Molecular Cancer Research
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
Molecular Cancer Research 5, 153-163, February 1, 2007. doi: 10.1158/1541-7786.MCR-06-0034
© 2007 American Association for Cancer Research
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Supplementary Data, Model, et al
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
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 HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Model, F.
Right arrow Articles by Lofton-Day, C.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Model, F.
Right arrow Articles by Lofton-Day, C.

Cancer Genes and Genomics

Identification and Validation of Colorectal Neoplasia–Specific Methylation Markers for Accurate Classification of Disease

Fabian Model1, Neal Osborn2, David Ahlquist2, Robert Gruetzmann3, Bela Molnar4, Ferenc Sipos5, Orsolya Galamb4, Christian Pilarsky3, Hans-Detlev Saeger3, Zsolt Tulassay4, Kari Hale1, Suzanne Mooney1, Joseph Lograsso1, Peter Adorjan6, Ralf Lesche6, Andreas Dessauer7, Joerg Kleiber7, Baerbel Porstmann7, Andrew Sledziewski1 and Catherine Lofton-Day1

1 Epigenomics, Inc., Seattle, Washington; 2 Division of Gastroenterology and Hepatology, Mayo Clinic, Rochester, Minnesota; 3 University Hospital Carl Gustav Carus, Technical University Dresden, Dresden, Germany; 4 Hungarian Academy of Sciences, Clinical Gastroenterology Research Unit, Budapest, Hungary; 5 Department of Medicine, Semmelweis University, Budapest, Hungary; 6 Epigenomics AG, Berlin, Germany; and 7 ICCU Division, Roche Diagnostics, Penzberg, Germany

Requests for reprints: Catherine Lofton-Day, Epigenomics, Inc., Suite 300, 1000 Seneca Street, Seattle, WA 98101. Phone: 206-883-2913; Fax: 206-254-9151. E-mail: clofton{at}us.epigenomics.com


   Abstract
Top
 Notes
 Abstract
Introduction
 Results
 Discussion
 Materials and Methods
 Acknowledgements
 References
 
Aberrant DNA methylation occurs early in oncogenesis, is stable, and can be assayed in tissues and body fluids. Therefore, genes with aberrant methylation can provide clues for understanding tumor pathways and are attractive candidates for detection of early neoplastic events. Identification of sequences that optimally discriminate cancer from other diseased and healthy tissues is needed to advance both approaches. Using well-characterized specimens, genome-wide methylation techniques were used to identify candidate markers specific for colorectal neoplasia. To further validate 30 of these candidates from genome-wide analysis and 13 literature-derived genes, including genes involved in cancer and others with unknown functions, a high-throughput methylation-specific oligonucleotide microarray was used. The arrays were probed with bisulfite-converted DNA from 89 colorectal adenocarcinomas, 55 colorectal polyps, 31 inflammatory bowel disease, 115 extracolonic cancers, and 67 healthy tissues. The 20 most discriminating markers were highly methylated in colorectal neoplasia (area under the receiver operating characteristic curve > 0.8; P < 0.0001). Normal epithelium and extracolonic cancers revealed significantly lower methylation. Real-time PCR assays developed for 11 markers were tested on an independent set of 149 samples from colorectal adenocarcinomas, other diseases, and healthy tissues. Microarray results could be reproduced for 10 of 11 marker assays, including eight of the most discriminating markers (area under the receiver operating characteristic curve > 0.72; P < 0.009). The markers with high specificity for colorectal cancer have potential as blood-based screening markers whereas markers that are specific for multiple cancers could potentially be used as prognostic indicators, as biomarkers for therapeutic response monitoring or other diagnostic applications, compelling further investigation into their use in clinical testing and overall roles in tumorigenesis. (Mol Cancer Res 2007;5(2):153–63)


   Introduction
Top
 Notes
 Abstract
 Introduction
Results
 Discussion
 Materials and Methods
 Acknowledgements
 References
 
Although colorectal cancer is the second most common cause of malignant death in industrialized countries, the mechanisms and pathways of the neoplastic events associated with this complex disease are not well understood. Genetic alterations in colorectal neoplasia have been studied extensively as candidate markers for detection and analysis of the disease (1-5), but much less is known about epigenetic changes, including aberrant methylation of genes. Several genes have been shown to be preferentially hypermethylated in both colorectal cancer and premalignant adenomas with dramatic effects on the expression of their resultant proteins (6-10), indicating that silencing of tumor suppressor genes or other genes in tumor pathways can occur both from mutation events and/or aberrant methylation. Application of expression-based microarray profiling has proved effective in distinguishing RNA profile differences between tumor types and classes, providing information for understanding of tumor pathways (11, 12). Recently, this technology has been adapted to methylation-based microarray profiling, which can distinguish the epigenetic methylation profile of samples from large groups of patients (13, 14). This type of analysis detects methylation ratios at CpG positions that have been amplified by PCR from bisulfite-modified genomic DNA. By evaluating modified DNA from different patient populations, these arrays have been used to identify methylation markers that distinguish among types of tumors, differentiate tumors from normal tissue, and predict clinical outcome (13). Because the methylation microarray requires larger amounts of DNA, it is not applicable to clinical situations where only low levels of DNA are available from samples such as biopsies or body fluids. To achieve sensitive detection of DNA methylation from such sources, real-time PCR methods (e.g., MethyLight) can be used to distinguish patient profiles (15).

From a clinical perspective, more accurate detection markers are needed to improve the effectiveness and efficiency of both the screening and surveillance of colorectal neoplasia. Aberrantly methylated genes represent attractive candidate markers for this purpose because cancer-specific methylation changes occur early in tumorigenesis (16), seem to be stable, yield a positive amplifiable signal, and can be assayed with high analytic sensitivity. Unfortunately, many or the more commonly described methylation markers in the literature, such as ER, MGMT, MLH1, and CDKN2a, have not been adequately tested for specificity to a target cancer by simultaneously analyzing methylation status across multiple tumor types and normal tissue. As a result, many of these most widely investigated markers are not suitable for specific detection of a particular disease. For example, methylation of the gene CDKN2a (p16) has been reported to be found in blood from patients with numerous types of cancer including oral cancer, gastric cancer, melanoma, non–small-cell lung cancer, hepatocellular cancer, and bladder cancer in a number of independent studies (17-22). Clearly, methylation of this gene is important in neoplastic progression, but its usefulness as a specific marker for a single cancer in a screening application is questionable. Furthermore, CDKN2a has been shown to be methylated in blood from individuals with noncancerous diseases, albeit at a lower rate (21, 23).

Because of the genetic heterogeneity of colorectal neoplasia, multiple genetic markers may be required for acceptable tumor detection rates (24, 25). Because methylation occurs early and in distinct genomic areas, it might be possible to achieve high clinical sensitivity with fewer methylated DNA markers (6). Feasibility studies have shown that aberrantly methylated DNA markers can be assayed from serum or plasma (16, 26-31) and from stool (32-34) to detect colorectal cancer. However, robustly conducted genome-wide searches are needed to identify methylated DNA sequences that optimally discriminate colorectal neoplasia from other tissues and normal blood components.

In this study, we report the use of a genome-wide PCR-based discovery process to identify sequences that are differentially methylated between colorectal neoplasia, normal colon tissue, and peripheral blood lymphocytes (PBL) from healthy age-matched individuals. We provide validation of these differential methylation markers via use of both methylation microarrays and real-time PCR for discrimination of colorectal neoplasia compared with healthy mucosa and age-matched healthy PBLs and also with other disease states, including actively inflamed epithelia and malignant tissues. The markers identified are consistent with the concept that hypermethylation is an important proponent of tumorigenesis because several of the candidates found in our genome-wide screening have recently been implicated as being involved in the neoplastic process, and several candidates from our literature-based search that were previously reported to be involved in cancer were verified in this study. The high accuracy of these markers suggests that the sensitive, methylation-specific real-time PCR assays described in this study may be useful for detection of disease at early stages in blood and for interrogation of neoplastic pathways. Based on our comprehensive analysis of these candidate markers in diverse tissue types, we suggest potential applications for the markers.


   Results
Top
 Notes
 Abstract
 Introduction
 Results
Discussion
 Materials and Methods
 Acknowledgements
 References
 
Genome-Wide Discovery
The discovery process resulted in more than 500 unique sequences that were potential candidates for colorectal cancer biomarkers. The differentially methylated sequences identified by methylation-specific arbitrarily primed PCR (35) and methylated CpG island amplification (36) were scored and prioritized using the following scoring variables: (a) appearance using multiple discovery methods; (b) appearance in multiple pools of like samples; (c) located within a CpG island; (d) located within the promoter region of a gene; (e) located near or within predicted or known genes; (f) known to be associated with disease; (g) class of gene (transcription factor, growth factor, tumor suppressor, oncogene); and (h) repetitive element. Under this scoring schema, a sequence received a point for each of the above criteria and received a score of –8 for having repetitive sequence content >50%. Therefore, the highest score possible was 7; the lowest was –8. Scores were automatically calculated for each sequence using genomic annotations from the Ensembl database.8 Scoring results for all differentially methylated sequences identified by genome-wide discovery are shown in Supplementary Table S1. Using the scoring criteria above along with manual review of the sequences, 30 sequences were selected for microarray analysis (Table 1 ). Sequences with significant (>50%) repetitive element content were eliminated from consideration. Our comprehensive database of sequences derived from internal genome-wide discovery experiments allowed us to also eliminate sequences found using other previously tested tumor types. Selected sequences scored ≥1, with the majority scoring ≥3.


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

  		TABLE 1. Sequences Selected from Genome-Wide Discovery or from Literature for Validation on Oligonucleotide Microarray

 
Gene Array Study
For additional confirmation of the methylation state of the potential markers, we constructed a methylation-specific gene array containing oligonucleotides representing the 30 selected genome-wide discovery sequences and also 13 potential methylation biomarkers from the literature (Table 1). Additional genes were chosen from a previous microarrray study of literature-derived sequences and selected based on involvement in neoplasia and discrimination of colorectal cancer versus pathologically normal colon tissue (data not shown). In our discovery experiments, the exon 1 region of the TMEFF2 gene was identified as being hypermethylated. Because the promoter region of this gene had been described as differentially methylated in the literature (37) and was also shown to discriminate between colorectal cancer and healthy colon in the previous microarray study (data not shown), this region was included as a candidate sequence. TMEFF2 methylation measurements from promoter and exon 1 region are highly correlated (between-amplificate correlation R = 0.76) and were therefore aggregated and treated as one locus for further analysis.

We determined the ability of the 42 differentially methylated gene regions to discriminate between colorectal cancer and other tissues using a large, highly diverse sample set containing colorectal cancer tissue and tissue samples from other types of cancers, colon inflammatory conditions, colon polyps, and numerous histopathologically determined normal tissues.

Hierarchical Clustering
To identify systematic similarities in the overall methylation patterns of samples and genes, we did a hierarchical clustering on the entire gene set and the set of 204 colon-derived tissue DNA samples (Fig. 1 ). The majority of normal and inflammatory colon samples fall into a cluster that shows no methylation on most genes (cluster N: 25 normal, 29 inflammatory, 12 colon polyp, and 16 colorectal cancer samples). The other cluster (cluster C) consists predominantly of neoplastic samples and is clearly separated into two subclusters (C1 and C2), which show different degrees of hypermethylation. The subcluster with the strongest methylation is composed only of neoplastic tissue (cluster C1: 28 colon polyp and 38 colorectal cancer samples). The other subcluster shows an intermediate degree of methylation and includes some histologically normal and inflammatory samples (cluster C2: 4 normal, 2 inflammatory, 15 colon polyp, and 35 colorectal cancer samples). There is no significant association between the two neoplastic subclusters and tumor stage or grade. However, there are a significantly higher number of adenomas >1 cm in the subcluster C1 than in the subclusters N and C2 (C1: 14 colon polyps ≥1 cm, 15 colon polyps <1 cm; C2: 2 colon polyps ≥1 cm, 13 colon polyps <1 cm; N: 2 colon polyps ≥1 cm, 11 colon polyps <1 cm; P < 0.01).


Figure 1
View larger version (65K):
[in this window]
[in a new window]

  		FIGURE 1. Hierarchical clustering of all 204 colon-derived tissue samples in the training set and all 42 loci from the combined gene array. Columns, patient samples; rows, genomic loci. Column annotations give the gene name. The class information was unknown to the clustering algorithm. The average degree of methylation of each genomic locus in each sample is represented by the decadic logarithm of the methylation proportion ranging from <1% methylated alleles (green) to methylation of all alleles (red). There are three main tissue clusters labeled as N, C1, and C2. Cluster N composition: 25 (30%) normal colon, 29 (35%) inflamed colon, 12 (15%) colon polyp, and 16 (20%) colon cancer samples. Cluster C1 composition: 28 (42%) colon polyp and 38 (58%) colon cancer samples. Cluster C2 composition: 4 (7%) normal colon, 2 (4%) inflamed colon, 15 (27%) colon polyp, and 35 (63%) colon cancer samples. Between cluster comparison: C1 has the highest degree of methylation and contains 46% of the neoplastic and preneoplastic samples. C2 contains 10% of the nontumor and 35% of the neoplastic and preneoplastic samples. N contains 90% of the nontumor samples, 18% of the colorectal cancer samples, and 22% of the colon polyp samples.

 
As can be expected from the directed selection of candidate sequences for the microarray study, the overall clustering results show a clear separation between normal and inflammatory samples in cluster N on the one side and polyp and colorectal cancer samples in cluster C on the other side. Cluster N contains 90% of the nonneoplastic samples. Cluster C contains 81% of the neoplastic and preneoplastic samples. The majority of discriminatory markers are hypermethylated in polyp and colorectal cancer samples from cluster C and show a typical CpG island methylator phenotype (CIMP; ref. 38). Polyp and colorectal cancer samples in cluster N, on the other hand, are samples not methylated for the majority of discriminatory markers tested and appear to be a CIMP-negative population. The observed strong similarity between the colorectal cancer and colon polyp samples is supported by previous studies that show early alterations in methylation in precancerous conditions of the colon (39, 40). Based on the clustering results, all subsequent analyses of the data combine the colorectal cancer and colon polyp samples for comparison to normal tissue, other cancers, and inflammatory bowel disease.

Individual Marker Performance
To quantify the influence of non–colon-derived tissues on the classification performance of individual markers, we analyzed the data set in two different ways. First, we looked at the complete sample set. Here, the negative class consisted of 214 samples from normal colon (29), inflammatory colon (31), PBL (14), and other normal (10 liver, 7 stomach, and 7 esophagus) and noncolorectal cancer tissues (28 breast cancer, 38 lung cancer, 9 liver cancer, 10 pancreatic cancer, 10 bile duct cancer, 10 stomach cancer, 5 prostate cancer, and 6 esophageal cancer). The positive class was composed of 144 colorectal cancer and polyp samples. Thirty markers were highly significant with P < 0.0001. Ten markers showed a very strong class separation with an area under the receiver operating characteristic curve (AUC) of ≥0.8 (Table 2 ). The sensitivity of these strong markers ranged between 35% and 52% at a specificity level of 95%. In a second analysis, we looked only at colon-derived tissues. In this case, the negative class consisted of 60 samples from normal and inflammatory colon. The positive class was again composed of 144 colon cancer and polyp samples. Despite the lower sample number as compared with the full data set, 29 markers were highly significant with P < 0.0001. Nineteen markers showed a very strong class separation with an AUC of ≥0.8 (Table 2). The sensitivity of these strong markers ranged between 44% and 81% at a specificity level of 95%. The omission of non–colon-derived tissues resulted in a strong increase of {Delta}AUC ≥ 0.05 for 17 markers and a strong decrease of {Delta}AUC ≤ –0.05 for five markers. Classification results of all individual markers are summarized in Table 2.


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

  		Table 2. Single-Marker Classification Performance for Microarray (Left) and MethyLight Assays (Right)

 
Marker Panel Performance
Using a panel of markers does not significantly improve performance over the best single marker, TMEFF2. The best two-marker panel is TMEFF2 plus NGFR. This panel has a sensitivity of 55% (confidence interval, 44-68%) at 95% specificity in the classification of all samples (+5% compared with TMEFF2 alone). The sensitivity in classifying only colon-derived tissue samples is 85% (confidence interval, 75-93%) at 95% specificity (+4% compared with TMEFF2 alone).

Paneling does not significantly increase sensitivity of the markers for colorectal cancer over TMEFF2 alone because all of our markers detect the same subset of colorectal cancer and polyp samples. The CIMP-positive cancer cluster shown in Fig. 2 (cluster C) includes 81% of the colorectal cancer and polyp samples in the study. TMEFF2 alone is heavily hypermethylated on 67% of these CIMP-positive samples (78 of 116 CIMP samples with TMEFF2 methylation >10%). Only 11% of the remaining CIMP-negative colorectal cancer and polyp samples show TMEFF2 hypermethylation (3 of 28 CIMP-negative samples with TMEFF2 methylation >10%). Because no additional marker shows significant hypermethylation on the CIMP-negative samples or significantly higher methylation levels than TMEFF2 on the CIMP-positive samples, overall marker complementarity is minimal.


Figure 2
View larger version (35K):
[in this window]
[in a new window]

  		FIGURE 2. Methylation levels of different tissue classes. For each marker gene (rows), the distribution of methylation levels in the major tissue classes is visualized by a box plot. Left column, methylation levels from microarray analysis. Horizontal axis, percent methylation with 1% to 100% methylation scale. Right column, methylation levels from real-time MethyLight analysis. Horizontal axis, 0.01% to 100% methylation. Individual box plots, middle 50% of the data; middle line, median; whiskers extend to the most extreme data point which is no more than 1.5 times the interquartile range from the box. Methylation measurement values outside the whisker range are plotted as individual points.

 
Distribution of Methylation Frequencies
To further understand the behavior of the strongest markers on different tissue types, we looked at the distribution of methylation frequencies on all 358 samples grouped into five major tissue classes. For this analysis, colorectal cancer and polyp samples, as well as normal and inflammatory colon tissue samples, were combined because their respective methylation rate distributions were similar. Figure 2 shows box plots of the major tissue classes for all markers from the gene array with an AUC of ≥0.8. Median methylation levels of all tissue subclasses are shown in Fig. 3 and detailed box plots for all tissue subclasses can be found in Supplementary Fig. S1. The overall low degree of methylation of markers GSK3B, RNF4, and CD44 is a result of the poor correlation between different CpG positions within the same amplicon (median between CpG correlations: GSK3B, R = 0.27; RNF4, R = 0.28; CD44, R = 0.41; all other 17 most discriminating markers, R ≥ 0.47) and indicates a lack of comethylation within the CpG island.


Figure 3
View larger version (31K):
[in this window]
[in a new window]

  		FIGURE 3. Relative methylation levels of normal and noncolorectal cancer tissue classes in comparison with colorectal cancer and polyps. For each tissue class (rows) and each marker gene (columns), the median methylation level is plotted as fold change over the median colorectal cancer methylation level. Fold changes are restricted to a range of 2-fold hypermethylation and 8-fold hypomethylation over the median colorectal cancer methylation level. See Supplementary Fig. S1 for box plots of all subclasses.

 
Generally, all markers, with the exception of APC, show hypermethylation of the colorectal cancer class compared with the healthy colon and the PBL class. However, the methylation patterns of our markers differ considerably with regard to the noncolonic healthy tissues and noncolonic cancer classes. Markers TMEFF2, ZDHHC22, SLITRK1, SLC32A1, DLX5, GSK3B, NGFR, PCDH17, N33, and BCOR differentiate colon neoplasia very well from the majority of other tissues (AUC ≥ 0.8). All show strong hypermethylation of colorectal cancer compared with other tissues with varying differences among the other tissue classes. Other markers such as RNF4, SIX6, CD44, CSPG2, CDH13, GPR7, EYA4, ALX4, and APC show only small or no differences between colorectal cancer and the noncolonic cancer class. N33 not only shows significant hypermethylation of colorectal cancer compared with normal colon but also gives a very strong discrimination between colon tissue (normal colon and colorectal cancer) and most other tissues. All of our markers show some degree of hypermethylation in stomach and esophageal cancer tissue and, to some lower extent, in normal stomach tissue.

Marker Validation with MethyLight Assays
We developed 11 real-time MethyLight assays for markers that were designated as having strong to poor performance on the gene array. Nine of the markers had very high performance (AUC ≥ 0.8) in the colon tissue only classification (TMEFF2, ZDHHC21, SLITRK1, SSLC32A, NGFR, N33, RNF4, EYA4, and ALX4). Two markers with poorer performance (BCL6 and SMAD7) were tested because, although the array results were not strong, original discovery scoring of these sequences was high (6 and 4, respectively), and this information would also allow us to further correlate array performance results with real-time assay results. For TMEFF2, the real-time assay was designed in the promoter region of the gene. Classification performance of the MethyLight assays was estimated on an independent sample set with a negative class of normal and normal adjacent colon (46 samples), inflammatory colon (9 samples), and other cancerous tissues (15 breast cancer and 15 liver cancer). The positive class was composed of 39 colon cancer and adenoma samples. Of the nine MethyLight assays for the strongest gene array markers, five (TMEFF2, ZDHHC21, NGFR, N33, and EYA4) were highly significant (P < 0.0001) and showed a substantial class separation with an AUC of ≥0.8 (Table 2). Three assays (SLITRK1, SSLC32A, and ALX4) showed a significant but weaker class separation (P < 0.009; AUC > 0.72). RNF4, a strong candidate from the gene array, could not be reproduced using real-time PCR analysis because almost all amplifications yielded no product. This is likely due to a lack of significant comethylation of CpG sites within the assay region but was not further investigated. The two poorer performing gene array markers (BCL6 and SMAD7) showed poor results with their corresponding real-time MethyLight assays, confirming results obtained with the gene array. The sensitivity of the five strongest markers ranged between 55% and 83% at a specificity level of 95%. Figure 2 shows the methylation frequency distributions of the nine MethyLight assays for the most discriminative gene array markers. The scale of methylation level for these real-time assays is extended to 0.01% methylation as compared with 1% methylation used for the gene array data because of the increased sensitivity of real-time PCR. At this level of analytic sensitivity, TMEFF2, ZDHHC22, and NGFR are completely negative on PBL and show high specificity with regard to other tissues, indicating that these markers may be excellent candidates for blood-based early detection applications. TMEFF2, ZDHHC22, and EYA4 all have minimal overlap of methylation levels between colorectal cancer and normal and inflammatory colon tissue, making them potential candidates for stool-based assays or molecular classification tests; however, quantitation would be necessary for these analyses.


   Discussion
Top
 Notes
 Abstract
 Introduction
 Results
 Discussion
Materials and Methods
 Acknowledgements
 References
 
Using the combined approach of genome-wide methylation discovery with candidate marker identification, followed by microarray analysis and real-time PCR verification, we identified a set of highly methylated sequences that are present in colorectal neoplasia. We identified markers such as TMEFF2, NGFR, and ZDHHC22 that have high specificity in the diverse sample set and may be useful in clinical applications, such as noninvasive screening for detection of colorectal neoplasia in either blood- or stool-based tests. We also found markers that discriminate between normal colon and tissue with early neoplastic changes, which have potential for use in molecular classification of colon tissue to more accurately determine early neoplastic changes, tumor aggressiveness, or treatment response. For example, TMEFF2, ZDHHC22, and EYA4 could be useful for molecular classification of early stages of disease in applications such as inflammatory bowel disease surveillance.

Many genes identified during our discovery and validation process have not been reported to be methylated in the setting of cancer biology and may provide insight into gene regulation. BCOR has been shown to be associated with genes involved in cancer or regulation of cell growth. Recent studies indicate that BCOR is a transcriptional repressor of BCL6, a proto-oncogene, and is an important transcriptional regulator of embryogenesis (41, 42). Inactivation of this gene by methylation in the promoter region could provide a selective advantage for malignant cell growth. NGFR, also known as p75 (NTR), was recently identified as a tumor suppressor gene that induces apoptosis in malignant cells (43). No association with methylation in the promoter region of this gene or inhibition of this gene by methylation has previously been described. Other identified genes such as SLITRK1 and SLC32A1 have neither been associated with cancer nor reported as having aberrant methylation in their promoter regions. Interestingly, another solute carrier family member, SLC5A8, has been implicated as a tumor suppressor and was shown to be methylated in both gastric cancer cell lines and primary gastric cancers (44). Clearly, these genes warrant further investigation into their roles in malignant transformation.

Because >90% of the marker candidates identified in the methylation array study could be validated by real-time PCR (MethyLight) analysis, these data support the use of our process to identify and confirm methylation biomarkers. By using a broad genome-wide method to identify initial candidates along with a systematic selection system to differentiate those candidates with characteristics most likely to be biologically important, followed by verification on methylation microarrays and finally validation by real-time PCR, we have clearly shown that valuable biomarkers for oncological diagnostic applications, such as TMEFF2, ZDHHC22, and NGFR, can be found.

It is also evident that the markers identified in this study do not identify all colorectal cancer tissues. The lack of increase in sensitivity with paneling of the markers and the inability to identify all colorectal tumors with these panels are thought to be due to the manner in which our markers were identified and also potentially due to biology. Because, at all stages in our process, we identified and tested our markers in relation to healthy samples and other cancers, we have eliminated many markers that are methylated to any degree in these tissues. For example, GSK3B, EYA4, and APC were not identified in our discovery process and, although very highly methylated in colorectal cancer and adenomas, they are also methylated in other cancers and healthy tissues. Due to the use of pooling in our initial genome-wide discovery experiments, we also introduced a bias towards markers that show hypermethylation in a majority of colorectal cancer samples. The signal of a hypothetical marker having hypermethylation only on a small subclass of colorectal cancer samples would have been effectively diluted out by the pooling procedure. However, biologically, one can question whether methylation changes occur in all colorectal tumors. Indeed, we observed that many of the tumors with increased methylation in one marker exhibit increased methylation in multiple regions, as also reported by Issa (45). Are the remaining samples a CIMP-negative population? Follow-on marker identification studies will therefore be focused on studying the tissues that are methylation negative for the current marker set to answer this question.

In addition, further analysis of these candidate marker genes with close attention to their association with clinical variables, such as age, sex, colonic location, smoking history, family history, and others, which have been shown to be key predictors of cancer phenotype and clinical outcomes, could provide additional insight into their potential as biomarkers. Further prospective studies of these markers, based on real-time PCR assays, either in a remote sample amenable to population screening or in biopsy samples from longitudinal studies, are indicated.


   Materials and Methods
Top
 Notes
 Abstract
 Introduction
 Results
 Discussion
 Materials and Methods
Acknowledgements
 References
 
Patient Samples
Institutional review boards at all participating sites approved this study. A complete list of all patient samples used in this study is found in Supplementary Table S2.

Genome-Wide Discovery. Differentially methylated sequences were identified using pathologically verified colonic tissue samples obtained from the National Disease Research Interchange (Philadelphia, PA), the Cooperative Human Tissue Network (Nashville, TN), and ILSbio (Chestertown, MD). These included 25 adenocarcinomas, 6 adenomas, and 42 tumor-free control tissues. Normal blood for peripheral blood lymphocyte isolation was obtained from Puget Sound Blood Center (Seattle, WA).

Gene Array. Pathologically verified tissues were obtained from surgical procedures or endoscopic biopsies done at the Mayo Clinic (Rochester, MN), Semmelweis University Clinic (Budapest, Hungary), or University Hospital Carl Gustav Carus (Dresden, Germany). All normal tissues were obtained from patients endoscopically verified as absent of lesions and without a history of neoplasia. The total sample set included 358 patient DNAs and 2 control DNAs. The patient DNAs were extracted from 29 normal colon samples; 31 inflammatory bowel disease; 55 colon polyps (45 polyps <1 cm, 10 polyps ≥1 cm); 89 colorectal cancers (30 Dukes A/B, 56 Dukes C/D, 1 unknown, 2 high-grade polyps ≥1 cm); 116 noncolonic cancer samples from liver (9), bile duct (10), pancreas (10), lung (squamous and adenocarcinoma; 38), breast (28), prostate (5), esophagus (6), stomach (10), and normal tissue from sites other than colon: PBL (14); esophagus mucosa (7), gastric mucosa (7), and liver (10).

Additionally, one control sample of unmethylated human DNA (Molecular Staging, New Haven, CT) and one control sample of enzymatically methylated DNA (SssI; New England Biomedical) were included. All colon and lung tissues were matched by age/sex as well as location in the colon and the lung (central and peripheral).

MethyLight Assays. Pathologically verified tissues were obtained from surgical procedures or endoscopic biopsies done at the Mayo Clinic (Rochester, MN), Semmelweis University Clinic (Budapest, Hungary), or University Hospital Carl Gustav Carus (Dresden, Germany), or by commercial sample collections done by Asterand (Detroit, MI), Integrated Lab Services (Research Triangle Park, NC), and Clinomics (Pittsfield, MA) in accordance with a provided specimen collection protocol. The total sample set included 149 patient DNAs from normal colon tissue (18), pathologically normal colon tissue adjacent to tumor (28), normal PBLs (25), inflammatory bowel disease (9), colon polyps (11), colorectal cancers (28), breast cancer (15), and liver cancer (15). Not all assays were run on all samples because of limited DNA amounts.

DNA Extraction
DNA extraction of snap-frozen surgical samples for discovery was done using Genomic Tip-500 columns (Qiagen, Valencia, CA). Extraction for the microarray and real-time PCR assays was optimized by sample type including tissue sections from snap-frozen tissue, frozen surgical specimens, and snap-frozen small biopsies. Surgical specimens from University Hospital Carl Gustav Carus were extracted using Genomic Tip-100 columns. Frozen tissue sections from Mayo Clinic were extracted using a MagNa Pure device (Roche Applied Science, Indianapolis, IN). DNA from biopsies done at Semmelweis University Clinic was prepared using Qiagen buffers and the High Pure PCR Template Preparation Kit (Roche Applied Science).

Genome-Wide Identification of Differentially Methylated Sequences
To identify markers with high specificity for colon cancer, we used pooled genomic DNA from colonic normal, adenoma, and adenocarcinoma tissues and analyzed them using the previously described methods, methylation-specific arbitrarily primed PCR (35) and methylated CpG island amplification (36).

Patient samples used in these experiments were divided into three age groups: >65, 50 to 65, and <50 years. Samples were also divided into four types depending on the extent of disease: (a) normal adjacent tissue (>6 cm from tumor) or no disease, (b) adenomas, (c) cancer with no nodal involvement or metastasis (N0M0), and (d) advanced disease with nodal involvement (N1-2M0) and/or metastasis (N1-2M1). For each of these age and disease groups, three to five patient samples were combined into one pool. In addition, methylation patterns of all cancerous and precancerous conditions from all age groups were compared with age-matched normal peripheral blood lymphocytes.

Gene Array
The microarray was done as previously specified (13) with oligonucleotides covering regions of 43 discovery experiment– and literature-derived genes and 2 control genes. For the discovery experiment–derived genes, primer pairs and oligonucleotides were designed around the identified differentially methylated sequence whenever possible. Multiple primer pairs and oligonucleotides were designed for some genes for a total of 54 amplicons and a total of 248 oligonucleotide pairs. Each oligonucleotide contained two to three CpG sites. Hybridization conditions allowed the detection of single nucleotide differences. Additionally, eight negative control oligonucleotides with random sequences were included to facilitate estimation of unspecific background hybridization. The methylation proportion of each oligonucleotide was estimated from four spot repetitions per microarray and, on average, four hybridization repetitions per sample using a maximum likelihood algorithm.9 Unmethylated human DNA (Molecular Staging) and enzymatically methylated control DNAs (SssI; New England Biomedical) were used to calibrate the data. Amplicons for all discovery genes, candidate genes, and control genes used in the combined array are shown in Supplementary Table S3.

MethyLight Assays
The MethyLight assays were done on the ABI Prism 7900 (Applied Biosystems) using standard TaqMan chemistry as previously reported by Eads et al. (15).

Standard curves for each assay were established using CpGenome Universal Methylated DNA (Millipore/Chemicon, Temecula, CA) at concentrations between 31.6 pg/µL and 31.6 ng/µL DNA. Sample DNA was diluted to 2 ng/µL and aliquoted into strip tubes for three assays. Ten nanograms of DNA per reaction were tested in duplicate for each assay. A methylation-unspecific assay for ß-actin was used to determine total bisulfite-treated DNA concentration for each sample. Assay components and running conditions are shown in Supplementary Table S4.

Statistical Analysis
Analysis of the gene array was done on log10-transformed methylation proportions averaged over all CpG positions from the same gene by computing the mean.

Hierarchical clustering of the gene array data was done by using the simple two-norm as distance metric between samples and between genes. Samples and genes were clustered using Ward's minimum variance method (46). Fisher's exact test was used to analyze the association between clustering results and phenotypes.

AUC values were estimated using the trapezoidal rule. P values were computed with a Wilcoxon test. A simple cutoff classifier was used for classification. Sensitivity and specificity were estimated by 200 bootstrapping runs that randomly divided the data set into training set (about two thirds of the samples) and test set (about one third of the samples). For every bootstrap run, the cutoff was set to 95% specificity on the respective training set. Sensitivity and specificity were then computed from the respective test set. We report median sensitivity and specificity values from the 200 bootstrap runs as well as 90% confidence intervals (5% and 95% quantiles of the bootstrap estimates). For the two-marker panel analysis, the reported panel value for each sample was computed by taking the maximum of the two individual marker measurement values.

MethyLight analysis was done on the ratio of methylated DNA (measured by the respective marker assay) to total bisulfite DNA (measured by the ß-actin assay). Ideally, this ratio results in a number in the range [0, 1] and represents the proportion of methylated DNA in the respective sample. DNA amounts were estimated from the respective standard curves by linear regression. Replicate marker measurements were averaged.


   Acknowledgements
Top
 Notes
 Abstract
 Introduction
 Results
 Discussion
 Materials and Methods
 Acknowledgements
References
 
We thank Lori Tonnes-Priddy, Karen Cardon, Tamas Rujan, and Jana Papassotiriou for their technical contributions; and Kathy Obermeyer and Jennifer Maas for their valuable assistance in the management of the diverse and extensive samples sets.


   Notes
Top
 Notes
Abstract
 Introduction
 Results
 Discussion
 Materials and Methods
 Acknowledgements
 References
 
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Note: Supplementary data for this article are available at Molecular Cancer Research Online (http://mcr.aacrjournals.org/).

8 http://www.ensembl.org Back

9 F. Model. Statistical analysis of microarray based DNA methylation data, Ph.D. Thesis, Technische Universität Berlin, 2007. Back

Received 2/ 6/06; revised 12/21/06; accepted 12/22/06.


   References
Top
 Notes
 Abstract
 Introduction
 Results
 Discussion
 Materials and Methods
 Acknowledgements
 References
 

  1. Itzkowitz SH. Colonic polyps and polyposis syndromes. In: Felman LFM, Sleisenger M, editors. Sleisenger and Fordtran's gastrointestinal and liver disease: pathophysiology, diagnosis, management. 7th ed. Vol. II. Philadelphia: Saunders; 2002. p. 2175–214.
  2. Vogelstein B, Fearson E, Hamilton SR, et al. Genetic alterations during colorectal-tumor development. N Engl J Med 1988;319:525–32.[Abstract]
  3. Fearon ER, Vogelstein B. A genetic model for colorectal tumorigenesis. Cell 1990;61:759–67.[CrossRef][Medline]
  4. Chung D. The genetic basis of colorectal cancer: insights into critical pathways of tumorigenesis. Gastroenterology 2000;119:854–65.[CrossRef][Medline]
  5. Jass J, Whitehall V, Young J, Leggett B. Emerging concepts in colorectal neoplasia. Gastroenterology 2002;123:862–76.[CrossRef][Medline]
  6. Herman JG. Hypermethylation pathways to colorectal cancer: implications for prevention and detection. Gastroenterol Clin North Am 2002;31:945–58.[CrossRef][Medline]
  7. Herman JG, Umar A, Polyak K, et al. Incidence and functional consequences of hMLH1 promoter hypermethylation in colorectal carcinoma. Proc Natl Acad Sci U S A 1998;95:6870–5.[Abstract/Free Full Text]
  8. Jones PA, Baylin SB. The fundamental role of epigenetic events in cancer. Nat Rev Genet 2002;3:415–28.[Medline]
  9. Kane MF, Loda M, Gaida GM, et al. Methylation of the hMLH1 promoter correlates with lack of expression of hMLH1 in sporadic colon tumors and mismatch repair-defective human tumor cell lines. Cancer Res 1997;57:808–11.[Abstract/Free Full Text]
  10. Jones PA. DNA methylation and cancer. Cancer Res 1986;46:461–6.[Free Full Text]
  11. Golub TR, Slonim DK, Tamayo P, et al. Molecular classification of cancer: class discovery and class prediction by gene expression monitoring. Science 1999;286:531–7.[Abstract/Free Full Text]
  12. Alizadeh AA, Eisen MB, Davis RE, et al. Distinct types of diffuse large B-cell lymphoma identified by gene expression profiling. Nature 2000;403:503–11.[CrossRef][Medline]
  13. Adorjan P, Distler J, Lipscher E, et al. Tumour class prediction and discovery by microarray-based DNA methylation analysis. Nucleic Acids Res 2002;30:e21.[Abstract/Free Full Text]
  14. Gitan RS, Shi H, Chen CM, Yan PS, Huang TH. Methylation-specific oligonucleotide microarray: a new potential for high-throughput methylation analysis. Genome Res 2002;12:158–64.[Abstract/Free Full Text]
  15. Eads CA, Danenberg KD, Kawakami K, et al. MethyLight: a high-throughput assay to measure DNA methylation. Nucleic Acids Res 2000;28:e32.[Abstract/Free Full Text]
  16. Herman JG, Baylin SB. Gene silencing in cancer in association with promoter hypermethylation. N Engl J Med 2003;349:2042–54.[Free Full Text]
  17. Nakahara Y, Shintani S, Mihara M, Hino S, Hamakawa H. Detection of p16 promoter methylation in the serum of oral cancer patients. Int J Oral Maxillofac Surg 2006;35:362–5.[Medline]
  18. Ulivi P, Zoli W, Calistri D, et al. p16INK4A and CDH13 hypermethylation in tumor and serum of non-small cell lung cancer patients. J Cell Physiol 2006;206:611–5.[CrossRef][Medline]
  19. Ichikawa D, Koike H, Ikoma H, et al. Detection of aberrant methylation as a tumor marker in serum of patients with gastric cancer. Anticancer Res 2004;24:2477–81.[Medline]
  20. Marini A, Mirmohammadsadegh A, Nambiar S, Gustrau A, Ruzicka T, Hengge UR. Epigenetic inactivation of tumor suppressor genes in serum of patients with cutaneous melanoma. J Invest Dermatol 2006;126:422–31.[CrossRef][Medline]
  21. Chu HJ, Heo J, Seo SB, et al. Detection of aberrant p16INK4A methylation in sera of patients with liver cirrhosis and hepatocellular carcinoma. J Korean Med Sci 2004;19:83–6.[Medline]
  22. Valenzuela MT, Galisteo R, Zuluaga A, et al. Assessing the use of p16(INK4a) promoter gene methylation in serum for detection of bladder cancer. Eur Urol 2002;42:622–8; discussion 628–630.[CrossRef][Medline]
  23. Holst CR, Nuovo GJ, Esteller M, et al. Methylation of p16(INK4a) promoters occurs in vivo in histologically normal human mammary epithelia. Cancer Res 2003;63:1596–601.[Abstract/Free Full Text]
  24. Ahlquist DA, Skoletsky JE, Boynton KA, et al. Colorectal cancer screening by detection of altered human DNA in stool: feasibility of a multitarget assay panel. Gastroenterology 2000;119:1219–27.[CrossRef][Medline]
  25. Dong S, Traverso G, Johnson C, Geng L. Detecting colorectal cancer in stool with the use of multiple genetic targets. J Natl Cancer Inst 2001;93:858–65.[Abstract/Free Full Text]
  26. Zou HZ, Yu BM, Wang ZW, et al. Detection of aberrant p16 methylation in the serum of colorectal cancer patients. Clin Cancer Res 2002;8:188–91.[Abstract/Free Full Text]
  27. Nakayama H, Hibi K, Taguchi M, Takase T. Molecular detection of p16 promoter methylation in the serum of colorectal cancer patients. Cancer Lett 2002;188:115–9.[CrossRef][Medline]
  28. Esteller M, Sanchez-Cespedes M, Rosell R, Sidransky D, Baylin SB, Herman JG. Detection of aberrant promoter hypermethylation of tumor suppressor genes in serum DNA from non-small cell lung cancer patients. Cancer Res 1999;59:67–70.[Abstract/Free Full Text]
  29. Brabender J, Usadel H, Danenberg KD, et al. Adenomatous polyposis coli gene promoter hypermethylation in non-small cell lung cancer is associated with survival. Oncogene 2001;20:3528–32.[CrossRef][Medline]
  30. Usadel H, Brabender J, Danenberg KD, et al. Quantitative adenomatous polyposis coli promoter methylation analysis in tumor tissue, serum, and plasma DNA of patients with lung cancer. Cancer Res 2002;62:371–5.[Abstract/Free Full Text]
  31. Grady WM, Rajput A, Lutterbaugh JD, Markowitz SD. Detection of aberrantly methylated hMLH1 promoter DNA in the serum of patients with microsatellite unstable colon cancer. Cancer Res 2001;61:900–2.[Abstract/Free Full Text]
  32. Ahlquist DSA, Klatt KK, Harrington JJ, Cunningham JM. Novel use of hypermethylated DNA markers in stool for detection of colorectal cancer: a feasibility study. Gastroenterology 2002;122 (Suppl 1): A40.
  33. Müller H, Oberwalder M, Fiegi H, et al. Methylation changes in faecal DNA: a marker for colorectal cancer screening? Lancet 2004;363:1283–5.[CrossRef][Medline]
  34. Chen WD, Han ZJ, Skoletsky J, et al. Detection in fecal DNA of colon cancer-specific methylation of the nonexpressed vimentin gene. J Natl Cancer Inst 2005;97:1124–32.[Abstract/Free Full Text]
  35. Liang G, Salem CE, Yu MC, et al. DNA methylation differences associated with tumor tissues identified by genome scanning analysis. Genomics 1998;53:260–8.[CrossRef][Medline]
  36. Toyota M, Ho C, Ahuja N, et al. Identification of differentially methylated sequences in colorectal cancer by methylated CpG island amplification. Cancer Res 1999;59:2307–12.[Abstract/Free Full Text]
  37. Liang G, Robertson KD, Talmadge C, Sumegi J, Jones PA. The gene for a novel transmembrane protein containing epidermal growth factor and follistatin domains is frequently hypermethylated in human tumor cells. Cancer Res 2000;60:4907–12.[Abstract/Free Full Text]
  38. Issa JP. CpG-island methylation in aging and cancer. Curr Top Microbiol Immunol 2000;249:101–18.[Medline]
  39. Sabbioni S, Miotto E, Veronese A, et al. Multigene methylation analysis of gastrointestinal tumors: TPEF emerges as a frequent tumor-specific aberrantly methylated marker that can be detected in peripheral blood. Mol Diagn 2003;7:201–7.[CrossRef][Medline]
  40. Bai AH, Tong JH, To KF, et al. Promoter hypermethylation of tumor-related genes in the progression of colorectal neoplasia. Int J Cancer 2004;112:846–53.[CrossRef][Medline]
  41. Huynh KD, Fischle W, Verdin E, Bardwell VJ. BCoR, a novel corepressor involved in BCL-6 repression. Genes Dev 2000;14:1810–23.[Abstract/Free Full Text]
  42. Ng D, Thakker N, Corcoran CM, et al. Oculofaciocardiodental and Lenz microphthalmia syndromes result from distinct classes of mutations in BCOR. Nat Genet 2004;36:411–6.[CrossRef][Medline]
  43. Tabassum A, Khwaja F, Djakiew D. The p75(NTR) tumor suppressor induces caspase-mediated apoptosis in bladder tumor cells. Int J Cancer 2003;105:47–52.[CrossRef][Medline]
  44. Ueno MTM, Akino K, Suzuki H, et al. Aberrant methylation and histone deacetylation associated with silencing of SLC5A8 in gastric cancer. Tumour Biol 2004;25:134–40.[CrossRef][Medline]
  45. Issa JP. The epigenetics of colorectal cancer. Ann N Y Acad Sci 2000;910:140–53; discussion 153–145.[Abstract/Free Full Text]
  46. Venables WN, Ripley BD. Modern applied statistics with S-PLUS. 3rd ed. New York: Springer; 2002.
  47. Strausberg RL, Feingold EA, Grouse LH, et al. Generation and initial analysis of more than 15,000 full-length human and mouse cDNA sequences. Proc Natl Acad Sci U S A 2002;99:16899–903.[Abstract/Free Full Text]
  48. Nagase T, Nakayama M, Nakajima D, Kikuno R, Ohara O. Prediction of the coding sequences of unidentified human genes. XX. The complete sequences of 100 new cDNA clones from brain which code for large proteins in vitro. DNA Res 2001;8:85–95.[Abstract]
  49. Beckers M, Gabriels J, van der Maarel S, et al. Active genes in junk DNA? Characterization of DUX genes embedded within 3.3 kb repeated elements. Gene 2001;264:51–7.[CrossRef][Medline]
  50. Ota T, Suzuki Y, Nishikawa T, et al. Complete sequencing and characterization of 21,243 full-length human cDNAs. Nat Genet 2004;36:40–5.[CrossRef][Medline]
  51. Suzuki M, Shigematsu H, Shames DS, et al. DNA methylation-associated inactivation of TGFß-related genes DRM/Gremlin, RUNX3, and HPP1 in human cancers. Br J Cancer 2005;93:1029–37.[CrossRef][Medline]
  52. Hacker BM, Tomlinson JE, Wayman GA, et al. Cloning, chromosomal mapping, and regulatory properties of the human type 9 adenylyl cyclase (ADCY9). Genomics 1998;50:97–104.[CrossRef][Medline]
  53. Bos R, van Diest PJ, van der Groep P, et al. Protein expression of B-cell lymphoma gene 6 (BCL-6) in invasive breast cancer is associated with cyclin D1 and hypoxia-inducible factor-1{alpha} (HIF-1{alpha}). Oncogene 2003;22:8948–51.[CrossRef][Medline]
  54. Nagase T, Kikuno R, Ishikawa K, Hirosawa M, Ohara O. Prediction of the coding sequences of unidentified human genes. XVII. The complete sequences of 100 new cDNA clones from brain which code for large proteins in vitro. DNA Res 2000;7:143–50.[Abstract/Free Full Text]
  55. Halder SK, Beauchamp RD, Datta PK. Smad7 induces tumorigenicity by blocking TGF-ß-induced growth inhibition and apoptosis. Exp Cell Res 2005;307:231–46.[CrossRef][Medline]
  56. Osborn NK, Zou H, Molina JR, et al. Aberrant methylation of the eyes absent 4 gene in ulcerative colitis-associated dysplasia. Clin Gastroenterol Hepatol 2006;4:212–8.[Medline]
  57. Szadkowski M, Jiricny J. Identification and functional characterization of the promoter region of the human MSH6 gene. Genes Chromosomes Cancer 2002;33:36–46.[CrossRef][Medline]
  58. Esteller M, Sparks A, Toyota M, et al. Analysis of adenomatous polyposis coli promoter hypermethylation in human cancer. Cancer Res 2000;60:4366–71.[Abstract/Free Full Text]
  59. Stallmach A, Wittig BM, Kremp K, et al. Down-regulation of CD44v6 in colorectal carcinomas is associated with hypermethylation of the CD44 promoter region. Exp Mol Pathol 2003;74:262–6.[CrossRef][Medline]
  60. Toyooka S, Toyooka KO, Harada K, et al. Aberrant methylation of the CDH13 (Hcadherin) promoter region in colorectal cancers and adenomas. Cancer Res 2002;62:3382–6.[Abstract/Free Full Text]
  61. Hart MJ, de los Santos R, Albert IN, Rubinfeld B, Polakis P. Down-regulation of ß-catenin by human Axin and its association with the APC tumor suppressor, ß-catenin and GSK3ß. Curr Biol 1998;8:573–81.[CrossRef][Medline]
  62. Nakayama T, Watanabe M, Suzuki H, et al. Epigenetic regulation of androgen receptor gene expression in human prostate cancers. Lab Invest 2000;80:1789–96.[Medline]
  63. Ahuja N, Li Q, Mohan AL, Baylin SB, Issa JP. Aging and DNA methylation in colorectal mucosa and cancer. Cancer Res 1998;58:5489–94.[Abstract/Free Full Text]
  64. Biswas S, Chytil A, Washington K, et al. Transforming growth factor ß receptor type II inactivation promotes the establishment and progression of colon cancer. Cancer Res 2004;64:4687–92.[Abstract/Free Full Text]
  65. Alonso ME, Bello MJ, Gonzalez-Gomez P, et al. Aberrant promoter methylation of multiple genes in oligodendrogliomas and ependymomas. Cancer Genet Cytogenet 2003;144:134–42.[CrossRef][Medline]
  66. Lin SY, Yeh KT, Chen WT, Chen HC, Chen ST, Chang JG. Promoter CpG methylation of caveolin-1 in sporadic colorectal cancer. Anticancer Res 2004;24:1645–50.[Medline]



This article has been cited by other articles:


Home page
 Clin. Chem.Home page
C. Lofton-Day, F. Model, T. DeVos, R. Tetzner, J. Distler, M. Schuster, X. Song, R. Lesche, V. Liebenberg, M. Ebert, et al.
DNA Methylation Biomarkers for Blood-Based Colorectal Cancer Screening
Clin. Chem., February 1, 2008; 54(2): 414 - 423.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Supplementary Data, Model, et al
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
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 HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Model, F.
Right arrow Articles by Lofton-Day, C.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Model, F.
Right arrow Articles by Lofton-Day, C.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
Cancer Research Clinical Cancer Research
Cancer Epidemiology Biomarkers & Prevention Molecular Cancer Therapeutics
Molecular Cancer Research Cancer Prevention Research
Cancer Prevention Journals Portal Cancer Reviews Online
Annual Meeting Education Book Cell Growth & Differentiation