Main

Well-differentiated neuroendocrine tumors including pancreatic endocrine tumors and carcinoid tumors of the lung and gastrointestinal tract are uncommon,1, 2 with an age-adjusted annual incidence of 2.5–4.5 per 100 000.3, 4 These indolent malignancies have site-specific differences among well-differentiated neuroendocrine tumors from lung, gastrointestinal tract, and pancreas including clinicopathological features, behavior and genetic alterations.

The molecular mechanisms of neuroendocrine tumorigenesis are poorly understood but have been the focus of many recent reports.5, 6, 7, 8, 9, 10, 11 We have previously shown methylation of p14 (ARF), p16 (INK4a) and O6-methyl-guanine methyltransferase (MGMT) genes in well-differentiated neuroendocrine tumors from pancreas and gastrointestinal tract.11

A chromosome 9p21 locus encodes two cell cycle inhibitors, p16 and p14, that are transcribed from two separate promoters and first exons (1α and 1β, respectively) joined through the same splice acceptor site to exon 2 coding sequences but in different reading frames.12, 13 p16 inhibition of cyclin-dependent kinase 4 (CDK4) causes retinoblastoma protein to stay in its active form and arrest the cell cycle at G1–S transition.14, 15 p14 also inhibits the cell cycle by blocking MDM2 inhibition of p53 activity.16 Inactivation of the p16 and p14 genes has been reported for a variety of tumors.15, 17, 18, 19 Both p16 and p14 can be inactivated through gene mutation, heterozygous or homozygous deletion, and methylation of the CpG island in the promoter region. In pancreatic and gastrointestinal well-differentiated neuroendocrine tumors, methylation of the promoter region is the major mechanism for inactivation of these two tumor suppressor genes.8

MGMT is a DNA repair gene that is responsible for removal of mutagenic adducts from the O6 position of guanine.20 In cells lacking MGMT activity, O6-alkyl-guanine mispairs with thymine during DNA replication and results in guanine–cytosine transition mutation. Methylation of CpG islands in the promoter regions of the MGMT gene can cause loss of transcription of the gene, and promoter methylation has been reported in many tumor types,20 including pancreatic well-differentiated neuroendocrine tumors.7

RAS-association domain family 1 has seven different isoforms that are produced by alternative splicing and transcription from two different promoters with CpG islands.21 RAS-association domain family 1, isoform A (RASSF1A) gene is a tumor suppressor gene in the RAS pathway that can regulate proliferation, induce apoptosis, and bind to and stabilize microtubules. RASSF1A gene is frequently methylated in a variety of tumors,21 including pancreatic, pulmonary and gastrointestinal well-differentiated neuroendocrine tumors.7, 22, 23, 24

In the present study, we studied well-differentiated neuroendocrine tumors from the pancreas, lung and gastrointestinal tract for methylation of RASSF1A, p14, p16 and MGMT genes, and associated the epigenetic alterations with clinicopathologic features including metastasis status. We found associations with tumor site, histopatholoic type, patient age and metastasis that have implications for the pathogenesis of the tumors and clinical characteristics.

Materials and methods

Characteristics of Specimens and Patients

Frozen tumor and non-neoplastic tissue were obtained from surgical specimens of patients undergoing resections for well-differentiated neuroendocrine tumors in the frozen section laboratory of the Department of Pathology, MD Anderson Cancer Center. The MD Anderson Cancer Center Surveillance Committee (institutional review board) approved this study. The patient records and histopathological findings were reviewed. The tumors were classified as benign well-differentiated neuroendocrine tumors, well-differentiated neuroendocrine tumors of uncertain malignant potential, and well-differentiated neuroendocrine carcinomas using established criteria.25 There were three benign well-differentiated neuroendocrine tumors, nine well-differentiated neuroendocrine tumors of uncertain malignant potential, and 35 well-differentiated neuroendocrine carcinomas (Figure 1). Of these well-differentiated neuroendocrine tumors, 16 were from pancreas (pancreatic endocrine tumors), and 31 were from lung and gastrointestinal tract (carcinoid tumors). Of the tumors, 16 were in the ileum. The nonileal tumors included five pulmonary, three gastric, two duodenal, one appendiceal, three cecal and one rectal tumor. The functional status of tumors was ascertained by serum measurements of hormones and/or clinical syndrome due to hormonal hypersecretion. There was one gastrinoma and one insulinoma. Two patients with pancreatic neuroendocrine tumors had multiple endocrine neoplasia type 1 (MEN1), and two patients with gastric well-differentiated neuroendocrine tumors had chronic atrophic gastritis of the body and fundus associated with Pernicious anemia.

Figure 1
figure 1

Methylation status of RASSF1A, p14, p16 and MGMT genes in pancreatic, nonileal and ileal neuroendocrine tumors, histopatholoic type, and lymph node and liver metastasis status.

Full size image

DNA Extraction

DNA from both tumor and non-neoplastic tissue was extracted from microdissected fresh-frozen specimens using a commercial kit (Qiagen DNA extraction kit, Qiagen Inc., Valencia, CA, USA), after a hematoxylin- and eosin-stained slide from a frozen block was reviewed. The tumor cell cellularity was at least 90% in all samples examined. The non-neoplastic tissue was from the pancreatic or lung parenchyma, or normal mucosa from the primary site of tumor and had a minor component of neuroendocrine cells or islets of Langerhans. The tumor tissue was obtained from the liver metastases in one patient with pancreatic and five patients with ileal well-differentiated neuroendocrine carcinomas and the non-neoplastic tissue from the primary site was unavailable from these patients (Figure 1).

Methylation of RASSF1A, p14, p16, and MGMT Genes

Methylation of the RASSF1A, p14, p16, and MGMT genes was evaluated by bisulfite treatment of DNA followed by methylation-specific polymerase chain reaction (MSP), as previously described.6, 20, 26, 27, 28 In brief, 2 μg of microdissected genomic DNA was denatured with 2 M NaOH at 37°C for 10 min, followed by incubation with 3 M sodium bisulfite, pH 5.0, at 50°C for 16 h in darkness. After treatment, DNA was purified using the DNA Cleanup Kit (Promega Corporation, Madison, WI, USA) as recommended by the manufacturer, incubated with 3 M NaOH at room temperature for 5 min, precipitated with 10 M ammonium acetate and 100% ethanol, washed with 70% ethanol, and finally resuspended in 20 μl of distilled water. RASSAF1 gene methylation was determined by PCR amplification of unmethylated alleles by using primer pair: 5′-TTTGGTTGGAGTGTGTTAATGTG-3′ and 5′-CAAACCCCACAAACTAAAAACAA-3′, and methylated alleles by using primer pair: 5′-GTGTTAACGCGTTGCGTATC-3′ and 5′-AACCCCGCGAACTAAAAACGA-3′.28 PCR was carried out using the following conditions: denaturation at 95°C for 15 min, 35 cycles (95°C for 30 s, 60°C for 30 s, 72°C for 45 s) and extension at 72°C for 10 min. The primers and PCR conditions for the p14, p16 and MGMT genes are as previously reported.20, 26, 27 The colon cancer cell line RKO (American Type Culture Collection, Manasas, VA, USA) and water were used as positive and negative controls, respectively. PCR products from methylated and unmethylated reactions were electrophoresed on 6% acrylamide gels and visualized by ethidium bromide staining (examples in Figure 2). Samples with low level of methylation (less than 5% methylated alleles compared to unmethylated alleles were designated as unmethylated based on previous studies.6 The methylation status of p14, p16 and MGMT genes of 12 of the tumors has been reported previously.6

Figure 2
figure 2

Methylation-specific PCR for RASSF1A, p16, p14 and MGMT genes using unmethylated (U) and methylated (M) primers in three representative samples. Sample numbers and molecular weight (MW) is on the top of the lanes. Sample 1 is unmethylated and samples 2 and 3 are methylated for RASSF1A gene, samples 1 is methylated and samples 2 and 3 are unmethylated for p14, sample 1 is methylated and samples 2 and 3 are unmethylated for p16; and samples 1 and 3 are methylated and samples 2 is unmethylated for MGMT.

Full size image

Statistical Analysis

All statistical analysis was performed using SPSS (SPSS Inc., Chicago, IL, USA). Comparisons of categorical variables were made using χ2 and Fisher's exact test. Continuous data, including age at diagnosis and tumor size, were evaluated by Student's T-test and one-way ANOVA.

Results

Clinicopathologic Features

The clinicopathological features of the 16 patients with pancreatic well-differentiated neuroendocrine tumors, 15 with nonileal well-differentiated neuroendocrine tumors and 16 with ileal well-differentiated neuroendocrine tumors have been previously reported.11 Liver metastasis was present in 69% of patients with ileal neuroendocrine tumors compared to 27% with nonileal neuroendocrine tumors and 25% with pancreatic neuroendocrine tumors (P=0.02). Age, gender, size of tumor, and vital status were not statistically different among the three groups.

Methylation of RASSF1A, p14, p16, and MGMT Genes

Methylation of the RASSF1A gene was present in non-neoplastic pancreatic parenchyma of 26% (4/15) patients with pancreatic neuroendocrine tumors, but was unmethylated in non-neoplastic lung parenchyma from five patients, and three gastric, two duodenal, eight ileal and four colorectal (including one from the patient with appendiceal neuroendocrine tumor) non-neoplastic mucosal samples from patients with nonileal and ileal neuroendocrine tumors. Methylation of p14, p16 and MGMT genes is infrequent in non-neoplastic pancreas or gastrointestinal mucosa, as reported in our previous study.6

Methylation of the RASSF1A gene was present in 57% (27/47) of tumors, of the p14 gene in 49% (23/47), of the p16 gene in 26% (12/47), and of the MGMT gene in 13% (6/47) of tumors (Figure 1 and Table 1). Pancreatic, nonileal and ileal neuroendocrine tumors had methylation of RASSF1A gene in 63, 40 and 69% of tumors, respectively (not significant), p14 gene in 44, 33 and 69% (not significant), p16 gene in 19, 33 and 25% (not significant), and MGMT gene in 13, 27 and 0% (P=0.04).

Table 1 Frequency of methylation of RASSAF1A, p14, p16 and MGMT genes in pancreatic, nonileal and ileal neuroendocrine tumors
Full size table

Epigenetic Alterations and Clinicopathologic Features

The mean age of patients whose non-neoplastic pancreatic parenchyma had RASSF1A gene methylation was 54.2±15.4 years, compared to 55.8±17.2 years for those whose non-neoplastic pancreatic parenchyma was unmethylated (not significant), and 58.7±13.1 years for those whose non-neoplastic pancreatic or lung parenchyma, or mucosa was unmethylated (not significant).

Methylation status of the RASSF1A, p14, p16 and MGMT genes and clinicopathologic features are compared in Table 2. CpG island methylation at RASSF1A was associated with histopathologic type of tumor and lymph node metastasis, and p16 methylation was associated with patient age and liver metastasis status. RASSF1A methylation was present in 66% of well-differentiated neuroendocrine carcinomas, 56% of well-differentiated neuroendocrine tumors of uncertain malignant potential but in 0% of benign well-differentiated neuroendocrine tumors (P=0.03). Lymph node metastasis was present in 78% of patients whose tumors had RASSF1A methylation compared to 35% in patients whose tumors were unmethylated at p16 (P=0.004). The mean age of patients whose tumors had p16 methylation was 68.3±8.7 years, compared to 54.8±13.0 years for those without p16 methylation (P=0.002). Liver metastasis was present in 67% of patients whose tumors had p16 methylation compared to 31% in patients whose tumors were unmethylated at p16 (P=0.04).

Table 2 Methylation of RASSF1A, p14, p16 and MGMT genes in pancreatic, nonileal and ileal neuroendocrine tumors and clinicopathologic associations
Full size table

The number of methylated genes and clinicopathologic features are compared in Table 3. Two or more genes were methylated in 53% (25/47) of tumors, one gene was methylated in 30% (14/47) of tumors, and all four genes were unmethylated in 17% (8/47) of tumors (Figure 2). The mean age of patients whose tumors had methylation of two or more gene was 61.4±12.7 years, compared to 60.6±10.7 years for those with methylation of one gene, and 45.7±14.9 years for those with no methylated gene (P=0.01). Liver metastasis was present in 52% of patients whose tumors had methylation of two or more genes, compared to 14% in patients whose tumors had methylation of one gene and 50% in those patients who had no methylation of any of the genes (P=0.044).

Table 3 Number of methylated genes in pancreatic, nonileal and ileal neuroendocrine tumors and clinicopathologic associations
Full size table

Some of these associations were also present when tumor subsites were considered (Tables 2 and 3). Patients whose pancreatic well-differentiated neuroendocrine tumors were methylated at RASSF1A gene had more frequent lymph node metastasis (P=0.002), and whose tumors had methylation of one or more genes were older compared to those whose tumors were unmethylated at all four genes (P=0.01). Patients whose nonileal well-differentiated neuroendocrine tumors were methylated at RASSF1A gene had more frequent lymph node metastasis (P=0.049), and whose tumors had p16 gene methylation had more frequent liver metastasis (P=0.04). The patients whose ileal well-differentiated neuroendocrine tumors had p16 methylation were older compared to those without p16 methylation (P=0.03).

Discussion

In the current study, we found RASSF1A gene to be frequently methylated in all three sites of well-differentiated neuroendocrine tumors: 63% in pancreatic tumors, 40% in nonileal tumors, and 69% in ileal tumors. It was the most frequently methylated gene in our study. Similarly, high frequency of RASSF1A gene methylation has been reported in pancreatic and gastrointestinal well-differentiated neuroendocrine tumors,7, 23, 24, 29 but a slightly lower frequency in pulmonary well-differentiated neuroendocrine tumors.22 The RAS-association domain family 1 gene has several isoforms including RASSF1A and RASSF1C that are transcribed from two different CpG island promoters.29, 30, 31 The hypermethylation of RASSF1A promoter is frequent in many human cancers, and there is an inverse correlation between RASSF1A silencing by methylation and K-RAS activation.21, 29, 30, 31 Previous studies have shown that well-differentiated neuroendocrine tumors lack K-RAS or BRAF mutations,11, 32, 33 but have methylation of RASSF1A gene in pancreatic, pulmonary and gastrointestinal well-differentiated neuroendocrine tumors.7, 22, 23, 24 This suggests that RAS pathway is involved in well-differentiated neuroendocrine tumors mostly by gene silencing of RASSF1A gene by methylation.

In our study, methylation of p16 was found in 19% of pancreatic tumors, 33% of nonileal tumors and 25% of ileal tumors as we reported previously.6 Similarly, methylation of p16 gene was reported in more than 50% of gastrinomas and nonfunctional pancreatic well-differentiated neuroendocrine in previous studies34 and in 40% of pancreatic well-differentiated neuroendocrine tumors in another study.7 In contrast, p16 methylation was infrequent in pulmonary neuroendocrine tumors.22

p14 methylation was most prevalent in gastrointestinal tumors compared to other types of human tumors.26 In our study, we found frequent methylation of p14 in well-differentiated neuroendocrine tumors. In contrast, we and others had previously reported a low frequency of p14 methylation in pancreatic well-differentiated neuroendocrine tumors.6, 7

In our study, MGMT methylation was not present in ileal well-differentiated neuroendocrine tumors and was infrequent in pancreatic and nonileal well-differentiated neuroendocrine tumors. We have previously shown that complete loss of chromosome 18 is frequently present in ileal well-differentiated neuroendocrine tumors but not in nonileal or pancreatic well-differentiated neuroendocrine tumors.11 These findings suggest that ileal well-differentiated neuroendocrine tumors have differences in genetic and epigenetic alterations compared to well-differentiated neuroendocrine tumors from other sites. The MGMT gene is responsible for removing alkylation of DNA at the O6 position of guanine. Methylation of CpG islands in the promoter region of MGMT can cause gene silencing.20 Frequent MGMT methylation is reported in brain tumors, colorectal cancers, lung cancers, head and neck cancers and high-grade lymphomas but is uncommon in pancreatic, renal, bladder, endometrial and breast carcinoma.20

CpG island methylator phenotype has been reported in cancers from a variety of sites and types including colon, stomach and pancreas.35, 36, 37 In our study, methylation of p16 gene and methylation of two or more genes were associated with liver metastasis, and methylation of RASSF1A gene with histopathologic type of tumor and lymph node metastasis. In another study, similar associations were reported between methylation of p16 gene and lymph node metastasis, and between methylation of multiple genes and lymph node or liver metastasis in patients with pancreatic neuroendocrine tumors.7

In our study, p16 gene methylation was more common in tumors from older patients. Age-related increase in methylation has been reported in normal tissue and cancer.38, 39, 40 In our study, RASSF1A gene methylation was present in four samples of non-neoplastic pancreatic parenchyma. However, in our study islets of Langherhans were a minor component in the samples of non-neoplastic pancreatic parenchyma, and it is possible that RASSF1A gene methylation may be present in the exocrine cells. Similarly, a previous study has reported RASSF1A gene methylation of non-neoplastic parenchyma.23 Age-related methylation of RASSF1A gene has been reported in a variety of tissues including pancreas41 but no association was present between age and RASSF1A gene methylation of non-neoplastic or tumor samples in our study.

In summary, our study shows that methylation profiles differ by tumor subsite, patient age and metastatic status. The findings in our study further support the concept that these three groups of tumor are fundamentally different and the tumorigenic process evolves along different pathways.