Hypermethylation of tumor suppressor genes (p16INK4A, p14ARF and APC) in adenocarcinomas of the upper gastrointestinal tract



Aberrant promoter methylation is an important mechanism for gene silencing. In the present study, 50 Barrett's esophagus-associated esophageal adenocarcinomas (ADC), 50 cardiac ADC and 50 gastric ADC were investigated by means of methylation-specific real-time PCR for hypermethylation in the tumor suppressor genes APC, p16INk4A and p14ARF. Additionally, expression of p16INK4A protein in the carcinomas was assessed using immunohistochemistry. Marked differences in hypermethylation were found between esophageal, cardiac and gastric ADC in the APC gene (78% vs. 32% vs. 84%) and in the p16INK4A gene (54% vs. 36% vs. 10%). Hypermethylation of p14ARF was absent from esophageal ADC and present infrequently in cardiac (2%) and gastric ADC (10%). Complete loss of p16INK4A protein expression was detectable in 45% of all tumors and was significantly associated with hypermethylation of the p16INK4A gene (p<0.0001, χ2-test). Our results suggest that hypermethylation of p16INK4A and APC are frequent findings in esophageal, cardiac and gastric ADC. Additionally, the data point to a tumor specific methylation pattern in upper gastrointestinal ADC. © 2004 Wiley-Liss, Inc.

Adenocarcinomas (ADC) of the esophagus and the stomach are characterized by important differences in etiological and clinical background. Whereas the incidence of esophageal ADC has increased in Western countries in the last 30 years, there has been a decrease in gastric ADC incidence during the whole 20th century.1 Chronic gastroesophageal reflux disease and subsequent development of Barrett's esophagus on the one side and dietary factors, bile reflux and Helicobacter-pylori-infection on the other side are among the most important risk factors for esophageal or gastric ADC, respectively.2, 3 The position of cardiac ADC in this context is less clear. Although traditionally considered as a gastric carcinoma, cardiac ADC shares a number of features with esophageal ADC such as profound male predominance, rising incidence and etiological association with chronic gastroesophageal reflux.4, 5 Moreover, recent investigations using the comparative genomic hybridization technique showed that genetic aberrations in cardiac ADC are probably more closely correlated to esophageal than to gastric ADC.6, 7 However, the hypothesis that esophageal and cardiac ADC are probably one entity5 has been contradicted by the results of others.8

Aberrant DNA methylation is a common feature of human cancer.9, 10, 11, 12 In recent years, a CpG island hypermethylation profile of tumors has emerged that indicates a tumor type specific methylation pattern at least for some tumor suppressor genes.13, 14 Concerning adenocarcinomas of the upper gastrointestinal tract, previous studies indicate that hypermethylation of the tumor suppressor genes APC and p16INK4A are prevalent findings in esophageal15, 16 as well as in gastric ADC.17, 18 In contrast, hypermethylation of p14ARF seems to be substantially less frequent in esophageal ADC16 than in gastric ADC.19, 20 However, since methylation patterns of esophageal and gastric ADC have not yet been compared within one study, reported differences between both tumor types may be influenced by differences in the methods used for the detection of hypermethylation. Moreover, the prevalence of hypermethylation of APC, p16INK4A and p14ARF in cardiac ADC has not been investigated so far. Thus, it is currently unclear whether significant differences in methylation pattern of these tumor suppressor genes exist between esophageal, cardiac and gastric ADC. Furthermore, it is not known whether distinct methylation patterns exist in correlation with the predominant histological subtypes of upper gastrointestinal tract adenocarcinomas (diffuse subtype vs. gland-forming subtype) as it has been recently shown for histological subtypes of lung cancer21 and breast cancer.22

In our study, we therefore investigated hypermethylation of APC, p16INK4A and p14ARF in 50 esophageal, 50 cardiac ADC and 50 gastric ADC with special reference to underlying histological subtype. Additionally, the expression of p16INK4A protein was determined immunohistochemically and compared to hypermethylation data.


Paraffin-embedded tumor samples from 50 Barrett's esophagus-associated esophageal ADC, 50 cardiac ADC and 50 gastric ADC that underwent surgical resection between January 1, 1987 and December 31, 2001 were investigated. The only selection criterion was that no preoperative radio- or chemo-therapy had been given. Differentiation between esophageal and cardiac ADC was performed according to standard criteria.23 The pT category and the pN category of the tumors were determined according to the current TNM classification,24 cardiac carcinomas were classified as tumors of the stomach. Furthermore, histological subtyping of all tumors was performed according to the WHO classification for gastric ADC3 and Lauren's classification.25 Esophageal and cardiac ADC were included in this approach as no generally accepted histological system exists for subtyping of these tumor types. Tumor grading for all carcinomas was determined according to the WHO classification.3 All tumors with a diffuse or signet-ring cell component comprising more than 5% of the tumor mass were categorized as G3. The patients' and tumors' characteristics are summarized in Table I.

Table I. Clinicopathological Parameters of 150 Adenocarcinomas (ADC) Under Investigation
 Esophageal ADCCardiac ADCGastric ADC
Age - median (range)62 (33–81)63 (38–79)70 (37–89)
Gender - male45 (90%)45 (90%)23 (46%)
 female5 (10%)5 (10%)27 (54%)
pT category   
 116 (32%)6 (12%)13 (26%)
 210 (20%)25 (50%)22 (44%)
 323 (46%)13 (26%)13 (26%)
 41 (2%)6 (12%)2 (4%)
pN category   
 021 (42%)13 (26%)18 (36%)
 1–329 (58%)37 (74%)32 (64%)
 12 (4%)1 (2.0%)2 (4%)
 220 (40%)14 (28%)15 (30%)
 328 (56%)35 (70%)33 (66%)
WHO classification   
Adenocarcinoma (tubular, papillary, mucinous)47 (94%)45 (90%)32 (64%)
Signet-ring cell carcinoma3 (6%)5 (10%)18 (36%)
Lauren's classification   
 Intestinal44 (88%)38 (76%)25 (50%)
 Diffuse01 (2%)13 (26%)
 Mixed6 (12%)11 (22%)12 (24%)

DNA preparation and real-time methylation-specific PCR

Genomic DNA from formalin-fixed, paraffin-embedded ADC tissue and corresponding tumor-free gastric smooth muscle tissue, which served as constitutive control, were prepared under light microscopic control as published before.26 For methylation-specific PCR (MSP), 1 μg of genomic DNA was modified using a CpG modification kit (CpG Genome™ DNA modification kit, Intergen, Purchase, NY) according to the manufacturer's protocol. Briefly, by bisulfite treatment, unmethylated cytosine bases are converted to uracil, whereas methylated cytosines remain unchanged. After treatment, methylated DNA sequence differs from unmethylated DNA, which is used to design methylation specific primers and probes. Real-time MSP was performed in a LightCycler (Roche Diagnostics, Mannheim, Germany) using the Taqman technology and previously published primers and probes.16

Before the analysis of the target genes, i.e., APC, p16INK4A and p14ARF, the presence of amplifiable DNA was determined in each sample by amplification of a fragment of MYOD1 as reference gene.27 Primers and probes of the MYOD1 gene were located in an area without CpG nucleotides; thus, amplification of MYOD1 by MSP occurs independent of CpG island methylation status. MSP for MYOD1 and subsequently also for APC, p16INK4A and p14ARF were performed in a final volume of 18 μl with 100 ng of bisulfite-treated DNA, 1.2 μl of each primer, 1.0 μl probe, 2.0 μl MgCl2 and 2.0 μl of a ready-to-use “hot” start reaction mix for PCR (LightCycler-Fast start DNA master Hybridization Probes, Roche Diagnostics, Mannheim, Germany). Sequences of primers and probes and the MSP conditions are summarized in Table II. PCR reaction and real-time-data collection were performed using the LightCycler software, Version 3.5 detection system (Roche Diagnostics, Mannheim, Germany). Each experiment included a positive control with known hypermethylation of the respective gene and a negative control that consisted of distillated water. Each sample that provided positive fluorescence signal using methylation specific primers and probes was considered as positive for hypermethylation.

Table II. Sets of Primers and Probes for the Investigated Gene Loci and Conditions of Methylation-Specific PCR (MSP)
LocusPrimer and probe set (probes are labeled 5′FAM and 3′TAMRA)MSP-Conditions
APCF: gAA CCA AAA CgC TCC CCA T1. 10 min 95°C,
 R: TTA TAT gTC ggT TAC gTg CgT TTA TAT2. 10 sec 95°C,
 3. 5 sec 58°C,
 Probe: CCC gTC gAA AAC CCg CCg ATT A4. 8 sec 72°C,
 5. →40 cycles
p16INK4AF: Tgg AgT TTT Tgg TTg ATT ggT T1. 10 min 95°C,
 R: AAC AAC ACC CAC ACC TCC T2. 10 sec 95°C,
 3. 5 sec 58°C,
 Probe: ACC CAA CCC CAA ACC ACA4. 8 sec 72°C,
 5. →40 cycles
p14ARFF: ACg ggC gTT TTC ggT AgT T1. 10 min 95°C,
 R: CCg AAC CTC CAA AAT CTC gA2. 10 sec 95°C,
 3. 5 sec 60°C,
 Probe: CgA CTC TAA ACC CTA CgC ACg CAA AA4. 8 sec 72°C,
 5. →40 cycles
MYOD1F: ggA TTT ATA TTT ATg Tgg Tgg gTg g1. 10 min 95°C,
 3. 5 sec 61°C,
 Probe: gTT Agg ggA TAg Agg gAg gTA TTT Agg TTg4. 8 sec 72°C,
 5. →40 cycles

p16INK4A immunohistochemistry

Slides from one representative tumor block per case were stained immunohistochemically using a monoclonal p16INK4A-antibody (Clone:16P07; Neomarkers, Westinghouse, CA). at a dilution of 1:50 according to a standard avidin-biotin-peroxidase protocol. The number of stained tumor cells was determined semiquantitatively; each sample was assigned to one of the following categories: 0 (0–4 positive tumor cells %), I (5–50%) or II (51–100%). Protein expression was determined blinded for the results of methylation analyses and vice versa.


The comparison between the frequency of methylation in APC, p16INK4A and p14ARF in the study groups was performed by means of the χ2-test; p-values < 0.05 were considered as statistically significant.


Hypermethylation of APC

Using real-time methylation-specific PCR, hypermethylation of APC was found significantly more often in esophageal ADC (78%) and in gastric ADC (84%) than in cardiac ADC (32%; p<0.0001 each; χ2-test) (Table III; Fig. 1c). Hypermethylation of APC in corresponding normal tissue was found only in 1 patient with esophageal ADC and in 1 patient with gastric ADC.

Table III. Frequency of APC, p14ARF and p16INK4A Gene Hypermethylation and Loss of p16INK4A Protein Expression in Esophageal, Cardiac and Gastric Adenocarcinomas (ADC)
Hypermethylation n (%)Loss of protein expression n (%)
Esophageal39 (78%)027 (54%)36 (72%)
 (n = 50)    
Cardiac16 (32%)1 (2%)18 (36%)21 (42%)
 (n = 50)    
Gastric42 (84%)5 (10%)5 (10%)10 (20%)
 (n = 50)    
Figure 1.

(a) Strong cytoplasmatic and nuclear expression of p16INK4A protein in a moderately differentiated adenocarcinoma of the cardia. (b) Partial loss of p16INK4A expression (immunohistochemical category I) in a poorly differentiated adenocarcinoma of the stomach (center and lower half). (c) Result of a methylation-specific real-time PCR for APC exhibiting hypermethylation in 2 carcinoma samples (CA-1 and CA-2) and in the positive control. No fluorescence signals due to a lack of hypermethylation in corresponding normal tissues (Normal-1 and Normal-2).

Considering the entirety of the tumors under investigation, no correlation was found between hypermethylation of APC and the parameters pT category, pN category, grading and Lauren's classification (data not shown). Regarding the WHO classification, signet-ring cell carcinomas showed a tendency for a higher prevalence of APC hypermethylation (80.8%) than the other types of adenocarcinomas (tubular, papillary and mucinous) (61.3%), which marginally failed to achieve statistical significance (p=0.0589).

Hypermethylation of p16INK4A

Hypermethylation of p16INK4A was found significantly more often in esophageal ADC (54%) and in cardiac ADC (36%) than in gastric ADC (10%; p<0.0001 and p=0.0020, respectively; Table III). Hypermethylation of p16INK4A was not detectable in any of the corresponding normal tissues,

Looking at the ADC as a whole, hypermethylation of p16INK4A was found significantly less frequent among signet-ring cell carcinomas (15.4%) as defined by the WHO classification than among other types of adenocarcinomas (37.1%; p=0.0327). Concerning Lauren's classification, hypermethylation of p16INK4A was found in none of 14 diffuse carcinomas but was found in 36.5% of intestinal carcinomas and in 37.9% of mixed carcinomas (p=0.0208). No correlation was found with the parameters pT category (pT1/pT2 vs. pT3/pT4), pN category (pN0 vs. pN+) and grading (data not shown).

Hypermethylation of p14ARF

Hypermethylation of p14ARF was absent in esophageal ADC and present infrequently in cardiac (2%) and gastric ADC (10%; Table III), only the differences between esophageal and gastric ADC being statistically significant (p=0.0218). In none of corresponding normal tissues, hypermethylation of p14ARF was detectable.

Considering the entirety of tumors under investigation, no correlation was found between hypermethylation of p14ARF and the parameters pT category, pN category, grading, WHO classification and Lauren's classification (data not shown).

Expression of p16INK4A protein

Expession of p16INK4A protein was considered to be present when nuclear and/or cytoplasmatic staining was detectable (Fig. 1a,b). Of the esophageal ADC, 72% were completely negative (category 0), 4% were weakly positive (category I, 5–50% positive cells) and 24% were strongly positive (category II, 51–100% positive cells). Among the cardiac ADC, 42% were in category 0, 12% in category I and 46% in category II. Twenty percent of the gastric ADC were in category 0, 14% in category I and 66% in category II (Table III). Among nonneoplastic cells of the esophagus and the stomach, expression of p16INK4A protein was found in scattered cells of the basal and parabasal layer of the esophageal squamous epithelium and in the epithelia of the glands of the fundic mucosa, as well as in scattered lymphocytes, fibroblasts and endothelial cells of the gastric and the esophageal wall.

Correlation between p16INK4A protein expression and gene hypermethylation

Complete absence of p16INK4A protein expression (category 0) was present in a total of 67 of 150 ADC under investigation. Of these 67 cases, 39 (58.2%) exhibited concomitant hypermethylation of the p16INK4A gene. In contrast, only 10 of 83 cases (12.1%) with p16INK4A protein expression exhibited concomitant hypermethylation of the p16INK4A gene (p<0.0001).


Our study showed significant differences in the prevalence of promoter hypermethylation in APC, p16INK4A and p14ARF between esophageal, cardiac and gastric ADC. Thus, hypermethylation of APC was found significantly more often in esophageal ADC and in gastric ADC than in cardiac ADC, whereas hypermethylation of p16INK4A was more prevalent in esophageal ADC and in cardiac ADC than in gastric ADC. Furthermore, a marginally significant higher prevalence of hypermethylation of p14ARF was detected in gastric ADC than in esophageal ADC. These data provide evidence that the 3 tumor types are probably characterized by a specific methylation pattern and add to the currently held view that they represent 3 distinct entities. However, esophageal, cardiac and gastric ADC in themselves are not homogeneous diseases. Especially, gastric ADC is recognized for marked morphological heterogeneity and the 2 histological main types of gastric cancer, i.e., diffuse and intestinal cancer, are characterized by marked differences in etiological and epidemiological background.3 We therefore compared the methylation data with the underlying tumor subtypes as defined by the most recognized classification systems for gastric ADC, i.e., the WHO classification and Lauren's classification. This analysis showed that hypermethylation of APC is found more often among signet-ring cell carcinomas as defined by the WHO classification than among other types of ADC, although the difference marginally failed to achieve statistical significance. In contrast, hypermethylation of p16INK4A is relatively prevalent in gland-forming 'intestinal' carcinomas and not found in pure diffuse carcinomas according to Lauren's classification. Owing to the relative rarity of p14ARF hypermethylation, the correlation analyses showed no significant differences with regard to histological subtype whereas our findings confirm previously published data that p14ARF hypermethylation is more prevalent in gastric than in esophageal ADC.16, 19, 20 Currently, only few studies have addressed the question whether different histological subtypes of cancer originating within the same organ differ in the hypermethylation pattern. Thus, p16INK4A hypermethylation has been correlated with the squamous cell type of lung cancer21 and hypermethylation of the DAP kinase gene with the invasive lobular subtype of breast cancer.22 In contrast to our study, To et al.20 found no correlation between p16INK4A hypermethylation and Lauren's classification in a series of 31 gastric ADC. This contradicting result may be due to the relatively low number of cases under investigation or to differences in etiological and genetic background of the 2 study populations. However, our findings fit into the currently considered concept that significant differences in the molecular aberrations exist between the diffuse and glandular subtypes of gastric cancer.3, 28

In our study, hypermethylation of p16INK4A was correlated with complete loss of p16INK4A protein expression with high statistical significance, a finding that underlines the functional relevance of p16INK4A hypermethylation in upper gastrointestinal ADC. However, we found exceptions in both directions, i.e., hypermethylation-negative tumors exhibiting complete loss of p16INK4A protein expression as well as hypermethylation-positive tumors exhibiting retained protein expression. The first finding can probably be explained by alternative mechanisms of p16INK4A inactivation, such as loss of heterozygosity, point mutations or homozygous deletions.29 Retained protein expression in hypermethylation-positive ADC on the other side may be due to partial methylation or hemimethylation of the p16INK4A gene.10, 11

In conclusion, our study provides evidence that the prevalence of CpG island hypermethylation in ADC of the upper gastrointestinal tract is correlated with the location of the primary tumor and with the underlying histological subtype.


The expert technical assistance of Mrs. H. Huss, Mrs. C. Pawlik, Mrs. S. Schneeloch and Mrs. C. Feldhoff is greatly appreciated.