Abstract
Background: Recent data indicate that gastroenteropancreatic neuroendocrine tumours (GEP-NETs) have a hypomethylated long interspersed element (LINE1) promoter. To answer the question, of whether LINE1 may be of value in assessing the malignant potential of GEP-NETs, we analysed LINE1 methylation in different organs. Materials and Methods: A total of 58 GEP-NETs of gastric (n=14), pancreatic (n=15), small intestine (n=17), appendix (n=8), colorectal (n=4) and non-neoplastic tissues were analysed using DNA isolation, bisulphite-treatment and pyrosequencing. Results: LINE1 hypomethylation was detected in 50% of gastric, 100% pancreatic, 82% small intestine, 87.5% appendix and 100% colorectal NETs. G1 (p<0.001) and G2 (p<0.05) colorectal, and G1 (p<0.001) and G2 (p<0.001) pancreatic NETs exhibited significant LINE1 hypomethylation compared with non-neoplastic tissues. Higher rates of LINE1 hypomethylation in G2 pancreatic NETs than in G1 NETs (p<0.05) were observed. NETs exhibited a significantly lower frequency of hypomethylation in cases with lymph node metastases (p<0.05). Conclusion: LINE1 hypomethylation may serve as a marker of tumour grade and lymph node metastasis.
Gastroenteropancreatic neuroendocrine tumours (GEP-NETs) are heterogeneous and rare tumours of epithelial origin occurring in the gastrointestinal (GI) tract exhibiting neuroendocrine differentiation (1-3). Numerous classification systems were used before the World Health Organization (WHO) classifications in 2000, subsequently revised in 2004 (4). The European Neuroendocrine Tumour Society (ENETS) proposed a TNM staging system based on tumour thickness and/or size, lymph node involvement and metastatic disease, further supplemented by a grading system (5, 6). These guidelines for GEP-NETs were adopted in the seventh edition of the TNM Classification in 2010 (7-10). Presently, apart from the recent TNM staging, NETs are graded according to their Ki-67 proliferation index or mitotic rate into G1 and G2, and neuroendocrine carcinoma (NEC) (4). Such a classification and grading has been reported to more accurately predict outcomes (11, 12). However, in individual cases, the assessment of the actual malignant potential is difficult and uncertain (13).
In addition to clinical or histopathological markers, distinct abnormalities, such as point mutations, deletions, chromosomal losses/gains and, most importantly, methylation, have also been reported as markers of differentiation and biological aggressiveness of GEP-NETs (14-19). Among these, hypomethylation of long interspersed nucleotide element type 1 (LINE1) has been previously implicated in well-differentiated NETs (16). LINE1, the most common family of non-long terminal repeats (LTR) retrotransposons in the human genome, comprises about 17% of the genome and exhibits a distinctive pattern of hypomethylation in various neoplasms (20-22). LINE1 elements retain their competency to be active in the human genome, and their hypomethylation can lead to transcriptional activation of intact LINE1 elements, induce retrotransposition, and facilitate genetic instability (23-25). In GEP-NETs, relative hypomethylation of LINE1 in tumours has been reported to be more common in patients whose tumours had loss of chromosome 18 and methylation of RASSF1A, while this was associated with lymph node metastasis (16).
In the current study, we analysed LINE1 methylation patterns in a large series of GEP-NETs of different locations, stages of disease and proliferative activity.
Materials and Methods
Patients and tissue samples. A collective of 58 archival formalin-fixed paraffin-embedded (FFPE) biopsies and resected GEP-NETs samples from 58 patients diagnosed at the Institute for Pathology, Ruhr University of Bochum between 2002 and 2008 were collected for the retrospective study. The following tumour sites were analyzed: stomach (n=14), pancreas (n=15) and small intestine/colorectum (n=29). The primary site of small intestine colorectum tumours was the duodenum (n=3), ileum (n=4), small intestine not otherwise specified (n=4), appendix (n=8), rectum (n=2) and one patient for each sigmoidal colon (n=1), ampulla of vater (n=1) and colon not otherwise specified (n=1).
The clinical data, as well as localisation, TNM and ENET stage and grade, are shown in Table I. The corresponding non-neoplastic mucosa from patients were defined as normal.
The study was approved by the local Ethics Committee of the Ruhr University Bochum, Faculty of Medicine (register no. 3980-11).
Classification and histopathological grading. The GEP-NET specimens were subjected to histological examination by two independent pathologists (A.T, I.S) for confirmation of the grading, WHO classification of NETs and staging according to the ENETs and TNM system (4). The grading system based on mitotic rate and Ki-67 labelling index was as recently proposed by ENETS and adopted by TNM 2010: G1 [mitotic count <2 per 10 high power fields (HPF), Ki-67 index up to 2%], G2 (mitotic count 2-20 per 10 HPF, Ki-67 index 3-20%), G3 (mitotic count >20 per 10 HPF, Ki-67 index up to >20%). By convention, NETs of the appendix are not described by the grading system. The discrimination of typical and atypical carcinoids is based on a mitotic count of up to 1 per 10 HPF and 2-10 per 10 HPF respectively. Nevertheless a grading of the 8 tumours was performed.
Immunohistochemical analysis. Immunostaining was performed using a Leica BOND-MAX™ autostainer and Bond Polymer Refine Red Detection™ system (Leica Microsystems, Wetzlar, Hesse, Germany) according to the manufacturer's specifications. Briefly, 1-2 micron-thin sections were subjected to heat treatment for 20 min and then incubated with pre-diluted anti-Ki-67 (#M7240, MIB-1, mouse monoclonal; Dakocytomation, Hamburg, Germany) or anti-synaptophysin 67 (#AM363-5M, mouse monoclonal; Biogenex, Fremont, CA, USA), for 20 min. Recommended positive and negative controls were used. Slides were evaluated using a Zeiss Axioplan 2 imaging system (objective, ×40). All tumour cell areas on the slide that stained positively were included as part of the evaluations regardless of the degree of staining. Cases with Ki-67 immunoreactivity of less than 1% were scored “0”.
DNA isolation, bisulphite pyrosequencing and LINE1 analysis. The histopathological lesions of interest were micro-dissected and used for DNA extraction. For patient tumour samples, 10 μm sections were made from archived FFPE tissues. Tumour tissues were microdissected using a scalpel, with the corresponding HE-slide as a template. After heating for 1 h at 62°C, the samples were deparaffinized and rehydrated by passing them through first xylene and then a graded isopropanol series. DNA isolation was performed using the Qiagen FFPE tissue kit (Qiagen, Hilden, North Rhine-Westphalia Germany), following the instructions of the manufacturer. DNA quantity was assessed spectrophotometrically. One microgram of genomic DNA was subjected to bisulphite conversion using the EpiTect Bisulphite kit (Qiagen) according to the supplier's protocol. The eluted DNA (20 μl volume) was used for the methylation-sensitive PCR analysis.
Pyrosequencing was performed in a PSQ HS96 instrument (Qiagen) using previously described methylation-specific primers for quantitative sequencing (pyrosequencing) of LINE1 (26). The mean of % methylation at 4 individual CpG sites within the sequenced region was taken as % LINE1 methylation for each sample. The primers were High-performance liquid chromatography (HPLC)-grade purified and manufactured by Metabion international AG (Martinsried, Bavaria, Germany).
PCR conditions for pyrosequencing. Fifty nanograms of bisulphite-treated DNA were used in the 25 μl PCR reaction with 400 pmol/l forward and reverse primers, respectively. PCR conditions for LINE1 were, 95°C for 15 min (95°C for 40, 55°C for 40, and 72°C for 40), 40 cycles and 1×72°C for 10 min using HotStar Taq (Qiagen).
Statistical analysis. All statistical analyses were carried out with the Statistica version 9 software package. Significance testing with continuous variables was performed using the Student t-test (for comparison of two groups).
Results
ENETs and TNM staging, grading and other clinicopathological features. The ENETs and TNM staging, grading, Ki-67 labelling index and clinicopathological features of 58 patients with GEP-NETs are summarized in Table I. Representative staining of G1 NET, G2 NET and NEC tumours for standard HE and immunohistochemical staining with Synaptophysin and Ki-67 are shown in Figure 1. Data on positive lymph node involvement was available for 11 patients (pN1=11).
Tumour site-specific LINE1 hypomethylation. LINE1 methylation was assessed using the primers described by Tellez et al., and a mean of four CpG sites was taken as an estimate of LINE1 methylation (26). When tumours exhibited a significant decrease in % LINE1 methylation in comparison to normal tissues, they were defined to be hypomethylated.
Levels of LINE1 methylation tumour samples and those of normal tissues are shown in Figure 2. The median LINE1 methylation in pancreatic GEP-NET samples was 64% compared with 73% in normal samples (p<0.001) and hence was statistically significant (Figure 2C). The median LINE1 methylation in small intestine-NET samples was 63% compared with 69.5% in normal samples (p<0.05) and hence was statistically significant (Figure 2D). The median LINE1 methylation in all 21 small intestine and colorectal NET samples excluding the 8 appendix NETs was 63.5% compared with 69.5% in normal samples (p<0.001) and hence was statistically significant (Figure 2F). In contrast, the median LINE1 methylation in gastric NETs and appendix NET samples in comparison to their respective normal samples were statistically insignificant (Figure 2B, E).
All grades of GEP-NET tumour samples (G1 NET, G2 NET and NEC) were significantly less methylated than normal tissues at LINE1 for all sites except the stomach (Figure 2G). The median LINE1 methylation in small intestine/colonic G1 NET (p<0.001) and G2 NET (p<0.05) samples was 62% and 64%, respectively, compared to 69% in normal samples (Figure 2H). Median LINE1 methylation in pancreatic G1 NET and G2 NET samples was 66.5% and 63%, respectively, compared to 73% in normal samples (p<0.001) and hence was statistically significant (Figure 2I).
Clinicopathological features of the study cohort including European Neuroendocrine Tumour Society (ENETS) and Tumour lymph node metastasis (TNM) staging, grading and Ki-67 labelling index.
Staining of G1 NET, G2 NET and neuroendocrine carcinoma (NEC) tumours. HE and immunohistochemical staining of synaptophysin and Ki-67 in G1 NET, G2 NET and NEC tumour samples.
LINE1 hypomethylation and grading. In pancreatic G2 NETs (n=5), LINE1 methylation (range=56%-64%) was significantly lower compared with G1 NETs (n=10) (range=61%-72%). The median LINE1 methylation in G2 pancreatic NET samples was 63% compared with 66.5% in G1 pancreatic NET samples (p<0.05) (Figure 3). No significant differences in LINE1 methylation between G1 NETs, G2 NETs and NECs of gastric, small intestine/colorectal or appendiceal NETs could be observed.
Site-specific LINE1 hypomethylation of GEP-NETs in relation to pT and pN category of TNM and ENETS classification and tumour size. There was no significant correlation between LINE1 methylation and tumour stage and pT category (defined by UICC and ENET). Classified by TNM 2010, median LINE1 methylation from 22 pT1 (range=57%-72%), 15 pT2 (range=54-73%) and 7 pT3 tumours (range=49%-67%) was 66%, 63% and 64% respectively. LINE1 in 1 pT4 tumour was methylated by 76%. No significant methylation differences between pT1 and pT2 (p=0.121417) and between pT2 and pT3 tumours (p=0.43855) were observed. No significant differences were observed using the ENET classification (19 pT1, 15 pT2 and 10 pT3 tumours; pT1 vs. pT2: p=0.285, pT2 vs. pT3: p=0.265).
According to the pN category, LINE1 methylation in the primary tumours was significantly lower in pN1 samples (median methylation 62%), compared with pN0 samples (median methylation 67%) (p<0.05) (Figure 4).
LINE1 hypomethylation did not significantly correlate with tumour size. The Pearson product-moment correlation coefficient (r) of LINE1 methylation vs. tumour diameter for gastric, colorectal and pancreatic GEP-NETs were −0.02, −0.31 and −0.28, respectively. The percentage explained variance (r2 ×100) calculated for these correlations were 0.07%, 10.09% and 8.24% respectively and was, hence, insignificant.
Long interspersed element (LINE1) methylation in healthy tissues and gastroenteropancreatic neuroendocrine tumour (GEP-NET) samples: A: all sites (p<0.001); B: gastric NET; C: pancreatic NET (p<0.001), D: small intestine NET (p<0.05); E: appendiceal NET; and F: small intestine and colorectal NET without appendix (p<0.001). Comparison of LINE1 methylation in healthy tissues and G1 NET, G2 NET and neuroendocrine carcinoma (NEC) tumours: G: all sites (each p<0.001); H: small intestine (each p<0.05); and I: pancreatic tumours (each p<0.001).
Discussion
Genome-wide losses of DNA methylation have been regarded as a common epigenetic event in several malignancies including GEP-NETs and may play an essential role in tumour development and progression. GEP-NETs exhibit a wide and complex spectrum of clinical behaviours with site-specific or growth site-dependent functional modulations in their behavior (27). Although various epigenetic alterations, including LINE1 hypomethylation have been reported for GEP-NETs, their tumour site-specific relevance remained unexplored. Our study, to our knowledge, for the first time, demonstrated that LINE1 hypomethylation is a predictor of tumour progression in small intestine/colorectal and pancreatic NETs but not in gastric and appendiceal NETs. Previous reports from us and other groups have shown that in general, gastric NETs and appendiceal NETs are less malignant among GEP NETs (28-30). Such site-specific variation in LINE1 hypomethylation in GEP-NETs is consistent with a model of site-specific functional interactions between mesenchymal and neuroendocrine cells contributing to the malignancy of GEP-NETs, depending on their growth site. Although GEP-NETs may possess a malignant potential, their biological behaviour differs depending on their type and hence their malignancy can be predominantly defined on the basis of cellular proliferation rate/grade (mitotic count or Ki-67 index). Such a grading based on the Ki-67 index, has been reported not only to more accurately predict outcomes but also to be an independent predictor for survival (12, 31, 32). Therefore, the current study investigating tumour site-specific LINE1 hypomethylation also encompassed its diversity across progressive grades. Our study revealed that although both small intestine/colorectal and pancreatic NETs demonstrate LINE1 hypomethylation in G1 and G2 grades, the pattern of hypomethylation is more progressive during successive grades of pancreatic NET. Furthermore, the current study also demonstrated that LINE1 hypomethylation did not correlate with size or depth of GEP-NETs but was associated with lymph node metastasis. Based on these results, we hypothesize that this site-specific diversity of global hypomethylation in NETs may explain the site-specific differences in genomic instability and loss of heterozygosity (LOH) of chromosomes 11q, 16q and 18 observed in neuroendocrine tumours (33). Interestingly, an association between global hypomethylation and 18q LOH was demonstrated in colorectal carcinomas (34, 35), an association also characteristic of NETs (16). Similarly, previous studies have demonstrated that in prostatic adenocarcinomas, genome-wide DNA hypomethylation of LINE1 was associated with alterations of chromosome 8 (36). Furthermore, grade-dependent variation in LINE1 hypomethylation, as well as significantly greater hypomethylation observed in N1 tumours relative to N0 tumours strongly indicate that global hypomethylation may also contribute towards clonal changes and clonal evolution in
Relation between long interspersed element (LINE1) methylation in pancreatic G1 neuroendocrine tumour (NET) and G2 NET: Comparison of LINE1 methylation in G1 NET and G2 NET (p<0.05).
Relation between long interspersed element (LINE1) methylation in pN0 and pN1 tumours. Comparison of LINE1 methylation in the primary tumours with known pN0 and pN1 status showing a significantly lower hypomethylation rate in pN1 tumours (p<0.05).
GEP-NETs. Along similar lines, Choi et al. also observed that LINE1 hypomethylation was associated with lymph node metastasis, and was independent of RASSF1A methylation and inversely correlated with O-6-methylguanine-DNA methyltransferase (MGMT) gene methylation (16). Such a model is consistent with the role of epigenetic plasticity in tumour progression (37-40).
In summary, the current study, to our knowledge, for the first time, demonstrates a site- and grade-specific diversity of LINE1 methylation pattern in GEP-NETs.
Acknowledgements
We thank Forschungsförderung Ruhr-Universität Bochum Medizinischen Fakultät (FoRUM), and Protein Research Unit Ruhr within Europe (PURE), Ruhr University, Bochum for their kind financial support.
Footnotes
-
Conflict of Interest
The Authors declare that no conflict of interest exists.
- Received May 24, 2012.
- Revision received July 20, 2012.
- Accepted July 23, 2012.
- Copyright© 2012 International Institute of Anticancer Research (Dr. John G. Delinassios), All rights reserved
