Late diagnosis is one of the major reasons associated with high mortality in LC. Current detection methods based on low-dose computed tomography and minimally invasive cytology show low positive predictive value and low sensitivity, respectively. Therefore, there is an urgent necessity to implement molecular noninvasive biomarkers to improve LC detection and prolong survival. This study was designed to transfer and evaluate the utility of our previously identified epigenetic signature in lung tumoral tissue and bronchial aspirates into blood samples. We used ddPCR to quantify DNA methylation ratio of the 3 remaining genes in cfDNA, but only BCAT1 showed significant and robust results. It is worth stressing that the methylation status of BCAT1 alone determined in blood yielded a notable discrimination capacity (AUC = 0.85), with sensitivity and specificity profiles comparable to those obtained in our previous study using the 4-gene signature in bronchial fluids (AUC = 0.91) [5].

Several studies have been published in plasma samples, reporting differentially methylated genes when comparing LC patients versus control donors [7]. The seminal study reported by Kneip et al. in 2011 validated the diagnostic performance of SHOX2, which showed an AUC = 0.78 using real-time PCR [8]. Later, the analysis of a combination of SHOX2 and PTEGR4 methylation levels in blood demonstrated significant discriminatory performance in distinguishing patients with LC from subjects without malignancy (AUC = from 0.86) [9]. Other genes have been found to be differentially methylated in plasma samples when comparing LC patients and healthy controls, including RASSF1A and RARB2 [10] or an epigenetic signature as an adjunct to low-dose CT scan screening [11]. Our study takes advantage of the ultrasensitive technique ddPCR, instead of qPCR, with the advantage of evaluating one single biomarker with high AUC value.

Recently, Chen et al. reported an approach based on methylation microarrays and whole genome bisulfite sequencing (WGBS) directly in cfDNA, which identified an epigenetic signature, called PanSeer, for cancer detection [12]. Despite these promising epigenomic results, the implementation in the clinic might be a long and costly process. Therefore, the evaluation of reduced candidate genes, such as BCAT1, may currently be a more feasible and affordable strategy for noninvasive detection of LC.

This study presents some limitations, despite the 4 genes BCAT1, CDO1, ZNF177 and TRIM58 being promising candidates in our previous study, we were unable to amplify TRIM58 by ddPCR, and CDO1 and ZNF177 showed poor performance. In the case of TRIM58, we believe, that being located the CpGs of interest in a very high-density CpG island, involves that designing probes in this type of regions may be quite difficult and challenging. We expect the development of new probe design tools, specific for ddPCR, to overcome these difficulties. Furthermore, despite the excellent performance of BCAT1 in stages I–III, we are aware that the number of early stage samples in our cohort is low. We also included smoking status as a covariate in our logistic regression model. This was motivated by a previous meta-analysis study in bibliography reporting an association between cigarette smoking and DNA methylation in 1405 genes, including BCAT1 [13]. Therefore, a future study using a large cohort in a prospective screening would be helpful.

In conclusion, our study suggests BCAT1 as a potential noninvasive epigenetic biomarker for LC detection and might also be very helpful to monitor therapeutic efficacy or to define more precise screening programs. However, future clinical trials and validation studies in other laboratories with larger cohorts of patients should be carried out. Furthermore, combination studies to test potential synergistic effects among BCAT1 and other lung cancer biomarkers, such as SHOX2, PTEGR4, could also be considered.