure in no. 2 presented irregular shapes showing bigger and zigzag lumens of gland, and elongated cell nuclei with crowded and multilayer arrangement, hence, these crypts are called aberrant crypt foci (ACF) which is consist of a certain number of accumulated aberrant crypts with or without atypical hyperplasia (19). Significantly enlarged crypts in dimension are demonstrated in no. 3 whose closest distance to the CRC lesions is about 2.0 cm, the aberrant crypts revealed that these nuclei were larger than those of normal cells, the arrangement of the cells was more crowded and multilayered, and polarity of cells had disappeared, which suggest that histologically dramatic alteration appears in colorectal mucosa in the process of canceration.

Tumor microenvironment (TME), a contributor of carcinogenesis, includes not only tumor tissues within TME, but also other tissues and stoma cells and large groups of cytokines produced by the cells. Therefore, in addtion to the exploration of histological alteration in colon epithelium, study of several stroma cells within tissue microenvironment was also performed by our group.

Cyclin D1 (CD1), a marker of tumor cells and a mediator of cell cycle, can regulate transformation of G1/S phase. However, excessive cell division can be caused by overexpressed CD1 through regulation of cell cycle resulting in uncontrolled cell growth, leading to formation of neoplasm. Thus, many studies have demonstrated CD1 overexpression in various forms of cancer (20–22). CD133, a member of cytomembrane protein superfamily, is deemed as a marker of tumor stem cells of CRC (23,24). Increasing tumor cells and stem cells during colorectal cancerization can be presented by the expression of CD1 and CD133. Cytoskeleton, a kind of dominant tissue structure, is responsible for supporting cell stucture and accommodating components in cells, and associated with cell behavior, signal pathway, and apoptosis (25), and alteration of cytoskeleton can directly contribute to oncogenesis. Cytokeratin 18 (CK18), a component of epithelial cytoskeleton, served as a specific biomarker for the epithelium (26). Vimentin, a kind of intermediate filament, is also an important epithelial cytoskeleton taking part in various biological functions including cell growth, differentiation, signal transduction, and regulation of interaction between cytoskeleton proteins and cell adhesion molecules (27). In addition, its expression is located in mesenchyme and hence acted as marker of fibroblasts. Therefore, CD133, CD1, CK18, and vimentin were selected in our study.

To analyze the expression level of the four biological makers, we performed IHC staining. In this study, as shown in Fig. 2, we found that the expression of these markers was dysregulated, for instance, CK18 had a high expression in no. 1 whose microenvironment may be normal, namely, its expression was higher than that of no. 2 (P<0.01) and no. 3 (P>0.05), showing alteration of cytokeleton in no. 3, while its corresponding mesenchymal marker vimentin revealed a low expression in no. 1, but its upregulation was found in no. 3 and in mesenchyme of crypts whose microenvironment is likely to be abnormal and tumorigenic responsible for colorectal carcinogenesis, thus leading to high expression of CD133 and CD1 which were higher than that of no. 1 (P<0.05). CD133 expression was mainly in the interior of crypts instead of the mesenchyme. On the contrary, CD1 expression was mainly in mesenchyme between crypts instead of in crypts. Hence, the results demonstrated increased tumor stem cells and their more extensive distribution. Besides, no. 2 represents a kind of transitional tissue which is between normal and cancerous, hence presenting an inconclusive variation.

To examine DNMT expression under different tissue microenvironment, we respectively carried out IHC staining and western blotting for protein quantification. Our results, as demonstrated in Fig. 3A, revealed that expression of DNMT1, DNMT3A, and DNMT3B was mainly found in aberrant tissues (no. 3), and weak staining was observed in normal tissue (no. 1), implying that alteration of DNA methylation existed in the aberrant tissues leading to abnormal DNA methylation within the promoter regions of genes, consequently, inactivated tumor suppressor gene (TSG) or other related genes including CK18 and reactiveted oncogenes including CD1, vimentin, and CD133 in tumor cells are caused, which are considered to be the most well-defined epigenetic signature in cancer (28,29). All negative controls revealed negligible background staining. These DNMT levels were significantly higher compared with those in the no. 1 (P<0.05) shown by OD value (Fig. 3B). The higher the degree of dysplasia, the higher the expression level of DNMTs, DNMT1 expression mainly existed in mesenchyme between crypts, but in no. 2, its expression was more and located in both mesenchyme and crypts. High expressed DNMT3A in no. 2 and no. 3 was mainly in mesenchyme, while high DNMT3B expression was in both mesenchyme and crypts (Fig. 3A). To further acquire protein quantification, western blotting demonstrated by Fig. 3C and D revealed the expression profile of the DNMTs similar to IHC results. Substantial research has shown that DNA methylation appears in the cancerous and para-cancerous tissues (30), these data are in agreement with our present study.

To screen differentially significantly expressed miRNAs in no. 1, no. 2 and no. 3, we performed a miRNA microarray. After volcano plot filtering (Fig. 4), the differentially significant miRNAs were displayed. As shown in Fig. 5, no. 2 vs. no. 1, 13 miRNAs were filtered, among them, 5 miRNAs were upregulated, 8 miRNAs were downregulated. No. 3 vs. no. 1, 45 miRNAs were differently expressed, 13 miRNAs were upregulated, 32 miRNAs were downregulated. No. 3 vs. no. 2, 13 miRNAs were upregulated, 7 miRNAs were downregulated. Previous studies have discovered that the degree of epigenetic difference between cancer cells and normal cells was considerably higher than that of normal cells with different phenotypes and even different germ cells, such as fibroblasts and epithelial cells (2). In this study, we found a cluster of upregulated and downregulated miRNAs in no. 2 and no. 3 tissues compared to no. 1 with normal microenvironment, implying differences of DNA methylation pattern in different tissues. These results reveal that cancer cells are different from normal cells in epigenetics, consistent with the previous studies. In addition, complexity and frequency of epigenetic alterations observed in cancer cells appear to be able to overturn the explanation that epigenetic alterations rely on single event (2), on the contrary, these pathological alterations may be caused by multiple mechanisms. Our results included in total 78 dramatically expressed miRNAs, and many upregulated or downregulated miRNAs also emerged in the same tissues, implying that some co-regulatory mechanisms may exist between different miRNAs or the same miRNAs may be co-regulated by several distinctive mechanisms.

Four main DNMT isoforms are present in mammalian cells. DNMT1, the maintenance methyltransferase, is responsible for the conservation of methyl marks during DNA replication, while, DNMT3A and DNMT3B are contributed mainly to de novo DNA methylation, DNMT1, the catalytically inactive isoform, has a synergy with DNMT3A and DNMT3B (31), although a study appears to disagree with this rigid classification (32). These DNMTs play an important role in DNA methylation pattern of cellular processes, for example, differentiation of stem cells during embryonic development (33,34). However, abnormal DNA methylation profile caused by DNMT mutations or overexpression is likely to be an early epigenetic event during oncogenesis (33). Other epigenetic regulators including miRNAs are also involved in aberrant mediation of DNA methylation profile, and their combination participates in tumor transformation (35–38). Our results found the aberrant changes of DNMTs in the precancerous lesions of CRC, revealing abnormal accumulation of methylated DNA.

To determine the miRNAs targeting DNMTs, we adopted predictive information from the following three databases, mirbase, miranda and targetscan (http://www.ebi.ac.uk/enright-srv/microcosm/htdocs/targets/v5/, http://www.microrna.org/microrna/home.do, http://mir-db.org/miRDB/), and overlapping parts of the three predictive results served as the final result prediction. As a result, three target miRNAs were screened by the database, and other three target miRNAs were obtained through previous literature. Finally, six target miRNAs including hsa-miR-185-5p, hsa-miR-1237, hsa-miR-429, hsa-miR-200a-5p, hsa-miR-497-5p, and hsa-miR-625-3p were selected. These miRNAs target DNMT1, DNMT3A, and DNMT3B (as shown in Table II). Because of better sample clustering after removing c and e (c and e represent number of each sample), hierarchical clustering of six target miRNAs for 3 samples of every location was performed (Fig. 6). As shown in Fig. 6 and Table II, expression level of hsa-miR-200a-5p and hsa-miR-429 in no. 3 was higher than that of no. 1 and no. 2 (P-value of hsa-miR-429 <0.05), indicating that low expressed DNA methylation profile probably participates in the formation of certain oncology mechanisms. For example, global DNA hypomethylation mainly occurs in dinucleotide of cytosine guanine (CpG) within mobility and repetitive genome components, including satellite DNA sequence (pericentromeric regions), repetitive sequence of long interspersed nuclear element (LINE) and retrotransposon (39). Hypomethylation profile results in destruction of normal dinucleotide sequence and damage of chromosome stability through accumulating chromosome damage and inducing the reactivation of previously silenced genetic elements (retrotransposon), leading to the formation of various forms of cancer (40). A study has reported that miR-429 can target phosphatase and tensin homolog (PTEN) in endometrial cancer as a TSG and contribute to oncogenesis (41). Furthermore, in addition to genomic instability resulted from DNA hypomethylation, positive transcriptional regulation of proto-oncogenes is also a stimulative process of tumor initiation (42,43). Recently, studies have found that mutation of TSG adenomatous polyposis coli (APC) is a driving event in initiation of coloretal cancer and seems to control expression and activity of DNMTs, then leading to demethylation of numerous genes and forcing intestinal epithelial cells to maintain a undifferentiated and progenitor-like status (44). Similar to APC mutation, whether the aberrant miRNAs observed in our study are involved in this mechanism needs to be thoroughly investigated in the future.

Furthermore, DNMTs mediated by miRNAs play an important role in hypermethylation of CpG islands. Downregulated hsa-miR-185-5p, hsa-miR-497-5p, hsa-miR-625-3p and hsa-miR-1237 were observed in no. 3 tissue, implying increased level of DNMTs. These results suggest that CpG islands have an inclination to hypermethylation, because CpG islands of colon epithelial cells are not methylated under normal conditions (35,45), therefore, CpG island methylation pattern in tissue cells with inclination to canceration have altered. For example, the expression levels of hsa-miR-625-3p and hsa-miR-1237 had a decreasing tendency from normal tissues to precancerous lesions, although their difference was not statistically significant. While the expression levels of hsa-miR-185-5p and hsa-miR-497-5p of no. 2 location were more than that of no. 1 (hsa-miR-185-5p had no statistical difference), implying that possibly some mechanisms are involved in these changes where miR-185-5p and miR-497-5p of no. 2 tissue were upregulated.