• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptNIH Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Cancer Biol Ther. Author manuscript; available in PMC Oct 19, 2009.
Published in final edited form as:
PMCID: PMC2763640
NIHMSID: NIHMS107839

Genome-wide profiling of methylated promoters in pancreatic adenocarcinoma

Abstract

Many genes undergo aberrant methylation in human cancers, and microarray platforms enable more comprehensive profiling of aberrant DNA methylation patterns.

Results

1,010 of 87,922 probes on the 88 K promoter array (606 genes) had a higher signal (log2 > 2) in the pancreatic cancer line, Panc-1 compared to the non-neoplastic pancreatic duct line, HPDE. Using this cut-off, bisulfite sequencing and/or MSP confirmed differential methylation of all 27 genes (66 probes) predicted to be methylated by the MCA array. More than half of the genes aberrantly hypermethylated in Panc-1 were not expressed in the pancreatic duct (HPDE) by expression array analysis. Using the 244 K CpG island array, 1,968 CpG islands were differentially methylated in MiaPaca2 compared to normal pancreas. The MCA method was more likely to identify hypermethylation within CpG islands than a cocktail of methylation sensitive restriction enzymes. DNA methylation profiles using 10 ng of DNA were highly correlated with those obtained using 5 µg of DNA (R2 = 0.98). Analysis of 57 pancreatic cancers and 34 normal pancreata using MSP identified MDFI, hsa-miR-9-1, ZNF415, CNTNAP2 and ELOVL4 as methylated in 96%, 89%, 86%, 82% and 68% of the cancers vs. 9%, 15%, 6%, 3% and 9% of normal pancreata, respectively.

Methods

We used methylated CpG island amplification (MCA) and Agilent promoter and CpG island microarrays to identify differential DNA methylation patterns in pancreatic cancer vs. normal pancreas. We examined MCA array reproducibility, compared it to methylation profiles obtained using a cocktail of methylation-sensitive restriction enzymes and examined gene expression of methylated genes.

Conclusion

Promoter and CpG island array analysis finds aberrant methylation of hundreds of promoters and CpG islands in pancreatic cancer cells.

Keywords: DNA methylation, pancreatic cancer, methylated CpG island amplification (MCA), promoter microarray, epigenetics

Introduction

Aberrant CpG island methylation is an important cause of altered gene function in human cancer. This altered gene function arises from both gene silencing associated with aberrant hypermethylation of promoter CpG islands and induction of gene expression associated with gene hypomethylation.1 Aberrant DNA methylation contributes to pancreatic cancer development and progression210 and the detection of aberrant DNA methylation is being evaluated as a strategy to improve the diagnosis of pancreatic cancer.11 While the causes of aberrant CpG island methylation during the cancer development are not well understood,1216 the identification of differentially methylated CpG islands in cancer relative to normal tissues may lead to the development of cancer-specific markers of cancer and may also identify important pathways that merit therapeutic targeting.17 As the extent of DNA methylation alterations become more apparent, robust high-throughput technologies are being evaluated that can efficiently interrogate genome-wide DNA methylation profiles. Several genome-wide strategies have been developed to interrogate methylated CpG islands2,1821 and to identify genes whose expression is under epigenetic control.2,22 Newer array platforms provide an opportunity to more efficiently interrogate promoters for evidence of differential methylation.

In this study, we employed a methylated CpG island amplification (MCA) strategy which uses the methylation sensitive restriction enzyme, SmaI and its isoschizomer, Xma1 and applied it to a genome-wide promoter microarray platform (Agilent) as a method of detecting differential methylation patterns between pancreatic cancer and normal pancreas samples. We also compared the results obtained using the MCA microarray method to those obtained using a cocktail of methylation sensitive restriction enzymes. We evaluated the reproducibility and accuracy of the MCA microarray method and identified novel genes that are aberrantly methylated in pancreatic cancer.

Results

DNA methylation profiles using MCA microarrays

The principle underlying MCA involves the amplification of closely spaced methylated SmaI sites to enrich for methylated CpG islands.25 The amplified DNA derived from MCA was hybridized to promoter or CpG island arrays. Equal amounts of input DNA representing equal numbers of genome equivalents are used in the MCA procedure and equal amounts of MCA amplicon are hybridized to the Agilent arrays. The MCA technique identifies differences in the methylation status of the SmaI restriction site (CCCGGG) between samples. Importantly, ~70–80% of CpG islands contain at least two closely spaced (<1 kb) SmaI sites (CCCGGG) so that most CpG islands can be interrogated with this assay. In addition, the Sma1/Xma1 combination creates an overhang that helps ensure selective amplification of methylated templates.

An overview of the experimental approach we used in this paper is provided in Figure 1. We first performed MCA on DNA from the pancreatic cancer line, Panc-1 and the non-neoplastic pancreatic epithelial cell line, HPDE and hybridized labeled amplicons to the Agilent 44 K promoter array. We used these cell line DNAs to ensure that the MCA results would represent methylation profiles of pure pancreatic cancer cells and normal pancreatic duct cells, respectively. After global Loess normalization and inter-array normalization with Chip analytics 1.3 software, we generated a list of genes with differential probe signals between these two cell lines (Fig. 2 and Suppl. Table 1). To confirm the accuracy of our MCA microarray data, we performed bisulfite sequencing and methylation specific PCR (MSP). We first evaluated the methylation status of the top 20 probes (in 16 genes) with the highest differential Panc-1/HPDE signal ratio by bisulfite sequencing and/or MSP (Table 1). We also randomly selected 11 additional genes for analysis that had significant differential signals in multiple probes (log2 ratio 2–5 range) (Table 1 and Fig. 3). In addition, we also selected two genes for bisulfite sequencing that showed no difference in microarray signals between Panc-1 and HPDE and both cell lines were unmethylated. We chose a four-fold (log2 2) differential signal ratio as the lower cut-off for differential methylation between two samples so as to minimize assigning differential methylation to signals that might arise from variations in the assay (such as incomplete restriction digestion) or from differences in copy number between cancer and non-neoplastic samples. The accuracy of the MCA assay was verified by the finding that all 27 genes were extensively methylated in Panc-1 but largely unmethylated in HPDE and all of the SmaI sites sequenced showed methylation in Panc-1 but not in HPDE, as would be expected from the MCA assay (Table 1). Thus, bisulfite sequencing confirmed that all of the Sma1 sites examined that were predicted to be methylated by the MCA array were methylated in Panc-1 and were unmethylated in HPDE. The differentially methylated CpG islands were generally methylated at many other adjacent CpG sites, highlighting the fact that the methylation status of the SmaI site within a CpG island often predicts its overall methylation status (Table 1). In contrast, for several genes examined by bisulfite sequencing that had hypomethylated SmaI sites in Panc-1 relative to HPDE many adjacent CpGs did not show similar hypomethylation (Table 1).

Figure 1
An overview of experimental strategy to evaluate differential DNA methylation in pancreatic cancer cells using MCA microarrays.
Figure 2
(A and B) Magnitude versus amplitude (MA) plot of an MCA microarray hybridization. The x-axis represents the log2 intensity of the Panc-1 (Cy5) and HPDE (Cy3) channel, and the y-axis represents the log2 ratio of Panc-1/HPDE. The red and green line represents ...
Figure 3
(A) Reproducibility of the MCA microarray. The scatter plots represent the correlation of probe intensities on replicated arrays for Panc-1 (Cy5) and HPDE (Cy3). (B) DNA methylation profiles of chromosome 1 using the MCA Agilent array. Each circle indicates ...
Table 1
A list of genes validated by bisulfite sequencing or MSP

Additionally, we examined the microarray results for genes on the X chromosome. The Panc-1 cell line originated from a male while HPDE is from a female. Thus, HPDE will be affected by X chromosome inactivation and epigenetic silencing. Indeed, 119 of 3,621 probes (3.2% of total probes, 71 genes) on the X chromosome showed lower signal ratios indicating that they were likely to be methylated (Fig. 2C). In contrast, only 0.008% of the probes (772 of 87,922 probes) in the genome were detected as differentially hypermethylated in HPDE relative to Panc-1. The 71 genes included the genes DNASE1L1, EFNB1, ELF4, MTM1, PDK3, SAT, SLC9A6, SSR4, STS and TAZ (Fig. 1C), all of which have been identified as subject to X chromosome inactivation.28,29 These data further support the accuracy of the MCA microarray method.

Based on the accuracy of the MCA array for identifying differential methylation of probes with log2 of ≥ 2 (~4-fold) signal ratios, we used this cut-off to identify differentially methylated probes. By this criterion, we found 1,010 of the 87,922 probes (representing 606 genes) on the A and B chip of human Agilent promoter array were differentially methylated in the Panc-1 compared to HPDE. The genes identified as differentially methylated included ppENK and Pax5, genes that we previously identified by sequencing cloned amplicons using an MCA/RDA protocol,30 as well as methylated genes identified through other strategies (LHX2, CLDN5, NPTX2,TFPI2).2,4,31 However, most of the genes identified by the MCA array method did not overlap with differentially methylated genes we had identified previously in Panc-1 vs. HPDE. For example, we previously reported the identification of numerous aberrantly methylated genes in pancreatic cancer by comparing gene expression profiles of Panc-1 and other pancreatic cancer cell lines vs. the HPDE cell line before and after epigenetic treatment (4). We found that only 4 of the 73 genes on the Affymetrix U133A chip with a five-fold increase in expression after 5Aza-dC treatment were identified as methylated candidates using the MCA array method.

We also examined the expression pattern in normal pancreas of the 606 genes identified by the MCA array with promoter methylation in Panc-1 relative to HPDE. Interestingly, of the genes that were represented on the U133A array only 32.7% were expressed in HPDE (based on having a “present” call by the MAS 5.0 Software). Thus, the majority of genes identified as differentially methylated in Panc-1 relative to HPDE by the MCA array are not expressed in normal pancreatic duct.

Although a cut-off of signal ratio log2 two predicted differential methylation between two samples as confirmed by bisulfite sequencing, it is likely that some probes with somewhat lower signal ratios also reflect differentially methylated loci. We also examined the MCA array results for evidence of Panc-1 hypomethylation. There were 772 probes representing 547 genes with a reduced signal log2 ratio (<-2) in Panc-1 compared to HPDE suggesting hypomethylation of Sma1 sites in Panc-1 relative to HPDE. We confirmed the Sma1 site hypomethylation in 4 of 5 genes tested by bisulfite sequencing (Table 1).

DNA methylation profiles with the 244 K CpG island array

We next profiled DNA methylation using the Agilent 244 K CpG island array. This array uses the same probe design as the 44 K A and B promoter arrays but interrogates 27,800 CpG islands with ~8 probes per island. For this experiment we profiled the pancreatic cancer cell line, MiaPaca2 and compared it to microdissected normal pancreas to identify methylation profiles in other normal pancreas samples besides the non-neoplastic pancreatic duct cell line, HPDE. Using the same cut-offs established for the 44 K array, we found MiaPaca2 had 6,712 differentially hypermethylated probes representing 1,968 CpG islands and 5,514 hypomethylated probes representing 3,378 CpG islands compared to normal pancreas. We confirmed the aberrant methylation status of 10 differentially methylated genes in MiaPaca2 including ACTA1, BAI1, CNTNAP2, ELOVL4, EYA4, FAM84A, HOXA5, Hsa-mir-9-1, MDFI, PAX7, PKP1, SOX14, TLX3, TNFRSF18 and ZNF415.

Array reproducibility

To test the reproducibility of the MCA assay, we repeated the MCA experiment. DNA was re-extracted from Panc-1 and HPDE and these replicate samples were subjected to the MCA procedure. Amplicons were labeled and hybridized onto the Agilent human promoter array 44 K A-chip. After global Loess normalization, the probe intensity of each array was compared between the replicates. The experimental results were highly reproducible with Pearson correlations for probe intensities for the Cy5 and Cy3 of 0.928 and 0.955, respectively (Fig. 3). As a result 80% of the probes with a signal ratio of log2 ≥ 2 in the first array also had a log2 ≥ 2 in the replicate array.

MCA array performance using low-input DNA

We performed an MCA array experiment using 10 ng of input DNA (pancreatic cancer cell line Su8686). Several MCA digestions and amplifications were performed using 1, 10, 50 and 100 ng of input DNA and observing the patterns of amplicons on agarose gels (data not shown). MCA using 10 ng or more produced similar gel patterns to those obtained with 5 µg of input DNA. We therefore compared 5 µg of input DNA (labeled with Cy3) to 10 ng DNA (labeled with Cy5) on the same 244 K Agilent array. The relative intensities of Cy5 and Cy3 signals were similar and highly correlated (R2 = 0.9808 by Pearson correlation analysis) indicating that 10 ng of input DNA can generate accurate MCA array results (Suppl. Fig. 1).

Methylation array results using a cocktail of methylation sensitive restriction enzymes

We next used a cocktail of restriction methylation sensitive enzymes [HpaII, HhaI(Hin6I), HpyCH4IV, Hin1I and AciI]32 to compare DNA methylation in Panc-1 vs. HPDE using the Agilent 44 K B-chip. With the cocktail method DNA is initially cut at non CpG sites (with a GTAC cutter that leaves a TA overhang to facilitate adaptor PCR), resulting fragments can then be amplified unless they are cut at unmethylated restriction sites by one or more of the enzymes in the cocktail. Fragments that lack a methylation-sensitive restriction site are also amplified by adaptor PCR but are not differentially amplified between samples. The method will also detect differences between samples that differ in their restriction site sequences.28 The enzyme cocktail method identified fewer 44 K B chip probes with a statistically significantly elevated fold-change (log2 > 2 fold-change) in Panc-1 vs. HPDE than the MCA method (123 probes vs. 584 probes, respectively) (Chip Analytics Software, Whitehead Error Model) (Suppl. Fig. 3). Only 13 candidate methylated genes were identified as differentially methylated candidate genes in both experiments, despite the fact that many of the methylated promoters identified by the MCA method also contained one or more of the enzyme cocktail’s restriction sites. We suspected that the two enzyme strategies preferentially sample different regions of the genome. We expected the cocktail method to identify more differential methylation outside of extensively methylated CpG islands than the MCA method and might even preferentially amplify GC poor fragments over the GC rich fragments of CpG islands. We examined the CpG island status of the 50 probes with the highest fold-change in Panc-1 vs. HPDE by the MCA and the cocktail method. Thirty-one of the top 50 probes (62%) identified by the MCA method were located within a CpG island and 48 of these 50 probes (96%) had a CpG island within 1 kb either side of the probe. In contrast, only 17 of the top 50 (34%) probes identified using the cocktail method were located within a CpG island and only 28 of these 50 probes (56%) had a CpG island within 1 kb of the probe. To confirm that the MCA method preferentially identifies cancer associated hypermethylation in CpG islands, we also performed another cocktail array experiment comparing methylation using the Agilent 44 k B chip in Panc-1 to another normal pancreas sample in this case the hTERT immortalized pancreatic cell line, HPNE, the only other immortalized non-neoplastic pancreatic cell line available in the literature. Similar results were obtained (Suppl. Fig. 2). Only 11 of the top 50 (22%) probes identified using the cocktail method were located within a CpG island and only 24 of these 50 probes (48%) had a CpG island within 1 kb of the probe. Not surprisingly, many of the same genes were identified as differentially methylated (Log2 > 2) in the Panc-1 vs. HPNE experiment (162 probes representing 134 genes) as were found when Panc-1 and HPDE (123 probes representing 112 genes) were compared (45 genes in common).

Selection of candidate genes for MSP analysis

To prioritize candidate methylated genes that would be useful as markers of pancreatic cancer, we employed a gene expression filtering strategy. We examined the genes silenced in Panc-1 relative to HPDE using the Affymetrix Human genome U133A oligonucleotide microarray. Of 11,317 transcripts expressed in HPDE (had a “present” call in the U133A array), 1,188 transcripts were not expressed in Panc-1 (lacked a present signal). We then merged the expression data for Panc-1 vs. HPDE with the MCA microarray data. Among the genes underexpressed genes in Panc-1 relative to HPDE, 39 were identified by the MCA array experiment as differentially methylated in Panc-1. This list included genes known to be methylated in pancreatic cancer such as TFPI-2 and IGFBP3 (our unpublished data) and other genes such as Twist1. We then examined the U133A gene expression pattern of these 39 genes in four other pancreatic cancer cell lines to identify genes with under expression in other cell lines relative to HPDE. From this list we selected eight genes (BNC1, CNTNAP2, CYP2W1, DEF6, ELOVL4, MDFI, PKP1 and ZNF415) for MSP analysis that had not been previously investigated as targets for aberrant methylation. All of these genes were methylated in Panc-1 and in other pancreatic cancer cell lines but not in HPDE by MSP. We also used bisulfite sequencing to confirm that these four genes were differentially methylated between Panc-1 and HPDE (Table 1 and Fig. 2). We excluded four genes that were methylated by MSP in non-neoplastic pancreas as such genes would not serve as specific markers (PKP1, DEF6, CYP2W1 and BNC1 were methylated in the HPNE line). MSP analysis of the remaining four genes found them to be commonly methylated in the 10 pancreatic cancer cell lines but not methylated in the non-neoplastic pancreatic epithelial lines (Fig. 4). We also examined the methylation status of the 5' CpG island of hsa-miR-9-1, one of the probes with the highest differential methylation in Panc-1 vs. HPDE. We found that hsa-miR-9-1 was methylated in 89% of 57 mostly pancreatic cancer samples and in 15% of normal pancreata (Fig. 2A, B and Fig 4). Since this microRNA is not present on the U133A array we also examined the expression of miR-9 by RT-PCR, the mature form of hsa-mir-9-1 and found HPDE expressed miR-9, but most pancreatic cancer cells did not express this microRNA by qRT-PCR (data not shown).

Figure 4Figure 4
MSP analysis of select genes in pancreatic and periampullary cancers and normal pancreata. Black box indicates completely methylated, white box indicates completely unmethylated, gray box indicates partially methylated genes.

To better determine the potential utility of the four methylated genes (ELOVL4, ZNF415, MDFI and CNTNAP2) and one methylated microRNA precursor (hsa-miR-9-1) as markers of pancreatic ductal adenocarcinoma, we further analyzed the methylation status of these five markers in an additional set of 47 pancreatic and periampullary cancer xenografts and in 32 normal pancreas tissues and two non-neoplastic pancreas epithelial cell lines. As shown in Figure 4, we identified aberrant methylation at all of these loci with a prevalence of methylation of 68% to 96%. These genes were partially methylated in 3% to 15% of 34 normal pancreata. We did not observe any relationship between patient age or gender and methylation of these genes, nor was there any relationship with tumor size, grade or stage although a larger sample size would be needed to identify modest associations.

Discussion

In this study we examine the utility of using methylated CpG island amplification and the Agilent promoter microarray platform to identify differentially methylated genes in pancreatic cancer vs. normal pancreas. We find that the MCA array assay (1) identifies hundreds of differentially methylated CpG islands between pancreatic cancer and normal pancreas samples, with confirmation of all 27 genes (66 probes) evaluated by bisulfite sequencing and/ or MSP, (2) is highly reproducible (correlation between replicates of R2= 0.92), (3) identifies more differential methylation of CpG islands than a methylation sensitive restriction enzyme cocktail, (4) identifies methylation in many genes that are not normally expressed in the pancreatic duct by gene expression array analysis and (5) it can be used with small amounts of input DNA (similar results were obtained using 5 µg and 10 ng of DNA (Suppl. Fig. 2), a useful feature for assaying methylation in clinical samples.

Although the MCA assay interrogates Sma1 sites, all of the genes examined by bisulfite sequencing with high Panc-1/HPDE signals and hypermethylated SmaI sites also had extensive adjacent differential hypermethylation in Panc-1 (Table 1) suggesting that many of these ~600 hypermethylated genes were extensively methylated in their promoters. In contrast, for the genes found to have Sma1 site hypomethylation in Panc-1, CpGs adjacent to these Sma1 sites were often not hypomethylated in Panc-1 despite confirming by bisulfite sequencing hypomethylation in Panc-1 in probes-associated SmaI sites with low Panc-1/HPDE signals. This result suggests that in cancer cells such as Panc-1 extensive hypermethylation rather than hypomethylation of a normally methylated CpG island is more common and when gene hypomethylation does occur it may involve a few CpGs rather than all of the CpG islands. Indeed, we have previously found that gene hypomethylation in pancreatic cancers tends to occur in CpG poor promoters.3

Since the identification of hypermethylated genes in Panc-1 is limited by probes on the Agilent 88 K promoter array and the fraction of promoters measurable by the MCA assay (estimated to be ~70% of CpG islands),33 the number of differentially methylated promoters in Panc-1/HPDE is no doubt higher than 600 genes. Indeed, using the 244 K CpG island array, we identified 1968 differentially methylated CpG islands between MiaPaca2 and normal pancreas, including many CpG islands that are not known to be part of gene promoters. Thus, the large number of differentially methylated genes identified by the MCA array method concurs with the large number of differentially methylated genes in colon cancers determined by analyzing patterns of gene expression after 5Aza-dC and histone deacetylase inhibitor treatment.34 It also concurs with recent results by Issa et al. who used the MCA strategy with the Agilent promoter array platform35 and a 12 K CpG island microarray21 and also found the MCA-based methods can identify hundreds of differentially methylated CpG islands in colon cancers.21 The accuracy and reproducibility the MCA method and its ability topreferentially identifydifferential methylation at CpG islands, suggests that for identifying aberrant CpG island methylation in cancer the MCA method is preferable to the methylation cocktail method32,36 which mostly identified differentially methylated probe candidates outside of CpG islands.

There are now a variety of genome-wide approaches have been utilized to identify differentially methylated genes in cancer.20,22,3739 The 5-Aza-dC method is a very useful method to identify genes reactivated by treatment with DNA methylation inactivation;22 one can also prioritize reactivated genes by gene expression analysis to identify genes that are normally expressed but undergo silencing in cancers,4 or examine differential responses to DAC and HDAC inhibition.34 When we compared our experience with the 5Aza-dC and MCA array methods, we found only a few genes were identified as methylated in Panc-1 both by the MCA array and using 5Aza-dC treatment.4 This is not surprising. Some silenced genes re-expressed by 5Aza-dC do not contain promoter Sma1 sites and will not be identified by MCA array. Many genes identified as methylated in pancreatic cancer cells by MCA array analysis were not expressed in a normal pancreas line indicating that many genes that undergo aberrant methylation in cancer are not normally expressed. Many such genes probably do not undergo reexpression by 5Aza-dC. In addition, the MCA array can identify partially methylated genes that are still actively expressed. Thus, these two methods yield complementary results: MCA array-based methods are suitable for tissue samples, while 5Aza-dC-based methods also providing gene expression information.

Many of the genes we identified as methylated by our microarray approach have been described as methylated previously including numerous homeobox genes and other genes that are targets of the polycomb complex37,40,41 as well as several microRNAs (see Suppl, Table 1). Furthermore, many of the differentially methylated genes identified by MCA could be useful for molecular diagnostic strategies. We identified a new panel of sensitive and specific MSP markers of pancreatic cancer that could be useful for molecular diagnostic strategies. We and others have previously examined the utility of a panel of DNA methylation markers in pancreatic juice samples for the diagnosis of pancreatic cancer.4,11,42 The identification of additional aberrantly methylated genes in pancreatic cancers may help efforts to use these markers as diagnostic tools, particularly since using a panel of markers may help ensure high specificity.11

In conclusion, we find that the MCA array method can detect hundreds of differentially methylated genes in pancreatic adenocarcinoma relative to normal pancreas, is reproducible and can reliably detect DNA methylation profiles using small amounts of input DNA.

Materials and Methods

Cell lines and tissue samples

Ten human pancreatic cancer cell lines were used AsPC1, BxPC3, Capan2, CFPAC1, Hs766T, MiaPaCa2, Panc-1, SU8686, Panc1.28 and Panc2.5. An immortalized cell line derived from normal human pancreatic ductal epithelium (HPDE) and human pancreatic Nestin-expressing cells (HPNE) were generously provided by Dr. Ming-Sound Tsao (University of Toronto, Ontario Canada) and Dr. Michel M. Ouellette (University of Nebraska Medical Center, Omaha, NE), respectively.

Normal and neoplastic tissues were obtained from pancreatic adenocarcinomas resected at the Johns Hopkins Hospital. Normal pancreata were obtained from discarded stored frozen tissues from 11 patients who underwent a pancreatic resection for an intraductal papillary mucinous neoplasm (IPMN) without associated invasive adenocarcinoma, 7 patients for a neuroendocrine tumor and 14 patients for invasive ductal adenocarcinoma. These 32 patients were of similar age to the average age of patients with pancreatic ductal adenocarcinoma (mean age 63.6 years, 18 female, 31 Caucasian and 1 African-American). The histologically normal pancreata from these patients as well as the two non-neoplastic pancreas cell lines (34 samples in all) were chosen to control for possible age-related or field-effect changes in methylation in normal appearing pancreas among patients with a propensity to develop pancreatic cancer. Normal pancreatic tissue from each case was microdissected from glass slides after microscopic examination to obtain DNA from normal pancreas tissue. Genomic DNA was isolated from 47 cancer xenografts (43 pancreatic, 3 distal common bile duct and 1 duodenal cancer) (mean age 67.0, 30 Female, 37 Caucasian, 4 African-American, 2 Oriental, 4 unknown) established from the primary carcinomas as described previously.23 Four of these 43 pancreatic cancers arose in association with an IPMN. Briefly, discarded surgical specimens of patients operated at the Johns Hopkins Hospital were implanted subcutaneously to successive cohorts of mice to either obtain tumor material for these studies or carried over to subsequent mice. The institutional review committee on clinical investigation reviewed and approved the use of these patient samples for this study.

MCA procedure

MCA was performed as described previously with minor modifications.24,25 Briefly 5 µg of DNA was digested with SmaI and XmaI (New England Biolabs). Unmethylated SmaI sites are eliminated by SmaI digestion (which does not cut CCCGGG sites if they contain a methylated CpG), which leaves a blunt end fragment. Methylated SmaI sites are then digested with the non-methylation-sensitive SmaI isoschizomer XmaI, leaving a CCGG overhang (sticky end). Adaptors are ligated to these sticky ends, and PCR is performed to amplify the methylated sequences. The restriction fragments were ligated to an RMCA adapter and amplified by PCR in a 100 µl volume containing 200 pmol of RMCA 24-mer primer, 600 mM Tris-SO4, 2 mM MgSO4, 160 mM (NH4)2(SO4), 200 µM each dNTP, 2% v/v DMSO, 0.5 M betaine and 2 U of Platinum Taq Hifidelity polymerase (Invitrogen, CA). The reaction mixture was incubated at 72°C for 5 min and at 95°C for 3 min, and then subject to 25 cycles of 1 min at 95°C and 3 min at 77°C followed by a final extension of 10 min at 77°C.

For the duplicate MCA experiment DNA was again extracted from the pancreatic cancer cell line; Panc-1 and the non-neoplastic pancreatic cancer cell line; HPDE, respectively and the MCA procedure repeated.

For the low-input (10 ng) DNA experiment, we made additional modifications to the MCA procedure. The main modification was to change the polymerase to Titanium Taq. The rationale for this was to optimize the amplification step and ensure that any PCR-induced errors in amplification associated with low-input DNA were minimized. Briefly, 10 ng or 5 µg of SU8686 DNA was digested with 10 units of SmaI for 16 h and then digested with two units of XmaI for 6 h. DNA fragments were then precipitated with ethanol and ligated to 0.05 nmol of RMCA adaptor. PCR amplifications were carried out using 30 µl of the whole ligation mix as a template in a 200 µl volume containing 1.0 M GC-melt, 4 µl of 100 µM RMCA24 primer, 28 µl 1154 of each 2.5 mM dNTP, 20 µl of 10x Titanium Taq PCR buffer, and 4 µl of 50x Titanium Taq DNA polymerase (Clontech, Palo Alto, CA). The reaction mixture was incubated at 72°C at 5 min and at 95°C for 3 min. Samples were then subjected to 25 cycles of amplification consisting of 1 min at 95°C and 3 min either at 77°C in a thermal cycler. The final extension time was 10 minutes.

Five-enzyme methylation-sensitive restriction enzyme cocktail

We employed a previously described method of enriching for methylated sequences using a 5 methylation-sensitive enzyme cocktail.28 Genomic DNA was cleaved by Csp6I (which recognizes GTAC) (10 U/ml in G+/−buffer, Fermentus Life Sciences, Lithuania) for 8 h at 37°C. The digested DNA was purified using QIAquick PCR Purification Kit (Qiagen, Valencia, CA) and ligated to TA-overhang specific TA-1 adaptor. The ligation reaction was carried out at 22°C for 18 h, followed by a heat-inactivation step at 65°C for 5 min. The unmethylated ligation products are eliminated from the reaction by cleavage with a cocktail of methylation-sensitive restriction enzymes of HpaII (C/CGG), HhaI (Hin6I) (GCG/C), HpyCH4IV (A/CGT), Hin1I (GR/CGYC) and AciI (CCGC) in 2xY+/Tango buffer (Fermentus) for 8 h at 37°C. Since not all fragments restricted by Csp6I will contain one of these five enzyme sites, not all unmethylated fragments are removed by this step.

Ligated fragments were then amplified by PCR in a 100 µl volume containing 200 pmol of TA-1b primer, 600 mM Tris-SO4, 2.5 mM MgSO4, 160 mM (NH4)2(SO4), 200 µM each dNTP, 2% v/v DMSO, 0.5 M betaine and 2 U of Platinum Taq Hifidelity polymerase (Invitrogen, CA). The reaction mixture was incubated at 72°C for 5 min and at 95°C for 2 min, and then subject to 25 cycles of 30 sec at 94°C and 2 min at 67°C followed by a final extension of 10 min at 67°C. After purification, 2 µg of amplicon was subjected to microarray hybridization.

Agilent human promoter and CpG island microarrays

Agilent’s Human promoter chip-on-chip microarray was used as the main platform for assessing promoter methylation. This array platform has 87,922 probes arrayed on two 44 K slides. The A slide contains probes for promoters from chromosome 1–10 and the B slide 10–23, X and Y. Array hybridization was performed by the Sidney Kimmel Cancer Center Microarray Core Facility at Johns Hopkins. Briefly, 2 µg of MCA amplicon was labeled with Cy3-dUTP or Cy5-dUTP (Perkin Elmer) by using BioPrime DNA Labeling System (Invitrogen). These dye-labeled amplicon mixtures were then mixed and co-hybridized to each of the 44 K human promoter ChIP-on-chip microarrays. After the hybridization, the microarray slides were washed, dried and scanned using an Agilent G2505B scanner. During the course of these studies, the Agilent 244 K CpG island microarray became available which was then used for the low-input DNA experiment and to compare methylation profiles of MiaPaca2 to normal pancreas. This 244 K chip contains 195 K CpG island probes and 50 K non-CpG island probes and uses the same probe design and array characteristics as the 44 K promoter array chip. Data were extracted with Agilent Feature Extraction 9.1 software. Methylation-specific sites were called using Agilent’s ChIP Analytics 1.3 software, incorporating the Whitehead Error Model, which has a false-positive rate of approximately 0.5% and a false-negative rate of approximately 20%.26

Bisulfite sequencing

The methylation status of the 5' CpG islands of each gene was determined by bisulfite sequencing as described previously (11). DNA samples were treated with sodium bisulfite (Sigma) for 3 h at 70°C. After purification with the Wizard DNA clean-up system (Promega, Madison, WI), 3 µl of bisulfite-treated DNA was amplified by PCR with RDA buffer. PCR conditions were as follows: (a) 95°C for 5 min; (b) 45 cycles of 95°C for 20 s, 61°C for 20 s and 72°C for 30 s; and (c) a final extension of 4 min at 72°C. Amplicons were purified with the QIAquick PCR purification kit (Qiagen, Germantown, MD). Sequence analysis was carried out at the Johns Hopkins Core Sequencing Facility using automated DNA sequencers (Applied Biosystems). PCR primer sequences for all genes analyzed in this study are available on request.

Methylation specific PCR (MSP)

After the bisulfite-treatment, 1 µl of treated DNA was amplified using primers specific for either methylated or unmethylated DNA. Primers were designed to detect the sequence differences between methylated and unmethylated DNA as a result of bisulfite modification, and each primer pair contained 5–8 CpG sites to provide optimal specificity. Primer sequences are available on request. PCR conditions were as follows: (a) 95°C for 5 min; (b) 40 cycles of 95°C for 20 s, 60°C–62°C for 20 s and 72°C for 30 s; and (c) a final extension of 4 min at 72°C. Five µl of each PCR product were loaded onto 3% agarose gels and visualized by ethidium bromide staining. MSP results were defined as completely methylated if there was no amplification of the corresponding unmethylated alleles and completely unmethylated if there was no amplification of methylated alleles. Samples were regarded as partially methylated if both a methylated and an unmethylated band were amplified.

Oligonucleotide array data

Gene expression in the cell lines Panc-1 and HPDE was examined using data we previously generated using the Affymetrix U133A oligonucleotide arrays (Affymetrix, Santa Clara, CA).27 Signal intensity for each transcript was normalized and analyzed with the Affymetrix Expression Console Software 1.0 MAS 5.0 algorithm.

Statistical analysis

Values reported are means ± SD. All data were normally distributed and underwent equal variance testing. Statistical analysis was determined by SPSS program at 11.0.1J Windows version. Mean differences between two subgroups were compared using Student’s t-test. p < 0.05 was considered as statistically significant. The correlation between replicated microarray results was determined using a Pearson correlation model.

Supplementary Material

CBT GMP

Note

Supplementary materials can be found at: www.landesbioscience.com/supplement/OmuraCBT7-7-Sup.pdf

Acknowledgements

We thank Dr. Wayne Yu of the Johns Hopkins Sidney Kimmel Comprehensive Cancer Center Microarray Core for his helpful discussions.

Supported by the NCI grants Specialized Programs of Research Excellence in Gastrointestinal Malignancies (CA62924), the Jimmy V. foundation, the Michael Rolfe Foundation.

References

1. Jones PA, Baylin SB. The epigenomics of cancer. Cell. 2007;128:683–692. [PMC free article] [PubMed]
2. Sato N, Fukushima N, Chang R, Matsubayashi H, Goggins M. Differential and epigenetic gene expression profiling identifies frequent disruption of the RELN pathway in pancreatic cancers. Gastroenterology. 2006;130:548–565. [PubMed]
3. Sato N, Maitra A, Fukushima N, van Heek NT, Matsubayashi H, Iacobuzio-Donahue CA, Rosty C, Goggins M. Frequent hypomethylation of multiple genes overexpressed in pancreatic ductal adenocarcinoma. Cancer Res. 2003;63:4158–4166. [PubMed]
4. Sato N, Fukushima N, Maitra A, Matsubayashi H, Yeo CJ, Cameron JL, Hruban RH, Goggins M. Discovery of novel targets for aberrant methylation in pancreatic carcinoma using high-throughput microarrays. Cancer Res. 2003;63:3735–3742. [PubMed]
5. Sato N, Goggins M. Epigenetic alterations in intraductal papillary mucinous neoplasms of the pancreas. J Hepatobiliary Pancreat Surg. 2006;13:280–285. [PubMed]
6. Sato N, Goggins M. The Role of Epigenetic Alterations in Pancreatic Cancer. J Hepatobiliary Pancreat Surg. 2006;13:286–295. [PubMed]
7. Sato N, Matsubayashi H, Abe T, Fukushima N, Goggins M. Epigenetic downregulation of CDKN1C/p57KIP2 in pancreatic ductal neoplasms identified by gene expression profiling. Clin Cancer Res. 2005;11:4681–4688. [PubMed]
8. Fukushima N, Sato N, Ueki T, Rosty C, Walter KM, Yeo CJ, Hruban RH, Goggins M. Preproenkephalin and p16 gene CpG island hypermethylation in pancreatic intraepithelial neoplasia (PanIN) and pancreatic ductal adenocarcinoma. Am J Pathol. 2002;160:1573–1581. [PMC free article] [PubMed]
9. Sato N, Ueki T, Fukushima N, Iacobuzio-Donahue CA, Yeo CJ, Cameron JL, Hruban RH, Goggins M. Aberrant Methylation of CpG Islands in Intraductal Papillary Mucinous Neoplasms of the Pancreas Increases with Histological Grade. Gastroenterology. 2002;123:1365–1372. [PubMed]
10. Sato NFN, Hruban RH, Goggins M. CpG island methylation profile of pancreatic intraepithelial neoplasia Mod Pathol. 2007;21:238–244. [PMC free article] [PubMed]
11. Matsubayashi H, Canto M, Sato N, Klein A, Abe T, Yamashita K, Yeo CJ, Kalloo A, Hruban R, Goggins M. DNA methylation alterations in the pancreatic juice of patients with suspected pancreatic disease. Cancer Res. 2006;66:1208–1217. [PubMed]
12. Issa JP, Ottaviano YL, Celano P, Hamilton SR, Davidson NE, Baylin SB. Methylation of the oestrogen receptor CpG island links ageing and neoplasia in human colon. Nat Genet. 1994;7:536–540. [PubMed]
13. Matsubayashi H, Skinner H, Iacobuzio-Donahue C, Abe T, Sato N, Riall TS, Yeo CJ, Kern SE, Goggins M. Pancreaticobiliary cancers with deficient methylenetetrahydrofolate reductase genotypes. Clin Gastro Hepatol. 2005;3:752–760. [PubMed]
14. Matsubayashi H, Sato N, Brune K, Blackford AL, Hruban RH, Canto M, Yeo CJ, Goggins M. Age- and Disease-Related Methylation of Multiple Genes in Non-neoplastic Duodenal tissues. Clin Cancer Res. 2005;11:573–583. [PubMed]
15. Di Croce L, Raker VA, Corsaro M, Fazi F, Fanelli M, Faretta M, Fuks F, Coco FL, Kouzarides T, Nervi C, Minucci S, Pelicci PG. Methyltransferase Recruitment and DNA Hypermethylation of Target Promoters by an Oncogenic Transcription Factor. Science. 2002;295:1079–1082. [PubMed]
16. Fraga MF, Ballestar E, Paz MF, Ropero S, Setien F, Ballestar ML, Heine-Suner D, Cigudosa JC, Urioste M, Benitez J, Boix-Chornet M, Sanchez-Aguilera A, et al. Epigenetic differences arise during the lifetime of monozygotic twins. Proc Natl Acad Sci USA. 2005;102:10604–10609. [PMC free article] [PubMed]
17. Ting AH, McGarvey KM, Baylin SB. The cancer epigenome—components and functional correlates. Genes Dev. 2006;20:3215–3231. [PubMed]
18. Costello JF, Fruhwald MC, Smiraglia DJ, Rush LJ, Robertson GP, Gao X, Wright FA, Feramisco JD, Peltomaki P, Lang JC, Schuller DE, Yu L, et al. Aberrant CpG-island methylation has non-random and tumour-type-specific patterns [In Process Citation] Nat Genet. 2000;24:132–138. [PubMed]
19. Keshet I, Schlesinger Y, Farkash S, Rand E, Hecht M, Segal E, Pikarski E, Young RA, Niveleau A, Cedar H, Simon I. Evidence for an instructive mechanism of de novo methylation in cancer cells. Nat Genet. 2006;38:149–153. [PubMed]
20. Rauch T, Li H, Wu X, Pfeifer GP. MIRA-Assisted Microarray Analysis, a New Technology for the Determination of DNA Methylation Patterns, Identifies Frequent Methylation of Homeodomain-Containing Genes in Lung Cancer Cells. Cancer Res. 2006;66:7939–7947. [PubMed]
21. Estecio MR, Yan PS, Ibrahim AE, Tellez CS, Shen L, Huang TH, Issa JP. High-throughput methylation profiling by MCA coupled to CpG island microarray. Genome Research. 2007;17:1529–1536. [PMC free article] [PubMed]
22. Suzuki H, Gabrielson E, Chen W, Anbazhagan R, van Engeland M, Weijenberg MP, Herman JG, Baylin SB. A genomic screen for genes upregulated by demethylation and histone deacetylase inhibition in human colorectal cancer. Nat Genet. 2002;31:141–149. [PubMed]
23. Rubio-Viqueira B, Jimeno A, Cusatis G, Zhang X, Iacobuzio-Donahue C, Karikari C, Shi C, Danenberg K, Danenberg PV, Kuramochi H, Tanaka K, Singh S, et al. An in vivo platform for translational drug development in pancreatic cancer. Clin Cancer Res. 2006;12:4652–4661. [PubMed]
24. Ueki T, Toyota M, Skinner H, Walter KM, Yeo CJ, Issa JP, Hruban RH, Goggins M. Identification and characterization of differentially methylated CpG islands in pancreatic carcinoma. Cancer Res. 2001;61:8540–8546. [PubMed]
25. Toyota M, Ho C, Ahuja N, Jair KW, Li Q, Ohe-Toyota M, Baylin SB, Issa JP. Identification of differentially methylated sequences in colorectal cancer by methylated CpG island amplification. Cancer Res. 1999;59:2307–2312. [PubMed]
26. Boyer LA, Lee TI, Cole MF, Johnstone SE, Levine SS, Zucker JP, Guenther MG, Kumar RM, Murray HL, Jenner RG, Gifford DK, Melton DA, et al. Core transcriptional regulatory circuitry in human embryonic stem cells. Cell. 2005;122:947–956. [PMC free article] [PubMed]
27. Sato N, Fukushima N, Maitra A, Matsubayashi H, Yeo CJ, Cameron JL, Hruban RH, Goggins M. Discovery of novel targets for aberrant methylation in pancreatic carcinoma using high-throughput microarrays. Cancer Res. 2003;63:3735–3742. [PubMed]
28. Carrel L, Cottle AA, Goglin KC, Willard HF. A first-generation X-inactivation profile of the human X chromosome. Proc Natl Acad Sci USA. 1999;96:14440–14444. [PMC free article] [PubMed]
29. Ke X, Collins A. CpG islands in human X-inactivation. Ann Hum Genet. 2003;67:242–249. [PubMed]
30. Ueki T, Toyota M, Skinner H, Walter KM, Yeo CJ, Issa JP, Hruban RH, Goggins M. Identification and characterization of differentially methylated CpG islands in pancreatic carcinoma. Cancer Res. 2001;61:8540–8546. [PubMed]
31. Sato N, Parker AR, Fukushima N, Miyagi Y, Iacobuzio-Donahue CA, Eshleman JR, Goggins M. Epigenetic inactivation of TFPI-2 as a common mechanism associated with growth and invasion of pancreatic ductal adenocarcinoma. Oncogene. 2005;24:850–858. [PubMed]
32. Schumacher A, Kapranov P, Kaminsky Z, Flanagan J, Assadzadeh A, Yau P, Virtanen C, Winegarden N, Cheng J, Gingeras T, Petronis A. Microarray-based DNA methylation profiling: technology and applications. Nucleic Acids Res. 2006;34:528–542. [PMC free article] [PubMed]
33. Toyota M, Ahuja N, Ohe-Toyota M, Herman JG, Baylin SB, Issa JP. CpG island methylator phenotype in colorectal cancer. Proc Natl Acad Sci USA. 1999;96:8681–8686. [PMC free article] [PubMed]
34. Schuebel KE, Chen W, Cope L, Glockner SC, Suzuki H, Yi JM, Chan TA, Neste LV, Criekinge WV, Bosch SV, van Engeland M, Ting AH, et al. Comparing the DNA Hypermethylome with Gene Mutations in Human Colorectal Cancer. PLoS Genet. 2007;3:157. [PMC free article] [PubMed]
35. Shen L, Kondo Y, Guo Y, Zhang J, Zhang L, Ahmed S, Shu J, Chen X, Waterland RA, Issa JP. Genome-wide profiling of DNA methylation reveals a class of normally methylated CpG island promoters. PLoS Genet. 2007;3:2023–2036. [PMC free article] [PubMed]
36. Taylor KH, Pena-Hernandez KE, Davis JW, Arthur GL, Duff DJ, Shi H, Rahmatpanah FB, Sjahputera O, Caldwell CW. Large-scale CpG methylation analysis identifies novel candidate genes and reveals methylation hotspots in acute lymphoblastic leukemia. Cancer Res. 2007;67:2617–2625. [PubMed]
37. Schlesinger Y, Straussman R, Keshet I, Farkash S, Hecht M, Zimmerman J, Eden E, Yakhini Z, Ben-Shushan E, Reubinoff BE, Bergman Y, Simon I, et al. Polycomb-mediated methylation on Lys27 of histone H3 pre-marks genes for de novo methylation in cancer. Nat Genet. 2007;39:232–236. [PubMed]
38. Bibikova M, Lin Z, Zhou L, Chudin E, Garcia EW, Wu B, Doucet D, Thomas NJ, Wang Y, Vollmer E, Goldmann T, Seifart C, Jiang W, Barker DL, Chee MS, Floros J, Fan JB. High-throughput DNA methylation profiling using universal bead arrays. Genome Res. 2006;16:383–393. [PMC free article] [PubMed]
39. Lopez-Serra L, Ballestar E, Ropero S, Setien F, Billard LM, Fraga MF, Lopez-Nieva P, Alaminos M, Guerrero D, Dante R, Esteller M. Unmasking of epigenetically silenced candidate tumor suppressor genes by removal of methyl-CpG-binding domain proteins. 2008 [PubMed]
40. Rauch T, Wang Z, Zhang X, Zhong X, Wu X, Lau SK, Kernstine KH, Riggs AD, Pfeifer GP. Homeobox gene methylation in lung cancer studied by genome-wide analysis with a microarray-based methylated CpG island recovery assay. Proc Natl Acad Sci USA. 2007;104:5527–5532. [PMC free article] [PubMed]
41. Ohm JE, McGarvey KM, Yu X, Cheng L, Schuebel KE, Cope L, Mohammad HP, Chen W, Daniel VC, Yu W, Berman DM, Jenuwein T, et al. A stem cell-like chromatin pattern may predispose tumor suppressor genes to DNA hypermethylation and heritable silencing. Nat Genet. 2007;39:237–242. [PMC free article] [PubMed]
42. Fukushima N, Walter KM, Ueki T, Sato N, Matsubayashi H, Cameron JL, Hruban RH, Canto M, Yeo CJ, Goggins M. Diagnosing pancreatic cancer using methylation specific PCR analysis of pancreatic juice. Cancer Biol Ther. 2003;2:78–83. [PubMed]

Formats:

Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...

Links

  • MedGen
    MedGen
    Related information in MedGen
  • PubMed
    PubMed
    PubMed citations for these articles

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...