JMD ASIP MEMBERSHIP
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
Originally published online as doi:10.2353/jmoldx.2007.060209 on August 9, 2007

Published online before print August 9, 2007
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow All Versions of this Article:
jmoldx.2007.060209v1
9/4/510    most recent
Right arrow Purchase Article
Right arrow View Shopping Cart
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Dejeux, E.
Right arrow Articles by Tost, J.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Dejeux, E.
Right arrow Articles by Tost, J.
Journal of Molecular Diagnostics 2007, Vol. 9, No. 4
Copyright © 2007 American Society for Investigative Pathology & Association for Molecular Pathology
DOI: 10.2353/jmoldx.2007.060209

Rapid Identification of Promoter Hypermethylation in Hepatocellular Carcinoma by Pyrosequencing of Etiologically Homogeneous Sample Pools

Emelyne Dejeux*, Virginie Audard{dagger}, Catherine Cavard{dagger}, Ivo Glynne Gut*, Benoit Terris{dagger}{ddagger} and Jörg Tost*

From the Laboratory for Epigenetics, * Centre National de Génotypage, Evry; the Department of Endocrinology, Metabolism and Cancer, {dagger} Institut Cochin, INSERM U567, Centre National de la Recherche Scientifique UMR-S 8104 and University Paris 5, Faculté de Médecine René Descartes, UM3, Paris; and the Service d’Anatomopathologie, {ddagger} Hôpital Cochin, Assistance Publique-Hôpitaux de Paris, Paris, France


   Abstract
Top
 Abstract
Introduction
 Materials and Methods
 Results
 Discussion
 References
 
Aberrant DNA methylation patterns have been identified in a variety of human diseases, particularly cancer. Pyrosequencing has evolved in recent years as a sensitive and accurate method for the analysis and quantification of the degree of DNA methylation in specific target regions. However, the number of candidate genes that can be analyzed in clinical specimens is often restricted by the limited amount of sample available. Here, we present a novel screening approach that enables the rapid identification of differentially methylated regions such as promoters by pyrosequencing of etiologically homogeneous sample pools after bisulfite treatment. We exemplify its use by the analysis of five genes (CDKN2A, GSTP1, MLH1, IGF2, and CTNNB1) involved in the pathogenesis of human hepatocellular carcinoma using pools stratified for different parameters of clinical importance. Results were confirmed by the individual analysis of the samples. The screening identified all genes displaying differential methylation successfully, and no false positives occurred. Quantitative comparison of the pools and the samples in the pool analyzed individually showed a deviation of ~1.5%, making the method ideally suited for the identification of diagnostic markers based on DNA methylation while saving precious DNA material.


   Introduction
Top
 Abstract
 Introduction
Materials and Methods
 Results
 Discussion
 References
 
DNA methylation occurring at the 5-position of cytosines in the context of the dinucleotide CpG is of particular importance for proper development and gene regulation and is strongly implicated in the pathogenesis of various diseases.1 Epigenetic deregulation, ie, aberrant DNA methylation levels, altered patterns of histone tail modifications, and chromatin structure, has now been recognized as a hallmark of cancer.2, 3, 4 An overall decrease in DNA methylation (global hypomethylation) is accompanied by a region- and gene-specific hypermethylation of CpG islands. Genes involved in DNA repair, detoxification, cell-cycle regulation, and apoptosis are thereby often inappropriately inactivated. Epigenetic changes occur early in the progression process and often precede malignancy, and epigenetic lesions in normal tissue might set the stage for neoplasia.5 Methylation patterns can be shared by different types of tumors as well as being tumor type-specific.6 Analysis of DNA methylation patterns has proven useful as biomarker for the early diagnosis, classification, prognosis, and therapy of human cancers.7, 8, 9 Realizing the important participation of DNA methylation in the pathogenesis of cancer and other diseases, a variety of techniques for the study of DNA methylation has been developed in the last few years.2, 7, 8

One of the methods that has recently received much attention for the simultaneous analysis and quantification of the degree of methylation at several CpG positions in close proximity is pyrosequencing.10, 11, 12, 13 The pyrosequencing technology is based on the luminometric detection of pyrophosphate that is released on nucleotide incorporation and converted into a light signal by a cascade consisting of four enzymes.14 Its ease-of-use, high reliability, and flexibility have made pyrosequencing an analysis platform that has been widely used for various diagnostic applications such as routine (multiplex) genotyping,15, 16 bacterial typing,17, 18, 19 and sensitive detection of mutations.20 One of its major strengths is the quantitative nature of the results. The bioluminometric response is linear (R2 > 0.99) for the sequential addition of up to five identical nucleotides (C, G, and T) or three {alpha}-S-dATPs. Pyrosequencing has therefore been used as a method to determine allele frequencies of SNPs in pools of samples. Deviations from results obtained by genotyping of individual samples were as low as 1.6% (determined from peak height) in a pool of 1126 samples, and the technology was found to be able to detect differences in allele frequencies of less than 2% between pools of DNA.21, 22, 23

Pyrosequencing is ideally suited for DNA methylation analysis after bisulfite treatment of DNA because it combines the ability of direct quantitative sequencing, reproducibility, speed, and ease-of-use and is becoming more and more used. Besides the identification of genes aberrantly silenced by promoter hypermethylation in cancer,24, 25 pyrosequencing has been used as a reference method for validation of newly developed methods for DNA methylation analysis,26 to monitor chemically induced demethylation in leukemia patients,27 as well as for a diagnostic test for aberrant methylation in the imprinting disorders Prader-Willi and Angelman syndromes.28 One of the major limitations for the identification of novel genes involved in tumorigenesis is the amount of available sample that is in most cases derived from primary tumors. The successive use of several sequencing primers on the same DNA template (serial pyrosequencing) significantly reduces cost, labor, and analysis time as well as saving precious DNA samples for the analysis of a specific region amplified in a single polymerase chain reaction (PCR).29 However, it still requires prior knowledge of the presence of epimutations in the target. Using a candidate gene approach, we found in various projects that only 5 to 10% of genes hypothesized to display aberrant methylation actually show differential methylation in a certain tumor type (J.T., unpublished data). Here, we present a novel method that enables rapid screening for differential methylation using etiologically homogeneous pools of samples. We analyzed quantitatively the DNA methylation levels at 112 CpGs in the promoter or differentially methylated region (DMR) of five genes that have previously been found to be implicated in the pathogenesis of hepatocellular carcinoma (HCC): the cell-cycle regulator CDKN2A (p16),30 glutathione S-transferase {pi} (GSTP1) involved in detoxification and drug resistance,31 the insulin-like growth factor 2 (IGF2),32 the DNA mismatch repair gene MLH1,33 and ß-catenin (CTNNB1), which is mutated in 26 to 40% of human HCCs and not regulated by methylation of its promoter.34, 35 Samples were grouped into pools after bisulfite treatment depending on the etiology (hepatitis B or C virus infection and alcohol consumption) and on the tumoral and nontumoral tissue status: control liver, noncirrhotic or cirrhotic adjacent nontumoral liver, hepatic adenoma, and HCC with and without ß-catenin mutation. All genes displaying variable methylation patterns were successfully detected in the pools, and no false-negative results occurred. The high quantitative concordance of the data accumulated by single sample analysis compared with the pooled sample amplification clearly demonstrates the ability of our method to preserve precious sample resources from clinical specimens without loss of accuracy. Our method therefore provides the possibility to screen a large number of potential target genes even with a limited amount of sample.


   Materials and Methods
Top
 Abstract
 Introduction
 Materials and Methods
Results
 Discussion
 References
 
Samples
Tumor samples were collected in accordance with French law and ethical guidelines, and their use was approved by the institutional ethics committee. Operative notes and pathology reports were subjected to exhaustive review. Samples used in this study were 27 HCCs and three hepatic adenomas and their nontumoral counterparts and 10 normal control livers without cirrhosis. DNA from peritumoral liver tissue, a hepatic adenoma, and the carcinoma was available from one patient. The degree of hepatic fibrosis in noncancerous liver was graded according to the METAVIR classification.36 Seventeen HCCs developed on extensive fibrosis or cirrhosis (F3 and F4). Etiologies of the tumors were hepatitis C virus (HCV) infection for 11 cases (41%), hepatitis B virus (HBV) infection for four cases (15%), and alcohol consumption for seven cases (26%). Five cases (18%) developed in the absence of known etiology. Fourteen HCC and the three hepatic adenomas displayed ß-catenin mutations after RNA isolation and further sequencing with the primers F1/R1, F1/R2, and F2/R2, as described previously.35

Preparation of DNAs
DNAs were extracted from frozen tissue samples using standard methods, and DNA concentrations of extracted DNA were determined using the Quant-iT dsDNA broad range assay kit (Invitrogen, Cergy Pontoise, France) on a SpectraMAX Gemini XPS microplate spectrofluorometer (Molecular Devices, St. Grégoire, France) and normalized to a concentration of 50 ng/µl. Unmethylated DNA was obtained by whole genome amplification of a DNA extracted from a lymphoblastoid cell line from the CEPH/Utah collection using the REPLI-g kit (Qiagen, Courtaboeuf, France) using 50 ng as input DNA.

Highly methylated DNA was obtained by treating genomic DNA of the same lymphoblastoid cell line with the methylase SssI (Ozyme, St. Quentin-en-Yvelines, France). In brief, 7.5 µl of NE-buffer 2, 10 nmol of S-adenosylmethionine (SAM), and 6 U of SssI were added to 4.5 µg of human genomic DNA in a final volume of 67.5 µl. The solution was incubated at 37°C in a water bath. After 3 hours and again after an additional 2 hours, 10 nmol of SAM and 6 U of SssI were added, and the reaction was incubated overnight at 37°C. The enzyme was inactivated at 95°C for 5 minutes and DNA stored at –20°C until further use.

Bisulfite Treatment and Adjustment of Concentration
One µg of DNA was bisulfite converted using the MethylEasy HT kit for Centrifuge (Human Genetic Signatures, North Ryde, NSW, Australia) according to the manufacturer’s instructions. Concentration of the bisulfite converted DNAs was determined using the NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE) and normalized to a concentration of 20 ng/µl. Pools were constructed from bisulfite-converted DNAs.

DNA Methylation Analysis by Pyrosequencing
Quantitative DNA methylation analysis of the bisulfite-treated DNA was performed by pyrosequencing or, in case of several sequencing primers, by serial pyrosequencing.29 Primers for PCR amplification and pyrosequencing were purchased from Biotez (Buch, Germany). Regions of interest were amplified using 20 ng of bisulfite-treated human genomic DNA and 5 to 7.5 pmol of forward and reverse primer, one of them being biotinylated. Sequences of the oligonucleotides for PCR amplification and pyrosequencing are given in Table 1Go Go . Reaction conditions were 1x HotStar Taq buffer supplemented with 1.6 mmol/L MgCl2, 200 mmol/L dNTPs, and 2.0 U of HotStar Taq polymerase (Qiagen) in a 25-µl volume. The PCR program consisted of a denaturing step of 15 minutes at 95°C followed by 50 cycles of 30 seconds at 95°C, 30 seconds at the respective annealing temperature (Table 1)Go Go , and 20 seconds at 72°C, with a final extension of 5 minutes at 72°C. Amplification products were purified and rendered single-stranded on a Pyrosequencing workstation (Pyrosequencing AB, Uppsala, Sweden). PCR products were incubated for 10 minutes at room temperature with 51 µl of binding buffer (10 mmol/L Tris, 2 mol/L NaCl, 1 mmol/L ethylenediamine tetraacetic acid, and 0.1% Tween 20, pH 7.6, adjusted with 1 mol/L HCl) and 4 µl of streptavidin-coated Sepharose beads (GE Health Care, Uppsala, Sweden). The binding mix was aspirated, and the template was successively washed with 70% ethanol, rendered single-stranded with 0.2 mol/L NaOH, and neutralized with washing buffer (10 mmol/L Tris, pH 7.6, adjusted with 4 mol/L acetic acid). Beads were released into 40 µl of annealing buffer (20 mmol/L Tris and 2 mmol/L magnesium acetate, pH 7.6, adjusted with 4 mol/L acetic acid) containing 15 pmol of the respective sequencing primer (Table 1)Go Go . Primers were annealed to the target by incubation at 80°C for 2 minutes. Quantitative DNA methylation analysis was performed on a PSQ 96MA system with the PyroGold SQA reagent kit (Pyrosequencing), and results were analyzed using the Q-CpG software (V1.0.9; Pyrosequencing). Stripping of the template strand for subsequent annealing of a new sequencing primer (serial pyrosequencing) was performed by adding 20 µl of binding buffer to the completed sequencing reaction and resuspending the Sepharose beads. The binding mix was then purified without further incubation and the biotinylated template strand rendered again single-stranded using the above-described purification protocol. This process completely removes all DNA strands that have been de novo synthesized during the last sequencing run as well as remaining sequencing primers.29


View this table:
[in this window]
[in a new window]

  		Table 1. Sequences of Primers Used for Amplification and Pyrosequencing Reactions, Including Genbank Accession Numbers and Nucleotides (Nt) Corresponding to the Amplified Fragments as Well as the Annealing Temperatures for the Respective PCR Amplifications

 

View this table:
[in this window]
[in a new window]

  		Table 1A. Continued

 
Statistics
Quantitative DNA methylation values are represented by boxplots in which the distribution of data points in a sample set is displayed without any assumption of statistical distribution. Lower and upper quartiles of the data delimit the box, the median is represented by the bold black dot in the box, and minimal and maximum values are indicated by the lines. A Mann-Whitney U-test was used to compare quantitative methylation values between two groups. For IGF2, the average of all samples was calculated, and for each sample, the absolute distance of the methylation value from this average was used to define values for ranking because both hypo- and hypermethylation occur in the tumors.


   Results
Top
 Abstract
 Introduction
 Materials and Methods
 Results
Discussion
 References
 
The rationale behind this study was to evaluate pyrosequencing-based DNA methylation analysis in pools of samples that are homogenized for distinct etiologies to enable rapid and cost-effective screening of multiple genes potentially displaying differential methylation in a candidate region. As a proof-of-principle, we chose to analyze the methylation patterns in the promoter regions of four genes (CDKN2A, GSTP1, MLH1, and CTNNB1) and the differentially methylated region 2 (DMR2) of IGF2 in a panel of 71 samples (10 control livers, 27 paired HCC samples, three paired adenomas, and two additional peritumoral livers). Aberrant methylation in hepatocellular carcinogenesis had previously been reported for CDKN2A, GSTP1, MLH1, and IGF2, but not for CTNNB1, which is mutated in a subset of HCCs.

Limit of Detection
We and others12, 21 have previously shown that pyrosequencing-based DNA analysis has a limit of detection of ~3% for the minor component of a quantitative signal and a quantitative resolution of 5% or better. However, because this parameter is of utmost importance for the approach, we chose to reverify the limit of detection using the CDKN2A amplicon as a model system. Completely in vitro methylated and unmethylated DNAs were bisulfite-treated and normalized to a concentration of 20 ng/µl using a NanoDrop spectrophotometer. The use of the more accurate fluorescent dyes as used in the Quant-iT kit is no longer possible because these are highly selective for double-stranded DNA, and strands are no longer complementary after bisulfite treatment. The maximal error in the determination of the concentration after normalization because of pipetting and other random fluctuations was estimated to be 20 ± 1.3 ng/µl (6.5%). Completely methylated DNA was diluted into the unmethylated DNA to create mixtures with a methylation degree of 0, 2, 5, and 10%. Figure 1Go confirms the limit of detection being at 2% and clearly demonstrates the ability of our approach to detect methylation differences as low as 2 to 5%. We therefore decided to divide the samples into pools consisting of a maximum of eight samples. This approach permits the identification of aberrant methylation if a single sample displays a methylation degree of 20 to 25% against a background of seven unmethylated samples. If more than one sample is methylated, methylation levels of ~10% are sufficient to be detected.


Figure 1
View larger version (40K):
[in this window]
[in a new window]

  		Figure 1. Pyrograms obtained by the analysis of mixtures with a known degree of methylation in the promoter of CDKN2A with 0% (A), 2% (B), 5% (C), and 10% (D) of methylation. E: The linearity of the signal for the sixth CpG position shown in A–D is demonstrated. Quantitative differences as low as 2 to 5% can be detected by pyrosequencing. Similar linear regression coefficients were obtained for the other five CpGs. CpG 1: y = 1.3 x + 2.3, R2 = 0.998; CpG 2: y = 1.57 x + 2.3, R2 = 0.997; CpG 3: y = 1.20 x + 1.0, R2 = 0.984; CpG 4: y = 0.88 x + 6.6, R2 = 0.978; CpG 5: y = 0.79 x + 4.2, R2 = 0.998. F: The linearity of the signal throughout the entire dynamic range (0 to 100%) for the sixth CpG position is shown. All assays were performed in triplicate. The resolution and linear correlation is similar for IGF2 as demonstrated in G.

 
To assess the variation induced by bisulfite treatment, two independent bisulfite treatments were performed, and quantitative DNA methylation levels were analyzed at the CDKN2A promoter. Methylation degrees at individual CpG positions between showed a high degree of reproducibility with a mean variation of 1.9% (range, –5.9 to +6.3 for individual CpG positions). This variability is in the same range as the variation observed in independent PCR amplifications. These results are in good concordance with recently published results demonstrating a SD of 3.5% between seven independent bisulfite treatments.37 Bisulfite treatment does therefore not have a major influence on quantitative methylation values if well standardized.

Construction of Pools
For proof-of-principle, samples were divided into 12 pools. Seven of the 10 normal liver tissues were grouped together into one pool and the six available peritumoral liver samples without known etiology into the second pool. A third pool was constructed from peritumoral noncirrhotic livers from alcoholic individuals (n = 3), and the fourth pool was built from DNA extracted from peritumoral noncirrhotic liver from HBV- or HCV-infected individuals (n = 4). Peritumoral cirrhotic liver tissues were divided into three pools, one for patients with increased alcohol consumption (n = 5) and two pools for cirrhotic peritumoral liver tissues from patients with HBV or HCV infection (n = 6 each). The three hepatic adenomas were pooled together for further analysis. HCCs were divided into four pools: one for HCCs without ß-catenin mutations from HBV- or HCV-infected individuals (n = 7); a second pool of HCCs carrying no mutations in ß-catenin regrouping carcinoma developed on alcohol consumption or no known etiology (n = 6); a third for HCCs with ß-catenin mutations from HBV- or HCV-infected individuals (n = 8); and the fourth consisting of HCCs with ß-catenin mutations regrouping carcinoma developed on alcohol consumption or no known etiology (n = 6).

Analysis of Methylation Pattern in Pools
Methylation patterns were analyzed at 112 CpGs in the four promoter regions and the DMR in all 10 pools (Table 2)Go . These results clearly provide evidence for variable methylation patterns between pools for CDKN2A, GSTP1, and IGF2, whereas all pools are consistently unmethylated for MLH1 and as expected for CTNNB1. Based on this analysis only the three variable regions would be carried on to further analysis of all individual samples, whereas MLH1 and CTNNB1 would not be further investigated. It should be pointed out that no conclusion about methylation prevalence and distribution in a distinct sample group can be drawn from the information obtained by the analysis of the pools except the presence or absence of variable methylation patterns between pools. This is exemplified by the difference that is found between pools with identical tumor status and etiology such as the two pools constructed from peritumoral cirrhotic liver from HBV- or HCV-infected individuals for GSTP1 and IGF2.


View this table:
[in this window]
[in a new window]

  		Table 2. Average DNA Methylation Values in Percentage for the Analysis of the Five Genes in the 10 Pools

 
Analysis of Individual Samples
To demonstrate the reliability of our approach, we chose to analyze the entire sample cohort in all five genes. As expected from the analysis of the pools all samples were unmethylated for MLH1 and CTNNB1, whereas methylation patterns varied in IGF2, CDKN2A, and GSTP1. Figure 2Go shows boxplot diagrams for the five genes and the level of statistical significance for the different distributions. Normal liver samples and peritumoral liver samples without cirrhosis were grouped together for easier visualization because no difference between these groups was found in any of the analyzed genes. CDKN2A methylation was found in none of the normal—noncirrhotic— livers, none of the three tested adenomas, and in one of the cirrhotic livers (6%). However, methylation was detected in 7 of 13 (54%) of human HCCs and in 11 of 14 (79%) ß-catenin-mutated HCCs with a very significant difference in the quantitative methylation distribution between the two groups of HCCs. Significant differences were also observed between all groups of normal and the cancerous samples. The analysis of promoter methylation of GSTP1 revealed that 22 of 25 (88%) normal and peritumoral livers and 11 of 14 (69%) cirrhotic livers had a significant level of methylation, and all (100%) of the cancerous samples including the adenomas were found to be methylated. Quantitative methylation differences were highly significant between normal samples and cancerous samples, and also the comparison of the quantitative distribution showed a significant difference between the ß-catenin adenomas and the mutated HCCs and also between mutated and nonmutated HCCs. To exclude a technical problem, we amplified the target region around the transcription start of GSTP1 with a second independent primer set (Table 1)Go Go . The presence of methylation in the normal samples was confirmed, and quantitative measures for the analyzed CpG positions varied on average by 2.6% (range, –3.08 to +7.46%) between the two primer sets. In addition, a recent publication by another research group used a very similar primer set for amplification as that used by us and found it well suited for methylation analysis.38 The region analyzed in IGF2 is a differentially methylated region displaying allelic-specific methylation. In the normal livers an average methylation of 46.8% was found. Cirrhotic livers showed a slight but consistent and statistically significant decrease in methylation of ~5%. For the HCC samples, two phenomena occur simultaneously. Eight of 13 HCC (61%) showed a marked hypomethylation of the entire region, and three samples (23%) showed a trend toward hypermethylation. Methylation values of ß-catenin-mutated HCCs showed a tendency toward more extreme values for the hypomethylated (43%) as well as for hypermethylated (21%) samples compared with the nonmutated HCCs, but this did not reach the level of statistical significance.


Figure 2
View larger version (31K):
[in this window]
[in a new window]

  		Figure 2. Graphical representation (boxplots) of the results obtained by the analysis of the individual samples for 112 CpGs in five genes (CDKN2A, GSTP1, MLH1, IGF2, and CTNNB1) and the level of statistical significance for differential methylation values. *P < 0.05, **P < 0.005, and ***P < 0.0005.

 
Comparison between Measure of Individual Samples and Pools
The negative results for CTNNB1 and MLH1 were successfully predicted by the analysis of the pools, and the variable DNA methylation levels between the different etiological groups were estimated with high precision for CDKN2A, GSTP1, and IGF2. Tables 3Go and 4Go give two examples for the analysis of DNA methylation patterns for a pyrosequencing primer in GSTP1 for the normal liver pool with methylation content of 23.7% averaged over eight CpGs and for a primer in IGF2 analyzing a HCC pool developed on alcohol consumption or without a known etiology. Comparison of the results obtained by the analysis of the pooled DNA sample to the average of the individual samples comprised in the pool show an average deviation of 1.4% (range, –2.2 to 1.1%) for GSTP1 CpGs 25 to 32 and 1.3% (range, –2.3 to 2.6%) for CpGs 6 to 15 of the IGF2 DMR2. The slight but consistent drop of methylation levels in normal livers compared with cirrhotic livers in the IGF2 DMR2 is another example of already successfully detected changes in the pools (Table 2)Go . Comparison with the individual samples also explained perfectly the differences between the two pools of peritumoral cirrhotic liver from HBV- or HCV-infected individuals for GSTP1 and IGF2. The random distribution of the samples resulted in a concentration of the samples displaying a high degree of DNA methylation in one of the two pools.


View this table:
[in this window]
[in a new window]

  		Table 3. Quantitative DNA Methylation Values for a Pyrosequencing Primer Analyzing Eight CpGs in the Promoter of GSTP1 in the Pool of Normal Livers and Analysis of the Seven Samples Comprised in the Pool Individually

 

View this table:
[in this window]
[in a new window]

  		Table 4. Quantitative DNA Methylation Values for a Pyrosequencing Primer Analyzing 10 CpGs in the IGF2 DMR2 in a Pool of Human Hepatocellular Carcinomas Developed on Alcohol Consumption or without Known Etiology and Analysis of the Six Samples Comprised in the Pool Individually

 
Quantitative Accuracy and Pool Size
To investigate further the sensitivity and specificity of our approach, we used several larger pools containing a different number of tumor samples independent of their ß-catenin mutation status and etiology. Three pools were constructed containing 10 samples, three pools with 15 samples, and one pool consisting of 20 samples. Methylation patterns were analyzed for CDKN2A, MLH1, and IGF2. No methylation was detected in any of the pools for MLH1 in concordance with the results obtained from the individual analysis. For CDKN2A, methylation was found in all pools, and comparison to the average of the individual samples showed a similar quantitative accuracy independent of the pool size (Table 5)Go . Quantitative analysis is of even greater importance for the hyper- or hypomethylation in the DMR2 of IGF2 because methylation is present in both normal and cancerous tissues. Again, the cancer-related changes were detected in all pools and coincided well with the data obtained from the individual samples (Table 6)Go . Differences were again not dependent of the pool size. From these data, we can conclude that our approach has 100% specificity and 100% sensitivity if aberrant methylation is a prevalent phenomenon in the analyzed region. The pooling of DNA without prior adjustment of the concentration would further reduce the required amount of DNA and streamline the analysis. The same DNAs as above were directly pooled together in equal quantities irrespective of their concentration after bisulfite treatment. The IGF2 DMR2 was analyzed in these samples (Table 6)Go , and results were compared with the individual samples and the pools in which concentration of the samples had been equalized. Differences in the nonadjusted pools were clearly increased for the smaller pools, probably attributable to some samples present at higher concentrations than others. For the larger pools (n = 20 and n = 26), this effect was less pronounced, approaching the same accuracy as the pools with equal concentrations. In summary, all genes displaying differential methylation were successfully detected, and no false-positive results were generated, demonstrating the reliability and precision of our approach.


View this table:
[in this window]
[in a new window]

  		Table 5. Quantitative DNA Methylation Values for Two Pyrosequencing Primers Analyzing Seven and Six CpGs in the CDKN2A Promoter in Different Pools of Human Hepatocellular Carcinomas

 

View this table:
[in this window]
[in a new window]

  		Table 6. Differences in Percentage between the Average DNA Methylation Values for a Pyrosequencing Primer Analyzing Five CpGs in the IGF2 DMR2 in Pooled DNA Samples Compared with the Individual Samples Contained in the Pool

 

   Discussion
Top
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
References
 
We have shown here a novel approach for the effective screening for differential methylation between etiological homogeneous pools of clinical specimens and validated successfully this approach by the subsequent analysis of all samples individually. Three of the five genes (CDKN2A, GSTP1, and IGF2) displayed epigenetic plasticity in the analyzed region, which was confirmed by the individual analysis of the samples. Methylation patterns of the pooled samples reflect with high precision the average of the DNA methylation patterns of the individual samples contained in the respective pool.

Two genes (MLH1 and CTNNB1) did not display any variation in the DNA methylation patterns or levels either between pools or in the individual samples. These results clearly demonstrate the ability of our approach to identify genes with variable methylation patterns with high reliability. The omission of these two genes in the individual samples would thereby have contributed to significant reduction in labor and cost of the analysis or would have permitted the inclusion of other genes.

The negative results for CTNNB1 were anticipated because this gene is expressed in a subtype of human HCCs attributable to a genetic mutation preventing degradation of the protein and is so far not known to be regulated by epigenetic mechanisms. It was included as a negative control in this study to test for the detection of false-positive results in the pools. Hypermethylation of the CpG island spanning the transcription start of MLH1 seems to be a rare event. One study found methylation in ~10% of human HCCs with hepatitis viral infection whereas normal livers, including noncancerous regions of livers of patients with HCC, were consistently unmethylated.33 The negative results in our study can be explained by the comparatively low number of virus-infected HCCs in our study, phenotypic differences between the samples of Japanese and French origin, respectively.

As a critical negative regulator of cell-cycle progression, CDKN2A is frequently inactivated by promoter methylation in a variety of human tumors. We found hypermethylation in 67% of HCCs with a significant difference toward higher methylation levels in HCCs with ß-catenin mutations. In contrast to other studies39 we did not find CDKN2A methylation in matched normal tissues infected with the hepatitis virus and adenomas, and only one of the cirrhotic livers was clearly methylated. The discrepancy is probably because of the choice of technology because most other studies used standard methylation-specific PCR, which might also amplify and detect very low levels of methylation that cannot be detected by pyrosequencing. On the contrary the quantitative methylation differences between the two groups of HCCs have to our knowledge not been previously described and would not have been detected using conventional methylation-specific PCR. GSTP1 was found to be methylated in nearly all cancer tissues and to a lower degree in a large proportion (80%) of normal liver and matched noncancerous liver tissue from HCC patients. The high prevalence of methylation in normal liver could be attributable to the fact that only low levels of GSTP1 mRNA are found in adult liver (http://symatlas.gnf.org/SymAtlas/, accessed October 7, 2007),40 and immunohistochemical analysis demonstrated that expression of the protein is restricted to bile duct epithelial cells whereas other cells types of the liver do not have detectable levels of the protein.31, 41 GSTP1 might therefore become partially methylated. A second possibility would be that the methylation could also be attributable to aging. Unfortunately, no information on this parameter was available. Analysis of the samples with a second independent primer set confirmed the high degree of methylation found in the normal liver samples, and normal breast and colon tissue that express GSTP1 showed no methylation (<4%; J.T., unpublished data). A recent study found a similar repartition with GSTP1 methylated in 100% of HCCs and 90.5% of matched noncancerous tissue, but also 75% of normal liver tissue.42 The high prevalence of DNA methylation in normal tissue makes methylation at the GSTP1 promoter an unsuitable marker for diagnostic purposes. Nonetheless for the purpose of the validation of our pooled analysis, excellent concordance of pooled and individual samples was achieved. The observed hypomethylation in the IGF2 DMR2 is in good concordance with previously published results studying the same region.32 However, the published results could only distinguish globally different methylation patterns in the DMR because of the use of denaturing high-performance liquid chromatography (dHPLC) as method for read-out rather than a method with quantitative single nucleotide resolution as presented here. The slight, but consistent and significant, reduction of DNA methylation levels in the DMR2 in cirrhotic liver samples compared with normal livers, which was already detected in the pooled samples, demonstrates the quantitative resolution and power of our approach. The methylation changes might be correlated to the decreased serum IGF2 levels that have been postulated to be a sensitive and effective indicator for liver dysfunction.43 However, further studies are necessary to analyze this reduction, especially with respect to differential promoter usage as well as to elucidate the two concurrent phenomena of hyper- and hypomethylation, which are evident in cancer samples and that might in an unfortunate situation counterbalance each other. In this case no change would be detected in the analysis of pooled samples. In the present study hypomethylation was much more prevalent, thereby not confusing the analysis and detection of variable methylation patterns. However, it might be necessary for the analysis of imprinted genes to add several individual samples already in the first screening step to detect the simultaneous occurrence of hyper- and hypomethylation.

Pyrosequencing-based DNA methylation analysis relies on a candidate gene approach thus requiring a hypothesis or preliminary data to choose a limited number of genes and/or gene regions of potential interest. It aims at the identification of nonrare methylation events in a specific pathology, ie, alterations that occur in a large number of cells within a tumor and might directly contribute to—or are a nonrandom result of—tumorigenesis rather than passenger epimutations. Besides cost and effort necessary for a study, the amount of available clinical samples imposes major restrictions on the number of genes that can be analyzed. Although a growing number of genes that are inactivated by DNA methylation during carcinogenesis are being discovered, etiological, environmental, or geographic differences often require verification in a sample collection under investigation. In addition, because ~10% of all genes are postulated to be aberrantly methylated in different types or stages of tumors,6 a large number of genes still needs to be identified. The use of a prescreening step such as the method proposed here slightly increases time and effort for a given project. However, this is easily compensated by the concentration on differentially methylated regions between pools, and most, if not all, subsequently analyzed genes will display some difference in their methylation profile that can be used for diagnostics and/or correlated to a variety of clinical parameters. This significantly increases quality and quantity of the results of a study. Pyrosequencing of sample pools for the detection of differential DNA methylation helps to prioritize target genes. Parameters such as pool size can be adapted to capture anticipated quantitative differences between pools. Larger pools containing DNA of up to 26 samples were well suited to analyze aberrant DNA methylation present in a large proportion of samples as demonstrated in this study for CDKN2A or IGF2. In this setting, the devised approach displayed a sensitivity and specificity of 100%. However, methylation events occurring at low frequency might be missed in large pools because the quantitative resolution is not be sufficient to identify the subtle changes attributable to a single or very few differentially methylated samples. The method works best if knowledge as detailed as possible is available on the samples used to classify them into clinically or pathologically homogeneous groups to answer a specific question. The composition of pools might therefore be different for distinct questions. Additional pools of identical etiology can increase the confidence in the results of the prescreen and might also detect less frequent methylation events. One important advantage of our method is that less DNA is consumed, permitting use of the same amount of available samples to analyze methylation patterns in more genes and concentrate on the genes that actually display differential methylation. If only frequently hyper- or hypomethylated genes are of interest, the additional step of adjusting the concentration of bisulfite-treated DNAs might be omitted. The gain of rapidity and saving of DNA material, however, has to be balanced against decrease in accuracy. The demonstrated method enables rapid and accurate identification of genes functionally implicated in carcinogenesis and genes whose methylation pattern present a valuable biomarker for the early diagnosis, classification, prognosis, and therapy of human cancers.


   Acknowledgments
 
We thank Hafida El Abdalaoui (Centre National de Génotypage) for the bisulfite treatment of the samples and Anne Boland-Auge (Centre National de Génotypage) for the whole genome amplifications and the measurement of DNA concentrations.


   Footnotes
 
Address reprint requests to Jörg Tost, Ph.D., Laboratory for Epigenetics, Centre National de Génotypage, Bâtiment G2, 2 rue Gaston Crémieux, CP 5721, 91057 Evry Cedex, France. E-mail: tost{at}cng.fr

Supported by the Ministère Délégué à la Recherche of the French Government, the European Union (framework 6 integrated project MolPAGE LSHG-CT-2004-512066), INSERM, the Ligue Nationale contre le Cancer, Centre National de la Recherche Scientifique-Caisse Nationale d’Assurance Maladie des Professions Indépendantes/Assistance Publique-Hôpitaux de Paris (to V.A.), and the Société Française de Pathologie (to V.A.).

E.D. and V.A. contributed equally and should be considered joint first authors.

Accepted for publication April 2, 2007.


   References
Top
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
 References
 

  1. Robertson KD: DNA methylation and human disease. Nat Rev Genet 2005, 6:597-610[CrossRef][Medline]
  2. Coleman WB, Rivenbark AG: Quantitative DNA methylation analysis: the promise of high-throughput epigenomic diagnostic testing in human neoplastic disease. J Mol Diagn 2006, 8:152-156[Free Full Text]
  3. Jones PA, Baylin SB: The fundamental role of epigenetic events in cancer. Nat Rev Genet 2002, 3:415-428[Medline]
  4. Laird PW: Cancer epigenetics. Hum Mol Genet 2005, 14 Spec no 1:R65-R76
  5. Cui H, Horon IL, Ohlsson R, Hamilton SR, Feinberg AP: Loss of imprinting in normal tissue of colorectal cancer patients with microsatellite instability. Nat Med 1998, 4:1276-1280[CrossRef][Medline]
  6. Costello JF, Fruhwald MC, Smiraglia DJ, Rush LJ, Robertson GP, Gao X, Wright FA, Feramisco JD, Peltomaki P, Lang JC, Schuller DE, Yu L, Bloomfield CD, Caligiuri MA, Yates A, Nishikawa R, Su Huang H, Petrelli NJ, Zhang X, O’Dorisio MS, Held WA, Cavenee WK, Plass C: Aberrant CpG-island methylation has non-random and tumour-type-specific patterns. Nat Genet 2000, 24:132-138[CrossRef][Medline]
  7. Brena RM, Huang TH, Plass C: Quantitative assessment of DNA methylation: potential applications for disease diagnosis, classification, and prognosis in clinical settings. J Mol Med 2006, 84:365-377[CrossRef][Medline]
  8. Laird PW: Early detection: the power and the promise of DNA methylation markers. Nat Rev Cancer 2003, 3:253-266[CrossRef][Medline]
  9. Yoo CB, Jones PA: Epigenetic therapy of cancer: past, present and future. Nat Rev Drug Discov 2006, 5:37-50[CrossRef][Medline]
  10. Colella S, Shen L, Baggerly KA, Issa JP, Krahe R: Sensitive and quantitative universal pyrosequencing methylation analysis of CpG sites. Biotechniques 2003, 35:146-150[Medline]
  11. Dupont JM, Tost J, Jammes H, Gut IG: De novo quantitative bisulfite sequencing using the pyrosequencing technology. Anal Biochem 2004, 333:119-127[CrossRef][Medline]
  12. Tost J, Dunker J, Gut IG: Analysis and quantification of multiple methylation variable positions in CpG islands by pyrosequencing. Biotechniques 2003, 35:152-156[Medline]
  13. Uhlmann K, Brinckmann A, Toliat MR, Ritter H, Nürnberg P: Evaluation of a potential epigenetic biomarker by quantitative methyl-single nucleotide polymorphism analysis. Electrophoresis 2002, 23:4072-4079[CrossRef][Medline]
  14. Ronaghi M, Uhlen M, Nyren P: A sequencing method based on real-time pyrophosphate. Science 1998, 281:363-365[Free Full Text]
  15. Langaee T, Ronaghi M: Genetic variation analyses by pyrosequencing. Mutat Res 2005, 573:96-102[Medline]
  16. van der Straaten T, Kweekel D, Tiller M, Bogaartz J, Guchelaar HJ: Multiplex pyrosequencing of two polymorphisms in DNA repair gene XRCC1. J Mol Diagn 2006, 8:444-448[Abstract/Free Full Text]
  17. Clarke SC: Pyrosequencing: nucleotide sequencing technology with bacterial genotyping applications. Expert Rev Mol Diagn 2005, 5:947-953[CrossRef][Medline]
  18. Gharizadeh B, Oggionni M, Zheng B, Akom E, Pourmand N, Ahmadian A, Wallin KL, Nyren P: Type-specific multiple sequencing primers: a novel strategy for reliable and rapid genotyping of human papillomaviruses by pyrosequencing technology. J Mol Diagn 2005, 7:198-205[Abstract/Free Full Text]
  19. Jordan JA, Butchko AR, Durso MB: Use of pyrosequencing of 16S rRNA fragments to differentiate between bacteria responsible for neonatal sepsis. J Mol Diagn 2005, 7:105-110[Abstract/Free Full Text]
  20. Ogino S, Kawasaki T, Brahmandam M, Yan L, Cantor M, Namgyal C, Mino-Kenudson M, Lauwers GY, Loda M, Fuchs CS: Sensitive sequencing method for KRAS mutation detection by pyrosequencing. J Mol Diagn 2005, 7:413-421[Abstract/Free Full Text]
  21. Gruber JD, Colligan PB, Wolford JK: Estimation of single nucleotide polymorphism allele frequency in DNA pools by using pyrosequencing. Hum Genet 2002, 110:395-401[CrossRef][Medline]
  22. Neve B, Froguel P, Corset L, Vaillant E, Vatin V, Boutin P: Rapid SNP allele frequency determination in genomic DNA pools by pyrosequencing. Biotechniques 2002, 32:1138-1142[Medline]
  23. Wasson J, Skolnick G, Love-Gregory L, Permutt MA: Assessing allele frequencies of single nucleotide polymorphisms in DNA pools by pyrosequencing technology. Biotechniques 2002, 32:1144-1150[Medline]
  24. Mirmohammadsadegh A, Marini A, Nambiar S, Hassan M, Tannapfel A, Ruzicka T, Hengge UR: Epigenetic silencing of the PTEN gene in melanoma. Cancer Res 2006, 66:6546-6552[Abstract/Free Full Text]
  25. Xinarianos G, McRonald FE, Risk JM, Bowers NL, Nikolaidis G, Field JK, Liloglou T: Frequent genetic and epigenetic abnormalities contribute to the deregulation of cytoglobin in non-small cell lung cancer. Hum Mol Genet 2006, 15:2038-2044[Abstract/Free Full Text]
  26. Schatz P, Distler J, Berlin K, Schuster M: Novel method for high throughput DNA methylation marker evaluation using PNA-probe library hybridization and MALDI-TOF detection. Nucleic Acids Res 2006, 34:e59[Abstract/Free Full Text]
  27. Yang AS, Doshi KD, Choi SW, Mason JB, Mannari RK, Gharybian V, Luna R, Rashid A, Shen L, Estecio MR, Kantarjian HM, Garcia-Manero G, Issa JP: DNA methylation changes after 5-aza-2'-deoxycytidine therapy in patients with leukemia. Cancer Res 2006, 66:5495-5503[Abstract/Free Full Text]
  28. White HE, Durston VJ, Harvey JF, Cross NC: Quantitative analysis of SNRPN gene methylation by pyrosequencing as a diagnostic test for Prader-Willi syndrome and Angelman syndrome. Clin Chem 2006, 52:1005-1013[Abstract/Free Full Text]
  29. Tost J, El Abdalaoui H, Gut IG: Serial pyrosequencing for quantitative DNA methylation analysis. Biotechniques 2006, 40:721-726[Medline]
  30. Liew CT, Li HM, Lo KW, Leow CK, Chan JY, Hin LY, Lau WY, Lai PB, Lim BK, Huang J, Leung WT, Wu S, Lee JC: High frequency of p16INK4A gene alterations in hepatocellular carcinoma. Oncogene 1999, 18:789-795[CrossRef][Medline]
  31. Zhong S, Tang MW, Yeo W, Liu C, Lo YM, Johnson PJ: Silencing of GSTP1 gene by CpG island DNA hypermethylation in HBV-associated hepatocellular carcinomas. Clin Cancer Res 2002, 8:1087-1092[Abstract/Free Full Text]
  32. Poirier K, Chalas C, Tissier F, Couvert P, Mallet V, Carrie A, Marchio A, Sarli D, Gicquel C, Chaussade S, Beljord C, Chelly J, Kerjean A, Terris B: Loss of parental-specific methylation at the IGF2 locus in human hepatocellular carcinoma. J Pathol 2003, 201:473-479[CrossRef][Medline]
  33. Matsukura S, Soejima H, Nakagawachi T, Yakushiji H, Ogawa A, Fukuhara M, Miyazaki K, Nakabeppu Y, Sekiguchi M, Mukai T: CpG methylation of MGMT and hMLH1 promoter in hepatocellular carcinoma associated with hepatitis viral infection. Br J Cancer 2003, 88:521-529[CrossRef][Medline]
  34. Buendia MA: Genetics of hepatocellular carcinoma. Semin Cancer Biol 2000, 10:185-200[CrossRef][Medline]
  35. de La Coste A, Romagnolo B, Billuart P, Renard CA, Buendia MA, Soubrane O, Fabre M, Chelly J, Beldjord C, Kahn A, Perret C: Somatic mutations of the ß-catenin gene are frequent in mouse and human hepatocellular carcinomas. Proc Natl Acad Sci USA 1998, 95:8847-8851[Abstract/Free Full Text]
  36. Bedossa P, Poynard T: An algorithm for the grading of activity in chronic hepatitis C. The METAVIR Cooperative Study Group. Hepatology 1996, 24:289-293[Medline]
  37. Ogino S, Kawasaki T, Brahmandam M, Cantor M, Kirkner GJ, Spiegelman D, Makrigiorgos GM, Weisenberger DJ, Laird PW, Loda M, Fuchs CS: Precision and performance characteristics of bisulfite conversion and real-time PCR (MethyLight) for quantitative DNA methylation analysis. J Mol Diagn 2006, 8:209-217[Abstract/Free Full Text]
  38. Rodriguez-Canales J, Hanson J, Tangrea M, Erickson H, Albert P, Wallis B, Richardson A, Pinto P, Linehan W, Gillespie J, Merino M, Libutti S, Woodson K, Emmert-Buck M, Chuaqui R: Identification of a unique epigenetic sub-microenvironment in prostate cancer. J Pathol 2007, 211:410-419[CrossRef][Medline]
  39. Jicai Z, Zongtao Y, Jun L, Haiping L, Jianmin W, Lihua H: Persistent infection of hepatitis B virus is involved in high rate of p16 methylation in hepatocellular carcinoma. Mol Carcinog 2006, 45:530-536[CrossRef][Medline]
  40. Su AI, Cooke MP, Ching KA, Hakak Y, Walker JR, Wiltshire T, Orth AP, Vega RG, Sapinoso LM, Moqrich A, Patapoutian A, Hampton GM, Schultz PG, Hogenesch JB: Large-scale analysis of the human and mouse transcriptomes. Proc Natl Acad Sci USA 2002, 99:4465-4470[Abstract/Free Full Text]
  41. Niu ZS, Wang M: Expression of c-erbB-2 and glutathione S-transferase-pi in hepatocellular carcinoma and its adjacent tissue. World J Gastroenterol 2005, 11:4404-4408[Medline]
  42. Lehmann U, Berg-Ribbe I, Wingen LU, Brakensiek K, Becker T, Klempnauer J, Schlegelberger B, Kreipe H, Flemming P: Distinct methylation patterns of benign and malignant liver tumors revealed by quantitative methylation profiling. Clin Cancer Res 2005, 11:3654-3660[Abstract/Free Full Text]
  43. Wu YL, Ye J, Zhang S, Zhong J, Xi RP: Clinical significance of serum IGF-I, IGF-II and IGFBP-3 in liver cirrhosis. World J Gastroenterol 2004, 10:2740-2743[Medline]




This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow All Versions of this Article:
jmoldx.2007.060209v1
9/4/510    most recent
Right arrow Purchase Article
Right arrow View Shopping Cart
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Dejeux, E.
Right arrow Articles by Tost, J.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Dejeux, E.
Right arrow Articles by Tost, J.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS