This study was approved by the ethics committee of Okayama University (receipt number: 2220). All patients enrolled in the study provided written informed consent for the use of tissue and blood sample. July 2014 through February 2015, we obtained 7-ml peripheral blood samples using EDTA-2K tubes from two healthy volunteers and 24 lung adenocarcinoma patients with EGFR mutations. These patients had been treated with EGFR-TKI and their EGFR T790M mutational status was examined through re-biopsy as a part of clinical care. Clinical EGFR mutation tests in tissue samples had been performed before entry into the study by an commercial clinical laboratory via mutant-enriched PCR assay using the peptide nucleic acid (PNA)-locked nucleic acids (LNA) PCR clamp method (LSI Medience Corp., Tokyo, Japan) (11,12). The approval of the Institutional Review Board and informed consent from each patient were obtained. After drawing blood, samples were immediately centrifuged at 2,330 × g for 10 min to collect plasma, and the plasma samples were centrifuged at 16,000 × g for 10 additional min. The supernatants were collected, frozen, and stored at −80°C. In addition, we used an immortalized normal human bronchial epithelial cell line (HBEC-5KT) harboring the wild-type EGFR gene and the NCI-H1975 cell line (H1975) harboring the EGFR mutations L858R and T790M as negative and positive controls, respectively (13,14). HBEC-5KT and H1975 were kindly provided by Dr Adi F. Gazdar (The University of Texas Southwestern Medical Center, Dallas, TX, USA).

After thawing the frozen plasma samples, DNA was extracted using QIAamp Circulating Nucleic Acid kit (Qiagen, Venlo, The Netherlands), and eluted in 20 µl of the kit's elution buffer. Control DNA from cell lines was extracted using DNeasy Blood and Tissue kit (Qiagen) and fragmented through sonication to obtain an average size of 200 bp using Covaris M220 Focused-ultrasonicator (Covaris Inc., Woburn, MA, USA) to closely mimic cfDNA in plasma (15,16). DNA size distribution was analyzed using Agilent 2200 TapeStation (Agilent Technologies, Santa Clara, CA, USA). Genomic DNA was quantified by Qubit 2.0 Fluorometer and Qubit dsDNA HS or BR assay kit (Thermo Fisher Scientific, Inc., Waltham, MA, USA).

The sequences of primers, probes, and blocking oligo used in this study are shown in Table I. S1 and AS1 primers and the M1 probe, which were designed in a previous study, were used as a validated reference assay (9). The primer melting temperatures (Tm) were calculated using Primer3Plus (http://primer3plus.com/). Probes, including LNA, were designed using IDT Biophysics software (http://biophysics.idtdna.com/) and synthesized by IDT-MBL KK (Nagoya, Japan). The blocking oligo using PNA was synthesized by Fasmac Co., Ltd. (Atsugi, Japan).

ddPCR was performed using QX200 Droplet Digital PCR system (Bio-Rad, Hercules, CA, USA). The final volume of the PCR mixture was 20 µl, containing 10 µl of ddPCR Supermix for Probes (No dUTP) (Bio-Rad), 1 µM of each primer, 0.25 µM of each probe, 200 µM of dNTP, and 5 µl of DNA (from approximately 2 ml of blood sample or control DNA), with or without 5 µM of blocking oligo. The following PCR conditions for ddPCR were used: 1) an initial denaturation step at 95°C for 10 min followed by 2) 45 cycles at 94°C for 30 sec, and 3) 45 cycles at 57°C for 1 min. Each ramp rate was 2°C per second. Annealing temperatures were optimized by gradient PCR (data not shown). PCR products were then subjected to analysis by the QX-200 droplet reader and QuantaSoft analysis software (Bio-Rad). Assays were considered positive if >3 droplets were over the threshold fluorescence of 8,000. The fluorescence threshold and cut off number for mutation detection were determined by assessing these data using positive and negative control DNA from cell lines.

The concentrations of target alleles were calculated using QuantaSoft software 9.2.1 (Bio-Rad) based on Poisson distribution.

The size distribution of cfDNA extracted from healthy donors and lung cancer patients was assessed (Fig. 1A). It showed similar distribution patterns between healthy donors and lung cancer patients, and two peaks were identified in both sets of patients. The first peak was distributed from 128 to 168 bp (median, 158 bp), and the 2nd peak was distributed from 526 to 797 bp (median: 681 bp). Next, we prepared control cfDNA mimic samples through sonication to evaluate the method of highly-sensitive detection with focus on cfDNA; the size of the samples was then confirmed. After sonication, the genomic DNA samples were fragmented to an average size of approximately 200 bp (Fig. 1B).

To examine the advantage of using the LNA probe, ddPCR was performed using different probes consisting of a regular oligonucleotide probe (M1) and an LNA probe (M2) and the same primers (S1 and AS1). These probes were designed to display similar Tms, and were based either on regular oligonucleotides or LNA. The match-mismatch Tm difference of the regular oligonucleotides and LNA probes were 5.6°C and 11.4°C, respectively. As shown in Fig. 2A, the fluorescence intensity of negative droplets was suppressed in the LNA probe group compared with the regular oligonucleotide probe group in HBEC-5KT cells as negative controls. In H1975 cells, the negative droplet suppression by the LNA probe contributed to better separation between both positive and negative droplet clusters. These results suggest that the LNA probe displayed an improvement in mismatch discrimination, which was thought to possibly result in better specificity.

Next, we assessed the amplicon size for better detection of cfDNA. Since cfDNA is highly fragmented, we hypothesized that ddPCR with a short-size amplicon improves the detection efficiency of cfDNA. We designed new primers (S2 and AS2), the amplicon size of which was 43 bp, whereas the amplicon size of the validated reference primers (S1 and AS1) was 93 bp. ddPCR was performed using either of the two primer sets and the same LNA probe (M2). The template DNA from either HBEC-5KT or H1975 was fragmented through sonication to approximately 200 bp as previously mentioned. The use of a short-size amplicon led to approximately a 1.6-fold increase in the number of detected positive droplets (Fig. 2B). On the other hand, there were no significant differences in the number of false positive droplets between the two primer sets. This approach with short-size amplicon may be useful to increase cfDNA detection sensitivity because cfDNA is highly fragmented.

Since plasma cfDNA contains abundant cfDNA released from normal cells, a highly-sensitive method which enables mutations to be selectively enriched in a large background of wild-type alleles is preferable. We prepared a blocking oligo using PNA to suppress PCR amplification of the wild-type alleles and 100 ng diluted DNA mixture with a mutant content of only 1% was analyzed with or without the PNA clamp. The median positive droplet fluorescence intensity was increased in the group with the PNA clamp compared to the one without the clamp, which resulted in better separation between positive and negative droplets clusters. The use of the PNA clamp led to approximately a 1.5-fold increase in the number of detected positive droplets (Fig. 2C). This approach appeared to enhance PCR efficiency for T790M alleles by blocking the amplification of wild-type alleles.

The detection limit and quantitative ability of this assay were determined using serially diluted DNA mixtures of H1975 and HBEC-5KT. In total, 50 ng of template DNA was used for this assay, and ddPCR was performed using LNA probe (M2), short amplicon primers (S2 and AS2), and a PNA clamp. Results indicated that the PNA-LNA-ddPCR clamp method can detect mutant alleles in the sample with a mutant content of 0.01%, and the concentration of detected-mutant alleles correlated with the applied control DNA mixture concentrations in a linear fashion (R²=0.997) (Fig. 3).

We examined T790M mutation by ddPCR using the PNA-LNA-ddPCR clamp method for clinical plasma samples. The patient characteristics are shown in Table II. All patients had EGFR-mutant NSCLC and had received treatment with an EGFR-TKI, and 9 patients were positive for T790M in tumor re-biopsy. Of the 59 clinical plasma samples, T790M alleles were detected via ddPCR in 10 samples. As 21 of 59 plasma samples were collected from 9 patients with T790M positive in re-biopsy sample, the sensitivity of ddPCR was 42.8%. Among 10 positive samples in the ddPCR assay, nine were concordant with the T790M mutational status of the re-biopsy samples, and one was disconcordant. ddPCR specificity was determined as 97.3%. Results are summarized in Table III. In patients diagnosed as T790M-positive in plasma, one patient was enrolled in this study before treatment with third-generation EGFR-TKI after acquiring gefitinib resistance. We compared cfDNA T790M mutant alleles in pre- versus post-treatment plasma (Fig. 4). It revealed that the number of T790M alleles from plasma cfDNA rapidly decreased after third-generation EGFR-TKI treatment. This decline in plasma cfDNA was consistent with the computed tomography (CT) images. Six weeks after the administration of third-generation EGFR-TKI treatment, contrast-enhanced CT examination revealed that the primary tumor in the left lower lobe of the lung and mediastinal lymph nodes had significantly shrunk.