Serum samples from 60 patients with histologically confirmed NSCLC and 68 healthy controls, were collected at the Shanghai Sixth People's Hospital East Campus between October 2014 and June 2016. The Ethics Committee of Shanghai Sixth People's Hospital East Affiliated to Shanghai University of Medicine & Health Sciences (Shanghai, China) approved this study. The NSCLC patient group consisted of 34 patients with primary lung cancer and 26 patients with recurrent lung cancer. The staging of all 60 NSCLC patients was performed according to the tumour node metastasis (TNM) classification of malignant tumours, and peripheral blood was collected from these patients for cfDNA extraction. Four patients underwent multiple blood collections over time to determine whether the levels of cfDNA, CEA and CYFRA21-1 correlate with disease progression. In NSCLC patients, progressive disease is defined as an increase in target lesion size of at least 20% or the appearance of new lesions (16). All participants provided written informed consent. Peripheral blood was collected from patients in separate gel pro-coagulation vacuum tubes and EDTA vacutainer tubes (BD Biosciences, Franklin Lakes, NJ, USA), and serum and plasma were then separated by centrifugation at 1,200 × g for 10 min at room temperature, transferred to new tubes and centrifuged at 16,000 × g for 10 min at 4°C to remove cell debris. Plasma and serum were stored at −80°C prior to DNA extraction. To determine the optimal sample preparation and storage conditions for cfDNA quantification, venous blood samples from 4 healthy volunteers (2 females and 2 males; 26–48 years) were simultaneously collected for the preparation of serum samples. Serum was separated and then stored at room temperature (25°C) for 2, 6, or 10 h; at 4°C for 24 h; at −25°C for 1 month; or at −80°C for 1, 3, or 6 months.

Circulating DNA was separated from 1 ml serum and plasma samples with the PME free-circulating DNA extraction kit (Analytik Jena AG, Jena, Germany), according to the manufacturer's instructions for up to 1-ml extractions with the lysis solution SE/binding solution SBS mechanism. The process consisted of several steps as follows: i) capturing cfDNA in the polymer by mixing serum/plasma with 30 µl of VCR-1 and 150 µl of VCR-2, followed by centrifugation at 18,500 × g for 3 min at room temperature; ii) washing the polymer/DNA complex with 1 ml of ddH₂O; iii) adding 400 µl of lysis solution SE to dissolve the pellet; iv) adding 50 µl of PK and incubating the samples for 15 min at 70°C; v) adding 400 µl of binding solution SBS and transferring the samples to spin filters; vi) washing the samples by adding 500 µl of washing solution GS and 650 µl of washing solution BS, followed by centrifugation at 13,500 × g for 1 min at room temperature; and vii) eluting cfDNA in 50 µl of water and storing the cfDNA at −80°C until further analysis. The presence of cfDNA and its fragment size distribution were evaluated by using the Agilent High Sensitivity DNA kit (Agilent Technologies, Inc., Santa Clara, CA, USA) on the 2100 Bioanalyzer.

Genomic LINE1 sequences are distributed over all chromosomes; a short fragment (97 bp) of LINE1 was amplified and quantified by absolute qPCR quantification, using the standard curve method to indicate the total amount of serum and plasma cfDNA (17). LINE1 primer sequences were as follows: forward, 5′-TGGCACATATACACCATGGAA-3′; and reverse, 5′-TGAGAATGATGGTTTCCAATTTC-3′. The housekeeping gene β-actin was amplified with forward primer 5′-CCACACTGTGCCCATCTACG-3′ and reverse primer 5′-AGGATCTTCATGAGGTAGTCAGTCAG-3′, producing products of 99 bp. The reaction mixture for qPCR included 10 µl of SYBR^® Premix Ex Taq II (Tli RNaseH Plus; Takara Biotechnology Co., Ltd., Dalian, China), 2 µl of cfDNA template, 0.4 mM primer, 0.4 µl of ROX II and 6 µl of double-distilled water in a final reaction volume of 20 µl. The qPCR was performed with an Applied Biosystems 7500 real-time PCR system (ABI; Thermo Fisher Scientific, Inc., Waltham, MA, USA), and was conducted by using the following protocol: 95°C for 8 min, followed by 40 cycles of 95°C for 5 sec and 60°C for 34 sec. A standard curve generated by serial dilution (from 1.0–10,000 ng/ml) of the LINE1-pCEP plasmid was used to determine the absolute equivalent amount of cfDNA in each sample. Melting curve analysis was conducted to confirm the amplification of a single peak. Each assay was performed in triplicate.

The levels of serum CEA and CYFRA21-1 were detected with a CEA and Cyfra21-1 test kit (Roche Diagnostics, Shanghai, China) by Cobas e601 Analyzer. The normal range was <4.7 ng/ml for CEA and <3.3 ng/ml for CYFRA21-1. A receiver operating characteristic (ROC) curve was created to quantify the diagnostic potential of serum cfDNA, CEA and CYFRA21-1. Youden's index (YI = sensitivity + specificity−1) was used to select cut-off values. The area under the ROC curve (AUC) was used to measure the discriminatory power of the test.

SPSS Statistics 19.0 (IBM Corp., Armonk, NY, USA) and MedCalc statistical software were used to conduct the statistical analyses. Figures were generated by using GraphPad Prism 6.0 software (GraphPad Software, Inc., La Jolla, CA, USA). The Mann-Whitney U test and Kruskal-Wallis H test were applied to compare serum cfDNA concentrations between groups. Dunn's post hoc tests were carried out on each pair of groups following Kruskal-Wallis test. The diagnostic values of cfDNA, CEA and CYFRA21-1 for NSCLC were assessed by using ROC curves. A univariate Cox regression model was used to analyse survival data for progression-free survival (PFS). The Kaplan-Meier method was used to estimate the PFS, which was evaluated by the log-rank test. Results were considered significant at two-tailed P<0.05.

To compare cfDNA concentrations between plasma and serum, we extracted cfDNA from 1 ml each of serum and plasma from the same donor, and quantified the amount of cfDNA by LINE1 qPCR. LINE1 is a transposable element belonging to the group of LINEs found in all mammalian DNA, comprising ~17% of the human genome. cfDNA concentrations were ~1-to 8-fold higher in serum samples than those in plasma samples (Fig. 1A). By fragment analysis, we observed an enrichment of plasma cfDNA fragments at 166 bp and 350 bp in NSCLC patients (Fig. 1B; upper plot); in serum cfDNA, there were multiple additional peaks within the range of 580-3,200 bp (Fig. 1B; lower plot), showing that the fragment distribution of cfDNA is more complex in serum than that in plasma.

Strict standardization of blood collection and serum preparation methods is mandatory, before serum cfDNA concentration can be used as a potential marker of clinical disorders. The time and temperature of blood storage after centrifugation significantly impacted the serum cfDNA concentration (Fig. 1C and D). Compared with the cfDNA concentration in serum prepared immediately after venepuncture (0 h at room temperature), the cfDNA concentrations in serum decreased after storage at room temperature for 2 or 10 h. However, the changes in cfDNA concentration in serum after storage at 4°C for 24 h, −25°C for 1 month, or −80°C for 1, 3, or 6 months were not significant. Based on these findings, serum samples from NSCLC patients or normal controls were frozen at −80°C immediately after collection.

To determine whether cfDNA concentration correlates with NSCLC status or any other clinical parameters, we quantified cfDNA in serum specimens from 60 NSCLC patients and 68 normal controls in good health by LINE1 qPCR. Analysis of cfDNA in different subgroups of NSCLC patients indicated that serum cfDNA is independent of sex and age: there was no significant difference (P>0.05) in serum cfDNA concentration between NSCLC patients aged 39–60 years or 61–85 years, or between NSCLC patients of different sex (35 males and 25 females) (Table I). When patients were stratified by NSCLC status, the results showed that the cfDNA concentration was significantly higher in NSCLC patients [1.06 (0.75–2.01) ng/µl] than that in normal controls [0.42 (0.27–0.65) ng/µl] (Fig. 2A). NSCLC patients with more advanced disease (stage III–IV) had considerably higher cfDNA concentrations than those in either normal controls or patients with stage I–II disease (Fig. 2B). In line with the finding that patients with stage III–IV disease have higher cfDNA levels; NSCLC patients with lymph node metastasis and distant metastases also had significantly higher cfDNA levels than patients without metastasis (Table I).

ROC curves were generated to distinguish NSCLC patients from normal healthy controls based on serum cfDNA, CEA or CYFRA21-1 levels (Fig. 3). CEA and CYFRA21-1 are established biomarkers for the diagnosis of NSCLC (18). The AUC of cfDNA for distinguishing NSCLC patients from normal controls was 0.848 (95% CI: 0.774–0.905), which was comparable to that of CEA [0.829 (95% CI: 0.753–0.890)] and CYFRA21-1 [0.833 (95% CI: 0.756–0.893)]. No obvious differences were found between the AUC values of each indicator (P>0.05). YI was the largest at 0.67 ng/µl cfDNA, which was defined as the cut-off value of cfDNA, and yielded a sensitivity of 81.67% and a specificity of 77.94%. By comparison, at the cut-off values of CEA and CYFRA21-1, the sensitivity of CEA was 61.67% with a specificity of 97.06%, and the sensitivity of CYFRA21-1 was 63.33% with a specificity of 91.18%. To determine whether cfDNA quantification could be used in combination with existing biomarkers to maximize the diagnostic efficiency for NSCLC patients, the ROC AUC was calculated after combining CEA and CYFRA21-1 with cfDNA. The integration of cfDNA with CEA and CYFRA21-1 (AUC 0.915; sensitivity 83.33%; and specificity 85.29%) resulted in an improved diagnostic efficiency for NSCLC (Table II).

To assess whether the cfDNA concentration is different in samples obtained from NSCLC patients with progressive rather than stable disease, the 60 NSCLC patients were divided into the progression and stable groups. The cfDNA concentration was higher in NSCLC patients with progressive disease than that in patients with stable disease (Fig. 4A). In addition, the Kaplan-Meier curve revealed that serum cfDNA ≥1.04 ng/µl was related to poorer PFS than serum cfDNA <1.04 ng/µl (median PFS: 7.0 vs. 3.0 months, hazard ratio [HR], 0.43; 95% confidence interval [CI]: 0.19–0.96, P<0.05) (Fig. 4B).

To identify the predictive value of cfDNA in NSCLC progression, cfDNA, CEA and CYFRA21-1 levels were analysed in serum samples from 4 NSCLC patients with progressive disease at several time-points. The line charts were plotted to describe the changes in cfDNA, CEA and CYFRA21-1 levels during cancer progression. As shown in Fig. 5, the time-dependent changes in the three markers in patient 2 were nearly identical, but were inconsistent with the changes in the other 3 NSCLC patients. During the course of the disease, there was a general trend of cfDNA concentration decreasing or remaining stable before periods of disease stability, and increasing before periods of disease progression. Interestingly, cfDNA levels often peaked before a period of disease progression, and the lead time was 1–7 months compared with clinical medical imaging. Alterations in serum levels of CEA and CYFRA21-1 occurred in line with disease progression, and did not show such peaks.