Abnormal hypermethylation of CpG islands associated with tumor suppressor genes can lead to repression of gene expression and contribute significantly to tumorigenesis. Esophageal squamous cell carcinoma (ESCC) is thought to be developed through a multi-stage process, which involves basal cell hyperplasia (BCH), dysplasia (DYS), carcinoma in situ (CIS) and carcinoma. In the present study, we studied the hypermethylation of 10 selected genes in biopsies from normal individuals and resected tissues from ESCC patients. Tumor and neighboring normal and precancerous tissues including BCH, DYS and CIS were microdissected from the resected tissues by laser capture microdissection. Hypermethylation of CpG islands was examined in these samples for 10 genes: p16INK4a, p15INK4b, p14ARF, human leukocyte antigen (HLA)-A, -B, -C, hMLH1, E-cadherin (E-cad), fragile histidine triad and von Hippel-Lindau (VHL). Methylation of two Alu sequences, which neighbor E-cad and VHL, respectively, was used as control to verify the procedure of DNA extraction and chemical modification. In 48 biopsy samples with BCH or DYS, the most frequent hypermethylated genes were p16INK4a (18.8%) and p14ARF (14.6%). Seventeen out of these 48 samples (35.4%) contained hypermethylation of at least one gene. In the resected tissues, 52% of the BCH and 81% of the tumors showed hypermethylation of at least one gene. Genes hypermethylated in earlier stage lesions were always found hypermethylated at the later stage lesions in the same patient. All of the genes were methylated at some stages and they were clustered into four groups according to their frequencies. The first group of genes, which consisted of p16INK4a and p14ARF, was most frequently hypermethylated in all stages, and the frequencies increased from normal epithelial (0%) to BCH, to displasia/carcinoma in situ and ESCC. Other genes were hypermethylated less frequently. Our results suggest that hypermethylation of key genes, such as p16INK4a, p14ARF and hMLH1, may be used in combination with other molecular changes, such as p53 mutation, in the development of biomarkers for predicting the risk for ESCC.

Introduction

Esophageal squamous cell carcinoma (ESCC) is the sixth most common cancer worldwide (1). Because of difficulties in early diagnosis and poor efficacy of treatment, the 5-year survival rate of ESCC is <10%. Human ESCC is believed to develop through a multi-step process. Understanding of the molecular mechanisms in this process will not only provide biomarkers for early detection, but also enable us to improve treatment modalities. The high incidence of ESCC in Linzhou (formally named Linxian) in the Henan province of China offers an opportunity to study the molecular changes that take place during the development of this disease (2). Epidemiological studies have suggested tobacco, alcohol, nitrosamines, mycotoxins, physical injury and chronic inflammation as major risk factors in ESCC (25). In our previous studies, frequent alterations in the p53 and Rb tumor suppressor pathways, including mutation of p53, loss of heterozygosity of Rb, hypermethylation and homozygous deletion of p16INK4a, p15INK4b and p14ARF genes, have been identified in the ESCC samples from Linzhou (610). p53 Mutations (10) and p16INK4a and p14ARF hypermethylation (11) are likely to occur at the early stage of carcinogenesis, but more studies on these events are needed.

In addition to p16INK4a, p15INK4b and p14ARF, many other genes such as human leukocyte antigen (HLA) class I genes and fragile histidine triad (FHIT) gene have also been reported to be methylated in ESCC (11,12) and many more have been observed in other cancers (13,14). DNA methylation, especially 5′-CpG methylation, is an important mechanism in silencing the expression of genes (1317). The 5-methylcytosine protrudes into the major groove of the DNA helix (18) and possibly interferes with the binding of transcription factors (19). A group of methyl-CpG-binding proteins, which preferentially bind to methylated CpG dinucleotides, is also thought to be involved in DNA methylation mediated transcription inactivation (20,21). One of the methyl-CpG-binding proteins, MeCP2, interacts with Sin3A which is involved in histone deacetylation (22), suggesting that DNA methylation may induce chromosome remodeling through histone deacetylation resulting in transcriptional repression. The maintenance DNA methyltransferase, DNMT1, can establish a repressive transcription complex by binding to histone deacetylase and DNMT1 associated protein (23). Aberrant DNA methylation has been found in various genes, including the putative tumor suppressor genes, Rb and p16INK4a, leading to their down-regulation in tumors (24). Recently, Costello et al. (25) conducted a global examination of CpG methylation in a large group of tumors, and observed tumor type-specific patterns.

To elucidate the molecular mechanism of carcinogensis and develop biomarkers for early cancer detection, genome scale screens for genetic alterations have been conducted at both DNA and RNA levels by many scientists (2630). In order to provide insight into the epigenetic abnormality in ESCC carcinogenesis, we studied herein DNA hypermethylation of multiple genes in ESCC and their precancerous lesions. We selected 10 genes for the methylation analysis in biopsy samples and resected ESCC tissues. The 9p21 gene cluster, including p14ARF, p15INK4b and p16INK4a, has been shown in our laboratory to be frequently inactivated by promoter hypermethylation in ESCC. The frequencies of hypermethylation were as high as 40, 12.5 and 15% for p16INK4a, p15INK4b and p14ARF, respectively (8). The 6p21 gene cluster, harboring HLA-A, -B and -C genes, has also been frequently methylated and transcriptionally inactivated in ESCC (11). The hypermethylation of HLA genes, however, was not observed in several precancerous lesions of ESCC where hypermethylation of p16INK4a occurred (11). The FHIT gene on 3p14.2 was suggested to be a tumor suppressor gene in various types of epithelial cancers (31). Loss of FHIT gene expression was associated with progression of ESCC (32). Hypermethylation of FHIT was detected in 30% of the esophageal cell lines and 14% of ESCC (12). Although not reported in ESCC, hypermethylation of E-cadherin (E-cad), von Hippel-Lindau (VHL) and hMLH1 were frequent in many other cancers (3335). In this study, we characterized the pattern of hypermethylation of these genes at different histopathological stages of ESCC development. p16INK4a, p14ARF and hMLH1 were frequently hypermethylated in precancerous lesions, suggesting the contribution of these events in esophageal carcinogenesis in some patients.

Materials and methods

Collection of esophageal biopsies from general population

In endoscopic screening for esophageal cancer, biopsies were taken from symptom-free individuals in Huojia, a county neighboring Linzhou City (formerly known as Linxian) of northern China, a well-recognized high-risk area for ESCC. The samples were frozen and stored in liquid nitrogen, on dry ice, or at –80°C until use. The biopsies were embedded with tissue freeze medium (OTC). Serial sections, 10 μm thick, were cryosected and stained with hematoxylin and eosin to determine the precancerous lesions. Out of the 108 samples (one from each subject), 46 biopsies were diagnosed to contain basal cell hyperplasia (BCH) and 2 with dysplasia (DYS). Premalignant cells were microdissected using a laser capture microdissection system (ARCTURUS, Mountain View, CA). About 200–500 cells were dissected.

Collection of surgically resected samples from ESCC patients

Twenty-five surgically resected samples were collected from patients in the Linzhou People’s Hospital. The samples were frozen in liquid nitrogen within 1 h after surgical resection and were stored in liquid nitrogen, on dry ice, or at –80°C until use. All specimens were dissected and embedded with OTC. Serial sections, 10 μm thick, were cryosected, and one from every 10 slides was stained with hematoxylin and eosin. The presence of tumor, carcinoma in situ (CIS)/DYS, BCH or normal tissues in the samples was confirmed histopathologically. Precancerous lesions were microdissected using a laser capture microdissection system for the genetic assay. About 500 cells were dissected for each lesion.

DNA extraction and methylation-specific PCR

DNA was extracted with a QIAGEN Tissue Kit (QIAGEN, Valencia, CA) following the manufacturer’s procedure. The DNA was aliquoted and stored at –20°C until use. The DNA was modified by bisulfite reaction using the procedure developed by Herman et al. (34) and optimized for microdissected tissues in our laboratory (11). Methylated-specific and unmethylated-specific primers were taken from the literature or designed by us previously (Table I). Methylation was determined by the presence of the methylated-specific PCR products and absence of unmethylated-specific PCR. High annealing temperatures were used to ensure the specificity of both methylated and unmethylated-specific PCR. After 13 min of heat activation, the reaction was incubated for 4 cycles of 2 min at 95°C, 2 min at 65°C and then 2 min at 72°C. The PCR reaction then underwent 35–45 cycles of 10 s at 95°C, 45 s at 62°C and 30 s at 72°C. A 30 min incubation at 72°C was used to finalize the PCR amplification.

Clustering of genes according to their methylation patterns

Frequencies at four different stages, including normal epithelia, BCH, DYS and CIS, and tumor, were chosen to represent the methylation status for each gene. A real-valued matrix was constructed where each entry in the matrix is the frequency of methylation of a gene (or Alu sequence) in one of the four stages. A similarity matrix was calculated to represent the distance between genes. The similarity between genes was measured based on their Euclidian distance of the frequencies. A hierarchical unsupervised clustering method using furthest neighbor was used to cluster the genes. At each step, the closest two genes, gene and group of genes, or groups of genes, were clustered to form a new group. The distance between gene and group was measured so as to represent the furthest distance between the gene and any gene from the group. The distance between groups was the furthest distance between two genes, one from each group. This clustering process was terminated when all of genes merged into one single group. After clustering, the genes were organized in correspondence to their clusters and distance and represented in a dendrogram. A cut-off point (a distance threshold) was chosen so that a proper number of clusters were achieved.

Results

Multiple gene DNA hypermethylation in esophageal biopsies

Forty-eight esophageal biopsies that contained either BCH or DYS were analyzed for the methylation status with a panel of 10 different genes and two Alu sequences that are adjacent to E-cad and VHL genes. Examples of the results are shown in Figure 1. Except E-cad, all of the other genes were methylated in some cases (Table II) indicating that DNA hypermethylation of these genes can occur as early as BCH or DYS in the development of ESCC. Among the 10 genes, p16INK4a and p14ARF were most frequently methylated, observed in 19 and 15%, respectively, of the samples examined. Three samples (nos 6035, 6414 and 7883) contained hypermethylation at all three of p16INK4ap15INK4b and p14ARF genes and two samples (nos 6013 and 6062) contained hypermethylation at two of them. The high concurrency of hypermethylation at p16INK4a, p15INK4b and p14ARF genes is consistent with what we observed previously in ESCC surgical samples (8). In the present study, p16INK4a and p14ARF were methylated independently in some cases, but the hypermethylation of p15INK4b was always accompanied by the hypermethylation of either p16INK4a or p14ARF or both, suggesting that p15INK4b hypermethylation is a bystander of the hypermethylation of the other two genes. Hypermethylation of other genes was not frequent. FHIT was methylated in three cases; hMLH1, VHL, HLA-A and HLA-B were each methylated in two cases; HLA-C was methylated in one case and E-cad was not methylated in any of the 48 biopsies. Out of the 48 biopsy samples, which contained either BCH or DYS, 17 (35%) samples harbored hypermethylation of at least one of the genes. Nine (19%) samples harbored hypermethylation of at least two of the genes. The Alu sequences were methylated in all of the samples showing that the DNA was of good quality and the chemical modification was successful.

Gene hypermethylation in resected ESCC samples

Cells in areas with different stages of histopathological lesions were collected from surgically resected esophageal samples from 25 patients. Cells were obtained from the normal epithelial of all of the samples, areas with BCH in 21 samples, areas with DYS in 13 samples, areas with CIS in six samples and tumors in 21 samples (Table III). Like biopsy samples, methylation of the Alu sequences was observed in all of the cases, including normal tissues. None of the 10 genes were methylated in any of the normal tissues. Eleven out of 21 (52%) BCH tissues contained hypermethylation and three of them (15%) had hypermethylation at two genes. Nine out of 13 (69%) DYS samples contained hypermethylation and five of them (34%) had hypermethylation at two or more genes. Five out of six (83%) CIS samples contained methylation and four of them (67%) were with two or more genes. Since DYS and CIS are pathologically similar, we may combine them into one group. In the 14 samples containing either DYS or CIS or both, nine cases (64%) contained hypermethylation of one gene and five (36%) had hypermethylation of two or more genes. Seventeen out of the 21 (81%) tumor tissues had hypermethylation and 12 of them (57%) contained hypermethylation of two or more genes. The hypermethylation frequency increased from BCH (52%) to DYS/CIS (64%) and finally reached 81% in tumors. In addition, once a gene was hypermethylated in the early stage lesions, the hypermethylation of the same gene was always observed in the later stage lesions in the same esophageal specimen (Figure 1, Table III). These results suggest that DNA hypermethylation does not reverse and the frequency of hypermethylation increases during the development of ESCC. Compared with the biopsy samples, the resected tissues contained hypermethylation at a more orderly mode with fewer genes methylated, at relatively higher frequencies.

Clustering of genes according to their hypermethylation frequencies in different stage of lesions

The occurrence of genetic alterations in the development of ESCC may occur in a certain order (36). The earlier and more frequently a genetic alteration occurs, the more likely it contributes to the development of cancer. A mathematical approach, furthest neighbor unsupervised hierarchy clustering method, was used to cluster the genes into different groups based on their hypermethylation frequencies in different stages (Figure 2). In the beginning of the clustering analysis, the 10 genes and two Alu sequences were listed separately. Then two genes with the smallest distance merged to form a group. It was then followed by a new round of merging genes or groups with the smallest distance. This clustering process was terminated when all the genes merged into one single group (Figure 2, left panel). After clustering, all the genes were organized in a dendrogram where genes with similar patterns were close to each other (Figure 2, left panel). The relative distance between genes is indicated by the width of the connective stacked shapes. As shown in Figure 2 (right panel), the two Alu sequences (group 0) were methylated in all cases and thus form a standalone cluster. Other than the Alu sequences, p16INK4a and p14ARF genes (group 1) were most frequently methylated in all stages. Group 2 contains hMLH1 gene only, which was fairly frequently methylated in BCH, but the hypermethylation frequencies did not change much in later stages. Group 3 includes FHIT, HLA-B, HLA-C and p15INK4b genes. The hypermethylation frequencies of group three genes were quite low in BCH but significantly increased in DYS/CIS. Group 4 genes, including HLA-A, E-cad and VHL, were rarely methylated except in tumors.

If DNA hypermethylation is going to be used as biomarkers, it would be more efficient to use a smaller number of genes without decreasing the sensitivity. As shown in Figure 3, five genes, p16INK4a, p14ARF, hMLH1, HLA-B and FHIT, exhibited similar overall hypermethylation frequencies as the 10 genes did. Seventeen out of the 48 biopsy samples contained methylation of at least one of the 10 genes and 16 contained methylation of at least one of these five genes. In surgically resected tissues, all of the BCH (11 samples), DYS (nine samples) and CIS (five samples) that contained DNA hypermethylation contained methylation of at least one of the five genes.

Discussion

Ten genes, which had been shown to be frequently methylated and inactivated in ESCC or other cancers, were selected as the candidates for methylation screening in samples at different stages of development of ESCC. Among the 10 genes, p16INK4a, p15INK4b, p14ARF (7,8), HLA-A, HLA-B, HLA-C (11) and FHIT (12) have been reported to be methylated in ESCC. hMLH1, E-cad and VHL were reported to be inactivated by hypermethylation in a wide range of human cancers (37). When methylated, all of the genes were suppressed at the transcriptional level. We observed that hMLH1 was hypermethylated at appreciable frequencies in ESCC and even in precancerous lesions. Hypermethylation of E-cad and VHL were also observed in these samples but at lower frequencies. Methylation of both Alu sequences was detected in all biopsies suggesting the good quality of DNA and completeness of chemical modification. In this study, the average number of microdissected precancerous cells was 200–500 and the sample was used for the analysis for 10 genes. The result shows that the presently used methylation detection method is suitable for microdissected samples.

Hypermethylation of all the genes except E-cad was detected in the biopsy samples. Among them, p16INK4a (19%) and p14ARF (15%) were most frequently methylated. Hypermethylation of p16INK4a and p14ARF occurred in ESSC and was thought to be important events of ESCC carcinogenesis, contributing to inactivation of p53 and Rb tumor suppressor systems (8). The occurrence of p16INK4a and p14ARF hypermethylation in the precancerous lesions in the biopsies strengthens this idea. The frequencies of the hypermethylation of other genes were relatively low. The combination of these genes provided a much higher frequency of hypermethylation at the precancerous stage; i.e. 35% of the samples contained hypermethylation of at least one of the genes. If we consider hypermethylation of two or more genes, which may represent molecular changes closer to the development of cancer, there are nine (17%) such biopsy samples.

Surgically resected ESCC tissues often contain precancerous lesions of BCH, DYC and CIS. Our data from such samples showed that DNA hypermethylation occurred in all these stages but with different frequencies. Five of the 10 genes, including p16INK4a, p14ARF, HLA-B, hMLH1 and FHIT, were found methylated in BCH, but others were not. These five genes were also methylated in DYS, CIS and ESCC. Other than these genes, HLA-C was methylated in CIS. In ESCC, methylation was detected for all of the genes, including HLA-A, E-cad and VHL that were not methylated in earlier stages.

It is generally accepted that removal of methyl group from methylcystosine is an unlikely reaction under physiological conditions (15). In theory, loss of DNA hypermethylation can be achieved through DNA duplications and inhibition of methylation. We have observed that DNA methyltransferase inhibitor, 5-aza-2′-deoxy-cytodine induced DNA demethylation of HLA-B in esophageal cell line, which resulted in re-expression of the gene (11). However, in the samples analyzed herein, once a gene was methylated in a sample in one stage, it was always methylated in later stages, suggesting the maintenance DNA hypermethylation persists during the entire carcinogenesis process.

The BCH in the biopsies and the BCH in surgically resected tissues showed slightly different frequencies of gene hypermethylation. Hypermethylation in BCH from resected tissues showed a more ordered pattern with less genes methylated and at a higher frequency than the biopsy samples. This difference may be interpreted in different ways. The precancerous lesions in the ESCC patients may more closely resemble the ESCC progenitor cells than the precancerous lesions in biopsy samples, because many of the biopsied subjects may not develop cancer. However, the genes that were hypermethylated in biopsies with BCH and also hypermethylated in DYS and ESCC may be important in carcinogenesis. The reason that this event did not occur in the resected samples with BCH may imply that the samples analyzed may not be derived from the same clones as the carcinoma. A third possibility is that, because of the small sample size analyzed, the difference may not be statistically significant.

We clustered the genes as well as the two Alu sequences into five distinct groups based on their methylation frequencies at different stages of cancer development. The two Alu sequences fell into group 0 serving as a good control for both methylation analysis and mathematical clustering. Groups 1 through 4 were arranged in the order that genes that were methylated at a higher frequency and at earlier stages were given smaller numbers. Therefore, genes in smaller numbered groups, such as p16INK4a, p14ARF, hMLH1, HLA-B and HLA-C, may contribute more to the development of ESCC.

A goal of this project is to identify genes whose hypermethylation occurs early in ESCC development and is likely to contribute to the development of ESCC. Of the 10 genes analyzed, the hypermethylation and inactivation of five selected genes are more likely to contribute to the development of ESCC. We hypothesize that the more these genes that are hypermethylated, the more likely the individual will develop ESCC. Additionally, the hypermethylation of genes in lower numbered groups is likely to be more predictive than that of the genes in higher numbered groups. This hypothesis as well as the usefulness of these possible hypermethylation biomarkers, in combination with other molecular markers, such as p53 mutation, for predicting early outset of ESCC is being tested in a follow-up study in the Henan province, China.

Table I.

Primers for methylation-specific PCR

Target  Methylated specific (MS)  Un-methylated specific (US)  Position (size)/Genebank Accession no.  Reference 
aPrimers re-designed targeting at the DNA sequences specified in the references indicated. 
bPrimers directly taken from the indicated references. 
p14a  GTCGAGTTCGGTTTTGGAGG  TGAGTTTGGTTTTGGAGGTGG  MS: 225–389 (165)  (8
  GAAAACCACAACGACGAACG  AACCACAACAACAAACACCCCT  US: 227–386 (160)   
      L41934   
p15a  GATCGGTCGTTCGGTTATTG  GTTGTTTGGTTATTGTATGGG  MS: 269–475 (207)  (34
  CTTATTCTCCTCGCGCATTC  CCCTTATTCTCCTCACACAT  US: 274–477 (204)   
      S75756   
p16a  TGGAAAGATATCGCGGTTTT  GTTTGGAAAGATATCGCGGT  MS: 1035–1224 (190)  (34
  CCGCCGACTCCATACTACTC  CCACCAACTCCATACTACTC  US: 1038–1224 (187)   
      X94154   
HLA-Ab  TAGGTTTCGAAGGCGGTGTA  GGTTTTGAAGGTGGTGTATG  MS: 280–535 (256)  (11
  CCATAACGACCATCCTCGAC  CACCATAACAACCATCCTCA  US: 282–537 (256)   
      X55710   
HLA-Bb  TTTGTAATGGGGAGGCGTAG  TGTAATGGGGAGGTGTAGTG  MS: 379–627 (249)  (11
  AACGATTCGAAACGCCATAA  CAATTCAAAACACCATAACC  US: 381–625 (245)   
      D83956   
HLA-Cb  TTGGGTGTCGGGTTTTAGAG  TTGGGTGTTGGGTTTTAGAG  MS: 501–755 (255)  (11
  ACTCCCCGAATCCTACGC  AACTCCCCAAATCCTACAC  US: 501–756 (256)   
      D83957   
FHITa  GGTTTTTACGCGCGTTAGGT  GGTTTTTATGTGTGTTAGGT  MS: 205–304 (100)  (12
  CCTCGTAAAACGAAAAATACTCG  CCTCATAAAACAAAAAATACTCA  US: 205–304 (100)   
      U76263   
hMLH1a  ACGGGTAAGTCGTTTTGACG  GGTTATGGGTAAGTTGTTTTG  MS: 201–345 (145)  (38
  TAATCTATCGCCGCCTCATC  CCTAATCTATCACCACCTCATC  US: 197–347 (151)   
      U26559   
E-cada  GTGGCGTCGGAATTGTAAAG  GTGGTGTTGGAATTGTAAAGT  MS: 1023–1176 (154)  Island3 (33
  ACGCCGAAAAACTACGACTC  ACACCAAAAAACTACAACTCC  US: 1023–1176 (154)   
      L34545   
VHLa  AGCGCGTACGTAGTTTCGTT  TGGGAGTGTGTATGTAGTTTT  MS: 636–780 (145)  VHL-Exon1 (33
  CTCTTCGACGCCTACCTCCT  CTCTTCAACACCTACCTCCT  US: 632–780 (149)   
      AF010238   
E-Alub  TAGGTCGTTCGAGCGAGAGTG  AGGTTGTTTGAGTGAGAGTGTAG  MS: 24–195 (172)  Alu1 (33
  TACAAACGTATACCACCACACCG  AAACATATACCACCACACCAA  US: 25–193 (168)   
      L34545   
V-Alub  ACGTCGGTATATTGCGCG  TTGTTAATGATGTTGGTATATTGTGTG  MS: 88–466 (378)  VHL-Alu1A (33
  TTTTTTCACCCCTCTAAAATTTAATA  TTTTTTCACCCCTCTAAAATTTAATA  US: 80–466 (387)   
      U19763   
Target  Methylated specific (MS)  Un-methylated specific (US)  Position (size)/Genebank Accession no.  Reference 
aPrimers re-designed targeting at the DNA sequences specified in the references indicated. 
bPrimers directly taken from the indicated references. 
p14a  GTCGAGTTCGGTTTTGGAGG  TGAGTTTGGTTTTGGAGGTGG  MS: 225–389 (165)  (8
  GAAAACCACAACGACGAACG  AACCACAACAACAAACACCCCT  US: 227–386 (160)   
      L41934   
p15a  GATCGGTCGTTCGGTTATTG  GTTGTTTGGTTATTGTATGGG  MS: 269–475 (207)  (34
  CTTATTCTCCTCGCGCATTC  CCCTTATTCTCCTCACACAT  US: 274–477 (204)   
      S75756   
p16a  TGGAAAGATATCGCGGTTTT  GTTTGGAAAGATATCGCGGT  MS: 1035–1224 (190)  (34
  CCGCCGACTCCATACTACTC  CCACCAACTCCATACTACTC  US: 1038–1224 (187)   
      X94154   
HLA-Ab  TAGGTTTCGAAGGCGGTGTA  GGTTTTGAAGGTGGTGTATG  MS: 280–535 (256)  (11
  CCATAACGACCATCCTCGAC  CACCATAACAACCATCCTCA  US: 282–537 (256)   
      X55710   
HLA-Bb  TTTGTAATGGGGAGGCGTAG  TGTAATGGGGAGGTGTAGTG  MS: 379–627 (249)  (11
  AACGATTCGAAACGCCATAA  CAATTCAAAACACCATAACC  US: 381–625 (245)   
      D83956   
HLA-Cb  TTGGGTGTCGGGTTTTAGAG  TTGGGTGTTGGGTTTTAGAG  MS: 501–755 (255)  (11
  ACTCCCCGAATCCTACGC  AACTCCCCAAATCCTACAC  US: 501–756 (256)   
      D83957   
FHITa  GGTTTTTACGCGCGTTAGGT  GGTTTTTATGTGTGTTAGGT  MS: 205–304 (100)  (12
  CCTCGTAAAACGAAAAATACTCG  CCTCATAAAACAAAAAATACTCA  US: 205–304 (100)   
      U76263   
hMLH1a  ACGGGTAAGTCGTTTTGACG  GGTTATGGGTAAGTTGTTTTG  MS: 201–345 (145)  (38
  TAATCTATCGCCGCCTCATC  CCTAATCTATCACCACCTCATC  US: 197–347 (151)   
      U26559   
E-cada  GTGGCGTCGGAATTGTAAAG  GTGGTGTTGGAATTGTAAAGT  MS: 1023–1176 (154)  Island3 (33
  ACGCCGAAAAACTACGACTC  ACACCAAAAAACTACAACTCC  US: 1023–1176 (154)   
      L34545   
VHLa  AGCGCGTACGTAGTTTCGTT  TGGGAGTGTGTATGTAGTTTT  MS: 636–780 (145)  VHL-Exon1 (33
  CTCTTCGACGCCTACCTCCT  CTCTTCAACACCTACCTCCT  US: 632–780 (149)   
      AF010238   
E-Alub  TAGGTCGTTCGAGCGAGAGTG  AGGTTGTTTGAGTGAGAGTGTAG  MS: 24–195 (172)  Alu1 (33
  TACAAACGTATACCACCACACCG  AAACATATACCACCACACCAA  US: 25–193 (168)   
      L34545   
V-Alub  ACGTCGGTATATTGCGCG  TTGTTAATGATGTTGGTATATTGTGTG  MS: 88–466 (378)  VHL-Alu1A (33
  TTTTTTCACCCCTCTAAAATTTAATA  TTTTTTCACCCCTCTAAAATTTAATA  US: 80–466 (387)   
      U19763   
Table II.

Hypermethylated genes in biopsy samples with BCH/DYSa

No.  Pathology  p14  p15  p16  HLA-A  HLA-B  HLA-C  E-cad  E-Alu  FHIT  hMLH1  VHL  V-Alu 
aThe column named pathology indicates the type of precancerous lesion cells contained in the biopsy. B, BCH; D, DYS; cColumns under the name of each gene indicates their methylation status. E-alu: the Alu sequence associated with E-cad; V-alu, the Alu sequence associated with VHL; M, methylated; rows in italic indicate that in that sample, at least one gene is methylated. 
6013  B  M    M          M        M 
6020                     
6035  B  M  M  M          M      M  M 
6042                     
6051                     
6062  B    M  M          M  M      M 
6104  D        M  M      M        M 
6109                     
6136                     
6155                     
6158                     
6174                     
6268  D      M          M        M 
6292                     
6302                     
6310  B          M  M    M  M      M 
6314                     
6348                     
6359                     
6371                     
6403  B  M              M        M 
6414  B  M  M  M          M        M 
7424                     
7444                     
7465                     
7470                     
7543  B      M          M    M    M 
7557                     
7563                     
7576  B  M              M        M 
7579                     
7584                     
7613  B                M  M      M 
7619                     
7629                     
7644  B        M        M      M  M 
7664                     
7728  B      M          M        M 
7735                     
7780  B                M    M    M 
7817                     
7856  B  M              M        M 
7859                     
7871  B      M          M        M 
7879                     
7883  B  M  M  M          M        M 
7885                     
7888                     
Total  7 (15%)  4 (8%)  9 (19%)  2 (4%)  2 (4%)  1 (2%)  0 (0%)  48 (100%)  3 (6%)  2 (4%)  2 (4%)  48 (100%)   
No.  Pathology  p14  p15  p16  HLA-A  HLA-B  HLA-C  E-cad  E-Alu  FHIT  hMLH1  VHL  V-Alu 
aThe column named pathology indicates the type of precancerous lesion cells contained in the biopsy. B, BCH; D, DYS; cColumns under the name of each gene indicates their methylation status. E-alu: the Alu sequence associated with E-cad; V-alu, the Alu sequence associated with VHL; M, methylated; rows in italic indicate that in that sample, at least one gene is methylated. 
6013  B  M    M          M        M 
6020                     
6035  B  M  M  M          M      M  M 
6042                     
6051                     
6062  B    M  M          M  M      M 
6104  D        M  M      M        M 
6109                     
6136                     
6155                     
6158                     
6174                     
6268  D      M          M        M 
6292                     
6302                     
6310  B          M  M    M  M      M 
6314                     
6348                     
6359                     
6371                     
6403  B  M              M        M 
6414  B  M  M  M          M        M 
7424                     
7444                     
7465                     
7470                     
7543  B      M          M    M    M 
7557                     
7563                     
7576  B  M              M        M 
7579                     
7584                     
7613  B                M  M      M 
7619                     
7629                     
7644  B        M        M      M  M 
7664                     
7728  B      M          M        M 
7735                     
7780  B                M    M    M 
7817                     
7856  B  M              M        M 
7859                     
7871  B      M          M        M 
7879                     
7883  B  M  M  M          M        M 
7885                     
7888                     
Total  7 (15%)  4 (8%)  9 (19%)  2 (4%)  2 (4%)  1 (2%)  0 (0%)  48 (100%)  3 (6%)  2 (4%)  2 (4%)  48 (100%)   
Table III.

Hypermethylated genes in different stages of ESCC developmenta

  Normal  BCH  DYS  CIS  Tumor 
a`—’ indicates the corresponding sample was not available. The single letter or number indicates the methylated status of the gene it represents. 4, p14ARF; 5, p15INK4b; 6, p16INK4a; A, HLA-A; B, HLA-B; C, HLA-C; E, E-cad; F, FHIT; M, hMLH1; V, VHL. The Alu sequences associated with E-cad and VHL were methylated in all cases and thus not listed in this table. 
008    6,M  5,6,B,C,M  5,6,A,B,C,M 
018      —  —  — 
144      —  — 
145    —  —  — 
146    —  —  4,6,E 
149    —  F,V 
150      —  —   
2451    4,M  4,M  —  4,B,C,M 
2583    —  —    B,F 
2597      —  —   
8849        —  — 
971124      4,6  —  4,6,F 
971725    4,6,  4,5,6  4,5,6  — 
971726        — 
971727    —  —  6,M 
971728      —  —   
971733        —  E,V 
971734    4,M  —  —  4,C,M 
971741    —  —  — 
971754    4,6  4,6,B  4,6,B,C 
971755    — 
971756    B,C  A,B,C,M 
978516      —  —   
SQH    —    — 
  ZYJ  —  5,6,F 
  Normal  BCH  DYS  CIS  Tumor 
a`—’ indicates the corresponding sample was not available. The single letter or number indicates the methylated status of the gene it represents. 4, p14ARF; 5, p15INK4b; 6, p16INK4a; A, HLA-A; B, HLA-B; C, HLA-C; E, E-cad; F, FHIT; M, hMLH1; V, VHL. The Alu sequences associated with E-cad and VHL were methylated in all cases and thus not listed in this table. 
008    6,M  5,6,B,C,M  5,6,A,B,C,M 
018      —  —  — 
144      —  — 
145    —  —  — 
146    —  —  4,6,E 
149    —  F,V 
150      —  —   
2451    4,M  4,M  —  4,B,C,M 
2583    —  —    B,F 
2597      —  —   
8849        —  — 
971124      4,6  —  4,6,F 
971725    4,6,  4,5,6  4,5,6  — 
971726        — 
971727    —  —  6,M 
971728      —  —   
971733        —  E,V 
971734    4,M  —  —  4,C,M 
971741    —  —  — 
971754    4,6  4,6,B  4,6,B,C 
971755    — 
971756    B,C  A,B,C,M 
978516      —  —   
SQH    —    — 
  ZYJ  —  5,6,F 
Fig. 1.

Methylation detection using methylation-specific PCR. The methylation status of the Alu sequences and the CpG islands of the selected genes was determined. U, unmethylated PCR product using unmethylated-specific primers. M, methylated PCR product using methylated-specific primers. Presence of methylated PCR product and absence of unmethylated product indicate that the gene is methylated, and vice versa. For each gene, the results from one biopsy sample and one resected esophagus with different stages of lesions are illustrated. For example, for p14ARF, biopsy sample no. 6035 and resected sample no. 2451 are shown.

Fig. 1.

Methylation detection using methylation-specific PCR. The methylation status of the Alu sequences and the CpG islands of the selected genes was determined. U, unmethylated PCR product using unmethylated-specific primers. M, methylated PCR product using methylated-specific primers. Presence of methylated PCR product and absence of unmethylated product indicate that the gene is methylated, and vice versa. For each gene, the results from one biopsy sample and one resected esophagus with different stages of lesions are illustrated. For example, for p14ARF, biopsy sample no. 6035 and resected sample no. 2451 are shown.

Fig. 2.

Methylation frequencies of selected genes in different stages during the development of ESCC. The genes were clustered according to the methylation frequencies as described previously. (Left) As indicated in the methylation frequency index, darker color corresponds to higher frequency and vice versa. Genes with similar patterns were organized close to each other and the distances between genes were shown as the width of the stacked shapes that connect them. The vertical dark line indicates the cut-off point where the five groups were obtained. (Right) The vertical boxes represent the methylation frequencies of genes in specific stages. The numbers on the top indicate the group number from the clustering. The two Alu sequences were clustered into group 0 for their complete methylation in all stages. p16INK4a and p14ARF had the highest methylation frequencies in all stages. They were the earliest methylated genes and were clustered into group 1. hMLH1 was clustered into group 2; FHIT, HLA-B and HLA-C were clustered into group 3; p15INK4b, HLA-A, E-cad and VHL were rarely hypermethylated in precancerous lesions and clustered into group 4.

Fig. 2.

Methylation frequencies of selected genes in different stages during the development of ESCC. The genes were clustered according to the methylation frequencies as described previously. (Left) As indicated in the methylation frequency index, darker color corresponds to higher frequency and vice versa. Genes with similar patterns were organized close to each other and the distances between genes were shown as the width of the stacked shapes that connect them. The vertical dark line indicates the cut-off point where the five groups were obtained. (Right) The vertical boxes represent the methylation frequencies of genes in specific stages. The numbers on the top indicate the group number from the clustering. The two Alu sequences were clustered into group 0 for their complete methylation in all stages. p16INK4a and p14ARF had the highest methylation frequencies in all stages. They were the earliest methylated genes and were clustered into group 1. hMLH1 was clustered into group 2; FHIT, HLA-B and HLA-C were clustered into group 3; p15INK4b, HLA-A, E-cad and VHL were rarely hypermethylated in precancerous lesions and clustered into group 4.

Fig. 3.

. Selection of genes as possible biomarkers. At each specific stage (represented by different shape marked lines), the maximum overall hypermethylation frequency of a number of genes, with all possible combinations, was shown. The overall frequencies increased when more genes were counted. A goal of this project was to find the smallest number of genes that confer the highest overall hypermethylation frequency. As shown in the curves, when five genes were used, the overall frequencies at all four stages were only slightly lower than those when all 10 genes were counted. Such five genes are p16INK4a, p14ARF, hMLH1, FHIT and HLA-B.

Fig. 3.

. Selection of genes as possible biomarkers. At each specific stage (represented by different shape marked lines), the maximum overall hypermethylation frequency of a number of genes, with all possible combinations, was shown. The overall frequencies increased when more genes were counted. A goal of this project was to find the smallest number of genes that confer the highest overall hypermethylation frequency. As shown in the curves, when five genes were used, the overall frequencies at all four stages were only slightly lower than those when all 10 genes were counted. Such five genes are p16INK4a, p14ARF, hMLH1, FHIT and HLA-B.

3
To whom correspondence should be addressed Email: csyang@rci.rutgers.edu

We thank Dr Mingzhu Fang, Mr Chi So and Ms Yimin Wang for helpful discussion and critical reading of the manuscript. We are also grateful to Ms Dongxuan Jia for her assistance in preparing the frozen tissue samples for our analysis. Supported by NIH Grant CA65781 and facilities from NIEHS Center Grant ES 05022 and NCI Cancer Center Supporting Grant CA 72030.

References

1
Parkin,D.M., Pisani,P. and Ferlay,J. (
1993
) Estimates of the worldwide incidence of eighteen major cancers in 1985.
Int. J. Cancer
 ,
54
,
594
–606.
2
Yang,C.S. (
1980
) Research on esophageal cancer in China: a review.
Cancer Res.
 ,
40
,
2633
–2644.
3
Yang,C.S. (1982) Nitrosamines and other etiological factors in the esophageal cancer in Northern China. In Magee,P.N. (ed.) Nitrosamines In Human Cancer. Cold Spring Harbor Laboratory, New York, pp. 587–501.
4
Morgan,G.P. (
1995
) Epidemiology of oesophageal cancer.
J. Roy. Soc. Med.
 ,
88
,
119
.
5
Bollschweiler,E. and Holscher,A.H. (
2001
) [Carcinoma of the esophagus—actual epidemiology in Germany].
Onkologie
 ,
24
,
180
–184.
6
Xing,E.P., Yang,G.Y., Wang,L.D., Shi,S.T. and Yang,C.S. (
1999
) Loss of heterozygosity of the Rb gene correlates with pRb protein expression and associates with p53 alteration in human esophageal cancer.
Clin. Cancer Res.
 ,
5
,
1231
–1240.
7
Xing,E.P., Nie,Y., Wang,L.D., Yang,G.Y. and Yang,C.S. (
1999
) Aberrant methylation of p16INK4a and deletion of p15INK4b are frequent events in human esophageal cancer in Linxian, China.
Carcinogenesis
 ,
20
,
77
–84.
8
Xing,E.P., Nie,Y., Song,Y., Yang,G.Y., Cai,Y.C., Wang,L.D. and Yang,C.S. (
1999
) Mechanisms of inactivation of p14ARF, p15INK4b and p16INK4a genes in human esophageal squamous cell carcinoma.
Clin. Cancer Res.
 ,
5
,
2704
–2713.
9
Wang,L.D., Hong,J.Y., Qiu,S.L., Gao,H. and Yang,C.S. (
1993
) Accumulation of p53 protein in human esophageal precancerous lesions: a possible early biomarker for carcinogenesis.
Cancer Res.
 ,
53
,
1783
–1787.
10
Gao,H., Wang,L.D., Zhou,Q., Hong,J.Y., Huang,T.Y. and Yang,C.S. (
1994
) p53 tumor suppressor gene mutation in early esophageal precancerous lesions and carcinoma among high-risk populations in Henan, China.
Cancer Res.
 ,
54
,
4342
–4346.
11
Nie,Y., Yang,G.-Y., Song,Y., Zhao,X., So,C., Liao,J., Wang,L.-D. and Yang,C.S. (
2001
) DNA hypermethylation is a mechanism for loss of expression of the HLA class I genes in human esophageal squamous cell carcinomas.
Carcinogenesis
 ,
22
,
1615
–1623.
12
Tanaka,H., Shimada,Y., Harada,H., Shinoda,M., Hatooka,S., Imamura,M. and Ishizaki,K. (
1998
) Methylation of the 5′ CpG island of the FHIT gene is closely associated with transcriptional inactivation in esophageal squamous cell carcinomas.
Cancer Res.
 ,
58
,
3429
–3434.
13
Sugimura,T. and Ushijima,T. (
2000
) Genetic and epigenetic alterations in carcinogenesis.
Mutat. Res.
 ,
462
,
235
–246.
14
Momparler,R.L. and Bovenzi,V. (
2000
) DNA methylation and cancer.
J. Cell. Physiol.
 ,
183
,
145
–154.
15
Baylin,S.B., Herman,J.G., Graff,J.R., Vertino,P.M. and Issa,J.P. (
1998
) Alterations in DNA methylation: a fundamental aspect of neoplasia.
Adv. Cancer Res.
 ,
72
,
141
–196.
16
Robertson,K.D. and Jones,P.A. (
2000
) DNA methylation: past, present and future directions.
Carcinogenesis
 ,
21
,
461
–467.
17
Jones,P.A. (
2001
) Cancer. Death and methylation.
Nature
 ,
409
,
141
143–144.
18
Tate,P.H. and Brid,A.P. (
1993
) Effects of DNA methylation on DNA-binding proteins and gene expression.
Curr. Opin. Genet. Dev.
 ,
3
,
226
–231.
19
Kass,S.U., Pruss,D. and Wolffe,A.P. (
1997
) How does DNA methylation repress transcription?
Trends Genet.
 ,
13
,
444
–449.
20
Bird,A.P. and Wolffe,A.P. (
1999
) Methylation-induced repression—belts, braces and chromatin.
Cell
 ,
99
,
451
–454.
21
Hendrich,B. and Bird,A. (
2000
) Mammalian methyltransferases and methyl-CpG-binding domains: proteins involved in DNA methylation.
Curr. Top. Microbiol. Immunol.
 ,
249
,
55
–74.
22
Nan,X., Ng,H.H., Johnson,C.A., Laherty,C.D., Turner,B.M., Eisenman,R.N. and Bird,A. (
1998
) Transcriptional repression by the methyl-CpG-binding protein MeCP2 involves a histone deacetylase complex.
Nature
 ,
393
,
386
–389.
23
Rountree,M.R., Bachman,K.E. and Baylin,S.B. (
2000
) DNMT1 binds HDAC2 and a new co-repressor, DMAP1, to form a complex at replication foci.
Nature Genet.
 ,
25
,
269
–277.
24
Jones,P.A. and Laird,P.W. (
1999
) Cancer epigenetics comes of age.
Nature Genet.
 ,
21
,
163
–167.
25
Costello,J.F., Fruhwald,M.C., Smiraglia,D.J. et al. (
2000
) Aberrant CpG-island methylation has non-random and tumour-type-specific patterns.
Nature Genet.
 ,
24
,
132
–138.
26
Dong,G., Loukinova,E., Chen,Z., Gangi,L., Chanturita,T.I., Liu,E.T. and Van Waes,C. (
2001
) Molecular profiling of transformed and metastatic murine squamous carcinoma cells by differential display and cDNA microarray reveals altered expression of multiple genes related to growth, apoptosis, angiogenesis and the NF-kappaB signal pathway.
Cancer Res.
 ,
61
,
4797
–4808.
27
Jiang,H., Kang,D.C., Alexandre,D. and Fisher,P.B. (
2000
) RaSH, a rapid subtraction hybridization approach for identifying and cloning differentially expressed genes.
Proc. Natl Acad. Sci. USA
 ,
97
,
12684
–12689.
28
Velculescu,V.E., Zhang,L., Vogelstein,B. and Kinzler,K.W. (
1995
) Serial analysis of gene expression.
Science
 ,
270
,
484
–487.
29
Perou,C.M., Sorlie,T., Eisen,M.B., et al. (
2000
) Molecular portraits of human breast tumours.
Nature
 ,
406
,
747
–752.
30
Loeb,L.A. (
2001
) A mutator phenotype in cancer.
Cancer Res.
 ,
61
,
3230
–3239.
31
Huebner,K., Garrison,P.N., Barnes,L.D. and Croce,C.M. (
1998
) The role of the FHIT/FRA3B locus in cancer.
Annu. Rev. Genet.
 ,
32
,
7
–31.
32
Shimada,Y., Sato,F., Watanabe,G., Yamasaki,S., Kato,M., Maeda,M. and Imamura,M. (
2000
) Loss of fragile histidine triad gene expression is associated with progression of esophageal squamous cell carcinoma, but not with the patient’s prognosis and smoking history.
Cancer
 ,
89
,
5
–11.
33
Graff,J.R., Herman,J.G., Myohanen,S., Baylin,S.B. and Vertino,P.M. (
1997
) Mapping patterns of CpG island methylation in normal and neoplastic cells implicates both upstream and downstream regions in de novo methylation.
J. Biol. Chem.
 ,
272
,
22322
–22329.
34
Herman,J.G., Graff,J.R., Myohanen,S., Nelkin,B.D. and Baylin,S.B. (
1996
) Methylation-specific PCR: a novel PCR assay for methylation status of CpG islands.
Proc. Natl Acad. Sci. USA
 ,
93
,
9821
–9826.
35
Leung,W.K., Yu,J., Ng,E.K., To,K.F., Ma,P.K., Lee,T.L., Go,M.Y., Chung,S.C. and Sung,J.J. (
2001
) Concurrent hypermethylation of multiple tumor-related genes in gastric carcinoma and adjacent normal tissues.
Cancer
 ,
91
,
2294
–2301.
36
Montesano,R., Hollstein,M. and Hainaut,P. (
1996
) Genetic alterations in esophageal cancer and their relevance to etiology and pathogenesis: a review.
Int. J. Cancer
 ,
69
,
225
–235.
37
Baylin,S.B., Esteller,M., Rountree,M.R., Bachman,K.E., Schuebel,K. and Herman,J.G. (
2001
) Aberrant patterns of DNA methylation, chromatin formation and gene expression in cancer.
Hum. Mol. Genet.
 ,
10
,
687
–692.
38
Herman,J.G., Umar,A., Polyak,K., et al. (
1998
) Incidence and functional consequences of hMLH1 promoter hypermethylation in colorectal carcinoma.
Proc. Natl Acad. Sci. USA
 ,
95
,
6870
–6875