Characterization of CDKN2A(p16) methylation and impact in colorectal cancer: systematic analysis using pyrosequencing

Abstract

Background: The aim of this study is to analyse CDKN2A methylation using pyrosequencing on a large cohort of colorectal cancers and corresponding non-neoplastic tissues. In a second step, the effect of methylation on clinical outcome is addressed.

Methods: Primary colorectal cancers and matched non-neoplastic tissues from 432 patients underwent CDKN2A methylation analysis by pyrosequencing (PyroMarkQ96). Methylation was then related to clinical outcome, microsatellite instability (MSI), and BRAF and KRAS mutation. Different amplification conditions (35 to 50 PCR cycles) using a range of 0-100% methylated DNA were tested.

Results: Background methylation was at most 10% with ≥35 PCR cycles. Correlation of observed and expected values was high, even at low methylation levels (0.02%, 0.6%, 2%). Accuracy of detection was optimal with 45 PCR cycles. Methylation in normal mucosa ranged from 0 to >90% in some cases. Based on the maximum value of 10% background, positivity was defined as a ≥20% difference in methylation between tumor and normal tissue, which occurred in 87 cases. CDKN2A methylation positivity was associated with MSI (p = 0.025), BRAF mutation (p < 0.0001), higher tumor grade (p < 0.0001), mucinous histology (p = 0.0209) but not with KRAS mutation. CDKN2A methylation had an independent adverse effect (p = 0.0058) on prognosis.

Conclusion: The non-negligible CDKN2A methylation of normal colorectal mucosa may confound the assessment of tumor-specific hypermethylation, suggesting that corresponding non-neoplastic tissue should be used as a control. CDKN2A methylation is robustly detected by pyrosequencing, even at low levels, suggesting that this unfavorable prognostic biomarker warrants investigation in prospective studies.

Figure 1

Representative pyrograms with PyroMarkQ96 showing percentage of methylation at each of five CpG sites evaluated.A) Highly methylated tumor sample with an average methylation percentage of (82% + 87% + 79% + 82% + 82%)/5 = 82.4%. B) Corresponding non-methylated non-tumoral tissue from the same patient as in A with the following methylation pattern: 2%, 2%, 0%, 0%, and 0%.

Figure 2

A): Serial dilutions of 100% and 0% methylated DNA were combined and pyrosequencing used to analyse the measured and theoretical percentage of methylated DNA as a function of the number of PCR cycles. A strong correlation between the measured and theoretical percentage of methylation was observed, particularly for PCR cycle numbers of 35 or more. B) Influence on methylation background as a function of PCR cycle numbers. C) In contrast to pyrosequencing, only a poor correlation between the measured and theoretical percentage of methylation was observed using methylation-specific PCR (MSP) which also showed a considerable dependency on PCR cycle number. D) Comparison of measured and theoretical percentage of methylation for 40 and 50 PCR cycles obtained using pyrosequencing and MSP.

Figure 3

A) Mean percent-methylation ± standard deviation of each CpG site within the region analysed. The average methylation across all sites was 8.3% in normal mucosa and 17.6% in colorectal cancer. The percent-methylation at each CpG site is significantly (p < 0.0001, all) larger in the tumor compared with normal tissue. B) Distribution of methylation in normal tissues and tumor. 7.8% of all cases should a greater percentage of methylation in normal tissues than in tumor. Most cases showed slightly more methylation in tumor than normal while 20.6% of cases were considered methylation positive (at least 20% more methylation in tumor).

Figure 4

Kaplan-Meier survival curves showing the unfavorable prognostic impact of CDKN2A methylation positivity in (A) all patients and in cases with (B) microsatellite stable (MSS/MSI-L), (C) microsatellite instability-high (MSI-H), (D) BRAF wild-type (WT) and (E) BRAF mutation.