An epigenetic classifier for early stage lung cancer

Abstract

Background: Methylated genes detected in sputum are promise biomarkers for lung cancer. Yet the current PCR technologies for quantification of DNA methylation and diagnostic value of the sputum biomarkers are not sufficient to be used for lung cancer early detection. The emerging droplet digital PCR (ddPCR) is a straightforward means for precise, direct, and absolute quantification of nucleic acids. Here, we investigate whether ddPCR can sensitively and robustly quantify DNA methylation in sputum for more precise diagnosis of lung cancer.

Results: First, the analytic performance of methylation-specific ddPCR (ddMSP) and quantitative methylation-specific PCR (qMSP) is determined in methylated and unmethylated DNA samples. Second, 29 genes, previously proposed as potential sputum biomarkers for lung cancer, are analyzed by using ddMSP in a training set of 127 lung cancer patients and 159 controls. ddMSP has higher sensitivity, precision, and reproducibility for quantification of methylation compared with qMSP (all p < 0.05). A classifier comprising four sputum methylation biomarkers for lung cancer is developed by using ddMSP, producing 86.6% sensitivity and 90.6% specificity, independent of stage and histology of lung cancer (all p > 0.05). The classifier has higher accuracy compared with sputum cytology (88.8 vs. 70.6%, p < 0.01). The diagnostic performance is confirmed in a testing set of 89 cases and 107 controls.

Conclusions: ddMSP is a robust tool for reliable quantification of DNA methylation in sputum, and the epigenetic classifier could help diagnose lung cancer at the early stage.

Keywords: DNA methylation; Diagnosis; Lung cancer; Sputum; ddPCR.

Fig. 1

The dynamic ranges and sensitivities of ddMSP and qMSP for quantification of DNA methylation. a In methylated DNA serially diluted in sputum DNA of a healthy nonsmoker, ddMSP can detect levels of methylated 3OST2 as low as 0.04% (LOQ = 0.04%). A negative template control (NTC) sample was also tested. R² = 0.966 shows excellent linear correlation between measured concentration of methylated DNA and expected percentage of methylation. b qMSP can detect methylated 3OST2 at 0.10% (LOQ = 0.10%) in the same diluted samples with R² of 0.935. c In 100% methylated DNA serially diluted into water, ddMSP can detect the smallest amount of methylated DNA at 156 pg/μL (156 pg/μL) with R² of 0.959. d qMSP can detect the smallest amount of methylated DNA at 156 pg/μL (156 pg/μL) with R2 of 0.937

Fig. 2

ddMSP and qMSP analyses of DNA methylation in clinical sputum samples. a ddMSP analysis of sputum samples of 20 cancer-free controls (normal subjects, N) and 20 patients diagnosed with lung tumor (T) for DNA methylation of RASSF1A. Each well of the sputum samples contained at least 10,000 droplets, suggesting that the clinical sputum specimens could be successfully “read” by ddPCR. b ddMSP analysis of 29 genes in a training set of 127 lung cancer patients and 159 controls developed an epigenetic classifier consisting of four DNA methylation biomarkers. The epigenetic classifier produced 0.92 AUC with 86.6% sensitivity and 90.6% specificity for diagnosis of lung cancer. c The epigenetic classifier had higher accuracy (88.8 vs. 70.63%, p = 0.004) and sensitivity (86.6 vs. 44.8%, p < 0.001) compared with sputum cytology, whereas keeping a similar specificity (90.6 vs. 91.2%, p = 0.34)