Identification of a DNA methylation signature to predict disease-free survival in locally advanced rectal cancer

This is the authors' abstract. We don't add key points for this paper.

// Jochen Gaedcke 1 * , Andreas Leha 2 * , Rainer Claus 3, 4 , Dieter Weichenhan 3 , Klaus Jung 2 , Julia Kitz 5 , Marian Grade 1 , Hendrik A Wolff 6 , Peter Jo 1 , Jérôme Doyen 7 , Jean-Pierre Gérard 7 , Steven A. Johnsen 1 , Christoph Plass 3 , Tim Beißbarth 2 , Michael Ghadimi 1 1 Department of General and Visceral Surgery, University Medical Center Goettingen, Goettingen, Germany 2 Department of Medical Statistics, University Medical Center Goettingen, Goettingen, Germany 3 Division of Epigenomics and Cancer Risk Factors, German Cancer Research Center (DKFZ), Heidelberg, Germany 4 Dept. of Medicine, Div. Hematology/Oncology, University of Freiburg Medical Center, Freiburg, Germany (present address) 5 Department of Pathology, University Medical Center Goettingen, Goettingen, Germany 6 Department of Radiation Oncology, University Medical Center Goettingen, Goettingen, Germany 7 Radiotherapy Department, Cyclotron Biomédical, Centre Antoine-Lacassagne Nice, France * Both authors contributed equally Correspondence to: Dr. Jochen Gaedcke, e-mail: jochen.gaedcke@med.uni-goettingen.de Received: August 03, 2014      Accepted: August 11, 2014      Published: August 19, 2014 ABSTRACT In locally advanced rectal cancer a preoperative predictive biomarker is necessary to adjust treatment specifically for those patients expected to suffer relapse. We applied whole genome methylation CpG island array analyses to an initial set of patients (n=11) to identify differentially methylated regions (DMRs) that separate a good from a bad prognosis group. Using a quantitative high-resolution approach, candidate DMRs were first validated in a set of 61 patients (test set) and then confirmed DMRs were further validated in additional independent patient cohorts (n=71, n=42). We identified twenty highly discriminative DMRs and validated them in the test set using the MassARRAY technique. Ten DMRs could be confirmed which allowed separation into prognosis groups (p=0.0207, HR=4.09). The classifier was validated in two additional cohorts (n=71, p=0.0345, HR=3.57 and n=42, p=0.0113, HR=3.78). Interestingly, six of the ten DMRs represented regions close to the transcriptional start sites of genes which are also marked by the Polycomb Repressor Complex component EZH2. In conclusion we present a classifier comprising 10 DMRs which predicts patient prognosis with a high degree of accuracy. These data may now help to discriminate between patients that may respond better to standard treatments from those that may require alternative modalities.