The suggestion that RASSF1A
silencing by methylation is involved in the pathogenesis of PET has been supported by the presence of methylation in the CpG island A of the gene [1
]. However, all studies used the same qualitative method to assess methylation and RASSF1A
expression has never been analyzed in PET.
In this study we conducted an exhaustive analysis of RASSF1 methylation status in order to define its putative role as a possible tumor specific and transcription regulatory event occurring in PET.
We first analyzed RASSF1
methylation by the qualitative MSP assay because this was the method used in previous studies on PET, and found methylation in 80% of cases, a frequency similar to that reported in those studies [1
]. However, MSP is highly sensitive recognizing as little as 0.1% of methylated alleles [31
], thus classifying a sample as methylated on the basis of a minimal proportion of methylated target. We then performed qMSP that has been previously used for the evaluation of methylation status of RASSF1A
in tumors other than pancreatic [6
]. This assay analyzes a region of CpG island A that has never been investigated in PET. By this quantitative assay, all PET showed some degree of methylation but only 55% of cases had at least 20% of their alleles methylated.
In normal pancreas, RASSF1
methylation was found in 65% of cases by MSP, a value that fit within the range of published studies that used the same technique [2
]. By qMSP assay, 45% of normals had at least 20% of their alleles methylated.
Finally, we obtained a detailed mapping of RASSF1A
methylation by pyrosequencing, assessing the level of methylation of each of 51 CpGs within CpG island A, also encompassing those explored by MSP and qMSP approaches. Pyrosequencing provided a portrait of the complexity of the methylation pattern of tumor cells, where RASSF1A
methylation showed a high variability in terms of distribution and level within and among samples. Similar to pancreas, a very complex distribution of methylation of RASSF1A
was found in breast cancer [19
Our pyrosequencing data showed that most of normal and tumor samples had an average methylation levels of the CpG island A below 40%, with the exception of one normal (case 2) and seven tumors, having values above 40% (Table ). To classify samples as "methylated", we applied the same cut-off used for qMSP (an average methylation level higher than 20%). Methylated samples were 13/20 (65%) in PET and 12/20 (60%) in normal pancreas. In matched samples, the average methylation level of the CpG island A was higher in PET than in normal samples in 15/20 (75%) cases (P = 0.01). Among the 3 normal samples with higher degree of methylation than the neoplastic counterpart, case 2 showed the largest difference of average methylation level between PET and normal. We repeated the pyrosequencing analysis of this case with same results and did not find any particular features in the clinical profile of the patient to associate to the abnormal methylation data.
Although the overall CpG island A methylation level revealed by pyrosequencing was higher in PET than in normal tissue (Figure ), normal pancreas displayed considerable methylation levels. The common occurrence of methylation of RASSF1A
in normal pancreas suggests that this epigenetic event might represent a "field defect", consisting in widespread epigenetic changes arising early in the pancreas before tumor onset, a hypothesis previously suggested [32
When comparing the three techniques employed in this study to investigate the methylation status, MSP showed different results from those obtained by both qMSP and pyrosequencing. This is consistent with the qualitative (MSP) and quantitative (qMSP and pyrosequencing) nature of these approaches. Conversely, qMSP and pyrosequencing gave comparable results (correlation r = 0.78), thus supporting the choice of using the same cut-off for both quantitative results to classify a sample as methylated.
Whatever the method used to detect methylation, the majority of cases showed concordantly methylated or unmethylated tumor/normal pairs, and, when discordant, the methylation was higher in tumor in most cases, except case 2 (Table ). This raises the question of whether methylation affects the expression of RASSF1
gene. Indeed, none of the previous papers suggesting the inactivation of RASSF1A
in PET due to hypermethylation analyzed gene expression [2
]. In the present work, we evaluated the mRNA expression of RASSF1
variants and showed that: i) all PETs and their matched normal tissues expressed RASSF1A
; ii) the average expression of RASSF1A
in PET was 6.8 times lower than that in normal tissues (see Figure and Additional file 2
, Table S2); iii) the overall extent of RASSF1A
methylation in PET correlated inversely with its expression and the role of methylation of the first exon seems more important than that of the promoter region (see Figure ). Accordingly, a correlation between RASSF1A
expression and the average methylation of the 51 CpGs of island A was found in two of the three PET cell lines analyzed. In all untreated cell lines RASSF1A
was always expressed despite a strong methylation in CpG island A; in particular, BON cell line showed a higher level of RASSF1A
expression compared to QGP1 and CM. Moreover, the treatment with the demethylating agent 5'-aza-2'-deoxycytidine enhanced significantly RASSF1A
expression in QGP1 and CM, but not in BON. This difference in expression levels and response to demethylating treatment is consistent with a possibly different genetic background. Indeed, it has been recently shown that CM, QGP1, and BON harboured different gene mutations; in particular, they had mutations in FLT1
, and PIK3CA
, respectively [34
The expression of RASSF1A splicing isoforms D, E, F and G, and that of the major variants deriving from alternative promoter usage, RASSF1B and RASSF1C, has never been studied in PET. Here we report that RASSF1 isoforms D, E, F were rarely expressed in PET and normal pancreas; RASSF1G was never found; RASSF1B was always expressed in both PET and normal pancreas, with no significant difference; RASSF1C expression was averagely 11.4 times higher in PET than in normal tissue. Pyrosequencing analysis revealed that all the CpGs within CpG island C lacked methylation in both tumor and normal tissues. The same situation was found in PET cell lines, where CpG island C was never methylated, and RASSF1C was always expressed. Interestingly, treatment with 5'-aza-2'-deoxycytidine enhanced mRNA expression significantly in two of the three cell lines.
The finding of hyperexpression of RASSF1C
in PET is of great interest in the light of the recently reported role of RASSF1C
in inhibiting ß-catenin degradation [35
]. Thus, RASSF1C
overexpression may represent a pathogenetic event in PET contributing in sustaining Wnt signaling that has been recently shown to regulate proliferation of pancreatic ß-cells [36
]. Moreover, RASSF1C
has also been implicated in promoting cell migration and attenuating apoptosis in breast cancer cells [37