Clinicopathological Significance of CDKN2A Promoter Hypermethylation Frequency with Pancreatic Cancer.

abstract CC BY

The prognosis of pancreatic cancer patients is very poor, with a 5-year survival of less than 6%. Previous studies demonstrated that the loss of function of CDKN2A is mainly caused by the hypermethylation of CDKN2A gene promoter; however, whether or not it is associated with the incidence of pancreatic cancer still remains unclear. In this study, we systematically reviewed the association between CDKN2A promoter methylation and pancreatic cancer using meta-analysis methods. The pooled data were analyzed by Review Manager 5.2. Fourteen studies eligible studies, including 418 pancreatic cancer, 155 pancreatic intraepithelial neoplasia (PanINs) and 45 chronic pancreatitis (CP) patients were analyzed. We observed that the frequency of CDKN2A methylation was significantly higher in pancreatic cancer patients than in normal healthy controls, the pooled OR = 17.19, 95% CI = 8.72–33.86, P < 0.00001. The frequency of CDKN2A methylation was also significantly higher in PanINs patients than that in normal individual controls, OR = 12.35, 95% CI = 1.70–89.89, P = 0.01. In addition, CDKN2A methylation was associated with worse survival in pancreatic cancer, HR = 4.46, 95% CI = 1.37–14.53, P = 0.01. The results strongly suggest that CDKN2A methylation is correlated with an increased risk of pancreatic cancer. CDKN2A methylation plays a critical role in pancreatic carcinogenesis and may serve as a prognostic marker.

identification of relevant studies CC BY

Seventy publications were identified by the search method as described above. Fifty-six of those were excluded due to laboratory studies, non-original articles (review), lacking of matched controls or studies irrelevant to the current analysis. Eventually, there were 14 studies included in final meta-analysis (Fig. 1).

study characteristics CC BY

Fourteen studies published from 2002 to 2012 were eligible for meta-analysis1213141516171819202122232425. A total of 418 pancreatic cancer, 155 PanINs and 45 chronic pancreatitis (CP) patients from China, Singapore, Japan, Germany, England, and United States (USA) were enrolled. Their basic characteristics were summarized in Table 1.

the inactivation of CC BY

We observed that frequency of CDKN2A methylation was higher in chronic pancreatitis than that in normal individual controls, but it did not reach significant difference. The pooled OR from 3 studies including 45 patients with chronic pancreatitis and 29 healthy individuals is shown in Fig. 2a (OR = 32.08, 95% CI = 0.67–1525.38, P = 0.08). These findings indicate that although an increased risk was identified in CP patients for the development of pancreatic cancer, CDKN2A gene methylation is not the only determinant for its malignancy. Another study showed that CDKN2A methylation was detected in 10% of CP. We excluded the article because no healthy individual controls were available26.

the inactivation of CC BY

We observed that the frequency of CDKN2A methylation was significantly higher in PanINs patients than that in normal individual controls. The pooled OR from 2 studies including 155 patients with PanINs and 26 healthy individuals is shown in Fig. 2b (OR = 12.35, 95% CI = 1.70–89.89, P = 0.01). PanIN is considered the precursor lesion of invasive pancreatic cancer27. Our findings indicate that CDKN2A gene methylation could be one of the determinants for its malignancy.

the inactivation of CC BY

We observed that the frequency of CDKN2A methylation was significantly higher in pancreatic cancer patients than in normal healthy controls. The pooled OR from 13 studies including 358 pancreatic cancer patients and 201 normal individuals is shown in Fig. 3 (OR = 17.19, 95% CI = 8.72–33.86, P < 0.00001), indicating that CDKN2A inactivation through methylation plays an important role in the pathogenesis of pancreatic cancer. Previously, diagnostic accuracy of some biomarkers such as K-ras analysis for pancreatic carcinoma has been shown to be diverse in different samples;28 therefore, we stratified the analysis of CDKN2A methylation by specimen types (blood, pancreatic tissue and pancreatic juice) for pancreatic carcinoma. The pooled OR for blood analysis from 3 studies including 85 pancreatic cancer patients and 60 healthy controls is shown in Fig. 4a (OR = 6.72, 95% CI = 2.26–20.02, P = 0.0006). We further calculated the sensitivity, specificity of CDKN2A methylation and other parameters in blood samples from pancreatic cancer patients as described in published literature29. Based on Table 2 which is derived from Fig. 4a the results are as follows: sensitivity is 41.0% (a/a + c = 0.41/0.41 + 0.59), specificity is 73.0% (d/d + b = 0.73/0.27 + 0.73), positive predictive value (PPV) is 57.6% (a/a + b = 0.41/0.41 + 0.27), and negative predictive value (NPV) is 57.6% (0.73/0.59 + 0.73). As shown in Fig. 4b,c, the pooled OR for pancreatic tissues from 8 studies having 192 pancreatic cancer patients and 98 healthy controls was 15.98 (95% CI = 5.67–45.04, Z = 5.24, P < 0.00001) and for pancreatic juice from 4 studies including 135 pancreatic cancer patients and 51 normal individual controls was 24.96 (95% CI = 6.34–98.28, Z = 4.60, P < 0.00001). The overall methylation frequency of CDKN2A in blood, pancreatic tissue and juice for pancreatic carcinoma were more than that in normal controls, suggesting a potential role of CDKN2A methylation analysis in diagnosing pancreatic cancer. PDA remains one of the most deadly malignancies worldwide with extremely poor overall survival4. While PENs, an indolent neuroendocrine tumor which may secret neuropeptides causing clinical manifestations, is rare, the estimated incidence of PENs in USA has increased by almost 10 times over the past decades711. The prognosis of PENs cannot be reliably predicted from histopathological assessment because of neuropeptides secretion19. This is the justification for us to make meta-analysis of CDKN2A methylation in PDA and PENs separately. As shown in Fig. 5a, PENs patients display significantly enhanced CDKN2A methylation frequency (OR = 34.96, 95% CI: 6.27–194.87, Z = 4.05, P < 0.0001) compared to controls. As shown in Fig. 5b, the same observation was obtained for PDA patients when compared to corresponding controls (OR = 14.33, 95% CI = 6.83–30.07, Z = 7.04, P < 0.00001). These data indicated that loss of CDKN2A gene expression through epigenetic modification correlated with both types of aforementioned pancreatic cancers.

prognostic values of CC BY

Only two included studies1822 estimated the relationship between overall survival (OS) in PENs and CDKN2A methylation, the pooled results (Fig. 6a) showed the presence of prognostic impact of CDKN2A gene methylation on PENs patients (OR = 4.52, 95% CI = 1.25–16.35, Z = 2.30, P = 0.02). The hazard ratio is shown numberically in the fifth column, the confidence interval of the summary of hazard ratio does not include 1.0 (it is 1.25–16.35) suggesting that the association is statistically significant. Ohtsubo et al.24 reported that the survival period was significantly shorter in patients with pancreatic carcinoma with CDKN2A hypermethylation than those with a normal CDKN2A gene expression. Combined survival data from all three studies (Fig. 6b) showed OS tended to be shorter in pancreatic cancer patients with epigenetic abnormalities of CDKN2A than in PCs with normal expression of CDKN2A gene (OR = 4.46, 95% CI = 1.37–14.53, Z = 2.48, P = 0.01). In detail, the hazard ratio is shown numberically in the fifth column, the confidence interval of the summary of hazard ratio does not include 1.0 (it is 1.37–14.53) suggesting that the association is statistically significant. Another study30 was excluded from the analysis because only the narrative description can be found, and we were unable to calculate the pooled HR for OS.

association between smoking and CC BY

Cigarette smoking has been considered to increase the risk of pancreatic cancer;31 therefore, we evaluated the relationship between methylation of CDKN2A and smoking status, in another word, the changes in frequency of CDKN2A alterations by smoking status. The pooled OR from two studies including 17 pancreatic cancer patients and 45 controls was 1.04 (95% CI = 0.32–3.43, Z = 0.07, P = 0.95), indicating no correlation between smoking and hypermethylation status of CDKN2A gene (Fig. 7).

sensitivity analyses and publication bias CC BY

A sensitivity analysis, in which one study was removed at a time, was conducted to assess the result stability. The pooled ORs were not significantly changed, indicating the stability of our analyses. The funnel plots were largely symmetric suggesting there were no publication biases in the meta-analysis of CDKN2A gene methylation/expression and clinicopathological features (Fig. 8).

discussion CC BY

Located in the retroperitoneum of individuals who show non-specific symptoms, pancreatic carcinoma is unlikely to be detected until it has reached an advanced stage in most of patients32. It has remained one of the most devastating and difficult tumors to diagnose and treat. Due to the absence of disease-specific manifestations, there is an urgent need for reliable biomarkers and new therapeutic target(s) in pancreatic carcinoma. Although the tumor suppressor genes K-ras, p53, CDKN2A and SMAD4 have found to be the central molecular genetic pathways in pancreatic cancer33343536, the gained survival advantage targeting aforementioned pathways remain limited. Other molecular events such as epigenetic changes have recently been identified to contribute to the initiation and progression of pancreatic cancer37. In the current study, we concluded that (1) CDKN2A inactivation through methylation plays an important role in the pathogenesis of pancreatic cancer (both in PDA and PENs), and could be one of the determinants for its malignancy as supported by higher CDKN2A methylation frequency in premalignant lesion, PanINs than in normal controls, (2) all type of samples such as blood, pancreatic tissue and juice have a potential role for CDKN2A methylation analysis in diagnosing pancreatic cancer, (3) overall survival tend to be shorter in pancreatic patients with epigenetic abnormalities of CDKN2A than in PCs with normal expression of CDKN2A gene, and (4) no correlation exists between smoking and hypermethylation status of CDKN2A gene. As the differential diagnosis between pre-malignant/malignant diseases and normal/benign lesions, the detection of specific tumor markers in blood and pancreatic tissues is a convenient and attractive diagnostic tool. In addition to qualitative analysis, Li et al. further quantitatively evaluated methylation levels using the SIRPH (SNuPE combined with ion pair reverse phase HPLC)3839 protocol and compared the DNA methylation levels in peripheral blood and cancer tissue for a panel of genes in pancreatic cancer. They found three different groups of methylation patterns. The first group of genes presented higher methylation levels in cancer tissues than in blood DNA (CDKN2A, APC, and DAPK1). In the second group, methylation levels were approximately equal (BCL2, CD44 and TNFRSF10), while in the third group, the DNA methylation levels of ACIN1 were lower in pancreatic tissues than in blood. Overall, methylation alterations in blood could provide a promising approach for early detection of pancreatic cancer12. The sensitivity, specificity and other parameters based on Table 2 demonstrated that sensitivity is 41.0% (a/a + c = 0.41/0.41 + 0.59), which indicated that the probability of being test positive when disease present; specificity is 73.0% (d/d + b = 0.73/0.27 + 0.73), which indicated the probability of being test negative when disease absent; positive predictive value (PPV) is 57.6% (a/a + b = 0.41/0.41 + 0.27), which indicated the probability of patient having disease when test is positive; and negative predictive value (NPV) is 57.6% (0.73/0.59 + 0.73), which indicated the probability of patient not having disease when test is negative. More interestingly, the methylation rates detected in pancreatic secretions endoscopically retrieved from the pancreatic duct proved to be somewhat higher than those detected in pancreatic tissues21. In parallel, some studies have investigated the alterations of molecular biomarkers other than methylation patterns in pancreatic secretions from pancreatic cancer patients. For instance, K-ras mutations in pancreatic juice of pancreatic cancer patients have been considered as a potential diagnosis tool for pancreatic cancer with acceptable specificity and sensitivity28. Chronic pancreatitis is an important predisposing condition resulting in pancreatic malignancy40. The analysis obtained by us displayed that CDKN2A hypermethylaion in chronic pancreatitis is higher than those of normal pancreatogram and lower than those of pancreatic carcinoma21, suggesting a specific role of CDKN4A in the development of malignant pancreactic diseases, although the difference of CDKN2A methylation frequency between chronic pancreatitis patients and controls in the present study did not reach statistical significance. In another word, CDKN2A changes, especially promoter hypermethylation might imply high-risk precursors in chronic pancreatitis that might develop to cancer26. However, larger studies are needed to be carried out to explore the true situation. A study published by Moore et al.41 has addressed that distinct molecular pathways may be involved in exocrine and endocrine tumorigenesis of the pancreas. In this context, exclusively exocrine pancreatic adenocarcinomas of ductal origin (PDA) and endocrine origin (PENs) have been included, de novo CDKN2A promoter hypermethylation was detected and shown to contribute to the tumorigenesis in both types. In support of our conclusion that overall survival tends to be shorter in pancreatic patients with epigenetic abnormalities of CDKN2A than in PCs with normal expression of CDKN2A gene, Gerdes et al.30 found reduced survival in patients with CDK2A alterations indicating CDKN2A is a prognostic marker in resected ductal pancreatic cancer. This article was excluded from the study because of no matched control. It needs to be emphasized that the epigenetic alterations other than CDKN2A promoter hypermethylation also contribute to the development of pancreatic cancer. The other well known epigenetic mechanisms are histone modifications [histone deacetylation by histone deacetylases (HDACs) and histone methylation by histone methyltransferases (HMT)] and microRNAs (miRNAs). TGFBR242 and CDH143 are examples which can be regulated by HDACs while EZH244 and SUV39H13745 are regulated by HMT in pancreatic cancers. An increasing number of miRNAs have been shown to be associated with pancreatic tumors4647. The most well-known one is miR-21, which is upregulated in pancreatic cancer and targets many apoptosis related genes including PTEN and PDCD4, resulting in inhibited apoptosis and consequently, increased tumorigenicity. No single technique of DNA methylation detection is appropriate for every application. Some limitations exist in this current analysis. First, DNA methylation at specific loci are dependent on modification of DNA by sodium bisulfate. Although the studies used quantitative gene-specific methylation analysis to link DNA methylation to functional outcomes, bisulfate sequencing is the ideal standard for mapping allele-specific methylation across CpG locations. Without allele-specific measurement, the difference between the status of mosaic methylation of individual alleles or complete methylation of a subpopulation of alleles cannot be distinguished48. Second, the current protocol is unable to differentiate 5-methylcytosin (5mc) from 5-hydroxy-cytosine (5hmc)4950. Third, similar to gene expression microarrays improving the study of transcriptional regulation, locus-specific DNA methylation on a genome-wide scale will revolutionize and facilitate the DNA methylation analysis50. It would be possible to map-out genome-wide DNA methylation patterns from distinct subtype of pancreatic cancer patients. This analysis would result in new insight into the pathogenesis of cancer and open up new avenues of drug discovery and targeted therapies for all kinds of cancer patients. Taken all together, due to the poor prognosis of pancreatic cancer, understanding the molecular events such as epigenetic changes in pancreatic cancer that drive this aggressive disease is the core for development of early diagnostic tools and more effective therapeutic strategies.

conclusion CC BY

The results of our study strongly suggest that CDKN2A methylation is correlated with an increased risk of pancreatic cancer. CDKN2A methylation plays a critical role in pancreatic carcinogenesis and may serve as a prognostic marker.

search strategy CC BY

Medline, Pubmed, Web of Science, Scopus and Embase were searched in August 2014 using the search terms: ‘p16’, ‘p16INK4a’, “CDKN2A”, ‘methylation’, ‘pancreatic cancer’, “pancreatic carcinoma” and ‘clinical studies’. Investigations identified through the search approach as described above were screened by titles first, then by abstracts of the publications. After exclusion of non-relevant publications and identifications of duplicates from the different databases, the remaining papers were evaluated in the full text version for in- and exclusion criteria and for relevant articles in the reference lists. All clinical studies except case reports were chosen, for instance, randomized controlled trials (RCTs), cohort studies, case-controls studies and case series. The language of publication was restricted to English. All searched data were retrieved. Authors’ bibliographies and references of selected studies were also searched for other relevant studies. The most complete study was chosen to avoid duplication if same patient populations were reported in several publications.

selection criteria CC BY

We collected all eligible articles about the relationship between CDKN2A methylation and/or expression and clinicopathological features and clinical outcomes in pancreatic cancer patients for this meta-analysis. Studies meeting the following inclusion criteria were included: (1) CDKN2A methylation and/or expression which were evaluated in the circulation, pancreatic juice, and/or pancreatic tissues, (2) researches which revealed the relationship between CDKN2A methylation and/or expression and pancreatic cancer clinicopathological parameters and prognosis, (3) CDKN2A methylation and/or expression which were examined by methylation specific polymerase chain reaction (MSP), (4) articles which were published as a full papers in English, (5) articles which provided sufficient information to estimate hazard ratio (HR) about overall survival (OS) and 95% confidence interval (CI) and probabilities for overall survival (OS) where applicable. The exclusion criteria included the following: (1) letters, reviews, case reports, conference abstracts, editorials, expert opinion, and non-English language papers; (2) articles having no information on OS or those that could not calculate the HR about OS from the given information; and (3) all publications regarding in vitro/ex vivo studies, cell lines and human xenografts were also excluded.

data extraction CC BY

Two investigators independently extracted data from eligible studies. Disagreements were resolved by discussion till consensus were achieved. Two investigators reviewed all of the articles that fit inclusion and exclusion criteria. The following information was recorded for each study: the name of first author, year of publication, sample source, number of cases, clinicopathological parameters, stage, CDKN2A methylation and/or expression, and patient survival. Data for study characteristics and clinical response were summarized and the data were turned into table format. Heterogeneity of investigation was evaluated to determine whether or not the data of the various studies could be analyzed in a meta-analysis.

statistical analysis CC BY

Analysis was conducted using the Stata 12.0 (Stata Corporation,TX, USA) and Review Manager 5.2 (Cochrane Collaboration, Oxford, UK). Comparisons of dichotomous measures were done by pooled estimates of odds ratios (ORs) as well as their 95% CIs. P value of <0.05 was considered to be statistically significant. Heterogeneity was examined by a chi-square test with significance being set at P < 0.10; the total variation among studies was estimated by I square. We used I square statistic to assess heterogeneity. The I square value is an estimate of variance due to between-study heterogeneity rather than chance (the Cochran Q statistics). Substantial heterogeneity exists when I square exceeding 50%. If there was heterogeneity among studies, we used a random effect model to pool the ORs; otherwise, a fixed effect model was selected. The database search generated 70 articles from Medline, Pubmed, the Web of Science, Scopus and Embase. After initial screening of all titles, abstracts and eligibility, 14 full-text studies were retrieved for a more detailed assessment. The search of the article references did not produce additional publications. Eventually, 14 publications met the inclusion criteria for qualitative study and meta-analysis. The article search and study selection are depicted in Fig. 1.

additional information CC BY

How to cite this article: Tang, B. et al. Clinicopathological Significance of CDKN2A Promoter Hypermethylation Frequency with Pancreatic Cancer. Sci. Rep. 5, 13563; doi: 10.1038/srep13563 (2015).