ZNF695 methylation predicts a response of esophageal squamous cell carcinoma to definitive chemoradiotherapy

Abstract

Purpose: Definitive chemoradiotherapy (dCRT) is one of the standard treatments for esophageal squamous cell carcinoma. Patients with a response to dCRT have a better prognosis than those resistant to dCRT while survival benefits for patients with residual tumors are limited. Nevertheless, few molecular markers to predict the response to dCRT are currently available. Here, we aimed to establish a DNA methylation marker to predict the response to dCRT.

Methods: A total of 104 patients were divided into screening (n = 43) and validation (n = 61) sets. A genome-wide DNA methylation analysis was performed using an Infinium HumanMethylation450 BeadChip array. Methylation levels were measured by quantitative methylation-specific PCR and normalized by the fraction of cancer cells in a sample.

Results: The genome-wide methylation analysis of seven responders and eight non-responders identified 18 genomic regions specifically (un)methylated in the responders. Among these, methylation of the promoter CpG island of ZNF695 was significantly associated with the response to dCRT in the screening set (P = 0.004), and a cutoff value was determined. In the validation set, the association was successfully validated (P = 0.021), and a high specificity (90 %) for the prediction of responders was obtained using the prefixed cutoff value. In addition, a multivariate analysis showed that ZNF695 methylation was an independent predictive factor for the response to dCRT (OR 7.55, 95 % CI 2.12-26.9, P = 0.002).

Conclusion: ZNF695 methylation was significantly associated with the response to dCRT and is a promising predictive marker for the response to dCRT.

Fig. 1

Selection processes of genomic regions whose methylation statuses were associated with the response to dCRT. From the 482,421 CpG sites on the Infinium HumanMethylation450 BeadChip array, those on autosomes and unmethylated in non-cancerous mucosae were selected. As specifically methylated in responders, 543 CpG sites, resulting in 16 genomic regions, were isolated. As specifically unmethylated in responders, 41 CpG sites, resulting in two genomic regions, were isolated. An unmethylated CpG site was defined when its β value was <0.2. A methylated CpG site was defined when its β value was more than or equal to half of the fraction of cancer cells in a sample [β value ≥ the fraction of cancer cells in the sample (%)/200]

Fig. 2

Normalized methylation levels of the candidate genomic regions in the screening set. Methylation levels of six genomic regions were measured by qMSP in 23 responders and 20 non-responders in the screening set. A normalized methylation level was calculated using the fraction of cancer cells in a sample [the normalized methylation level = 100 × (the measured methylation level (%))/(the fraction of cancer cells in the sample (%))]. The horizontal dotted line shows a cutoff value of 8.0 adopted for ZNF695

Fig. 3

Normalized methylation levels of ZNF695 in the validation set. Methylation levels of ZNF695 were measured by qMSP in 41 responders and 20 non-responders in the validation set. A normalized methylation level was calculated as described in the legend of Fig. 2. The horizontal dotted line shows the cutoff value of 8.0 prefixed in the screening set