Ann Surg Oncol DOI 10.1245/s10434-014-4277-2
ORIGINAL ARTICLE – TRANSLATIONAL RESEARCH AND BIOMARKERS
Clinical Relevance of Plasma DNA Methylation in Colorectal Cancer Patients Identified by Using a Genome-Wide HighResolution Array Pei-Ching Lin, MD1, Jen-Kou Lin, MD, PhD2,3, Chien-Hsing Lin, PhD4, Hung-Hsin Lin, MD2,3, Shung-Haur Yang, MD, PhD2,3, Jeng-Kai Jiang, MD, PhD2,3, Wei-Shone Chen, MD, PhD2,3, Chih-Chi Chou, PhD5, Shih-Feng Tsai, MD, PhD4,5, and Shih-Ching Chang, MD, PhD2,3 Department of Clinical Pathology, Yang-Ming Branch, Taipei City Hospital, Taipei, Taiwan; 2Division of Colon and Rectal Surgery, Department of Surgery, Taipei Veterans General Hospital, Taipei, Taiwan; 3Faculty of Medicine, School of Medicine, National Yang-Ming University, Taipei, Taiwan; 4Division of Molecular and Genomic Medicine, National Health Research Institutes, Zhunan, Taiwan; 5Department of Life Sciences and Genome Research Center and Institute of Genome Sciences, National Yang-Ming University, Taipei, Taiwan 1
ABSTRACT Background. DNA methylation is a potential tumor marker for several cancers, including colorectal cancer (CRC), because of its heritable and stable characteristics. Methods. Using a high-resolution, genome-wide approach, we epigenotyped [450,000 CpG sites in tumor and adjacent non-tumor tissues from 23 microsatellite instability (MSI)/microsatellite stability (MSS) CRC cases. Using matrix-assisted laser desorption ionization–time of flight mass spectrometry, the methylation status of five frequently hypermethylated genes were confirmed in 75 independent CRC series and 353 CRC patients with available plasma. Results. Compared with non-tumor tissues, 13 MSI tumors had 34,836 (7 %) aberrant methylation sites, 87 % of which were hypermethylated. In contrast, only 9,806
Chien-Hsing Lin, Pei-Ching Lin, and Jen-Kou Lin have contributed equally to this manuscript.
Electronic supplementary material The online version of this article (doi:10.1245/s10434-014-4277-2) contains supplementary material, which is available to authorized users. Ó Society of Surgical Oncology 2014 First Received: 9 September 2014 S.-F. Tsai, MD, PhD e-mail: [email protected] S.-C. Chang, MD, PhD e-mail: [email protected]
(2 %) differentially methylated sites were identified in ten MSS cases (62 % hypermethylated). In both MSI and MSS, 228 promoter-associated CpG islands were hypermethylated, with AGBL4, ZNF625, MDFI, TWIST1, and FLI1 being most frequently hypermethylated. In an independent set of 35 MSI and 40 MSS cases, the methylation status of these five genes significantly differed between tumor and adjacent non-tumor tissues. Of 353 CRC patients, 230 (65.2 %), 232 (65.7 %), and 247 (70.0 %) had AGBL4, FLI1, and TWIST1 promoter hypermethylation in circulating cell-free DNA, respectively. In patients without metastasis, the sensitivity of any two or three hypermethylation markers was 52.8–57.8 and 27.9–38.9 %, respectively. The sensitivity of any two or three markers was significantly high in patients with stage IV disease (73.0 and 55.6 %, respectively). The prognostic value of these epimarkers was inconclusive. Conclusion. DNA methylation patterns differed in CRC subtypes. The identified hypermethylation markers in CRC patients may have good sensitivity in different CRC stages.
Colorectal cancer (CRC) is among the leading causes of cancer deaths worldwide and has been the most common cancer in Taiwan since 2007.1 The etiology of CRC from benign neoplasms, such as polyps, to malignant tumors can be explained by two independent genetic pathways. More than two-thirds of CRC cases demonstrate chromosomal instability characterized by alterations in tumor suppressor genes and oncogenes, including APC, TP53, and K-RAS.2,3 Microsatellite instability (MSI) is present in 10–15 % of
P.-C. Lin et al.
CRC cases and results from germline mutations in the mismatch repair system or somatic hypermethylation in the promoter region of MLH1.4,5 In addition to these genetic alterations, epigenetic mechanisms, including DNA methylation, that regulate heritable changes in gene expression without alterations in DNA sequences are associated with cancer initiation and promotion. A subgroup of CRCs with the clinicopathologic features of sporadic MSI and BRAF mutations has the CpG island methylator phenotype identified by a subset of cancer-specific CpG island markers.6–10 In addition to abnormal DNA methylation being more frequent in cancer compared with normal tissue, and considering region-specific DNA methylation-mediated gene silencing, the development of the methylation-specific polymerase chain reaction technique allows for the detection of aberrant methylation, even in small amounts of tumor-derived DNA, making methylation a better choice as a tumor biomarker. Recently, many differentially methylated biomarkers have been identified in circulating cell-free DNA (cfDNA) in the blood of cancer patients compared with healthy controls by using methylation microarray followed by analysis of methylation-sensitive restriction enzymes and real-time PCR.11–17 The level of mSEPT9 in plasma DNA has been widely studied and is approved for use in the CRC screening program in Europe.18,19 Several case-control studies examining the level of mSEPT9 in CRC patients have demonstrated moderate sensitivity (50–70 %) and high specificity ([90 %) for this biomarker.15,18,19 However, differences in DNA methylation patterns among races have been identified, partly due to environmental factors as well as genetic variance.20,21 With advances in molecular technologies, fine-scale epigenomic profiling of cfDNA from cancer patients by highresolution, genome-wide analysis can enhance our knowledge of tumorigenesis and identify novel tumor suppressor genes and/or biomarkers. In the current study, we employed the Illumina HumanMethylation450 BeadChip (Illumina, CA, USA), which covers over 450,000 methylation sites for epigenotyping [99 % of RefSeq genes in 23 CRC patients. Compared with other studies on the CRC cancer genome that focused on few genes or used low-resolution arrays,22,23 we achieved a much higher resolution, with an average of 17 CpG sites on 21,103 genes. Genome-wide profiling of the two subtypes of sporadic CRCs revealed significant differences in the patterns of DNA methylation in the cancer genome. We selected the top five CRCcommon methylation markers for replication in 35 microsatellite stability (MSS) and 40 MSI cases, and further evaluated three epimarkers in circulating cfDNA from 353 clinical cases.
MATERIALS AND METHODS Patients and Tissue Collection Initially, clinical data and tissue samples of 568 CRC patients who underwent surgery between 2008 and 2009 were obtained from the Biobank of the Taipei Veterans General Hospital, a prospectively established biobank. After excluding patients who underwent preoperative chemoradiotherapy (n = 45), neoadjuvant chemotherapy (n = 36), emergent operative procedures (n = 32), or evidence of familial adenomatous polyposis (n = 4), 451 CRC patients were enrolled in this study (electronic supplementary Fig. 1). Before clinical data and sample collection, written informed consent was obtained from all patients. Plasma was collected before surgery. After surgery, samples were collected from different tumor quadrants, with meticulous dissection. Tissue fragments and plasma were immediately frozen and stored in liquid nitrogen. Sections of cancerous tissue and corresponding normal tissue were reviewed by a senior gastrointestinal pathologist. Clinical data, including age, sex, personal and family medical history, location, TNM stage, differentiation, pathological prognostic features, and follow-up conditions, were prospectively collected. Following surgery, patients were monitored quarterly for the first 2 years and semiannually thereafter. The follow-up protocol included physical examination, digital rectal examination, carcinoembryonic antigen (CEA) analysis, chest radiography, abdominal sonography, and computed tomography, if needed. Positron emission tomography or magnetic resonance imaging was arranged for patients with an elevated CEA level but an uncertain site of tumor recurrence. DNA Isolation and Quantification After approval by the Institutional Review Board of the Taipei Veterans General Hospital (2013-11-013CCF), tissue and plasma samples for this study were obtained from the biobank. DNA from tissue specimens was extracted using the QIAamp DNA Tissue Kit (Qiagen, Valencia, CA, USA) according to the manufacturer’s recommendations; DNA quantity was confirmed using the Nanodrop 1000 Spectrophotometer (Thermo Scientific, Waltham, MA, USA). cfDNA from 1 mL plasma from each case was extracted using the QIAamp Minelute Virus Kit (Qiagen) according to the manufacturer’s recommendations; DNA quantity was confirmed using PicoGreen (Life Technologies, Carlsbad, CA, USA) and a VICTOR3 V Spectrophotometer (PerkinElmer, Inc., Waltham, MA, USA).
Plasma DNA Methylation in CRC
Microsatellite Instability Analysis
Genome-Wide Gene Expression Analysis
Five reference microsatellite markers were used, according to international criteria, for determining MSI: D5S345, D2S123, BAT25, BAT26, and D17S250. Primer sequences were obtained from GenBank (www.gdb.org). Detection of MSI was performed as previously described.24 Briefly, DNA was amplified using fluorescent PCR. PCR products were denatured and analyzed by electrophoresis on 5 % denaturing polyacrylamide gels, and results were analyzed using the GeneScan Analysis software (Applied Biosystems, Carlsbad, CA, USA). Tumor samples that exhibited allele peaks different to the corresponding normal sample(s) were classified as MSI for that particular marker. Samples with C2 MSI markers were defined as having MSI, and those with 0–1 MSI markers were classified as MSS.
RNA samples from 13 MSI and 10 MSS tumors (the same cases as used in the methylation array analysis) were prepared using Qiagen’s RNeasy kit (Qiagen) and then assayed using an Agilent Systems Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA) to ensure that highquality RNA was used for the gene expression array experiments. The Illumina TotalPrep RNA amplification kit (Ambion, Austin, TX, USA) was used to amplify and generate biotinylated RNA. Illumina Human Ref-8 V3 arrays were processed and scanned at medium photomultiplier settings as recommended by the manufacturer, and were analyzed using the GenomeStudio software (Illumina). After subtracting the background, array data was normalized using the quantized method; a detection p value \0.01 was used to ensure that only expressed genes were included in the analysis.
Genome-Wide and Target-Specific Methylation Analysis Thirteen MSI and ten MSS CRC patients who underwent a comprehensive molecular analysis in a previous study25 were enrolled for genome-wide methylation analysis. A total of 600 ng of genomic DNA from tumors and adjacent non-tumor tissues underwent genome-wide epigenotyping using the Illumina HumanMethylation450 array according to the manufacturer’s instructions. After bisulfite conversion (EpiTect Fast 96 DNA Bisulfite Kit, Qiagen), whole-genome amplification, enzymatic fragmentation, precipitation, resuspension, and hybridization, the intensity data were acquired with an Illumina HiScan scanner. The image was processed using the GenomeStudio Methylation module (Illumina) to obtain the b value of each CpG site.26 Hypomethylation was defined as a DiffScore less than or equal to -30 and a delta b value less than -0.2. Hypermethylation was defined as a DiffScore C30 and a delta b value [0.2.22 To replicate the findings from the genome-wide methylation array, methylation of the AGBL4, ZNF625, MDFI, TWIST1, and FLI1 genes were quantified in tumor and normal tissues of 75 independent CRC patients using the Sequenom MassCLEAVE base-specific cleavage method and matrix-assisted laser desorption ionization–time of flight (MALDI-TOF) mass spectrometry (Sequenom, San Diego, CA, USA). In addition, three primer pairs were designed to amplify short DNA fragments (\100 bp) of the AGBL4, TWIST1, and FLI1 in cfDNA of 353 CRC patients. Primers were designed using the online EpiDesigner software; the quantitative methylation data for each CpG site or aggregates of multiple CpG sites obtained from the MassARRAY system were analyzed using the EpiTYPER software (Sequenom).
Statistical Analysis The statistical endpoint of the analyses was disease-free survival (DFS) and overall survival (OS) from the date of surgery. The group distributions for each clinicopathologic trait were compared using a two-tailed Fisher’s exact test and a Chi-square test. Numerical values were compared using Student’s t test. Data are expressed as mean ± standard deviation. Kaplan–Meier survival curves were plotted and compared using the log-rank test, and multivariate analysis was performed using the Cox proportional hazard model. Statistical analyses were performed using SPSS for Windows, version 13.0 (SPSS Inc., Chicago, IL, USA), and statistical significance was defined as p \ 0.05. RESULTS To characterize the DNA methylation patterns of CRC at the genome-wide scale, Illumina HumanMethylation450 BeadChip profiling was performed in 10 MSS and 13 MSI clinical samples (both tumor and adjacent normal DNA samples). Promoter hypermethylation of the MLH1 gene, resulting in DNA mismatch repair system deficiency, is frequently found in MSI tumor samples. MSI tumors had MLH1 hypermethylation in 9 of the 13 cases, as identified by array-based bisulfite-converted Sanger sequencing. The average methylation level of CpG sites in tumor tissue obtained from MSI-high cases differed significantly from those of normal tissue obtained from MSI-high cases, as well as tumor tissues from MSS cases and normal tissues of MSI-high cases (Fig. 1a). We compared the methylation data between MSS- and MSI-high tumors and found that 34,836 CpG sites (7 %) were differentially methylated in
P.-C. Lin et al.
MSI 1130 T 2664 T 1506 T 890 T 1974 T 1140 T 1478 T 1120 T 1204 T 1604 T 2442 T 1148 T 2448 T 1566 T 1818 T 1914 N 1478 N 1148 N 830 N 798 N 1532 N 1546 N 1120 N 1204 N 1140 N 1662 N 2448 N 2664 N 1604 N 1818 N 1974 N 1566 N 1662 T 1506 N 2442 N 1130 N 890 N 1710 N 1246 T 1546 T 1532 T 798 T 1914 T 830 T 1246 N 1710 T
HCT116 Nic...
HCT116 DM...
A
Genome-wide methylation analysis in CRC subtypes
B MSI
MSS 98%
93%
Equally methylated CpG Differentially methylated CpG 7%
2% (34,836)
(9,806)
38%
(30,306) 87% 13%
(6,116) 62%
Hypermethylated CpG Hypomethylated CpG
24%
25%
Gene-associated
41% 59%
Intergenic 75%
Gene-associated
Gene-associated
Intergenic
Intergenic
76%
46%
54%
Methylation analysis of promoter-associated CpG Island
C MSI
MSS 98%
89% Equally methylated CpG island Differentially methylated CpG island
2%
11% (1,078)
(249)
100%
100% Hypermethylated CpG island Hypomethylated CpG island
0%
MSI-high cases (Fig. 1b). Of these, 87 % (30,306 CpG sites) were hypermethylated and 13 % (4,530 CpG sites) were hypomethylated. For MSS tumors, only 9,806 (2 %)
0%
of CpG sites were differentially methylated, with 62 % hypermethylated and 38 % hypomethylated. Importantly, *75 % of hypermethylated CpG sites in both MSI-high
Plasma DNA Methylation in CRC b FIG. 1 Genome-wide methylation analysis in MSI and MSS CRC
cases. a Clustering analysis by genome-wide methylation analysis. Tumors and adjacent non-tumor tissues of 13 MSI and 10 MSS cases of CRC were epigenotyped and analyzed using the Illumina HumanMethylation450 BeadChip. The methylation patterns of MSI tumors were distinct from those of MSS tumors. The methylation of HCT116 was used as an internal control. b Genome-wide methylation analysis by CRC subtype. c Genome-wide methylation characterizations of promoter-associated CpG islands in CRC subtypes. MSI microsatellite instability, MSS microsatellite stability, CRC colorectal cancer
and MSS samples were located in or nearby functional genes. We further focused on the methylation patterns of promoter-associated CpG islands and found that 1,078 and 249 promoter-associated islands were hypermethylated in MSI and MSS tumors, respectively (Fig. 1c). To further explore the effect of methylation on expression levels on a genome-wide scale, we performed regression analysis of each gene in all samples. On Illumina expression and methylation arrays, there was at least one CpG site located in or nearby 12,424 expressed genes. Among these genes, the expressional levels of 5,166 (41.6 %) genes correlated with the methylation levels of CpG sites (p \ 0.05). For example, the correlation coefficient for methylation versus expression levels of the MLH1 gene was 0.49 (electronic supplementary Fig. 2). After
adjusting the p value cut-off with false discovery rate (FDR) correction, the expression levels of 4,538 (87.8 %) genes maintained a high correlation with the methylation levels of CpG sites. To identify common CRC methylation markers with potential clinical value, we identified 228 promoter-associated CpG islands that were hypermethylated in both MSI and MSS tumors (electronic supplementary Table 1), but did not find any common hypomethylated CpG sites (electronic supplementary Fig. 3). To replicate the results from genome-wide methylation analysis, we selected the top five CpG islands—AGBL4, ZNF625, MDFI, TWIST1, and FLI1—from 228 candidates, for epigenotyping in 35 MSS and 40 MSI CRC tumors. ZNF625 hypermethylation was replicated in these independent cases with a p \ 0.0001 (Fig. 2); [80 % of the cases were hypermethylated at these CpG sites (electronic supplementary Fig. 4). The amount of cfDNA in plasma is related to the OS rate in advanced cancer patients, and can serve as a target for methylation assays. We focused on the above five genes for developing a methylation assay for cfDNA, but only TWIST1, FLI1, and AGBL4 genes were detectable in the test phase. The methylation status of cfDNA in these three genes was determined by methylation-specific PCR in 353
Replication of ZNF625 methylation in independent samples Chr19 ZNF625
Covered Not covered
primer #1 100
120
140
160
180
200
220
240
260
280
300
Average of methylation ratio (seven covered CpG sites)
1
p
ORIGINAL ARTICLE – TRANSLATIONAL RESEARCH AND BIOMARKERS
Clinical Relevance of Plasma DNA Methylation in Colorectal Cancer Patients Identified by Using a Genome-Wide HighResolution Array Pei-Ching Lin, MD1, Jen-Kou Lin, MD, PhD2,3, Chien-Hsing Lin, PhD4, Hung-Hsin Lin, MD2,3, Shung-Haur Yang, MD, PhD2,3, Jeng-Kai Jiang, MD, PhD2,3, Wei-Shone Chen, MD, PhD2,3, Chih-Chi Chou, PhD5, Shih-Feng Tsai, MD, PhD4,5, and Shih-Ching Chang, MD, PhD2,3 Department of Clinical Pathology, Yang-Ming Branch, Taipei City Hospital, Taipei, Taiwan; 2Division of Colon and Rectal Surgery, Department of Surgery, Taipei Veterans General Hospital, Taipei, Taiwan; 3Faculty of Medicine, School of Medicine, National Yang-Ming University, Taipei, Taiwan; 4Division of Molecular and Genomic Medicine, National Health Research Institutes, Zhunan, Taiwan; 5Department of Life Sciences and Genome Research Center and Institute of Genome Sciences, National Yang-Ming University, Taipei, Taiwan 1
ABSTRACT Background. DNA methylation is a potential tumor marker for several cancers, including colorectal cancer (CRC), because of its heritable and stable characteristics. Methods. Using a high-resolution, genome-wide approach, we epigenotyped [450,000 CpG sites in tumor and adjacent non-tumor tissues from 23 microsatellite instability (MSI)/microsatellite stability (MSS) CRC cases. Using matrix-assisted laser desorption ionization–time of flight mass spectrometry, the methylation status of five frequently hypermethylated genes were confirmed in 75 independent CRC series and 353 CRC patients with available plasma. Results. Compared with non-tumor tissues, 13 MSI tumors had 34,836 (7 %) aberrant methylation sites, 87 % of which were hypermethylated. In contrast, only 9,806
Chien-Hsing Lin, Pei-Ching Lin, and Jen-Kou Lin have contributed equally to this manuscript.
Electronic supplementary material The online version of this article (doi:10.1245/s10434-014-4277-2) contains supplementary material, which is available to authorized users. Ó Society of Surgical Oncology 2014 First Received: 9 September 2014 S.-F. Tsai, MD, PhD e-mail: [email protected] S.-C. Chang, MD, PhD e-mail: [email protected]
(2 %) differentially methylated sites were identified in ten MSS cases (62 % hypermethylated). In both MSI and MSS, 228 promoter-associated CpG islands were hypermethylated, with AGBL4, ZNF625, MDFI, TWIST1, and FLI1 being most frequently hypermethylated. In an independent set of 35 MSI and 40 MSS cases, the methylation status of these five genes significantly differed between tumor and adjacent non-tumor tissues. Of 353 CRC patients, 230 (65.2 %), 232 (65.7 %), and 247 (70.0 %) had AGBL4, FLI1, and TWIST1 promoter hypermethylation in circulating cell-free DNA, respectively. In patients without metastasis, the sensitivity of any two or three hypermethylation markers was 52.8–57.8 and 27.9–38.9 %, respectively. The sensitivity of any two or three markers was significantly high in patients with stage IV disease (73.0 and 55.6 %, respectively). The prognostic value of these epimarkers was inconclusive. Conclusion. DNA methylation patterns differed in CRC subtypes. The identified hypermethylation markers in CRC patients may have good sensitivity in different CRC stages.
Colorectal cancer (CRC) is among the leading causes of cancer deaths worldwide and has been the most common cancer in Taiwan since 2007.1 The etiology of CRC from benign neoplasms, such as polyps, to malignant tumors can be explained by two independent genetic pathways. More than two-thirds of CRC cases demonstrate chromosomal instability characterized by alterations in tumor suppressor genes and oncogenes, including APC, TP53, and K-RAS.2,3 Microsatellite instability (MSI) is present in 10–15 % of
P.-C. Lin et al.
CRC cases and results from germline mutations in the mismatch repair system or somatic hypermethylation in the promoter region of MLH1.4,5 In addition to these genetic alterations, epigenetic mechanisms, including DNA methylation, that regulate heritable changes in gene expression without alterations in DNA sequences are associated with cancer initiation and promotion. A subgroup of CRCs with the clinicopathologic features of sporadic MSI and BRAF mutations has the CpG island methylator phenotype identified by a subset of cancer-specific CpG island markers.6–10 In addition to abnormal DNA methylation being more frequent in cancer compared with normal tissue, and considering region-specific DNA methylation-mediated gene silencing, the development of the methylation-specific polymerase chain reaction technique allows for the detection of aberrant methylation, even in small amounts of tumor-derived DNA, making methylation a better choice as a tumor biomarker. Recently, many differentially methylated biomarkers have been identified in circulating cell-free DNA (cfDNA) in the blood of cancer patients compared with healthy controls by using methylation microarray followed by analysis of methylation-sensitive restriction enzymes and real-time PCR.11–17 The level of mSEPT9 in plasma DNA has been widely studied and is approved for use in the CRC screening program in Europe.18,19 Several case-control studies examining the level of mSEPT9 in CRC patients have demonstrated moderate sensitivity (50–70 %) and high specificity ([90 %) for this biomarker.15,18,19 However, differences in DNA methylation patterns among races have been identified, partly due to environmental factors as well as genetic variance.20,21 With advances in molecular technologies, fine-scale epigenomic profiling of cfDNA from cancer patients by highresolution, genome-wide analysis can enhance our knowledge of tumorigenesis and identify novel tumor suppressor genes and/or biomarkers. In the current study, we employed the Illumina HumanMethylation450 BeadChip (Illumina, CA, USA), which covers over 450,000 methylation sites for epigenotyping [99 % of RefSeq genes in 23 CRC patients. Compared with other studies on the CRC cancer genome that focused on few genes or used low-resolution arrays,22,23 we achieved a much higher resolution, with an average of 17 CpG sites on 21,103 genes. Genome-wide profiling of the two subtypes of sporadic CRCs revealed significant differences in the patterns of DNA methylation in the cancer genome. We selected the top five CRCcommon methylation markers for replication in 35 microsatellite stability (MSS) and 40 MSI cases, and further evaluated three epimarkers in circulating cfDNA from 353 clinical cases.
MATERIALS AND METHODS Patients and Tissue Collection Initially, clinical data and tissue samples of 568 CRC patients who underwent surgery between 2008 and 2009 were obtained from the Biobank of the Taipei Veterans General Hospital, a prospectively established biobank. After excluding patients who underwent preoperative chemoradiotherapy (n = 45), neoadjuvant chemotherapy (n = 36), emergent operative procedures (n = 32), or evidence of familial adenomatous polyposis (n = 4), 451 CRC patients were enrolled in this study (electronic supplementary Fig. 1). Before clinical data and sample collection, written informed consent was obtained from all patients. Plasma was collected before surgery. After surgery, samples were collected from different tumor quadrants, with meticulous dissection. Tissue fragments and plasma were immediately frozen and stored in liquid nitrogen. Sections of cancerous tissue and corresponding normal tissue were reviewed by a senior gastrointestinal pathologist. Clinical data, including age, sex, personal and family medical history, location, TNM stage, differentiation, pathological prognostic features, and follow-up conditions, were prospectively collected. Following surgery, patients were monitored quarterly for the first 2 years and semiannually thereafter. The follow-up protocol included physical examination, digital rectal examination, carcinoembryonic antigen (CEA) analysis, chest radiography, abdominal sonography, and computed tomography, if needed. Positron emission tomography or magnetic resonance imaging was arranged for patients with an elevated CEA level but an uncertain site of tumor recurrence. DNA Isolation and Quantification After approval by the Institutional Review Board of the Taipei Veterans General Hospital (2013-11-013CCF), tissue and plasma samples for this study were obtained from the biobank. DNA from tissue specimens was extracted using the QIAamp DNA Tissue Kit (Qiagen, Valencia, CA, USA) according to the manufacturer’s recommendations; DNA quantity was confirmed using the Nanodrop 1000 Spectrophotometer (Thermo Scientific, Waltham, MA, USA). cfDNA from 1 mL plasma from each case was extracted using the QIAamp Minelute Virus Kit (Qiagen) according to the manufacturer’s recommendations; DNA quantity was confirmed using PicoGreen (Life Technologies, Carlsbad, CA, USA) and a VICTOR3 V Spectrophotometer (PerkinElmer, Inc., Waltham, MA, USA).
Plasma DNA Methylation in CRC
Microsatellite Instability Analysis
Genome-Wide Gene Expression Analysis
Five reference microsatellite markers were used, according to international criteria, for determining MSI: D5S345, D2S123, BAT25, BAT26, and D17S250. Primer sequences were obtained from GenBank (www.gdb.org). Detection of MSI was performed as previously described.24 Briefly, DNA was amplified using fluorescent PCR. PCR products were denatured and analyzed by electrophoresis on 5 % denaturing polyacrylamide gels, and results were analyzed using the GeneScan Analysis software (Applied Biosystems, Carlsbad, CA, USA). Tumor samples that exhibited allele peaks different to the corresponding normal sample(s) were classified as MSI for that particular marker. Samples with C2 MSI markers were defined as having MSI, and those with 0–1 MSI markers were classified as MSS.
RNA samples from 13 MSI and 10 MSS tumors (the same cases as used in the methylation array analysis) were prepared using Qiagen’s RNeasy kit (Qiagen) and then assayed using an Agilent Systems Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA) to ensure that highquality RNA was used for the gene expression array experiments. The Illumina TotalPrep RNA amplification kit (Ambion, Austin, TX, USA) was used to amplify and generate biotinylated RNA. Illumina Human Ref-8 V3 arrays were processed and scanned at medium photomultiplier settings as recommended by the manufacturer, and were analyzed using the GenomeStudio software (Illumina). After subtracting the background, array data was normalized using the quantized method; a detection p value \0.01 was used to ensure that only expressed genes were included in the analysis.
Genome-Wide and Target-Specific Methylation Analysis Thirteen MSI and ten MSS CRC patients who underwent a comprehensive molecular analysis in a previous study25 were enrolled for genome-wide methylation analysis. A total of 600 ng of genomic DNA from tumors and adjacent non-tumor tissues underwent genome-wide epigenotyping using the Illumina HumanMethylation450 array according to the manufacturer’s instructions. After bisulfite conversion (EpiTect Fast 96 DNA Bisulfite Kit, Qiagen), whole-genome amplification, enzymatic fragmentation, precipitation, resuspension, and hybridization, the intensity data were acquired with an Illumina HiScan scanner. The image was processed using the GenomeStudio Methylation module (Illumina) to obtain the b value of each CpG site.26 Hypomethylation was defined as a DiffScore less than or equal to -30 and a delta b value less than -0.2. Hypermethylation was defined as a DiffScore C30 and a delta b value [0.2.22 To replicate the findings from the genome-wide methylation array, methylation of the AGBL4, ZNF625, MDFI, TWIST1, and FLI1 genes were quantified in tumor and normal tissues of 75 independent CRC patients using the Sequenom MassCLEAVE base-specific cleavage method and matrix-assisted laser desorption ionization–time of flight (MALDI-TOF) mass spectrometry (Sequenom, San Diego, CA, USA). In addition, three primer pairs were designed to amplify short DNA fragments (\100 bp) of the AGBL4, TWIST1, and FLI1 in cfDNA of 353 CRC patients. Primers were designed using the online EpiDesigner software; the quantitative methylation data for each CpG site or aggregates of multiple CpG sites obtained from the MassARRAY system were analyzed using the EpiTYPER software (Sequenom).
Statistical Analysis The statistical endpoint of the analyses was disease-free survival (DFS) and overall survival (OS) from the date of surgery. The group distributions for each clinicopathologic trait were compared using a two-tailed Fisher’s exact test and a Chi-square test. Numerical values were compared using Student’s t test. Data are expressed as mean ± standard deviation. Kaplan–Meier survival curves were plotted and compared using the log-rank test, and multivariate analysis was performed using the Cox proportional hazard model. Statistical analyses were performed using SPSS for Windows, version 13.0 (SPSS Inc., Chicago, IL, USA), and statistical significance was defined as p \ 0.05. RESULTS To characterize the DNA methylation patterns of CRC at the genome-wide scale, Illumina HumanMethylation450 BeadChip profiling was performed in 10 MSS and 13 MSI clinical samples (both tumor and adjacent normal DNA samples). Promoter hypermethylation of the MLH1 gene, resulting in DNA mismatch repair system deficiency, is frequently found in MSI tumor samples. MSI tumors had MLH1 hypermethylation in 9 of the 13 cases, as identified by array-based bisulfite-converted Sanger sequencing. The average methylation level of CpG sites in tumor tissue obtained from MSI-high cases differed significantly from those of normal tissue obtained from MSI-high cases, as well as tumor tissues from MSS cases and normal tissues of MSI-high cases (Fig. 1a). We compared the methylation data between MSS- and MSI-high tumors and found that 34,836 CpG sites (7 %) were differentially methylated in
P.-C. Lin et al.
MSI 1130 T 2664 T 1506 T 890 T 1974 T 1140 T 1478 T 1120 T 1204 T 1604 T 2442 T 1148 T 2448 T 1566 T 1818 T 1914 N 1478 N 1148 N 830 N 798 N 1532 N 1546 N 1120 N 1204 N 1140 N 1662 N 2448 N 2664 N 1604 N 1818 N 1974 N 1566 N 1662 T 1506 N 2442 N 1130 N 890 N 1710 N 1246 T 1546 T 1532 T 798 T 1914 T 830 T 1246 N 1710 T
HCT116 Nic...
HCT116 DM...
A
Genome-wide methylation analysis in CRC subtypes
B MSI
MSS 98%
93%
Equally methylated CpG Differentially methylated CpG 7%
2% (34,836)
(9,806)
38%
(30,306) 87% 13%
(6,116) 62%
Hypermethylated CpG Hypomethylated CpG
24%
25%
Gene-associated
41% 59%
Intergenic 75%
Gene-associated
Gene-associated
Intergenic
Intergenic
76%
46%
54%
Methylation analysis of promoter-associated CpG Island
C MSI
MSS 98%
89% Equally methylated CpG island Differentially methylated CpG island
2%
11% (1,078)
(249)
100%
100% Hypermethylated CpG island Hypomethylated CpG island
0%
MSI-high cases (Fig. 1b). Of these, 87 % (30,306 CpG sites) were hypermethylated and 13 % (4,530 CpG sites) were hypomethylated. For MSS tumors, only 9,806 (2 %)
0%
of CpG sites were differentially methylated, with 62 % hypermethylated and 38 % hypomethylated. Importantly, *75 % of hypermethylated CpG sites in both MSI-high
Plasma DNA Methylation in CRC b FIG. 1 Genome-wide methylation analysis in MSI and MSS CRC
cases. a Clustering analysis by genome-wide methylation analysis. Tumors and adjacent non-tumor tissues of 13 MSI and 10 MSS cases of CRC were epigenotyped and analyzed using the Illumina HumanMethylation450 BeadChip. The methylation patterns of MSI tumors were distinct from those of MSS tumors. The methylation of HCT116 was used as an internal control. b Genome-wide methylation analysis by CRC subtype. c Genome-wide methylation characterizations of promoter-associated CpG islands in CRC subtypes. MSI microsatellite instability, MSS microsatellite stability, CRC colorectal cancer
and MSS samples were located in or nearby functional genes. We further focused on the methylation patterns of promoter-associated CpG islands and found that 1,078 and 249 promoter-associated islands were hypermethylated in MSI and MSS tumors, respectively (Fig. 1c). To further explore the effect of methylation on expression levels on a genome-wide scale, we performed regression analysis of each gene in all samples. On Illumina expression and methylation arrays, there was at least one CpG site located in or nearby 12,424 expressed genes. Among these genes, the expressional levels of 5,166 (41.6 %) genes correlated with the methylation levels of CpG sites (p \ 0.05). For example, the correlation coefficient for methylation versus expression levels of the MLH1 gene was 0.49 (electronic supplementary Fig. 2). After
adjusting the p value cut-off with false discovery rate (FDR) correction, the expression levels of 4,538 (87.8 %) genes maintained a high correlation with the methylation levels of CpG sites. To identify common CRC methylation markers with potential clinical value, we identified 228 promoter-associated CpG islands that were hypermethylated in both MSI and MSS tumors (electronic supplementary Table 1), but did not find any common hypomethylated CpG sites (electronic supplementary Fig. 3). To replicate the results from genome-wide methylation analysis, we selected the top five CpG islands—AGBL4, ZNF625, MDFI, TWIST1, and FLI1—from 228 candidates, for epigenotyping in 35 MSS and 40 MSI CRC tumors. ZNF625 hypermethylation was replicated in these independent cases with a p \ 0.0001 (Fig. 2); [80 % of the cases were hypermethylated at these CpG sites (electronic supplementary Fig. 4). The amount of cfDNA in plasma is related to the OS rate in advanced cancer patients, and can serve as a target for methylation assays. We focused on the above five genes for developing a methylation assay for cfDNA, but only TWIST1, FLI1, and AGBL4 genes were detectable in the test phase. The methylation status of cfDNA in these three genes was determined by methylation-specific PCR in 353
Replication of ZNF625 methylation in independent samples Chr19 ZNF625
Covered Not covered
primer #1 100
120
140
160
180
200
220
240
260
280
300
Average of methylation ratio (seven covered CpG sites)
1
p
