Detection of EGFR mutation of pulmonary adenocarcinoma in sputum using droplet digital PCR

Patients with advanced non-small cell lung cancer (NSCLC) harboring epidermal growth factor receptor (EGFR) mutations can experience improved prognosis when treated with EGFR-tyrosine kinase inhibitor [1, 2]. EGFR-mutated lung cancer is reported to comprise approximately 45% of the NSCLC cases in Asia and approximately 15% in Europe and the United States [3‐6]. Approximately 90% of EGFR mutations are either EGFR exon 21 L858R point mutations (Ex21) or EGFR exon 19 deletion mutations (Ex19) [7]. Detection of EGFR mutations is the first process in deciding the treatment for patients with advanced NSCLC. Currently, bronchoscopic biopsy, computed tomography (CT)-guided biopsy, and surgical biopsy are performed to collect tumor samples. However, these methodologies have various complications and are invasive for patients, which can be especially problematic for those in poor condition [8‐11]. In addition to these biopsies, non-invasive liquid biopsy using blood (liquid biopsy) has recently been performed to detect EGFR mutations. However, a meta-analysis reported that the sensitivity and specificity of the detection of EGFR mutations in blood were 65% and 91%, respectively, with recurring low detection sensitivity [12]. Development of a methodology capable of more safely collecting tumor samples and more accurately detecting EGFR mutations in patients with lung cancer is needed.

Sputum can be collected non-invasively from patients. Whether EGFR mutation analysis can be performed accurately using sputum is unclear. The sensitivity of detecting lung cancer in sputum cytology is reported to be approximately 40 to 66% [13, 14]. However, sputum has been considered to be unsuitable for EGFR mutation analysis because it contains many normal cells, such as bronchial epithelium and inflammatory cells, with tumor cells comprising < 1% of the total number of cells in sputum [15]. Furthermore, primary lung adenocarcinoma, for which molecular analysis is especially important in deciding the treatment protocol, was reported to be less likely to be detected via sputum cytology compared to central squamous cell carcinoma [16]. Therefore, an ultrasensitive method for detecting tumor EGFR mutations in sputum sample is necessary. Furthermore, it is necessary to know the clinical characteristics of patients whose sputum contains a sufficient amount of tumor cells for the EGFR gene detection.

Droplet digital polymerase chain reaction (ddPCR) is a technology that can detect and quantify specific sequences with much higher sensitivity and specificity than conventional real-time PCR [17, 18]. ddPCR is based on the limiting dilution of DNA sample in microcompartments within droplets of a water-in-oil emulsion [17]. The sensitivity of ddPCR was reported to be approximately 0.03%-0.001% mutant DNA in the detection of EGFR T790M gene mutation and 0.0005% mutant DNA in the detection of KRAS gene mutation [19, 20]. Considering that it is difficult to detect ≤ 1% mutant DNA using conventional PCR [21, 22], ultrasensitive ddPCR is necessary to detect EGFR mutations in the tumor cells in sputum, because of the abundant normal cells or non-mutated DNA in the background of a sputum sample. However, to date, few reports have described the use of ddPCR in EGFR mutation analysis in sputum samples.

This study aimed to examine EGFR mutations of primary lung adenocarcinoma in sputum samples using ddPCR and compare it with an EGFR mutation in surgically resected lung cancer. This study also analyzed the clinicopathological features of patients in which sufficient tumor cells were detected in sputum for EGFR mutation testing via ddPCR.

Of the 118 patients who enrolled in this study, the number of patients with SC (+), SC (−), and NC was 13 (11.0%), 76 (64.4%), and 29 (24.6%), respectively. Table 1 summarizes the radiological and cytopathological findings, and the EGFR mutation status (sputum sample and surgical resected specimen) of 13 SC (+) cases. In addition, allel frequency in samples in which EGFR mutation were detected in sputum samples was also shown. Classes III, IIIa, IIIb, IV, and V included 2, 2, 3, 1, and 5 patients, respectively. STAS was detected in surgically resected specimens in 12 of 13 SC (+) cases. Ex21 and Ex19, and wild-type EGFR were observed in 2, 2, and 9 cases, respectively, based on ddPCR of the sputum. Among SC (+) samples, the EGFR mutation status of the main tumors in FFPE sections and in sputum were identical in 12 cases, and the sensitivity, specificity, and positive predictive value to detect EGFR mutations (Ex19 and Ex21) were 80.0%, 100%, and 100%, respectively (Table 2). In one discordant case (Case 4), the cytological examination of sputum was Class IIIa with suspected squamous cell carcinoma. Figure 1 shows the results of ddPCR data analysis (1-D plot). Ex21 (Case 1) and Ex19 (Case 7) were detected in sputum by ddPCR. The lowest allele frequency in the EGFR mutant SC (+) cases was 0.24% (Table 1).

Figure 2 presents the cytological findings of sputum and histological findings of resected specimen in Case 1 to 3 of the 13 SC (+) patients. In all three patients, the EGFR mutation status of the main tumors in FFPE sections and in sputum were identical. In all three patients, STAS was detected in FFPE sections of the surgically resected tumors.

Among 76 patients of SC (−) group, there were 9 cases in which ddPCR could not be performed because of an insufficient amount of DNA collected. There was one case in which the Ex19 was detected in sputum samples in the SC (−) group. The EGFR mutation status was identical to that of FFPE sections of the tumor specimen. The sensitivity, specificity, and positive predictive value to detect EGFR mutations (Ex19 and Ex21) were 3.1%, 100%, and 100%, respectively, in SC (−) cases (Table 2).

Table 3 presents the comparison of the clinicopathological features between 13 cases of SC (+) and 105 cases of SC (−) + SNC. Compared to the SC (−) + SNC group, the CT tumor size was larger and PET SUVmax was higher in SC (+) group. Patients with SC (+) were in a more advanced stage (clinically and pathologically) compared to patients with SC (−). STAS was detected more frequently in FFPE sections of resected tumor in the SC (+) group compared to the SC (−) group (92.3% vs 34.3, p < 0.001). There was no significant difference in the EGFR mutation status between the two groups (p = 0.902); there were 61.5% and 43.8% of wild-type EGFR in SC (+) and SC (−) + SNC group, respectively.

Figure 3 depicts the results of ROC curve analysis for discrimination of the SC (+) and SC (−) + SNC groups. The area under the ROC curve (AUC) regarding CT tumor size was 0.823 (95% confidence interval [CI]: 0.717–0.929). The AUC regarding PET SUVmax was 0.809 (95% CI 0.718–0.901). There was no significant difference between the AUC of both groups (p = 0.809). The cut-off value of CT tumor size was 29 mm in ROC curve analysis, and the sensitivity and specificity were 69.5% and 84.6%, respectively. The cut-off value of PET SUVmax was 3.02, and the sensitivity and specificity were 58.1% and 100%, respectively.

Multivariate analysis based on a logistic regression model revealed that CT tumor size (odds ratio = 10.6, 95% CI 1.85–61.0, p = 0.008) and STAS (odds ratio = 17.7, 95% CI 1.97–158, p = 0.010) were independent potential predictive factors for SC (+) (Table 4).

This is the first report of the detection of EGFR mutations of primary lung adenocarcinoma using ddPCR from prospectively collected sputum samples and compared it with an EGFR mutation in surgically resected lung cancer. EGFR mutations can be detected with high sensitivity by ddPCR from sputum sample if the sputum cytology is positive. Since a CT tumor size ≥ 29 mm is a potential predictive factor for sputum cytology positive, sputum should be collected in such cases for the EGFR mutation analysis.

Bronchoscopic, CT-guided, and surgical biopsies are currently performed in clinical practice to obtain tumor tissue for molecular analysis. However, these methodologies are invasive. An overall complication rate of 1.55% was reported for bronchoscopic biopsy, and included bleeding (0.63%) and pneumothorax (0.44%), with a mortality rate of 0.003% [8]. The reported rates of mortality and serious complications for CT-guided percutaneous needle biopsy were 0.07% and 0.75%, respectively, with complication rates of 35% for pneumothorax [9, 10]. Furthermore, surgical biopsy requires general anesthesia and is more invasive. The reported mortality rate due to surgery was 0.5%, and the rate of complications that included pneumonia, air leakage, and atelectasis was 9.6% [11]. Plasma can be collected with minimal invasion from lung cancer patients to detect EGFR mutations in cfDNA. However, there is a possibility that EGFR mutations cannot be detected in patients with EGFR mutant lung cancer if the amount of cDNA is below the threshold of detection sensitivity [12]. For this reason, liquid biopsy is performed if a tissue biopsy cannot be performed in patients with poor conditional status. Similar to plasma, sputum can be collected non-invasively, which is a greater advantage for cancer patients compared to other biopsies. The use of sputum may be another option to detect EGFR mutations in patients who are in poor condition.

Tumor cells in sputum specimens were reportedly detected in < 1% of the cells contained in sputum [15], and sputum has been considered unsuitable for molecular analysis. The present study demonstrates that ddPCR can detect EGFR mutations in primary lung adenocarcinoma with high sensitivity (80.0%) and high specificity (100%) in SC (+) cases. One discordant case in SC (+) (Case 4, invasive lepidic adenocarcinoma) was suspected to be squamous cell carcinoma (Class IIIa) according to the cytomorphologic features of sputum cytology. The discordance of EGFR mutation between sputum and surgical specimens might be due to the absence of tumor cells in the sputum. Case 4 had a history of bronchial asthma, obstructive airway disease, and lung cancer surgery. Bronchial cell hyperplasia, reactive/atypical bronchial cell, or squamous metaplasia that might be misinterpreted as malignancy, can often be observed in respiratory cytology specimens obtained from patients with asthma, chronic obstructive airway disease, inflammatory disease of the lung, and past history of lung disease chemotherapy/radiation/surgical treatment [26, 27]. When performing EGFR mutation analysis in combination with sputum cytology, it is necessary to understand the cytomorphologic features and past history of lung disease in these patients. In contrast, we observed that the sensitivity of EGFR mutation detection was as low as 3.1% in SC (−) cases, and it was considered irrelevant to perform EGFR mutation analysis unless the cytology was positive. Only one case in which the EGFR mutation (Ex19) was detected in the sputum sample in the SC (−) case. Malignant cells might not have been detected by sputum cytology. Sputum cytology involves abundant normal cells, which sometimes make it difficult to detect malignant cells. Endo et al. reported that cytotechnologists with less cytology experience under-diagnosed significantly more frequently than those with considerable experience [28]. When performing EGFR mutation testing in combination with sputum cytology, well-trained cytotechnologists are needed to make cytological diagnoses.

Hubers et al. reported that the sensitivity of EGFR mutation detection was 30 to 50% in 10 sputum samples using four different EGFR mutation analyses (Cycleave PCR, COLD-PCR, Pangaea Biotech SL Technology, and High Resolution Melting) [29]. Su et al. performed amplification refractory mutation system (ARMS)-PCR for 35 sputum samples containing tumor cells collected from stage III-IV lung cancer patients and reported a 90.9% sensitivity [30]. Wu et al. reported a 63% sensitivity when an EGFR sensitizing mutation was analyzed in 50 sputum samples using next-generation sequencing [31]. Presently, ddPCR analysis of 80 sputum samples revealed an 80.0% detection sensitivity for EGFR mutations if the sputum cytology was positive. Recently, Wang et al. reported that detection sensitivity and specificity for EGFR mutations of 46.2% and 100%, respectively, as detected using SuperARMS using sputum cell-free DNA from 102 sputum samples [32]. It is necessary to prospectively examine in large scale which methodology can detect EGFR mutations more accurately in sputum.

It is unclear which sputum collection method was appropriate for the analysis of EGFR mutations. Hubers et al. used Saccomanno's fixative (2% polyethylene glycol in 50% ethanol) for 3-day pooled sputum [29, 33]. Su et al. collected spontaneous sputum in a 1.5 mL container [30]. Wu et al. collected approximately 5 mL of spontaneous sputum in a mixed solution with an equal volume of Saccomanno's fixative and 0.005% dithiothreitol solution at a 1:1 ratio [31, 34]. We used YM fixative solution for 3-day pooled sputum. It has been reported that the detection sensitivity of lung cancer in sputum is increased by a longer duration of sputum collection and with the induction of sputum by nebulization with hypertonic saline [33, 35]. A future study should examine whether the sensitivity of EGFR mutation detection can be improved by different sputum collection methodologies.

This study showed that EGFR mutation analysis of sputum should be performed for SC (+). However, little is known about the risk factors for SC (+). Therefore, sub-analysis of risk factors for SC (+) in patients with primary lung adenocarcinoma was performed in this study. As shown in Fig. 3 and Table 4, we found that CT tumor size was strongly associated with SC (+). Patients with CT tumors ≥ 29 mm were considered good candidates for EGFR mutation analysis. The detection sensitivity of lung cancer in sputum cytology was reported to be 40 to 66% [13, 14]. Presently, the sensitivity was 11.0%. There are two reasons of the low sensitivity we observed. Firstly, all the patients had lung adenocarcinoma. Sputum cytology is highly effective for central type squamous cell carcinoma in patients with hemoptysis [16]. Sing et al. reported that the detection rate of sputum in adenocarcinoma in 64 patients was 25.0% [14]. Secondly, patients with early staged lung cancer were included. The sensitivity of sputum cytology for patients with advanced stages are reported to be higher than for early stage cancer [14]. The diagnostic sensitivity of bronchoscopic biopsy was reported to be 88% (78% for peripheral lung cancer) [36]. The sensitivity of percutaneous needle biopsy was 86.1% [37]. Although sputum can be collected non-invasively from lung cancer patients, the detection sensitivity of lung cancer was lower compared to other methodologies. In sub-analysis of this study revealed that the sensitivity increased to 69.5% in sputum cytology if the CT tumor size was ≥ 29 mm. Risse et al. similarly reported that the detection sensitivity of primary lung cancer was high when the tumor size exceeded 24 mm [16].

This is the first study to analyze the correlation between pathological findings of surgically resected specimens and SC (+). SC (+) status was strongly associated with STAS. The presence of STAS was higher in patients with SC (+) than in patients with SC (−) (92.3% vs. 34.3%). STAS was a potential predictive factor for SC (+) in multivariate analysis (Table 4). STAS is a risk factor for recurrence of primary lung adenocarcinoma and squamous cell carcinoma [38, 39]. Previously, we examined the morphology and EGFR mutation status of tumor cells in airway secretions collected from segmental or lobar bronchus of surgically resected specimens and compared the results with FFPE tumor tissue. The study demonstrated that STAS may be spread to the respiratory tract as far as segmental or lobar bronchus of the tumor [40]. Because STAS was a risk factor for SC (+) in the present study, we suggest that malignant tumors can be efficiently detected from sputum if STAS is predicted in preoperative radiological findings. Toyokawa et al. reported that the presence of notch and the absence of ground glass opacity were CT findings that were related to the presence of STAS [41]. Kim et al. reported that solid component ratio ≥ 90% in CT was a potential predictive factor of STAS [42]. Performing sputum cytology for tumors that display these CT findings may increase the detection sensitivity of sputum cytology in patients with primary lung cancer.

There were several limitations in our study. First, this study was conducted at a single institution with a small number of patients. Second, we have not investigated the efficacy of EGFR-tyrosine kinase inhibitor based on the EGFR status detected in sputum samples. Third, the methodology of detecting EGFR mutation differed between surgical resected tumor samples and sputum samples. The discordance of EGFR mutations between the two samples might have occurred if the number of tumor cells harboring EGFR mutations with mutations was too small to detect these mutations, due to heterogeneity in FFPE tumor sections by conventional PCR. Fourth, it is unclear whether performing EGFR mutation testing of sputum based on sputum cytology results is valid. Further studies are needed to investigate the clinicopathological factors of lung cancer patients who are clinically relevant for sputum-based EGFR mutation testing. Fifth, this study examined only the Ex21 and Ex19. Future studies will need to detect other EGFR mutations, such as T790M, and other driver gene mutations by ddPCR from sputum, and evaluate the usefulness of ddPCR for sputum in clinical practice. Detection of T790M for the indication of Osimertinib is especially important in clinical practice. The detection of T790M from sputum may play an important role similar to that of cell-free DNA in plasma. Further studies are needed to demonstrate the clinical usefulness of T790M detection in sputum using ddPCR in patients with primary lung cancer undergoing EGFR-TKI treatment. Further research also is necessary to compare the detection rates of EGFR mutations between sputum and cell-free DNA in plasma.