Epigenetic silencing of transcription factor SIP1 in pancreatic cancer cells is associated with elevated expression and blood serum levels of microRNAs miR-200a,b

Cell lines and tissue samples

Fifteen human pancreatic cancer cell lines including A32-1, A38-5, AsPC1, BxPC3, Capan1, Capan2, CFPAC1, MiaPaCa2, Panc-1, Panc2.5, Panc2.8, Panc3.014, Panc215, PL3 and PL8 and 7 pancreatic cancer associated fibroblasts (CAFs) including CAF15, CAF16, CAF18, CAF19, CAF20, CAF21 and CAF22 were studied and grown under recommended conditions. An immortalized cell line derived from normal human pancreatic ductal epithelium (HPDE) and human pancreatic Nestin-expressing cells (HPNE) were provided by Dr. Ming-Sound Tsao (University of Toronto) and Dr. Michel Ouellette (University of Nebraska Medical Center), respectively.

Normal and neoplastic tissues were obtained from pancreatic adenocarcinomas resected at the Johns Hopkins Hospital. Fresh-frozen pancreatic tissues from 9 patients who underwent a pancreatic resection for an intraductal papillary mucinous neoplasm (IPMN) without associated invasive adenocarcinoma, 7 patients for neuroendocrine tumors and 19 patients for invasive ductal adenocarcinoma were microdissected to obtain normal pancreatic tissue for DNA analysis. Their mean age ± SD were 64.0±13.0 years (18 female). DNA was analyzed from 36 pancreatic cancer xenografts (32 pancreatic, 3 distal bile duct, and 1 duodenal cancer) (mean age 67.3±8.8 years, 24 female) established from primary carcinomas (15). Fresh-frozen pancreatic cancer tissues (n=7, mean age 62.4±15.4 years, 4 female) and normal pancreas from patients who had undergone resection for a pancreatic neuroendocrine neoplasm, IPMN and mucinous cystadenoma (n=13, mean age 55.5±20.6 years, 5 female), were microdissected, placed in RNAlater immediately after dissection and stored at −80C°. Frozen pancreatic tissues were microdissected using the PALM microlaser system as previously described (9). We also microdissected formalin-fixed paraffin-embedded tissues for SIP1 and E-cadherin methylation analysis as previously described (35).

Eighty-eight serum samples from Johns Hopkins Hospital were analyzed including 45 preoperative samples from patients with pancreatic ductal adenocarcinoma (64.4±10.0 years, 19 females), 11 preoperative samples from patients with chronic pancreatitis undergoing surgical resection (64.0±12.3 years, 3 females) and 32 healthy controls (44.3±10.2 years, 18 females). All sera were collected using standard procedures and stored at −80°C until analysis. All specimens were collected and analyzed with the approval of the Johns Hopkins Committee for Clinical Investigation.

Bisulfite treatment, Methylation specific PCR (MSP) and Bisulfite-modified sequencing (BMS) were performed as previously described (23).

RNA isolation and real-time PCR

Total RNA was extracted using mirVana miRNA isolation kit (Ambion 1560) for tissues and mirVana PARIS kit (Ambion 1556) for serum (100–400µl), according to manufacturer’s instructions and RT-PCR performed as we previously described (36). Real-time PCR was performed in triplicate. MicroRNAs were amplified after specific reverse transcription using TaqMan microRNA assays and TaqMan® MicroRNA Reverse Transcription Kit (Applied Biosystems) according to manufacturer’s instructions (Applied Biosystems) and normalized against 18S rRNA for cell lines and miR-16 for serum and microdissected tissues(20, 22, 37, 38). Relative expression was determined using the delta-delta Ct method and a >35 Ct value indicated negative amplification. Primers are shown in Supplemental Table 1.

5-aza-2’-deoxycytidine (5-Aza-dC) and Trichloroacetic acid (TSA) treatment was performed as we previously described (17). For anti-miRNA transfection analysis, 2 days after being treated by 1µM 5-Aza-dC, cells were transfected for 1 day then retreated by 5-Aza-dC for another 24 hours.

Transfection of microRNA precursors and inhibitors

Cells were plated at 5 × 10⁴ cells per well in 12-well plates and transfected with pre-miR miRNA precursors or anti-miR™ miRNA inhibitors (has-miR-200a and has-miR-200b, Ambion) at 40nM for miR-200a, miR-200b and the combination (each 20nM) using siPORT™ NeoFX™ Transfection Agent (Ambion). Cy™3 dye-labeled Pre-miR™ Negative Control #1 and Cy™3 dye-labeled Anti-miR™ Negative Control #1 (Ambion) were used as controls for pre-miRNA and anti-miRNA transfection, respectively. Cells were collected for RT-PCR 48 hours post-transfection. For the multiple transfection experiment, cells were split and re-transfected with additional pre-microRNA every 3–4 days for 18 days.

Immunohistochemistry

The expression of e-cadherin protein was examined by immunohistochemical labeling of formalin-fixed, paraffin-embedded tissue microarray (TMA) sections of pancreatic cancers and normal pancreas from 328 patients who underwent pancreaticoduodenectomy at Johns Hopkins Hospital using previously described methods (32, 39). Tissue sections were incubated with an anti-E-cadherin mouse monoclonal antibody (clone HEDC-1, Zymed Laboratories, San Francisco, 1/10 dilution) for 60 minutes at room temperature. The age (mean±SD) of these patients was 66.5±10.5 years (46% female). Expression patterns were classified as intact, partial loss or total loss of expression with total loss of expression indicating less than 5% of cancer cells expressing and partial when less than 95% of cells expressing e-cadherin.

Statistics

Statistical analysis was performed using the SPSS Statistics 17.0 and Microsoft Excel statistics software. A two-tailed P value <0.05 indicated statistical significance.

MiR-200a and miR-200b are hypomethylated in pancreatic cancer

MiR-200a and miR-200b were identified as candidate hypomethylated genes after comparing CpG island methylation profiles of 6 pancreatic cancer lines to normal pancreas using the methylation restriction assay known as methylation CpG island amplification (MCA) and identifying methylation patterns using the Agilent 244K CpG island array (23). On the array a CpG island containing 183 CpG dinucleotides (CGI-183;chr1:1087907–1090447 bps) (2541bp; %GC=65.0%; observed/expected CpGs=69.0%) had probes with elevated log₂ Cy3/Cy5 ratios (normal/cancer) indicating hypomethylation in 5 of 6 pancreatic cancer cell lines relative to normal tissues (Figure 1A). This CpG island is located upstream of the putative TSS of the miR-200a/200b locus (Figure 1B) (24).

To validate the hypomethylation of miR-200a and miR-200b in pancreatic cancer, we examined its promoter methylation status in pancreatic cancer and non-neoplastic cell lines using MSP. Eleven of 15 (73.3%) pancreatic cancer cell lines were unmethylated whereas two microdissected normal pancreatic duct samples were predominantly methylated (Table 1 and Supplemental Figure 1). The non-neoplastic cell line, HPNE was completely methylated while HPDE immortalized by HPV16-E6E7 was unmethylated (Table 1 and Supplemental Figure 1A).

We next performed bisulfite sequencing to verify the hypomethylation of the miR-200 promoter. We found that 9/11 pancreatic cancer cell lines, 6/6 pancreatic cancer xenografts and the non-neoplastic cell line HPDE were completely unmethylated for the 36 CpG sites sequenced while the non-neoplastic cell line HPNE and two cancer lines MiaPaCa2 and Panc-1 were methylated in all sequenced CpG sites (Supplemental Figure 1B), consistent with MSP data. We next expanded our MSP assay to 36 xenografts of primary pancreatic cancers, 35 normal pancreatic tissues and found that 97.2% (35/36) of pancreatic cancer xenografts were unmethylated while all 35 normal pancreatic tissue samples were methylated. To further analyze normal pancreatic tissue methylation we performed bisulfite sequencing on 10 normal pancreas samples, 6 of which demonstrated predominant methylation at most CpG sites 2 samples with ~50% methylation, and two samples that were predominantly unmethylated (Supplemental Figure 1C).

MiR-200a and miR-200b are overexpressed in pancreatic cancer

To evaluate whether miR-200a and miR-200b are overexpressed in pancreatic cancers, we measured its expression in pancreatic cell lines and primary pancreatic tissues using qRT-PCR. As shown in Figure 2, compared with the non-neoplastic cell lines HPNE and HPDE, miR-200a and miR-200b were overexpressed in 11/14 pancreatic cancer cell lines. Ten of 11 overexpressing cell lines were completely unmethylated at 5’CpGs while the 12^th cell line, AsPC1, was partial methylated (Table 1). MiR-200a and miR-200b were also overexpressed in primary pancreatic cancers compared with normal pancreatic tissues (p<0.001) (Figure 3). Among pancreatic cancer cell lines A38-5, MiaPaCa2 and Panc-1 with low levels of miR-200a and miR-200b (Figure 2), treatment with 5-Aza-dC induced miR-200a and miR-200b by 3-fold or more (Figure 4). Similarly, 5-Aza-dC induced miR-200a and miR-200b expression in HPNE (Figure 4). In contrast, apart from A38-5, treatment of pancreatic cancer cell lines with the HDAC inhibitor, TSA did not alter miR-200a/200b expression (Supplemental Figure 2).

DNA methylation and expression of SIP1 and effect of miR-200a and miR-200b

We next focused on one of the targets of miR-200, SIP1. SIP1 is an important transcriptional repressor that down regulates multiple genes including E-cadherin and shows aberrant expression in several cancers. Since SIP1 showed hypermethylation in our MCA microarray data, we examined the methylation status of SIP1. We found that the SIP1 promoter was methylated in 73.3% (11/15) of pancreatic cancer cell lines and in 97.1% (34/35) of pancreatic cancer xenografts while 2 non-neoplastic cell lines and 94.3% (33/35) of normal pancreas samples were unmethylated (Table 1). 30 of 35 pancreatic cancer xenografts had complete SIP1 methylation. E-cadherin was also unmethylated in all but one of these xenografts. SIP1 generally displayed a methylation profile inverse to that of miR-200 (Table 1). By qRT-PCR (Figure 2), and by serial analysis of gene expression (SAGE) analysis (9), SIP1 (ZEB2) was expressed in HPNE, but was not expressed in most pancreatic cancer cell lines. Similarly, microdissected pancreatic cancer tissues had significantly lower SIP1 RNA by qRT-PCR than normal pancreas (Figure 3). In contrast, we found abundant SIP1 RNA in 7 of 7 pancreatic cancer associated fibroblasts, which also displayed inverse expression profile to miR-200a and miR-200b (Figure 2). Since Snail and Slug induce SIP1 expression and repress E-cadherin, we examined their expression in pancreatic cancers by SAGE and found SNAIL was minimally expressed and there was no evidence that SNAIL or SLUG expression repressed E-cadherin expression (data not shown).

We found that DNA methylation regulated the expression of SIP1 as treatment with 5-Aza-dC induced SIP1 expression in all 7 cell lines tested (Figure 4). Interestingly, in HPNE cells which have methylated miR-200, induction of miR-200a and miR-200b expression by 5-Aza-dC (Figure 4A/4B) was associated with a 5-fold reduction of SIP1 expression supporting the negative regulation of SIP1 by miR-200 (Figure 4). SIP1 recruits the transcriptional repressor CtBP to inhibit Smad-mediated transcription (40), and since SMAD4 is mutated in ~55% of pancreatic cancers, we determined if loss of SIP1 was more likely pancreatic cancers with wild-type SMAD4, but we did not find any correlation (data not shown).

To test if miR-200 downregulates SIP1 expression in pancreatic cancer cells, we transfected the cell line Panc-1, which is fully methylated at miR-200 and unmethylated at SIP1, with pre-miR-200a or/and pre-miR miR-200b precursors and examined its effects at Day 2 and Day 18 post infection. Compared to control transfected cells, SIP1 mRNA was significantly reduced by miR-200a (p<0.01, p<0.01), miR-200b (p<0.05, p<0.01), and combined miR-200a/200b treatment (p<0.01, p<0.01) both two days after transfection and after 18 days of repeated transfections, respectively (Supplemental Figure 3).

Although microRNAs are not thought to induce epigenetic silencing, since we observed miR-200 overexpression and DNA methylation of SIP1, we considered the possibility that epigenetic inactivation of the SIP1 promoter in pancreatic cancers could be caused by aberrant miR-200 expression. To test this possibility we maintained expression of miR-200 for two weeks by transiently transfecting Panc-1 cells ever y 3–4 days. However, induction of miR-200 expression did not induce any SIP1 promoter methylation by MSP analysis (data not shown).

We also examined if inhibition of miR-200 could increase SIP1 expression. First we treated AsPC-1 cells with antisense inhibitors to miR-200a, miR-200b or both. Since SIP1 is silenced in association with DNA methylation in AsPC-1 cells, treatment with miR-200a, and/or miR-200b did not induce expression of SIP1 in AsPC-1 cells (Supplemental Figure 3). To determine if miR-200a and miR-200b still has the ability to regulate SIP1 expression in cells in which the SIP1 promoter is methylated, we pretreated AsPC1 cells with 5-Aza-dC to induce SIP1 expression. Inhibitors of miR-200a, miR-200b, alone or in combination increased SIP1 expression by 31% (p=0.39), 82% (p<0.05), 128% (p<0.05) compared to negative control transfected cells, respectively (Supplemental Figure 4).

Effect of miR-200 on E-cadherin expression and EMT

Since miR-200 and SIP1 have been identified as regulators of EMT, we examined E-cadherin expression in pancreatic cancers and determined if it was upregulated by miR-200 in pancreatic cancer cells. E-cadherin was expressed in 11/14 pancreatic cancer cell lines at levels approximately at or above the control pancreatic duct line, HPDE, but was not expressed in 7 CAF lines (Figure 2). Of the 3 cell lines lacking E-cadherin expression, MiaPaca2 was completely methylated at the E-cadherin locus associated with gene silencing ((15) and Figure 5), while Panc-1 and A38 express SIP1 and lack miR-200 expression suggesting these pancreatic cancers have lost E-cadherin expression and have undergone EMT as a result of SIP1 expression. Similarly, by SAGE (9) E-cadherin was expressed at or above levels in pancreatic duct epithelial cells in 21 of 24 pancreatic cancers. Indeed, we did find pre-miR-200b and combined pre-miR-200a/200b treatment increased E-cadherin expression in Panc-1 cells (p<0.05, p<0.05) (p<0.05, p<0.01) both at day 2 and at day 18 (after multiple transfections), respectively (Supplemental Figure 3). E-cadherin expression was also reduced in AsPC1 cells both by anti-miR inhibitors of miR-200b and anti-miR-200a and miR-200b, and by 5-Aza-dC treatment (Supplemental Figure 4).

However, we didn’t observe any reversal of EMT morphology (i.e. reversal of fibroblastoid spindles or an increase cell-to-cell contacts) in Panc-1 cells even after sustained suppression of E-cadherin by repeated transfection of miR-200a and miR-200b precursors (Supplemental Figure 3c).

E-cadherin protein expression in primary pancreatic adenocarcinomas

Our results indicate that only a minority of pancreatic cancers cell lines lack E-cadherin expression. Furthermore, we also find that most xenografts of primary pancreatic adenocarcinomas have complete methylation of SIP1 suggesting that SIP1 is not expressed in most pancreatic cancers. But since other mechanisms of EMT could be active in pancreatic cancers, we examined the prevalence of low e-cadherin expression by immunohistochemistry in the resected pancreatic cancer tissues of 328 patients who had undergone pancreaticoduodenectomy. We found intact expression in 188, partial loss in 137, and total loss 7 of e-cadherin expression in 328 pancreatic ductal adenocarcinomas (Supplemental Figure 5). Loss of e-cadherin expression was observed in 2/8 well-differentiated (25%), 65/175 moderately-differentiated (37%), and 74/146 poorly-differentiated (51%) carcinomas (p=0.03, chi-square test).

To examine if DNA methylation was responsible for the loss of e-cadherin, we obtained additional pancreatic cancer tissue from 6 of the 7 primary pancreatic cancers with total loss of e-cadherin. These pancreatic cancers were microdissected and assayed for methylation of E-cadherin and SIP1 by MSP. E-cadherin was methylated in 5 (83.3%) of these 6 pancreatic cancers. In contrast, E-cadherin was partially methylated in only 1 of 6 pancreatic cancer xenografts that had retained e-cadherin expression in its primary pancreatic cancer. SIP1 methylation was detected in 11 of 12 of these pancreatic cancers. These results implicate E-cadherin methylation as the likely cause of absent e-cadherin expression in most of these pancreatic cancers. We did find that one of the 6 pancreatic cancers with absent e-cadherin expression had unmethylated SIP1 suggesting that in this case SIP1 expression likely caused E-cadherin silencing (Supplemental table 2).

Elevated serum miR-200a and miR-200b in patients with pancreatic cancer

We next determined if miR-200a and miR-200b could be detected in serum and if it was more abundant in patients with pancreatic cancer. We measured miR-200a and miR-200b concentrations in 45 patients with pancreatic cancer and 32 healthy controls. Both miR-200a and miR-200b were significantly elevated in sera of patients with pancreatic cancer compared with healthy controls (p<0.001, p<0.001, respectively, Mann-Whitney; Figure 5) (miR-16 was used as a reference). Receiver operating characteristic (ROC) curve indicated serum levels of miR-200a and miR-200b could differentiate patients with pancreatic cancer from healthy controls with ROC curve areas of 0.861 (95% CI=0.774–0.949) and 0.85 (95% CI=0.763–0.938), respectively (Figure 5C/5D). With the cutoff at 0.28 (relative expression value), miR-200a had an 84.4% sensitivity for pancreatic cancer and 87.5% specificity compared to healthy controls. At the cutoff of 0.5 (relative expression value), miR-200b had a sensitivity of 71.1% and specificity of 96.9%. The serum levels of miR-200a and miR-200b in serum samples were significantly correlated (R²= 0.745, p<0.0001, Spearman).

Pancreatitis can mimic pancreatic cancer and many markers of pancreatic cancer are also abnormal in patients with chronic pancreatitis. We measured miR-200a and miR-200b in the serum of 11 patients with chronic pancreatitis and found that serum levels were not significantly different from those observed in patients with pancreatic cancer, (p=0.322, p=0.933, respectively, Mann-Whitney Test) (Figure 5).