High methylation levels of PCDH10 predict poor prognosis in patients with pancreatic ductal adenocarcinoma

Samples from surgically resected primary PDAC were collected from a series of 23 patients recruited at the Department of Surgery of “Casa Sollievo della Sofferenza” Hospital, IRCCS San Giovanni Rotondo. Only patients with histologically proven primary PDAC were enrolled in the study. Exclusion criteria for patients were a previous diagnosis for PDAC and neoadjuvant treatment before surgery. Tumors were staged in accordance with the TNM classification [15]. Clinical features and tumor characteristics were reported in Table 1. Patients gave informed written consent and approval from the ethical committee of the “Casa Sollievo della Sofferenza” IRCCS, San Giovanni Rotondo was obtained. In DNA methylation analyses Capan-2 human pancreatic cancer cell line was used as a control fully methylated for PCDH10 [7]. For PCDH10 mRNA expression analysis we used PCDH10 fully methylated pancreatic (Capan-2, AsPC-1) and gastric (AGS) cancer cell lines, as well as PCDH10 unmethylated breast cancer cell line (MB-231) [16, 17].

Resected PDACs and adjacent non-neoplastic tissues from the same patients were taken separately, immediately frozen in liquid nitrogen and stored at − 80 °C until the nucleic acid extraction. These control tissues were verified as tumor-free by a pathologist.

Genomic DNA was isolated using the AllPrep DNA/RNA mini kit (Qiagen, Hilden, Germany) according to the manufacturer’s instructions. DNA concentration and purity were controlled by NanoDrop Spectrophotometer (Thermo Fisher, Waltham, MA, USA).

Bisulfite treatment was performed according to the manufacturer’s protocol (EpiTect Bisulfite Kit, Qiagen). The bisulfite-treated DNA was amplified with primers designed according to MethPrimer [18]. Primer sequences and PCR conditions are available in Table 2.

COBRA is a technique to semiquantitate the methylated and unmethylated DNA after sodium bisulfite treatment by using restriction enzyme cutting sites. PCR products containing CpG dinucleotides and at least 1 BstUI restriction site were digested with BstUI (New England BioLabs) that recognizes the sequence 5′-CGCG-3′, retained in the bisulfite-treated methylated DNA, but not in the unmethylated DNA. The DNA digests were separated by denaturing high-performance liquid chromatography (dHPLC) (Wave 1100, Transgenomic, Omaha, NE).

In case of methylated CpG dinucleotides, after enzymatic digestion, the 235 bp PCDHAC2 PCR product, encompassing the promoter region − 43 to + 192 bp from the transcription start site, provides two fragments of 210 and 25 bp, respectively; the 290 bp PCDHGC5 PCR product, encompassing the promoter region − 3287 to − 2997 bp upstream from the transcription start site, provides two fragments of 200 and 90 bp, respectively; the 196 bp PCDH10 PCR product, encompassing the promoter region − 1204 to − 1008 bp upstream from the transcription start site, has two cutting sites for BstUI and provides three fragments of 113, 52 and 31 base pairs, respectively. For PCDH10, the presence of two cutting sites for BstUI restriction enzyme hampered the interpretation of the analysis in case of partial methylation of the analyzed CpG islands. For this reason, we used BGS for this gene in all cases analyzed. Also for PCDHAC2 and PCDHGC5 DNA from a representative tumor and non-neoplastic sample were subjected to bisulfite genomic sequencing (BGS) to verify COBRA results independently.

Total RNA was extracted from Capan-2, AsPC-1, AGS, MB-231 cancer cell lines using Trizol reagent (Invitrogen Corp., Carlsbad, California, USA. Complementary DNA (cDNA) was synthesised as previously described [19] and amplified for PCDH10 gene with previously published primers [20]. PCDH10 cDNA RT-PCR amplified fragments were separated by dHPLC.

To select genes that may play a key role in PDAC, we analyzed PDAC data in The Cancer Genome Atlas (TCGA) provisional database (accessed January 17, 2014) to identify functionally related gene groups frequently mutated in this tumor. Protocadherins were among the most frequently mutated genes in PDAC samples analyzed by TCGA (Table 3). We then used DAVID bioinformatics resources (http://​david.​abcc.​ncifcrf.​gov) to identify enriched biological themes and functional-related gene groups among the top 43 genes with > 10 mutations in PDAC according to TCGA. This analysis indicated that “Cadherins” including PCDH10, PCDHGC5, PCDH15, PCDHAC2, CDH10 were among functional-related gene groups statistically enriched after Bonferroni (P = 0.014) and Benjamini (P = 0.004) corrections (Table 4). Notably, the enrichment of the term “Cadherins” was confirmed in a more recent analysis (April 18, 2018) in which we included the top 424 genes with > 10 mutations in PDAC, merging mutational data from different databases, including TCGA, International Cancer Genome Consortium (ICGC), Queensland Centre for Medical Genomics (QCMG), UTSouthwestern Medical Center (UTSW) (UP_SEQ_FEATURE domain: Cadherin 5; Bonferroni P = 1.8 10^− 8; Benjamini P = 8.1 10^− 10). Based on the above results, in this study we analyzed CpG methylation for three genes that appeared to be frequently mutated in PDAC, including PCDHGC5 and PCDHAC2 that had not been studied before for epigenetic modifications and PCDH10, whose promoter methylation had been previously suggested as a prognostic marker in other cancers [13].

An exploratory study of methylation analysis on PCDHAC2, PCDHGC5 and PCDH10 was carried out in 11 pancreatic adenocarcinomas.

For PCDHAC2, dHPLC analysis of COBRA showed in all cases only the presence of the full-length fragment in tumors, indicating that the fragment was not cut because the C in the cutting site was unmethylated and thus converted to T by bisulfite treatment abolishing the cutting site (Fig. 1a, left panel). The same pattern indicating the absence of CpG methylation was observed in the non-neoplastic pancreatic tissues analyzed. BGS showed the presence of a minor C peak, indicating modest methylation, in three of nine CpG dinucleotides sequenced both in tumor and non-neoplastic tissue (Fig. 1a, right panel). These results confirmed the monomorphic pattern of methylation in tumor and non-neoplastic tissues and the lack of relevant CpG methylation indicated by COBRA. As far as PCDHGC5, analysis by COBRA indicated that CpG dinucleotides were methylated in the amplicon analyzed both in tumor (Fig. 1b, left panel) and non-neoplastic tissues. BGS agreed with this finding (Fig. 1b, right panel). For PCDH10, in the exploratory BGS analysis, the pattern of methylation in tumors among cases was different, with six cases showing lack of methylation and five cases showing > 50% methylation of CpG dinucleotides, whereas non-neoplastic pancreatic tissues resulted unmethylated. Overall, in the exploratory analysis, both for PCDHAC2 and PCDHGC5 all cases had similar patterns of methylation (Table 5) providing no indicator that could be related to clinicopathological features. Conversely, for PCDH10 the 11 cases analyzed in the exploratory study had different patterns of methylation, providing an indicator that could be related to clinicopathological features. Therefore, we extended BGS analysis of PCDH10 to the whole series of 23 pancreatic adenocarcinomas available (Table 5). This extended analysis revealed that tumors derived from nine cases resulted not methylated and 14 methylated, with a percentage of methylation ranging from 8 to 91%, and with a mean ± SE of 55.0 ± 7.8 (Fig. 1c, left panel, Tables 5 and 6). In all non-neoplastic pancreatic tissues analyzed PCDH10 resulted unmethylated. Sequencing of the human pancreatic carcinoma cell line Capan-2 showed complete methylation of PCDH10 CpG dinucleotides, as expected for this control cell line (Fig. 1c, right panel).

Since RNA samples from tissues of the patients analyzed were not available, we analyzed cDNA from pancreatic (Capan-2, AsPC-1) and gastric (AGS) cancer cell lines fully methylated for PCDH10, as well from a breast cancer cell line (MB-231) unmethylated for PCDH10, to assess whether methylation status of PCDH10 CpG dinucleotides was associated with effects on the expression of the corresponding transcript. In line with methylation status, Capan-2, AsPC-1 and AGS cell lines fully methylated for PCDH10 did not express the corresponding mRNA, whereas the cell line MB-231 unmethylated for PCDH10 expressed the corresponding transcript (Fig. 2).

Fourteen out of 23 (60.9%) tumors showed PCDH10 methylation. In these cases, the percentages of methylation ranged from 8 to 91%, with a mean ± SE of 55.0 ± 7.8. The box-and-whisker diagram shows the PCDH10 methylation levels registered among 23 PDAC cases (Fig. 3).

By ROC curve analysis, cases were dichotomized according to PCDH10 methylation status: tumors with methylation levels above 52% (n = 7) were considered PCDH10^High, and those with methylation levels below the cut-off value were considered PCDH10^Low (n = 16). By chi-square test, PCDH10 methylation status was found inversely correlated with the clinical presentation of jaundice (P = 0.036) (Table 7).

A disease progression was observed in 85.7% (6/7) of patients with PCDH10^High and 56.3% (9/16) of those with PCDH10^Low tumors. Death rates were 85.7 and 75.0% for patients with high and low methylation of PCDH10, respectively. At Kaplan–Meier analysis, PCDH10^High was significantly associated with worse PFS rates (P = 0.008), but not with OS (Fig. 4).

Univariate analysis revealed that PCDH10^High was a prognostic factor influencing PFS (HR = 4.0: 95% CI, 1.3–12.3; P = 0.016), but not the OS (Table 7).