Extracellular vesicle-derived DNA for performing EGFR genotyping of NSCLC patients

Lung cancer results in the largest number of cancer-related deaths worldwide and non-small-cell lung cancer (NSCLC) accounts for more than 85% of all lung cancer cases [1]. Most patients are diagnosed at an advanced stage due to lack of efficient diagnostic approaches and asymptomatic characteristic of the disease leading to a poor prognosis [1, 2]. Recent development of target specific drugs such as epidermal growth factor receptor-tyrosine kinase inhibitors (EGFR-TKIs) have slightly improved survival rate, but easy and fast diagnostic assessment of mutation status is important for timely treatment of patients. At present, majority of EGFR genotyping is done through tissue biopsy while liquid biopsies using cell-free DNA (cfDNA) are used as supplement tests [3, 4]. The conventional tumor biopsy to assess mutation status can be problematic depending on the location and size of the tumor. Liquid biopsy, a noninvasive way to detect circulating tumor DNA (ctDNA) in the blood, are proposed as an alternative way to detect, evaluate and monitor tumor-drug relation [3, 5]. The integration of liquid biopsy into cancer treatment depends on the precision of detecting ctDNA in blood samples, but plasma cfDNA only contains roughly 1% of ctDNA [6]. Therefore, even with high specificity reported in using ctDNA, varied sensitivity is a problem. For example, some studied reported relatively high sensitivities ranging from 66% to 78%, while other studies resulted low sensitivities ranging from 28.8% to 46%. [7‐10].

The main reason for this high variability of sensitivity ctDNA lies on the unstable nature of cfDNA in the samples [7]. In contrast, DNA inside extracellular vesicle (EV) shed by tumor cells is well protected by dual lipid membranous coating and thus has inherent stability [5, 11, 12]. Along with abundant new discoveries in various tumor-derived EVs and EV-derived DNA (EV DNA), they have great potential as cancer biomarkers as well as platforms for personalized medicine [12‐14]. For example, Thakur BK, et al. have demonstrated that the majority of DNA associated with tumor exosomes is double-stranded in various cancer cell-lines and highlight the translational value of exosomal DNA for its potential usefulness as a circulating biomarker for cancer detection [12].

Previous studies report that measurements of EV size varies according to the methods of the isolation and measurement, but EVs from human body fluids (plasma, urine) are generally found to be in 20~ 300 nm range [15‐17]. This study identified the size of BALF EVs to be from 20 to 250 nm (Fig. 1a and Additional file 1: Figure S1) and the size of plasma EV to be smaller ranging from 5 to 15 nm (Fig. 1b and Additional file 1: Figure S2). Visualization of purified EV fraction showed round shape and heterogeneous in size, averaging in 106 nm (SD ± 34 nm) BALF EVs (Fig. 1c and Additional file 1: Figure S3). Image obtained from EM of plasma EVs showed other substances in the background, which are probably proteins, low-density lipoproteins (LDL), and high-density lipoproteins (HDL) that exist in plasma (Fig. 1d).

Results of immuno-electron microscopy (immune-EM) identified the presence of dsDNA in the intraluminal portion of EVs covered by a multilayered or single-layered membrane (Fig. 1e and Additional file 1: Figure S4). NanoDrop analysis showed that both EV DNA and cfDNA obtained from BALF had similar concentration and purity (Additional file 1: Table S2). Bioanalyzer analysis of dsDNA length showed that the size of both EV DNA and ctDNA are > 3 kb, but abundance of dsDNAs that are longer than 3 kb is greater in EV DNA compared to ctDNA (Fig. 1f, Additional file 1: Table S3 and Figure S5). We compared the sensitivity of EGFR mutation testing between BALF EV DNA and cfDNA by performing PCR. Results showed higher sensitivity of EGFR mutation testing when using EV DNA compared to using cfDNA. For example, a threshold cycle (Ct) value of EV DNA was 32, whereas that of cfDNA could not be determined (Fig. 1g and Additional file 1: Table S4).

At present, blood plasma DNA samples are widely used for liquid biopsies and the major concern is the instability of the DNA from the blood sample, which leads to decreased sensitivity. Presumably, using plasma EV DNA instead of cfDNA could increase sensitivity as they are shielded from the outer environment by lipid bilayer structure of EV. We performed EGFR genotyping with 20 plasma samples of NSCLC patients. Comparison of the results with tissue biopsy results showed that plasma cfDNA only had 30% accordance with tissue typing. Even though detection sensitivity with plasma cfDNA was relatively low, we were able to improve detection sensitivity to 55% by using plasma EV DNA (Additional file 1: Table S5). This finding suggests that liquid biopsy using EV DNA is advantageous to conventional use of cfDNA. Recently, Allenson et al. reported that liquid biopsy of exo-DNA was superior to that of cfDNA for detecting mutant KRAS in plasma samples of patients with pancreatic ductal adenocarcinoma [11]. However, isolation and purification of EVs from plasma is associated with some technical difficulties. Characteristics of lipoproteins, especially low-density lipoproteins (LDLs), present in the plasma are very similar to those of EVs; moreover, contamination of isolated EVs samples with LDLs interferes with analysis [18]. Components similar to LDLs were identified in our study after isolating EVs from plasma (Fig. 1d), which could explain the relatively low sensitivity of liquid biopsy when using plasma EV DNA.

Although, the standard for non-invasive cancer diagnostics is detection of biomarkers circulating in blood, it remains a challenge due to abundant non-cellular contents. Our approach to overcome this problem is to analyze more immediate biofluids such as bronchoalveolar lavage fluid (BALF) of NSCLC patients. Proximal biofluids display component specificity and in some cases, they are in direct contact with the site of the disease. Bronchoalveolar washing is not an entirely non-invasive procedure, but often ordered in individuals with and suspected of lung cancer during bronchoscopy.

Higher quantity and quality of EV DNA compared to BALF cfDNA suggested a potential for clinical application of EV DNA for liquid biopsy. Best way to test the specificity and sensitivity of liquid biopsy done with BALF EV DNA is comparing the results with conventional tissue biopsy. Therefore, we tested 23 BALF samples from NSCLC patients with proven EGFR genotyping from tissue biopsy (9 EGFR wild type and 14 EGFR-mutated). EGFR genotyping using BALF EV DNA showed 100% accordance with tissue typing, while detection sensitivity using ctDNA only yielded 71.4% (Table 1). Test results of BALF samples were significantly higher compared to the result of plasma samples in both ctDNA and EV DNA, demonstrating that proximal biofluids better represents tumor status. In addition, as we have seen with the plasma sample, EV DNA resulted in higher specificity and sensitivity. Specifically, Kappa coefficient for EGFR genotyping by using tumor tissue sample and BALF EV DNA was 1.0 (p < 0.01), which was higher than that for EGFR genotyping by using tumor tissue sample and BALF ctDNA (kappa = 0.705). Our results of liquid biopsy using BALF EV DNA was compatible enough with tissue genotyping that it can replace tissue biopsy when obtaining tissue sample is difficult. Additionally, liquid biopsy could significantly reduce turn-around time, which usually takes two or three weeks for EGFR genotyping using tissue biopsy.

Unfortunately, most patients prescribed with TKI after identification of mutations such as Exon 19 deletion and p.L858R acquire resistance to the drug after prolonged treatment. Hence, precise and swift identification of secondary mutations, p.T790 M, that represent acquired resistance is important for planning future treatment. We extended liquid biopsy using BALF to the patients who developed acquired resistance to EGFR-TKIs and need rebiopsy to further test specificity and sensitivity. Nine patients who developed resistance to TKIs were examined for the p.T790 M mutation by performing conventional tissue rebiopsy and liquid biopsy with BALF samples. Of these nine patients, suitable cancer cells to perform tissue rebiopsy were obtained from only six patients, while adequate tissues were unobtainable from three patients. Two patients out of six patients were identified to have the p.T790 M mutation from tissue rebiopsy (Fig. 2a and Additional file 1: Table S6).

When we performed liquid biopsy of BALF cfDNA three patients were identified to have p.T790 M mutation including two patients who yielded positive results from the tissue rebiopsy as well as one patient who were not identified by the tissue rebiopsy (Fig. 2a and Additional file 1: Table S6). Sensitivity of liquid biopsy increases even further when tested with BALF EV DNA with total of five patients identified to have p.T790 M mutation including two patients who yielded positive results from the tissue rebiopsy as well as three patients who were not identified by the tissue rebiopsy. Three additionally identified patients consists of two patients who could not provide adequate tissue sample for performing tissue rebiopsy and one patient who yielded negative result for p.T790 M mutation from tissue rebiopsy (Fig. 2a and Additional file 1: Table S6). Furthermore, two newly identified patients showed partial response to subsequent osimertinib treatment (Additional file 1: Table S6). We demonstrated for the first time that EVs isolated from BALF of NSCLC patients carry genomic dsDNA and specific mutant EGFR DNA inside the double layered membranous vesicles (Fig. 2b). Clinical usefulness of genotyping using BALF EV DNA was more prominent in the matter of detecting p.T790 M mutation for the prescription of the 3rd generation EGFR-TKIs such as osimertinib [19]. Although the present study did not include a large sample size, preliminary findings of this study suggest that liquid biopsy of BALF EV DNA can overcome limitations associated with tissue rebiopsy, which is widely performed for detecting p.T790 M mutation.

Our results show that EGFR mutation detection in NSCLC patients is possible through EGFR genotyping of EVs present in plasma and BALF. Liquid biopsy of BALF EV DNA is non-invasive, simple and faster testing method that is also high in accuracy and even surpass detection sensitivity compared to tissue biopsy. Sensitivity is shown to be especially high in acquired resistance patients. This study revealed a novel liquid biopsy method of using EV DNA for EGFR genotyping. It has demonstrated potential to serve as a diagnostic and prognostic method in NSCLC patients.