Research Article

Cancer-Predicting Gene Expression Changes in Colonic Mucosa of Western Diet Fed Mlh1+/- Mice

  • Marjaana Pussila equal contributor,

    equal contributor Contributed equally to this work with: Marjaana Pussila, Laura Sarantaus

    Affiliation: Department of Biosciences, University of Helsinki, Helsinki, Finland

  • Laura Sarantaus equal contributor,

    equal contributor Contributed equally to this work with: Marjaana Pussila, Laura Sarantaus

    Affiliation: Department of Biosciences, University of Helsinki, Helsinki, Finland

  • Denis Dermadi Bebek,

    Affiliation: Department of Biosciences, University of Helsinki, Helsinki, Finland

  • Satu Valo,

    Affiliation: Department of Biosciences, University of Helsinki, Helsinki, Finland

  • Nima Reyhani,

    Affiliation: Department of Information and Computer Science, Aalto University, Espoo, Finland

  • Saara Ollila,

    Affiliations: Department of Biosciences, University of Helsinki, Helsinki, Finland, Institute of Biotechnology, University of Helsinki, Helsinki, Finland

  • Essi Päivärinta,

    Affiliation: Department of Food and Environmental Sciences, University of Helsinki, Helsinki, Finland

  • Päivi Peltomäki,

    Affiliation: Department of Medical Genetics, University of Helsinki, Helsinki, Finland

  • Marja Mutanen,

    Affiliation: Department of Food and Environmental Sciences, University of Helsinki, Helsinki, Finland

  • Minna Nyström mail

    Affiliation: Department of Biosciences, University of Helsinki, Helsinki, Finland

  • Published: October 08, 2013
  • DOI: 10.1371/journal.pone.0076865


Colorectal cancer (CRC) is the second most common cause of cancer-related deaths in the Western world and interactions between genetic and environmental factors, including diet, are suggested to play a critical role in its etiology. We conducted a long-term feeding experiment in the mouse to address gene expression and methylation changes arising in histologically normal colonic mucosa as putative cancer-predisposing events available for early detection. The expression of 94 growth-regulatory genes previously linked to human CRC was studied at two time points (5 weeks and 12 months of age) in the heterozygote Mlh1+/- mice, an animal model for human Lynch syndrome (LS), and wild type Mlh1+/+ littermates, fed by either Western-style (WD) or AIN-93G control diet. In mice fed with WD, proximal colon mucosa, the predominant site of cancer formation in LS, exhibited a significant expression decrease in tumor suppressor genes, Dkk1, Hoxd1, Slc5a8, and Socs1, the latter two only in the Mlh1+/- mice. Reduced mRNA expression was accompanied by increased promoter methylation of the respective genes. The strongest expression decrease (7.3 fold) together with a significant increase in its promoter methylation was seen in Dkk1, an antagonist of the canonical Wnt signaling pathway. Furthermore, the inactivation of Dkk1 seems to predispose to neoplasias in the proximal colon. This and the fact that Mlh1 which showed only modest methylation was still expressed in both Mlh1+/- and Mlh1+/+ mice indicate that the expression decreases and the inactivation of Dkk1 in particular is a prominent early marker for colon oncogenesis.


Colorectal cancer (CRC) evolves as a multistep process, which requires a series of genetic and epigenetic alterations. The process is accelerated in individuals with inherited cancer predisposition, and interactions between genetic and environmental factors, including diet, seem to be in key position in its etiology [1,2]. The importance of epigenetic alterations such as DNA methylation changes in the initiation of CRC is now acknowledged [3,4], yet, the earliest events in normal colonic mucosa available for early detection and prevention of cancer development remain to be elucidated.

Although inherited mutations in the tumor suppressor gene (TSG) APC (adenomatous polyposis coli), an important component of the Wnt/β-catenin signaling pathway, and the mismatch repair (MMR) genes (e.g. MLH1, mutL homolog 1), which control the mutation rate in a cell [2] may confer a high lifetime risk of cancer with an early age at onset, colorectal cancer is clearly a disease of increasing age [3,4]. Accordingly, methylation changes of a small subset of TSGs have been detected in the aging colonic mucosa of normal healthy individuals, and this methylation involves genes which often become more substantially methylated in neoplastic cells [5-7], suggesting their role in cancer initiation and progression. Along with aging some exogenous compounds from dietary sources are important modifiers of methylation patterns in the colon [8], probably explaining why Western populations consuming considerable amounts of red meat, saturated fat and sugar, and only moderate amounts of dietary fiber, vitamins and minerals (e.g. calcium, folate, and vitamin D), as well as plant derived nutrients show the highest CRC incidences in the world [1] (World Cancer Research Fund, Epigenetic changes thus provide a potential link between nutrition and cancer [9] and emphasize the need to elucidate dietary effects on gene regulation in intestinal mucosa.

Animal models, particularly rodents, provide a valuable resource in the field of environmental epigenetics including studies on changes mediated through diet [10]. Results suggesting that Western-style diet (WD) induces gastrointestinal tumors in mouse models for familial intestinal cancer and even in wild-type (WT) mice without any carcinogen treatment [11-14] prompted us to study the effects of WD exposures on gene regulation and methylation in normal colon mucosa with and without inherited colon cancer susceptibility. We selected 94 growth-regulatory genes previously linked to human CRC and studied their expression in the heterozygote Mlh1+/- mice analogous to human Lynch syndrome (LS), and WT Mlh1+/+ littermates. During a long-term feeding experiment we used our own modification of Western-style diet (WD*) which highly resembles the New Western Diet (NWD) shown to induce benign and malignant neoplasms in the colon of normal C57Bl/6 mice after an 18 month feeding experiment [12]. The main difference between NWD and WD* is the fat source which in WD* was largely changed from oil to animal (milk) fat and thus resembling more fat consumed by Western populations. Here, proximal colon, the predominant site of cancer formation in LS [15], revealed several significant age-related changes in gene expressions. Some changes were potentiated by WD* and some were found only in mice with genetic cancer predisposition. We further demonstrated that the expression changes detected from the histologically normal mucosa were remarkably early occurring prior to APC inactivation and the second hit in Mlh1.



Heterozygote B6.129-Mlh1tm1Rak mice (Mlh1+/-) (strain 01XA2) were obtained from NCI-MMHCC; National Institutes of Health, Mouse Repository, NCI-Frederick, MD. In B6.129-Mlh1tm1Rak mice, exon 2 is missing in one of the two Mlh1 alleles leading to non-functional Mlh1 protein [16]. Mlh1+/- mice have an increased morbidity compared to their WT littermates and approximately one third develop tumors such as lymphomas as well as tumors of the small and large intestine and a number of other organs during their lifespan [17]. The Mlh1+/- and Mlh1+/+ mice were genotyped using genomic DNA extracted from earmarks according to the protocol published at the Mouse Repository website (​ID=01XA2&p_allele=Mlh1%3Ctm1Rak%3E&prot_​no=1)(See Materials and Methods S1).

Ethics statement

Mice were bred and treated according to the study protocol approved by the national Animal Experiment Board in Finland (ESLH-2008-06502/Ym-23). The mice were humanely euthanized with CO2 inhalation.


At the age of 5 weeks (time point 0, tp0), Mlh1+/- and Mlh1+/+ mice were randomly divided into two dietary groups (n = 8/group, including both sexes) fed with AIN-93G (AIN) control diet [18] or Western-style (WD*) diet (Harlan Teklad, Madison, WI) (Table 1, detailed description of diets and their nutritional compositions are in Table S1).

Fat Sources(g/kg)a(g/kg)b
Soybean Oil70-
Anhydrous Milkfat-133
Canola Oil-55
Sunflower Oil-12
Carbohydrate Sources
Protein Source200 (Casein)232 (Casein, vitamin free)
Total energy content (kcal/g)3.84.6
Kcal from fat (%)17.239.2
Kcal from carbohydrates (%)63.942.3
Kcal from protein (%)18.818.5
Vitamin D (IU/kg)1000100
Folic acid (mg/kg)20.2

Table 1. Diet nutritional information.

For detailed composition of the experimental diets see Table S1.
a If not stated differently
b If not stated differently

Sample preparation

Eight mice per each group at tp0 (Mlh1+/-, Mlh1+/+) and tp1 (12 months of age) (Mlh1+/+ AIN, Mlh1+/- AIN, Mlh1+/+ WD*, Mlh1+/- WD*) 48 mice in total were sacrificed and sampled. Histological studies were carried out at The Finnish Centre for Laboratory Animal Pathology (FCLAP), University of Helsinki, Finland.

For genomic DNA and total RNA extractions, the mucosa (6 x 4 mm) was separated from the underlying submucosa and musculature under a dissecting microscope. Samples for RNA extraction were stored in RNAlater (Qiagen, Valencia, CA) at -80°C.

RNA extraction and reverse transcription

The total RNA samples were prepared using the RNeasy Plus Kit (Qiagen, Valencia, CA) with an extra Dnase treatment (Qiagen, Valencia, CA). The RNA integrity was analyzed with the Agilent 2100 Bioanalyzer (Agilent technologies, Santa Clara, CA) and only high quality RNA (RNA integrity number RIN > 8) was used for cDNA synthesis reactions, which were ran as duplicates and pooled for the RT-qPCR reactions. Reverse transcription from either 200 ng (individual samples for StellARray) or 1600 ng (pools of eight samples, 200 ng each, from eight different mice belonging to each tp1 group for TaqMan RT-qPCR) of total RNA was achieved with M-MuLV RNase H+ (Thermo Scientific, Finland) or Superscript III (Life technologies, Carlsbad, CA), respectively, using random primers according to the manufacturers’ instructions.

RNA expression studies

The RNA expressions of histologically normal tissue samples from proximal colonic mucosa were analyzed by using a quantitative custom made StellARrayTM platform (Lonza Group Ltd, Bar Harbor BioTechnology). The StellARray included 94 genes (73 TSGs) previously associated with development of colorectal cancer and/or documented to exhibit CpG island (CGI) hypermethylation in CRC and other human cancers (Table S2). APC was included in the array as an indicator of ongoing carcinogenesis. Eight mice from each study group were separately analyzed for the expression of the 94 genes of interest (48 RT-qPCR arrays in total).

Each StellARray plate well was loaded with 20 µl of SYBR Green master mix containing 8µl of sample specific cDNA and modified Thermus brockianus DNA polymerase (Thermo Scientific, Finland). The RT-qPCR arrays were run on a Mx3000P cycler (Agilent technologies, Santa Clara, CA) using the following cycling parameters: 50°C for 2 min, 95°C for 7 min, 40 cycles of 95°C for 15 s, and 60°C for 1 min. Fluorescence data was acquired during the 60°C annealing/extension step. The primer specificity was monitored through a melting curve analysis. The expression differences between different mouse groups were analyzed using the Global Pattern Recognition™ (GPR) software (Bar Harbor Biotechnology, Trenton, ME)

The results of statistically significant expression changes in relation to WD* and/or inherited cancer predisposition were validated at tp1 using TaqMan assays (Table S3). In contrast to StellARray RT-qPCR, the TaqMan RT-qPCR analysis was done from pooled samples. Each pooled sample was assayed in triplicate for the target genes, as well as the endogenous reference genes using the following cycling parameters: 1 cycle of 95°C for 10 min, 40 cycles of 95°C for 15 s, and 60°C for 1 min. Thermal cycling and fluorescence data acquisition were performed with a StepOnePlus cycler (Life Technologies, Carlsbad, CA) and Cq values were called using the Data-assist v2.0 software (Life Technologies, Carlsbad, CA). The uniformly expressed reference genes (Hdac1 and Stk4) were selected using the GeNorm algorithm [19] from amongst the 94 genes included in the StellARray. Mlh1 expression levels at tp0 were also quantitated using TaqMan assay.

If the amount of template was too low to give reliable results, the RT-qPCR validation was performed using pre-amplified cDNA. For each tp1 group, pooled samples were multiplex pre-amplified using 16 ng of cDNA with TaqMan PreAmp Master Mix Kit (Life Technologies, Carlsbad, CA) following manufacturer’s instructions. The cycling parameters were as follows: 1 cycle of 95°C for 10 min and 10 pre-amplification cycles of 95°C for 15 s, and 60°C for 4 min.

DNA methylation analysis

For the DNA methylation analysis, only the genes which had shown a significant mRNA expression decrease (i.e. candidate hypermethylated TSGs) in association with inherited predisposition and/or WD* were selected, and the methylation levels of their CpG islands (CGIs) were assessed individually for each tp0 and tp1 mouse. Genomic DNA for the methylation analysis was extracted using DNeasy Blood & Tissue Kit (Qiagen, Valencia, CA) from tissue samples from the immediate vicinity of those used for RNA expression studies.

The quantitative methylation analysis was performed with Sequenom’s MassARRAY EPITYPERTM system (Sequenom GmbH, Hamburg, Germany) which is a bisulphate-based technology relying on base-specific cleavage of RNA and matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF-MS) to determine the relative extent of methylation in DNA fragments containing either one or several subsequent CpG sites, which are referred to as CpG units [20]. In the mass spectrum, a distinct signal pattern results from the methylated and non-methylated target sequence, and the EPITYPER software determines the individual methylation ratios (i.e. signal intensity ratios as percentage of methylated to non-methylated signals) for CpG sites/unit within a target sequence. The system is able to detect methylation levels as low as 5%.

The amplicons were primarily selected to cover CGIs annotated by the UCSC genome browser and to be overlapping with the transcription start site and the 5’ UTR or be in their close proximity. Altogether, twelve different amplicons (lengths between 174 and 499 bp) were analyzed of which eight belonged to the pre-validated Mouse Standard EpiPanel and four were custom DNA methylation assays designed using Sequenom’s EpiDESIGNER software (Sequenom GmbH,, for which primer sequences and target sites are provided in Table S4. In order to reduce methylation variability introduced during PCR [21], three replicate amplifications were performed and pooled for mass analysis. The initial methylation data was filtered by Sequenom to exclude poor quality measurements. CpG units that yielded data in greater than 75% of the samples passed the initial quality control. From these, samples that yielded data in greater than 80% for all CpG units within an amplicon were selected for that sample/amplicon pair. For further analysis, CpG units which had data available for less than 50% of all samples and samples which had data available for less than 50% of all CpG units were excluded. After quality control 78% of the CpG units analyzed (152 out of 196) and all the 48 samples were included in further analysis steps.

Statistical analyses

In StellARray RT-qPCR, a common threshold was set across all the 48 plates within the experiment. The expression differences between different groups were analyzed using the Global Pattern Recognition™ (GPR) software, which executes efficiency correction and reference gene and control group normalizations (​.GPR also takes advantage of biological replicates to extract significant changes in gene expression. The StellARray system does not work with user defined reference genes. Instead, the GPR software algorithm determines the best set of reference genes within the experiment by comparing the expression of all of the genes included in the assay between test and control samples, generates a global pattern of expression changes, and creates a ranked list of significant changes [22]. In this way the selection of reference genes is unbiased since it enables the experimental data to define the stably expressed reference genes.

In TaqMan assays, relative mRNA expression changes were analyzed using the comparative Ct (ΔΔCt) method, which presents the data as fold changes in gene expression normalized to endogenous reference genes and relative to the control group [23]. Data-assist v2.0 software (Life Technologies, Carlsbad, CA) was used for quality control and normalization of the quantification cycle (Cq) data, and a median permutation method [24] was used to determine the significance of the expression fold changes compared to the control group with significance level of P < 0.05.

The correlation between the StellARray expression patterns of 5 week old Mlh1+/+ and Mlh1+/- mice was studied using Pearson correlation analysis (R value) of the PASW Statistics 18 system.

Chipster software [25] was used to visualize and analyze the methylation data. The Non-metric Multi-Dimensional Scaling (NMDS) tool was used to produce two-dimensional maps based on sample dissimilarity calculated using Euclidean distance, and the Dendrogram tool was used to create dendrograms of samples using normalized data with Pearson correlation and average linkage method. For Chipster analyses, missing methylation values of individual CpG units were defined as the median value of the CpG unit within the particular mouse group.

Comparison of the average methylation levels between different mouse groups was performed using Mann-Whitney test of the PASW Statistics 18 system.


Colonic neoplasias develop predominantly in WD* mice

Overall, one proximal adenocarcinoma and five adenomas/hyperplastic polyps were observed in the 32 mice at tp1. The significance of WD* on CRC risk in our mouse series is highlighted by the fact that 5 out of 6 mice with neoplasias were fed with WD* and that WD* caused tumor development also in wild-type mice without inherited predisposition, i.e. the only carcinoma and one adenoma were found in the Mlh1+/+ WD* mice (Table 2, Figure S1). The single AIN fed mouse developing neoplasia (hyperplastic polyp) was a mutation carrier.

Mouse GroupDkk1 expressed (n = 16)Dkk1 not expressed (n = 16)
Mlh1+/+ AIN62
Mlh1+/- AIN3a5
Mlh1+/+ WD*4b4c
Mlh1+/- WD*35a,b,bd

Table 2. Dkk1 inactivation and colonic neoplasias in different mouse groups.

a hyperplastic polyp; b adenoma; c adenocarcinoma; d not histologically confirmed

Mlh1+/- and Mlh1+/+ mice show similar mRNA expression patterns at the outset

At the beginning of the experimental feeding period (tp0), the Mlh1+/- and Mlh1+/+ mice showed a remarkably good correlation (Pearson’s R = 0.989, P < 0.01) between the mRNA expression patterns of the 94 genes included in the custom StellARray RT-qPCR array (Figure S2A). However, distinct from the other genes, but in accordance with the different genotypes, the expression of Mlh1 was approximately 50% lower in the heterozygote mice than in the WT mice (Figure S2B). Overall, an isogenic background in the beginning made it justified to reason that expression differences that would appear between the different mice groups later were acquired and not the outcome of the inherited genetic constitution.

WD* and/or inherited predisposition are associated with differential expression of several genes at tp1

At the age of 12 months (tp1), the expression of several genes was found to be increased or decreased as compared to tp0 (Tables 3 and 4). All GPR results (P-values and fold changes) of expression differences between different mice groups for all the 94 genes are in Table S5. Nine out of the 94 genes showed statistically significant (P < 0.05) expression changes between tp0 and tp1 in all mice groups (Mlh1+/+ AIN, Mlh1+/- AIN, Mlh1+/+ WD*, Mlh1+/- WD*) (Table 3). Decreased mRNA levels were seen in Ccnd1 (Cyclin D1), Cdh1 (Cadherin 1), and Mal (Myelin and lymphocyte protein, T cell differentiation protein), while increased expression was seen in Axin2, Cdx1 (Caudal type homeobox 1), Fzd10 [Frizzled homolog 10 (Drosophila)], Mbd2 (Methyl-CpG binding domain protein 2), Mbd4 (Methyl-CpG binding domain protein 4), and Rasgrf2 (Ras protein-specific guanine nucleotide releasing factor 2). Since these changes did not depend on inherited predisposition or WD* they can be attributable to aging.

Mlh1+/+ AINMlh1+/- AINMlh1+/+ WDMlh1+/- WD
Downregulated GenesFold Change (P)Fold Change (P)Fold Change (P)Fold Change (P)
Ccnd13.3 (0.003)3.8 (0.000)4.3 (0.001)3.6 (0.000)
Cdh15.4 (0.002)4.9 (0.000)5.9 (0.002)8.4 (0.000)
Mal3.4 (0.025)3.4 (0.014)3.2 (0.034)2.7 (0.028)
Upregulated GenesFold Change (P)Fold Change (P)Fold Change (P)Fold Change (P)
Axin22.6 (0.005)3.0 (0.002)1.8 (0.039)1.6 (0.028)
Cdx14.1 (0.002)5.6 (0.000)3.4 (0.002)5.4 (0.000)
Fzd104.8 (0.002)7.6 (0.000)3.4 (0.021)5.0 (0.000)
Mbd23.3 (0.004)4.7 (0.000)3.2 (0.003)5.7 (0.000)
Mbd43.0 (0.005)6.0 (0.000)3.0 (0.004)3.3 (0.000)
Rasgrf22.3 (0.002)1.8 (0.006)2.3 (0.001)1.5 (0.040)

Table 3. Genes showing statistically significant (P < 0.05) mRNA expression differences between tp0 and tp1 in all four mouse groups; in the control group (Mlh1+/+ AIN) and the three study groups (Mlh1+/- AIN, Mlh1+/+ WD, Mlh1+/- WD).

Mlh1+/+ AINMlh1+/- AINMlh1+/+ WDMlh1+/- WD
Downregulate GenesFold Change (P)Fold Change (P)Fold Change (P)Fold Change (P)
Dkk12.0 (0.110)5.1 (0.012)6.2 (0.003)7.3 (0.003)
Slc5a81.3 (0.078)1.7 (0.041)1.3 (0.224)1.5 (0.032)
Hoxd11.1 (0.348)2.0 (0.075)2.1 (0.019)1.5 (0.191)
Socs11.6 (0.460)2.7 (0.067)1.0 (0.485)3.1 (0.026)
Upregulated GenesFold Change (P)Fold Change (P)Fold Change (P)Fold Change (P)
Dkk22.2 (0.108)2.2 (0.037)1.7 (0.329)1.0 (0.637)
Rprm1.4 (0.442)2.0 (0.017)1.1 (0.725)3.3 (0.019)
Acaa1b2.4 (0.120)1.8 (0.140)3.5 (0.124)9.4 (0.000)

Table 4. Genes showing statistically significant (P < 0.05) mRNA expression differences between tp0 and tp1 in at least one of the study groups but not in the control group (Mlh1+/+ AIN).

Statistically significant expression changes between tp0 and tp1 associated with inherited cancer predisposition (Mlh1+/-) and/or WD* were observed in seven genes (Table 4, Table S5); decreased expression in Dkk1 [Dickkopf homolog 1 (Xenopus laevis)], Slc5a8 [(Solute carrier family 5 (iodide transporter), member 8], Hoxd1 (Homeobox D1), and Socs1 (Suppressor of cytokine signalling 1), and increased expression in Dkk2 [Dickkopf homolog 2 (Xenopus laevis)], Rprm (Reprimo, TP53 dependent G2 arrest mediator candidate), and Acaa1b (acetyl-Coenzyme A acyltransferase 1B).

The strongest expression decrease (7.3 fold) was seen in the Wnt signaling suppressor gene Dkk1 among the Mlh1+/- mice fed with WD*. In heterozygote mice fed with AIN and in WT mice fed with WD*, the decrease was 5.1 and 6.2 fold, respectively. Interestingly, the homeobox gene Hoxd1 showed an age-related expression decrease (2.1 fold) only in the Mlh1+/+ WD* group, and the statistically significant decrease related to WD* was further observed when this group was compared with the control group (Mlh1+/+ AIN) at tp1 (1.9 fold) (Table S5). The expression decrease of the transporter of butyrate Slc5a8 associated only with the Mlh1 heterozygosity (1.5 and 1.7 fold in WD* and AIN groups, respectively), whereas Socs1, a cytokine signaling regulator, was significantly down regulated (3.1 fold) only if both risk factors were present, i.e. in the Mlh1+/- WD* group.

The expression of Dkk2, another Wnt signaling modulator, was significantly increased in the Mlh1+/- AIN group (Table 4) so that at tp1, Dkk2 expression among Mlh1+/- mice was significantly lower (2.2 fold) in the WD* group compared to the AIN group (Table S5). Rprm showed a statistically significant age-related expression increase in association with Mlh1 heterozygosity (3.3 and 2.0. fold in WD* and AIN, respectively). Acaa1b showed a significant expression increase in the Mlh1+/- WD* mice (9.4 fold) and the expression levels revealed a significant difference (5.0 fold) when the Mlh1+/- WD* mice were compared to the Mlh1+/- AIN group at tp1 (Table S5).

There were no available tissue samples/extracts to validate the results at the protein level. Therefore, TaqMan RT-qPCR was used to validate expression changes associated with the cancer-predisposing Mlh1 mutation and/or WD*. The isogenic background and our own results that mice showed similar mRNA expression patterns at the outset (Figure S2A, S2B) made it justified to combine the individual samples (n = 8) from each group into pools for a subsequent use in the validation experiments. Figure 1 illustrates differences observed between the study groups and the control group at tp1.


Figure 1. Validation of the mRNA expression changes that were associated with WD* and/or inherited Mlh1 mutation using TaqMan assays.

Relative expression in different study groups (Mlh1+/- AIN, Mlh1+/+ WD*, and Mlh1+/- WD*) is compared to the control group (Mlh1+/+ AIN). Each sample is a mixture of eight RNA samples from eight different tp1 mice belonging to each mouse group. Data is presented as mean ± SEM (standard error of the mean) (n = 3), *significant difference compared to the control group. Median permutation method, P < 0.05. (A) Mlh1 shows the same 50% expression difference between the genotypes as at the starting point of the study. (B) Dkk1 is significantly down regulated in the study groups with WD* and/or Mlh1 heterozygosity. (C) Slc5a8 does not show significant expression differences at tp1. (D) Hoxd1 is down regulated in association with WD* in both genotypes; the down regulation being especially strong in the Mlh1+/+ WD group. (E) Socs1 does not show significant expression differences at tp1. (F) Dkk2 is down regulated in association with WD* in both genotypes.


The TaqMan assays confirmed the observation that Dkk1 is one of the most prominent candidates with expression in colon mucosa altered in association with inherited cancer predisposition and WD*. Due to the low levels of Dkk1 mRNA, RT-qPCR was performed on pre-amplified cDNA (amplification uniformity value of -0.76). Dkk1 mRNA levels were significantly decreased as compared to the control group (Mlh1+/+ AIN) in all study groups [Mlh1+/+ WD*, 1.5 fold (range, 1.4–1.7); Mlh1+/- AIN, 1.4 (1.3–1.5); Mlh1+/- WD*, 1.2 (1.1–1.2)] (Figure 1B). The expression of Hoxd1 was significantly decreased not only in Mlh1+/+ WD* mice [1.9 (1.7–2.2)], a finding already observed in StellARray, but also in the Mlh1+/- WD* group [1.3 (1.2–1.3)] (Figure 1D), highlighting the effect of WD* on Hoxd1 expression. The TaqMan RT-qPCR also confirmed the importance of WD* on the expression of Dkk2, where in StellARray a statistically significant decrease was seen in Mlh1+/- WD* mice [1.2, (1.2–1.3)], but now also in WT mice fed with WD* [1.2 (1.2–1.3), P = 0.053] (Figure 1F). The expression of Acaa1b was also found to be significantly increased both in the Mlh1+/- WD* and Mlh1+/+ WD* mice [2.6 (2.3–2.8) and 1.7 (1.7–1.8), respectively] (Figure S3). The expression levels of Slc5a8 and Socs1 did not differ between the control group and the study groups (Figure 1C, E).

Notably, no significant age or diet related expression changes were found in Apc or Mlh1. Like at the start, the expression of Mlh1 was approximately 50% lower in the heterozygote mice compared to WT mice, indicating that the second inactivating hit of Mlh1 had not yet occurred by tp1 (Figure 1A). The expression of Apc was similar in both genotypes (StellARray).

Significant increases in DNA methylation levels accompany reduced expression of genes

The methylation levels of the CGIs of the genes which had shown statistically significant decrease in mRNA expression in association with Mlh1 heterozygosity and/or WD* (Dkk1, Slc5a8, Hoxd1, and Socs1; Table 4) were analyzed individually for each tp0 and tp1 mouse using Sequenom’s MassARRAY EPITYPERTM system. Additionally, Mlh1, and Sfrp1 which had shown age-related expression decrease of borderline significance (1.4 fold, P = 0.06) in the Mlh1+/- WD* group (Table S5) and is a well-known gene in CRC and Wnt signaling were studied for methylation. For Dkk1, no UCSC annotated CGI exists and the target area was defined using EMBOSS CpG Plot showing an area of unusual CG composition, which is largely the same area analyzed previously in mice [26].

Within all studied CGIs, the methylation ratios varied widely between the individual CpG units studied (Figure S4). We therefore calculated the mean methylation values for the entire target regions (i.e. average methylation values across all qualified CpG units at each CGI) for each mouse, and these values were used when comparing the methylation levels between different mice groups (Table 5). Although the mean methylation levels seemed to be low in general, with the exception of Mlh1, mean methylation levels were statistically significantly higher among tp1 mice as compared to tp0 mice, which is in accordance with the observation that all other genes but Mlh1 had shown age-related decrease in mRNA expression (Table 4). Overall, mean methylation levels varied at tp0 between 3% and 8%, and at tp1 between 6% and 15%. The mean methylation values were the lowest for Mlh1 (5,6%), whereas the highest mean methylation values at tp1 were seen in Dkk1 (14,7%) and Sfrp1 (14,9%).

GeneMouse GroupMeanSDMdnMinMaxRangetp0 vs. tp1aG1 vs. G2atp0 vs. G1atp0 vs. G2a
Dkk1tp0 (n =15)0.0340.0240.0300.
tp1 (n =32)0.1470.0750.1200.070.350.28
G1 (n =18)0.1870.0780.1600.100.350.25
G2 (n =14)0.0940.0130.0900.070.120.05
Slc5a8tp0 (n =15)0.0580.0170.0520.
tp1 (n =32)0.0900.0440.0680.050.210.15
G1 (n =11)0.1420.0370.1360.090.210.12
G2 (n =21)0.0630.0080.0620.050.080.03
Hoxd1tp0 (n =15)0.0370.0080.0400.
tp1 (n =32)0.0930.0320.0800.060.180.12
G1 (n =13)0.1190.0210.1200.080.150.07
G2 (n =18)0.0690.0060.0700.060.080.02
Socs1tp0 (n =15)0.0840.0120.0830.
tp1 (n =32)0.1070.0280.0930.080.170.10
G1 (n =10)0.1440.0210.1480.110.170.07
G2 (n =22)0.0910.0080.0910.080.110.04
Sfrp1tp0 (n =15)0.0710.0140.0670.
tp1 (n =32)0.1490.0400.1360.100.240.14
G1 (n =11)0.1890.0400.1980.120.240.12
G2 (n =19)0.1260.0160.1220.100.160.06
Mlh1tp0 (n =16)0.0400.0160.0360.
tp1 (n =31)0.0560.0290.0480.020.130.10
G1(n =9)0.0860.0230.0880.050.130.08
G2 (n =22)0.0430.0210.0370.020.110.08

Table 5. Descriptive and comparative statistics of the methylation data of the genes, Dkk1, Slc5a8, Hoxd1, Socs1, Sfrp1 and Mlh1, for mice at tp0 and tp1 and for Group 1 (G1) and Group 2 (G2) (higher and lower methylation cluster, respectively) at tp1 and comparison of the mean methylation values between the different mouse groups.

a P(Mann-Whitney test, Exact Sig., 2 tailed)

Mice cluster into high and low methylation groups depending on diet and/or inherited predisposition

The Non-metric Multi-Dimensional Scaling (NMDS) tool of the Chipster software [25] showed that the tp0 and tp1 mice segregated into different parts of the plots indicating differences in their methylation levels. Differential clustering of methylation at tp1 was confirmed using the Chipster’s Dendrogam tool in which two distinct methylation clusters (higher and lower methylation cluster, i.e. Group 1 (G1) and Group 2 (G2), respectively) were observed in each CGI (Figure 2). Except for one mouse in Sfrp1 and in Socs1 (B214 and B225, respectively) an identical set of 11 mice clustered into the higher methylation group at CGIs of Dkk1, SLc5a8, Hoxd1, Socs1, and Sfrp1 (Figure 2). Among them, only two belonged to the control Mlh1+/+ AIN group (B211, and B214), while the remaining nine mice had either the inherited predisposition to CRC (B201, B212, B225, and B232) were fed with WD* (B204, B219, B231) or had both risk factors (B215, B220). In addition to these 11 mice another 7 mice (B233, B236, B237, B243, B246, B248, and B252) belonged to the higher Dkk1 methylation group.


Figure 2. Methylation clusters and the NMDS analysis of the methylation data.

Using the Chipster’s Dendrogram tool, two distinct clusters, the higher (Group 1) and the lower methylation cluster (Group 2) were observed at tp1. Except for one mouse in Sfrp1 and in Socs1 (B214 and B225, respectively) the same 11 mice clustered into the higher methylation cluster at CGIs of Dkk1, Slc5a8, Hoxd1, Socs1, and Sfrp1. The neoplasias are marked with superscripts ( ahyperplastic polyp, badenoma, cadenocarcinoma, dnot histologically confirmed) and the mice groups with different colors (black; Mlh1+/+ AIN, brown; Mlh1+/- AIN, turquoise; Mlh1+/+ WD*, pink; Mlh1+/- WD*). Also according to the NMDS analysis (Chipster) the tp0 and tp1 mice segregate into different parts of the plots indicating differences in their methylation levels. Red refers to tp0 mice, green refers to Group 1 mice, blue refers to Group 2 mice, and turquoise refers to mice that did not belong to either group.


The mean methylation levels of the Group 1 mice differed significantly from those of the Group 2 mice as well as from tp0 mice (Table 5). Also G2 mice differed significantly from tp0 mice in regard to Dkk1, Slc5a8, Hoxd1, and Sfrp1, while for Socs1 and Mlh1 they had remarkably similar mean methylation levels at both time points. In Mlh1, G1 mice had, however, a significantly higher mean methylation level than mice in G2 and at tp0 (9%, 4%, and 4%, respectively) (Table 5).

Expression and methylation changes are associated with colonic neoplasia and pinpoint Dkk1 as a promising biomarker

One proximal adenocarcinoma (B241, Mlh1+/+ WD) and five adenomas/hyperplastic polyps (adenomas: B219, Mlh1+/+ WD; B220, B236, Mlh1+/- WD and hyperplastic polyps: B215 Mlh1+/- WD; B244, Mlh1+/- AIN) were observed in the 32 mice at tp1 and none of them occurred in the control mouse group. Four out of six neoplasias were found in the mice showing higher methylation at CGIs of all five genes Dkk1, Slc5a8, Hoxd1, Socs1 and Sfrp1 (B219, B220), of the four genes Dkk1, Hoxd1, Socs1 and Sfrp1 (B215), or of Dkk1 (B236) (Figure 2).

Dkk1 drew our particular attention, since 4 out of the 6 neoplasms were found in mice which showed Dkk1 inactivation (Table 2). For Dkk1, up to 18 mice altogether clustered into the Group 1, in which the mean methylation level was 18,7%, whereas in the Group 2 it was 9,4% (Table 5, Figure 2). The mRNA expression of Dkk1 was completely silenced in 16 mice (B201, B204, B214, B215, B220, B225, B232, B233, B234, B236, B241, B242, B249, B250, B252, and B253) and at least in 10 cases hypermethylation was a plausible reason for its inactivation (Figure 2).


Our study was based on the presumption that along with aging certain dietary components especially in Western-style diet are important modifiers of methylation patterns in the colon {Arasaradnam, 2008 #301} [8] and thus may affect the transcription of genes involved in tumorigenesis. We further reasoned that in mutation carriers requiring just a second hit of the inherited susceptibility gene for malignant transformation, it might be possible to detect the earliest changes, which might even precede the second hit, and distinguish these from alterations occurring later in oncogenesis. These hypotheses proved valid when tested on a mouse model for Lynch syndrome, the most common form of familial CRC.

The candidate gene approach was chosen instead of a genome wide expression analysis in order to focus on alterations already shown to be associated with human colon cancer. Altogether, nine out of the 94 genes in the expression array showed statistically significant age-related expression changes in all mice groups, whereas statistically significant expression changes related to Mlh1+/- and WD* were observed for seven genes, and among those including expression decreases in Dkk1, Slc5a8, Hoxd1, and Socs1 (Sfrp1 of borderline significance). Many of the genes, which showed the age-related expression and methylation changes in our study associate with Wnt/β-catenin signaling, although, their changes did not necessarily imply Wnt signaling activation as does the expression decrease of Dkk1 and Sfrp1. The Wnt/β-catenin signaling pathway is aberrantly activated in most human colon cancers [27,28]. However, it has been recently found that a number of Wnt signaling-related molecules are differently expressed in the proximal and distal colon and suggested that the proximal colon may constitute a unique signaling niche that is more sensitive to Dkk1 signaling [29]. Indeed in the present study on proximal colon mucosa, we did not observe changes in the expression of Ctnnb1 (encoding β-catenin) or Apc, a member of the β-catenin destruction complex in the cytosol [30] however, the expression of Cdh1 (encoding cadherin 1, also known as E- cadherin), which is a major component of adherent junctions and binds to free cytosolic β-catenin [31] was significantly decreased in all mouse groups, down regulation being strongest in the Mlh1+/- WD* group. Dkk1 as well as Sfrp1 are secreted Wnt signaling antagonists [32,33] and act as “epigenetic gatekeepers”, whose aberrant silencing may lock the cells into stem cell like states allowing time for genetic gatekeeper mutations in the downstream pathway genes to appear [32,34]. Here we observed a significant age-related increase in methylation levels of Dkk1, Sfrp1, Slc5a8, Hoxd1, and Socs1 CGIs but not of Mlh1, which was still well-expressed at tp1, suggesting that expression decreases of the other genes precede the involvement of Mlh1. Although Mlh1 methylation levels have been reported to increase with age in normal colonic mucosa (characteristic of ‘type A’ genes) [5,7,35-37], our results rather comply with studies [37,38] classifying Mlh1 as a ‘type C’ gene, with methylation specificity for neoplasia. Of the genes, for which we observed ‘type A’ methylation, Sfrp1 has previously been classified as a ‘type A’ gene, while Socs1 has been reported as a ‘type C’ gene [6].

Here, to our knowledge, the Wnt antagonist Dkk1 is for the first time shown to be silenced already in histologically normal colon mucosa. Contrary to human homologues DKK1 and DKK2, whose silencing by CGI hypermethylation is indicated both in colon cancer cell lines and gastrointestinal tumors [33,39-41], Dkk1 downregulation in mice has been suggested to occur through factors independent of DNA hypermethylation [26]. The fact that majority of the mice not expressing Dkk1 in their colon mucosa were Mlh1 heterozygotes (10/16) suggest a link to haploinsufficiency caused by a loss of function mutation in Mlh1. On the other hand, the decreased level of vitamin D3 (1.25(OH) 2D3), which has been shown to strongly regulate the human DKK1 expression [28,42], may have contributed to the decreased Dkk1 mRNA levels and colonic neoplasias in mice fed with WD*. The importance of vitamin D3 in Dkk1 expression and its regulation by epigenetic modification was further supported in a recent study where its intake was negatively associated with DKK1 methylation in a large cohort of CRCs [41]. In our study, the mRNA expression of Dkk1 was completely silenced in 16 mice and at least in 10 cases the reason could have been hypermethylation.

Hoxd1 which is a target gene for Wnt [43] and Polycomb Group (PcG) proteins [44] provides another illustrative example of dietary effects. Hoxd1 has previously been reported to be hypermethylated in a colon cancer cell line [45] and HOX gene clusters in human lung carcinomas and in noncancerous lung tissues [46]. Here, for the first time, we show that the expression and methylation of Hoxd1 is altered already in histologically normal colonic mucosa and especially in mice fed with WD*, suggesting that dietary effects may create selective pressure for its silencing.

Western diet has been reported to induce oxidative stress responses and exert a pro-inflammatory stimulus in the colon long before tumors occur [47]. Metabolic errors may in essence initiate the epigenetic switch that contributes to carcinogenesis and cancer progression [48]. Changes that we observed in e.g. Acaa1b and Socs1 may be related to inflammation and alterations in energy metabolism. Furthermore, a mild inflammation has been reported to accelerate colon carcinogenesis in Mlh1-deficient mice [49], a finding in line with our observation that most WD*-related alterations are linked to Mlh1 heterozygosity.

In summary, among 73 TSGs, a specific set of genes and Dkk1 in particular were identified as promising candidates for altered methylation and expression, detectable already in histologically normal mucosa and potentiated by an inherited MMR gene mutation and Western-style diet. Such changes, which are linked to human colon cancer and occur prior to the second hit in the predisposing gene, may be among the very earliest alterations in multistep tumorigenesis. Our results highlight the interplay between genome, epigenome, and environment in colon tumorigenesis and encourage studies to explore the potential of the respective genes and alterations as biomarkers for diagnostic, prognostic and therapeutic applications in humans.

Supporting Information

Materials and Methods S1.

B6.129-Mlh1tm1Rak mice genotyping. The Mlh1+/- and Mlh1+/+ mice were genotyped using genomic DNA extracted from earmarks. Briefly, PCR reaction contained primers M001 (5'-TGTCAATAGGCTGCCCTAGG-3'; 0.33µM), M002 (5'-TGGAAGGATTGGAGCTACGG-3'; 0.33µM), and M003 (5'-TTTTCAGTGCAGCCTATGCTC-3'; 0.3µM), dNTP mix (0.2mM), 1x reaction buffer, Dynazyme II (0.1 U/µL) (Thermo Scientific, Finland), 50 ng of gDNA template, and MQ water in a total volume of 20 µL. Cycling conditions were 94°C for 3 min, followed by 35 cycles of 94°C 1 min, 60°C 2 min, 72°C 1 min, and an elongation step 72°C 3 min. Primer combination M001/M002 produced a 500 bp fragment indicating the mutant allele, and combination M001/M003 produced a 350 bp fragment indicating the wild-type allele.



Figure S1.

Histological images of (A) A well-differentiated colonic polypoid adenocarcinoma with focal (early) invasion of the lamina propria (mouse B241 / Mlh1+/+ WD*) and (B) A colonic adenoma (mouse B219 / Mlh1+/+ WD*). (4x magnification, Olympus BX63).



Figure S2.

Comparison of mRNA expression patterns between the Mlh1+/+ and Mlh1+/- mice at tp0. (A) Correlation of the mean Cq values of the 94 genes included in the StellARray between the tp0 mice with different Mlh1 genotypes (n = 8). Pearson’s correlation, P = 0.000 (2-tailed). (B) Mlh1 expression of the Mlh1+/- mice is 50% of the expression level detected in the Mlh1+/+ mice using TaqMan assay. Each sample is a mixture of eight RNA samples from eight different tp0 mice with the same Mlh1 genotype. Samples were ran in triplicate. *Significant difference compared to the Mlh1+/+ group. Median permutation method, P < 0.05.



Figure S3.

Validation of the StellARray mRNA expression changes in Acaa1b that were associated with WD* and/or inherited Mlh1 mutation using TaqMan assays. Relative expression in different study groups (Mlh1+/- AIN, Mlh1+/+ WD*, and Mlh1+/- WD*) compared to the control group (Mlh1+/+ AIN). Each sample is a mixture of eight RNA samples from eight different tp1 mice belonging to each mouse group. Data is presented as mean ± s.e.m. (n = 3, each pooled sample were ran triplicate), *significant difference compared to the control group. Median permutation method, P < 0.05. Acaa1b shows strong upregulation connected with WD* in both genotypes.



Figure S4.

Mean methylation at each qualified CpG unit at CGIs of Mlh1, Dkk1, Slc5a8, Hoxd1, Socs1, and Sfrp1 for the tp0 and tp1 mice separately. The mean methylation value for each qualified CpG unit among the tp0 and tp1 mice is presented with 95% confidence intervals.



Table S1.

Compositions of the experimental diets. AIN-93G control diet is a semi-synthetic diet designed to meet the nutritional requirements of growing rodents, while WD* is a modified AIN diet, which contains high dietary fat (39% of total calories) and reduced contents of fiber, calcium, vitamin D, and three methyl-transfer donors (i.e. folic acid, methionine, and choline). WD* contains more sucrose and correspondingly less complex carbohydrates than the control diet and its fat consists mainly of milkfat, while in AIN-93G, the fat source is exclusively soy bean oil. WD* resembles the previously published Western-style diet [11,12], which was described to induce neoplasms in the colon of normal C57BL/6 mice. Notably the fat content of WD* is yet more similar to the actual fat content consumed by Western populations.



Table S2.

Genes included in StellARray.



Table S3.

TaqMan assays for studied genes.



Table S4.

Amplicon data for methylation analyses.



Table S5.

GPR results of expression differences between different mice groups for the 94 genes studied.




The authors wish to thank Dr Anja Kipar, Dr Mikael Niku and Ms Anu Heiman-Lindh for their help in sample collection, Dr Anja Kipar for histological analyses, and Drs Eija Korpelainen and Massimiliano Gentile for their help with the Chipster software.

Author Contributions

Conceived and designed the experiments: MP LS PP MM MN. Performed the experiments: MP LS DDB SV SO EP. Analyzed the data: MP LS NR MN. Contributed reagents/materials/analysis tools: PP MN. Wrote the manuscript: MP LS PP MN.


  1. 1. Bingham S, Riboli E (2004) Diet and cancer--the European Prospective Investigation into Cancer and Nutrition. Nat Rev Cancer 4: 206-215. doi:10.1038/nrc1298. PubMed: 14993902.
  2. 2. Fearon ER (2011) Molecular genetics of colorectal cancer. Annu Rev Pathol 6: 479-507. doi:10.1146/annurev-pathol-011110-130235. PubMed: 21090969.
  3. 3. Feinberg AP, Ohlsson R, Henikoff S (2006) The epigenetic progenitor origin of human cancer. Nat Rev Genet 7: 21-33. doi:10.1038/nrg1748. PubMed: 16369569.
  4. 4. Lao VV, Grady WM (2011) Epigenetics and colorectal cancer. Nat. Rev Gastroenterol Hepatol 8: 686-700. doi:10.1038/nrgastro.2011.173.
  5. 5. Ahuja N, Issa JP (2000) Aging, methylation and cancer. Histol Histopathol 15: 835-842. PubMed: 10963127.
  6. 6. Issa JP, Ottaviano YL, Celano P, Hamilton SR, Davidson NE et al. (1994) Methylation of the oestrogen receptor CpG island links ageing and neoplasia in human colon. Nat Genet 7: 536-540. doi:10.1038/ng0894-536. PubMed: 7951326.
  7. 7. Toyota M, Issa JP (1999) CpG island methylator phenotypes in aging and cancer. Semin Cancer Biol 9: 349-357. doi:10.1006/scbi.1999.0135. PubMed: 10547343.
  8. 8. Arasaradnam RP, Commane DM, Bradburn D, Mathers JC (2008) A review of dietary factors and its influence on DNA methylation in colorectal carcinogenesis. Epigenetics 3: 193-198. doi:10.4161/epi.3.4.6508. PubMed: 18682688.
  9. 9. Verma M (2012) Cancer control and prevention by nutrition and epigenetic approaches. Antioxid Redox Signal 17: 355-364. doi:10.1089/ars.2011.4388. PubMed: 22047027.
  10. 10. Tammariello AE, Milner JA (2010) Mouse models for unraveling the importance of diet in colon cancer prevention. J Nutr Biochem 21: 77-88. doi:10.1016/j.jnutbio.2009.09.014. PubMed: 20122631.
  11. 11. Newmark HL, Yang K, Kurihara N, Fan K, Augenlicht LH et al. (2009) Western-style diet-induced colonic tumors and their modulation by calcium and vitamin D in C57Bl/6 mice: a preclinical model for human sporadic colon cancer. Carcinogenesis 30: 88-92. PubMed: 19017685.
  12. 12. Newmark HL, Yang K, Lipkin M, Kopelovich L, Liu Y et al. (2001) A Western-style diet induces benign and malignant neoplasms in the colon of normal C57Bl/6 mice. Carcinogenesis 22: 1871-1875. doi:10.1093/carcin/22.11.1871. PubMed: 11698351.
  13. 13. Yang K, Kurihara N, Fan K, Newmark H, Rigas B et al. (2008) Dietary induction of colonic tumors in a mouse model of sporadic colon cancer. Cancer Res 68: 7803-7810. doi:10.1158/0008-5472.CAN-08-1209. PubMed: 18829535.
  14. 14. Yang K, Yang W, Mariadason J, Velcich A, Lipkin M et al. (2005) Dietary components modify gene expression: implications for carcinogenesis. J Nutr 135: 2710-2714. PubMed: 16251635.
  15. 15. Lynch HT, Lynch PM, Lanspa SJ, Snyder CL, Lynch JF et al. (2009) Review of the Lynch syndrome: history, molecular genetics, screening, differential diagnosis, and medicolegal ramifications. Clin Genet 76: 1-18. doi:10.1111/j.1399-0004.2009.01230.x. PubMed: 19659756.
  16. 16. Edelmann W, Cohen PE, Kane M, Lau K, Morrow B et al. (1996) Meiotic pachytene arrest in MLH1-deficient mice. Cell 85: 1125-1134. doi:10.1016/S0092-8674(00)81312-4. PubMed: 8674118.
  17. 17. Edelmann W, Yang K, Kuraguchi M, Heyer J, Lia M et al. (1999) Tumorigenesis in Mlh1 and Mlh1/Apc1638N mutant mice. Cancer Res 59: 1301-1307. PubMed: 10096563.
  18. 18. Reeves PG, Nielsen FH, Fahey GC Jr. (1993) AIN-93 purified diets for laboratory rodents: final report of the American Institute of Nutrition ad hoc writing committee on the reformulation of the AIN-76A rodent diet. J Nutr 123: 1939-1951. PubMed: 8229312.
  19. 19. Vandesompele J, De Preter K, Pattyn F, Poppe B, Van Roy N et al. (2002) Accurate normalization of real-time quantitative RT-PCR data by geometric averaging of multiple internal control genes. Genome Biol 3: RESEARCH0034. PubMed: 12184808.
  20. 20. Ehrich M, Nelson MR, Stanssens P, Zabeau M, Liloglou T et al. (2005) Quantitative high-throughput analysis of DNA methylation patterns by base-specific cleavage and mass spectrometry. Proc Natl Acad Sci U S A 102: 15785-15790. doi:10.1073/pnas.0507816102. PubMed: 16243968.
  21. 21. Coolen MW, Statham AL, Gardiner-Garden M, Clark SJ (2007) Genomic profiling of CpG methylation and allelic specificity using quantitative high-throughput mass spectrometry: critical evaluation and improvements. Nucleic Acids Res 35: e119. doi:10.1093/nar/gkm662. PubMed: 17855397.
  22. 22. Akilesh S, Shaffer DJ, Roopenian D (2003) Customized molecular phenotyping by quantitative gene expression and pattern recognition analysis. Genome Res 13: 1719-1727. doi:10.1101/gr.533003. PubMed: 12840047.
  23. 23. Livak KJ, Schmittgen TD (2001) Analysis of relative gene expression data using real-time quantitative PCR and the 2(-Delta Delta C(T)) Method. Methods 25: 402-408. doi:10.1006/meth.2001.1262. PubMed: 11846609.
  24. 24. Efron B, Tibshirani R, Storey JD, Tusher V (2001) Empirical Bayes analysis of a microarray experiment. J Am Stat Assoc 96: 1151-1160. doi:10.1198/016214501753382129.
  25. 25. Kallio MA, Tuimala JT, Hupponen T, Klemelä P, Gentile M et al. (2011) Chipster: user-friendly analysis software for microarray and other high-throughput data. BMC Genomics 12: 507. doi:10.1186/1471-2164-12-507. PubMed: 21999641.
  26. 26. Oyama T, Yamada Y, Hata K, Tomita H, Hirata A et al. (2008) Further upregulation of beta-catenin/Tcf transcription is involved in the development of macroscopic tumors in the colon of ApcMin/+ mice. Carcinogenesis 29: 666-672. PubMed: 18204079.
  27. 27. Gregorieff A, Clevers H (2005) Wnt signaling in the intestinal epithelium: from endoderm to cancer. Genes Dev 19: 877-890. doi:10.1101/gad.1295405. PubMed: 15833914.
  28. 28. Pendás-Franco N, Aguilera O, Pereira F, González-Sancho JM, Muñoz A (2008) Vitamin D and Wnt/beta-catenin pathway in colon cancer: role and regulation of DICKKOPF genes. Anticancer Res 28: 2613-2623. PubMed: 19035286.
  29. 29. Koch S, Nava P, Addis C, Kim W, Denning TL et al. (2011) The Wnt antagonist Dkk1 regulates intestinal epithelial homeostasis and wound repair. Gastroenterology 141: 259-268, e251-258 doi:10.1053/j.gastro.2011.03.043. PubMed: 21440550.
  30. 30. Burgess AW, Faux MC, Layton MJ, Ramsay RG (2011) Wnt signaling and colon tumorigenesis--a view from the periphery. Exp Cell Res 317: 2748-2758. doi:10.1016/j.yexcr.2011.08.010. PubMed: 21884696.
  31. 31. Schneider MR, Dahlhoff M, Horst D, Hirschi B, Trülzsch K et al. (2010) A key role for E-cadherin in intestinal homeostasis and Paneth cell maturation. PLOS ONE 5: e14325. doi:10.1371/journal.pone.0014325. PubMed: 21179475.
  32. 32. Baylin SB, Ohm JE (2006) Epigenetic gene silencing in cancer - a mechanism for early oncogenic pathway addiction? Nat Rev Cancer 6: 107-116. doi:10.1038/nrc1799. PubMed: 16491070.
  33. 33. González-Sancho JM, Aguilera O, García JM, Pendás-Franco N, Peña C et al. (2005) The Wnt antagonist DICKKOPF-1 gene is a downstream target of beta-catenin/TCF and is downregulated in human colon cancer. Oncogene 24: 1098-1103. doi:10.1038/sj.onc.1208303. PubMed: 15592505.
  34. 34. Suzuki H, Watkins DN, Jair KW, Schuebel KE, Markowitz SD et al. (2004) Epigenetic inactivation of SFRP genes allows constitutive WNT signaling in colorectal cancer. Nat Genet 36: 417-422. doi:10.1038/ng1330. PubMed: 15034581.
  35. 35. Ahuja N, Li Q, Mohan AL, Baylin SB, Issa JP (1998) Aging and DNA methylation in colorectal mucosa and cancer. Cancer Res 58: 5489-5494. PubMed: 9850084.
  36. 36. Kawakami K, Ruszkiewicz A, Bennett G, Moore J, Grieu F et al. (2006) DNA hypermethylation in the normal colonic mucosa of patients with colorectal cancer. Br J Cancer 94: 593-598. doi:10.1038/sj.bjc.6602940. PubMed: 16421593.
  37. 37. Worthley DL, Whitehall VL, Buttenshaw RL, Irahara N, Greco SA et al. (2010) DNA methylation within the normal colorectal mucosa is associated with pathway-specific predisposition to cancer. Oncogene 29: 1653-1662. doi:10.1038/onc.2009.449. PubMed: 19966864.
  38. 38. Issa JP, Ahuja N, Toyota M, Bronner MP, Brentnall TA (2001) Accelerated age-related CpG island methylation in ulcerative colitis. Cancer Res 61: 3573-3577. PubMed: 11325821.
  39. 39. Aguilera O, Fraga MF, Ballestar E, Paz MF, Herranz M et al. (2006) Epigenetic inactivation of the Wnt antagonist DICKKOPF-1 (DKK-1) gene in human colorectal cancer. Oncogene 25: 4116-4121. doi:10.1038/sj.onc.1209439. PubMed: 16491118.
  40. 40. Maehata T, Taniguchi H, Yamamoto H, Nosho K, Adachi Y et al. (2008) Transcriptional silencing of Dickkopf gene family by CpG island hypermethylation in human gastrointestinal cancer. World J Gastroenterol 14: 2702-2714. doi:10.3748/wjg.14.2702. PubMed: 18461655.
  41. 41. Rawson JB, Sun Z, Dicks E, Daftary D, Parfrey PS et al. (2012) Vitamin D Intake Is Negatively Associated with Promoter Methylation of the Wnt Antagonist Gene DKK1 in a Large Group of Colorectal Cancer Patients. Nutr Cancer 64: 919-928. doi:10.1080/01635581.2012.711418. PubMed: 22966878.
  42. 42. Aguilera O, Peña C, García JM, Larriba MJ, Ordóñez-Morán P et al. (2007) The Wnt antagonist DICKKOPF-1 gene is induced by 1alpha,25-dihydroxyvitamin D3 associated to the differentiation of human colon cancer cells. Carcinogenesis 28: 1877-1884. doi:10.1093/carcin/bgm094. PubMed: 17449905.
  43. 43. Janssens S, Denayer T, Deroo T, Van Roy F, Vleminckx K (2010) Direct control of Hoxd1 and Irx3 expression by Wnt/beta-catenin signaling during anteroposterior patterning of the neural axis in Xenopus. Int J Dev Biol 54: 1435-1442. doi:10.1387/ijdb.092985sj. PubMed: 20979027.
  44. 44. Boyer LA, Plath K, Zeitlinger J, Brambrink T, Medeiros LA et al. (2006) Polycomb complexes repress developmental regulators in murine embryonic stem cells. Nature 441: 349-353. doi:10.1038/nature04733. PubMed: 16625203.
  45. 45. Jacinto FV, Ballestar E, Ropero S, Esteller M (2007) Discovery of epigenetically silenced genes by methylated DNA immunoprecipitation in colon cancer cells. Cancer Res 67: 11481-11486. doi:10.1158/0008-5472.CAN-07-2687. PubMed: 18089774.
  46. 46. Shiraishi M, Sekiguchi A, Oates AJ, Terry MJ, Miyamoto Y (2002) HOX gene clusters are hotspots of de novo methylation in CpG islands of human lung adenocarcinomas. Oncogene 21: 3659-3662. doi:10.1038/sj.onc.1205453. PubMed: 12032868.
  47. 47. Erdelyi I, Levenkova N, Lin EY, Pinto JT, Lipkin M et al. (2009) Western-style diets induce oxidative stress and dysregulate immune responses in the colon in a mouse model of sporadic colon cancer. J Nutr 139: 2072-2078. doi:10.3945/jn.108.104125. PubMed: 19759248.
  48. 48. Hitchler MJ, Domann FE (2009) Metabolic defects provide a spark for the epigenetic switch in cancer. Free Radic Biol Med 47: 115-127. doi:10.1016/j.freeradbiomed.2009.04.010. PubMed: 19362589.
  49. 49. Taniguchi K, Kakinuma S, Tokairin Y, Arai M, Kohno H et al. (2006) Mild inflammation accelerates colon carcinogenesis in Mlh1-deficient mice. Oncology 71: 124-130. doi:10.1159/000100522. PubMed: 17347588.