High-efficiency EGFR genotyping using cell-free DNA in bronchial washing fluid

Abstract

Background

EGFR mutation testing is required for treatment of lung adenocarcinoma using epidermal growth factor receptor-tyrosine kinase inhibitor. However, the amounts of tumor tissue or tumor cells obtained by bronchoscopy are often insufficient. Bronchial washing fluid, obtained by lavage with saline after tumor biopsy or brushing, and the supernatant of bronchial washing fluid are thought to contain cell-free DNA that would be potentially applicable for EGFR testing.

Methods

From among patients with suspected adenocarcinoma or non-small cell lung carcinoma diagnosed from biopsy or surgical specimens at the University of Tsukuba Hospital between 2015 and 2019, cell-free DNAs from 80 specimens of supernatant of bronchial washing fluid (50 with EGFR mutation and 30 with wild type EGFR) and 8 blood serum samples were examined for EGFR mutation using droplet digital PCR.

Results

Among the 50 patients harboring EGFR mutation, the rate of positivity for cell-free DNA extracted from supernatant of bronchial washing fluid was 80% (40/50). In nine of the EGFR mutation-positive cases, tumor cells were not detected by either biopsy or cytology, but the mutation was detected in four cases (4/9, 44%). Comparison of the cell-free DNA mutation detection rate between supernatant of bronchial washing fluid and blood serum in six cases showed that mutations were detected from the former in all cases (6/6, 100%), but from the latter in only one case (1/6, 17%).

Conclusions

Using supernatant of bronchial washing fluid samples, the detection rate of EGFR mutation was high, and EGFR mutations were detectable even when no tumor cells had been detectable by biopsy or cytology. Supernatant of bronchial washing fluid might be an effective sample source for EGFR mutation testing.

We were able to detect EGFR mutation from bronchial washing fluid specimens expected to contain cfDNA, in cases where tumor cells were undetectable by biopsy or cytology.

Introduction

Gefitinib was developed in 2002 as an epidermal growth factor receptor-tyrosine kinase inhibitor (EGFR-TKI) and has shown high effectiveness against lung adenocarcinoma in female and Asian patients (1). In 2004, Lynch (2) and Paez (3) reported simultaneously that mutations in exons 18–21 of the EGFR kinase domain were related to the efficacy of EGFR-TKI. Since then, EGFR mutation testing has become essential before any treatment using EGFR-TKI. Many methods for EGFR mutation testing have been developed, such as direct sequencing, the amplification refractory mutations system and peptide nucleic acid-locked PCR clamping. For enrichment of the mutant allele by PCR, selective digestion of wild-type DNA templates has also been attempted using restriction endonucleases. For such analysis, small amounts of DNA extracted from formalin-fixed paraffin-embedded tumor tissue are generally employed, with strict regulation of fixation conditions and tumor cell content.

Although various cytological specimens can also be used for genetic testing (4), such as brush-washing materials (5–7), needle aspiration materials and sputum (8–11), cell blocks made from fluid pleural effusion are currently considered to yield the best results (12–14).

On the other hand, in cases where it is not possible to obtain tumor tissues and/or make cell blocks, liquid biopsy using circulating tumor DNA (ctDNA) in blood serum is a valid alternative for detecting abnormalities of targetable genes, especially in recurrent or advanced cases. However, a major disadvantage of this approach is that it yields only a low number of ctDNA targets in blood serum. In order to avoid false-positive and false-negative results, it is important to adopt and validate technologies with high sensitivity and specificity in the pre-analytical phase of blood sampling (15).

Endoscopists usually collect bronchial washing fluid (BWF) after transbronchial lung biopsy and/or brushing cytology, since such BWF has a high possibility of containing tumor cells. Generally, for diagnosis, cytologists centrifuge the BWF sample to make a pellet of the tumor cells it contains. The remaining BWF supernatant (sBWF), containing no cellular component, is usually discarded. However, sBWF might nevertheless contain cell-free DNA (cfDNA) from tumor cells.

Droplet digital PCR (ddPCR) is one PCR-based method for detecting the total amount of a viral genome, orphan allele and/or small amounts of tumor-related genes (16,17). This method can be applied for liquid biopsy to detect a tumor genome from blood serum or plasma.

In this study, we focused on sBWF remaining after cytological examination to determine whether the cfDNA it contained could be used for EGFR mutation testing by ddPCR.

Materials and methods

Specimen collection and processing

Among lung carcinoma patients treated at the University of Tsukuba Hospital (Ibaraki, Japan) between 2015 and 2019, we selected sBWF samples from those who had been diagnosed as having non-small cell carcinoma by bronchoscopic lung biopsy (BLB) or from materials resected during surgical treatment. These specimens comprised 50 with EGFR mutation and 30 with wild type EGFR. The details are shown in Fig. 1a. Bronchial washing (BW) is a method for washing a targeted lesion within the respiratory tract. In brief, after BLB and/or brushing/curettage cytology, 20 mL of saline is injected into the target bronchus once or twice with a 20-mL syringe and then collected using the same syringe. If biopsy, brushing or curettage is not possible, BW may be the only option. In clinical practice, the BWF is centrifuged to obtain a cell pellet and sBWF. The cell pellet is then smeared on a glass slide, and the supernatant is usually discarded.

Figure 1.

(a). Specimen collection. Among lung carcinoma patients treated at the University of Tsukuba Hospital (Ibaraki, Japan) between 2015 and 2019, sBWF samples were obtained from 1159, and EGFR mutation testing (PNA-LNA PCR Clamp Method or Cobas EGFR Mutation Test v2) was undertaken for 310 using tissue samples (biopsy or surgical specimens). From among these cases, 80 were selected. Among them, the ex21 L858R EGFR point mutation was detected in 26 cases and the ex19 E746-A750 deletion mutation in 24; EGFR mutations were undetectable in 30 cases. Blood serum samples were included in three, three, and two cases, respectively. (b) Comparison of EGFR genotyping using cfDNA from supernatant of bronchial washing fluid (sBWF) with that using tissue samples (biopsy or surgical specimens). The positivity rate for EGFR mutation using cfDNA extracted from sBWF was 80% (40/50, ex21 L858R in 20/26 cases and ex19 E746-A750 del in 20/24 cases). Thirty cases were negative for EGFR mutation, but one case showed >2 copies of EGFR mutation/μL (ex21 L858R).

In this study, after centrifugation of BWF at 800 × g for 5 min at room temperature, the pellets were smeared onto glass slides and used for cytological diagnosis, whereas the sBWF was stored at −80°C for between 6 months and 5 years until DNA extraction at Tsukuba Human Tissue Biobank Center (Ibaraki, Japan).

Blood serum specimens had also been collected and stored at −80°C at Tsukuba Medical Laboratory of Education and Research (Ibaraki, Japan) for between 6 months and 2 years 6 months.

Tissue samples were fixed with 10% neutral buffered formalin and used for histological diagnosis and EGFR mutation testing (PNA-LNA PCR Clamp Method or Cobas EGFR Mutation Test v2) (SRL Inc, Tokyo, Japan). Clinical stage was evaluated according to the UICC TNM Classification of Malignant Tumors, 8th ed. (18). This retrospective study was reviewed and approved by the institutional review board of University of Tsukuba Hospital (No.H30–031).

DNA extraction

DNA was extracted from sBWF and blood serum using a MagDEA^® Dx SV kit (Precision System Science Co Ltd, Chiba, Japan) with magLEAD^® 6gC. In this study, using 400 μL of sBWF or blood serum, cfDNA was extracted and dissolved in 50 μL of distilled water. The DNA concentrations were measured using Nano Veu (GE Healthcare Bio-Sciences, Uppsala, Sweden). The extracted cfDNA was quantified using Tapestation (Agilent Technologies, Inc, Santa Clara, CA) (data not shown).

EGFR mutation analysis by ddPCR using sBWF

Allele fractions (AFs) were measured using the QX200 ddPCR System (Bio-Rad, Hercules, CA) in accordance with the manufacturer’s instructions.

Commercially available EGFR exon 21 L858R point mutation (ddPCR^™ Mutation Assay: EGFR p.L858R c.2573 T > G, Human) and exon 19 E746-A750 deletion (ddPCR™ Mutation Assay: EGFR p.E746_A750delELREA c.2235_2249del15, Human) detection assays were used as positive controls. Approximately 10 ng of cfDNA was used for ddPCR. Positive control gDNAs of ex 21 L858R and ex19 E746-A750del contained 50% mutant copies (Horizon Discovery, Cambridge, UK). Before ddPCR, positive control gDNA was fragmented with the restriction enzymes RspRSII and HaeIII, respectively.

Data were analyzed using QuantaSoft version 1.7 analytical software (Bio-Rad). Results were expressed as copies of mutant allele per microliter of sBWF or blood serum.

Whole slide images

Cytology slides were prepared using BWF. The cutoff was 25 tumor cells/1 HPF, divided into tumor cell-rich or tumor cell-scant groups. There were five selected cases in the tumor cell-rich group and another five selected cases in the tumor cell-scant group. The cytology slides in each group were then scanned with a NanoZoomer-2.0HT scanner (Hamamatsu Photonics K.K., Shizuoka, Japan) in ×40 scan mode. Images of five tumor cell-rich specimens and five tumor cell-scant specimens were captured at 1-μm intervals with 21 layers in each case.

We counted the average number of tumor cells and non-tumor cells in 10 fields in both of the groups. The ratio of the average number of tumor cells to the total number of cells was calculated, and compared with the results of EGFR mutation analysis.

Results

Clinical characteristics of the patients

sBWF samples from 80 lung cancer patients were examined: 78 of the samples had been obtained after BLB, and 2 after brushing cytology. Blood serum specimens had also been obtained from 8 of these 80 patients. The patients’ clinical characteristics are summarized in Table 1a. The median patient age was 68 (35–87) years; 37 (46%) were men and 43 (54%) were women. After bronchoscopic examination, the patients were diagnosed histologically as having adenocarcinoma (n = 48), non-small cell carcinoma (n = 8), adenoid cystic carcinoma (n = 1), atypical cells (n = 7) and no tumor cells (n = 14). Among the 14 cases with no tumor cells, 12 were diagnosed as adenocarcinoma from surgical specimens and 2 as adenocarcinoma from rebiopsy specimens. With regard to clinical stage distribution, 35 patients were cStage I (44%), 5 were cStage II (6%), 9 were cStage III (11%) and 31 were cStage IV (39%).

Table 1a.

Clinical characteristics of the study patients

Patient characteristics	n = 80
Mean age (range), years	68 (35–87)
Gender, n (%)  Male  Female	37 (46) 43 (54)
cStage, n  I/II/III/IV	35/5/9/31
EGFR genotype, n  Exon 19 E746-A750 del  Exon 21 L858R  Wild type	24 26 30
Cytology  Positive  Suspicious  Negative	34 26 20
Histology (biopsies)  Adenocarcinoma  Non-small cell carcinoma  Adenoid cystic carcinoma  Atypical cells  No tumor cells  N/A	48 8 1 7 14 2
Smoking  Current/Former/Never	4/36/40
Mean DNA (sBWF) conc., μg/mL (range)	3.95 (−3.00–262.90)
Mean DNA (blood serum) conc., μg/mL (range)	4.40 (2.70–5.80)

sBWF, supernatant of bronchial washing fluid.

In this study, EGFR mutations (ex21 L858R and ex19 E746-A750 del) in the 80 patients were examined by the PNA-LNA PCR Clamp Method or the Cobas EGFR Mutation Test v2 using biopsy or surgically resected specimens. Mutations were detected in 50 cases (ex21 L858R in 26 cases and ex19 E746-A750 del in 24 cases) and the remaining 30 cases were EGFR wild type (Fig. 1a). The relationships between EGFR mutation, cStage and smoking are summarized in Table 1b.

Table 1b.

Relationships between EGFR mutation, cStage and smoking

	cStage					Smoking
	I	II	III	IV	Total	Current	Former	Never	Total
ex19 E746-A750 del	14	0	1	9	24	0	12	12	24
ex21 L858R	15	2	2	7	26	0	6	20	26
Wild type	6	3	6	15	30	4	18	8	30
Total	35	5	9	31	80	4	36	40	80

EGFR mutation analysis of sBWF and blood serum

The median concentration of cfDNA extracted from sBWF and/or blood serum was 3.95 μg/mL (3.00–262.90 μg/mL) and 4.40 μg/mL (2.70–5.80 μg/mL), respectively. Using the cfDNA from either source, EGFR mutation was analyzed using a QX200 ddPCR System. First, EGFR mutation-negative cases were examined to confirm the negative results. Based on these results, the threshold for positivity was set at >2 copies/μL for all assays.

The results of EGFR genotyping using tissue samples or sBWF from all 80 cases are summarized in Fig. 1b (upper row is tissue, lower row is sBWF). Comparison between tissue samples and sBWF among the 50 EGFR mutation-positive cases showed that 40 (80%) were also confirmed using sBWF. The cStage distribution in relation to sBWF and tissue positivity (sBWF/tissue) was: cStage I 72% (21/29), cStage II 50% (1/2), cStage III 67% (2/3) and cStage IV 100% (16/16).

Differences in the detection rate of EGFR mutation in the presence and absence of tumor cells at each cStage

We compared differences in the rate of detection of EGFR mutations in the presence or absence of tumor cells in biopsy or BW cytology specimens. Of the 50 EGFR mutation-positive cases, 48 had biopsy and/or brushing/curettage followed by BW, and 2 had BW alone. Thus, 48 cases were analyzed (Table 2). EGFR mutations were detected in 39 of those 48 cases (81%) using sBWF. Of the 27 patients whose BW cytology and tissue samples were both positive for tumor cell content, 26 (96%) were positive for EGFR mutations, but one had no detectable EGFR mutation. Interestingly, EGFR mutations were detected in four of the nine cases (44%) without tumor cells in BW cytology or biopsy specimens.

Table 2.

Differences in the detection rate of EGFR mutations in the presence and absence of tumor cells at each cStage

	Cytology (−) Biopsy (−)	Cytology (−) Biopsy (+)	Cytology (+) Biopsy (−)	Cytology (+) Biopsy (+)	Total
cStage I (n = 27a)	33 (2/6)	71 (5/7)	100 (1/1)	92 (12/13)	74 (20/27)
cStage II (n = 2)	0 (0/1)	0 (0/0)	100 (1/1)	0 (0/0)	50 (1/2)
cStage III (n = 3)	0 (0/0)	0 (0/0)	0 (0/1)	100 (2/2)	67 (2/3)
cStage IV (n = 16)	100 (2/2)	100 (1/1)	100 (1/1)	100 (12/12)	100 (16/16)
Total	44 (4/9)	75 (6/8)	75 (3/4)	96 (26/27)	81 (39/48)

Biopsy (−), no tumor cells; biopsy (+), adenocarcinoma, non-small cell carcinoma, atypical cells; cytology (−), benign; cytology (+), malignancy, suspicious of malignancy, atypical cells.

cStage I case number excluded two cases in which biopsy was not performed.

^aThe ratio of EGFR mutation-positive cases (EGFR mutation positive cases (sBWF)/EGFR mutation positive cases (tissue)).

Among the 30 carcinomas for which surgical or biopsy specimens had been negative for EGFR mutation, 29 were also negative for EGFR mutation by ddPCR using sBWF. However, in one case, triplicate assays detected >2 copies/μL of EGFR mutation with ex21 L858R (Fig. 1b).

Whole slide images

Using whole-slide imaging (WSI), we investigated the relationship between the tumor cell content of BW cytology specimens and the rate of detection of EGFR mutation from sBWF (Table 3, Fig. 2). From 31 cases diagnosed as malignant or suspicious for malignancy by cytology, we selected 5 that were tumor cell-rich and 5 that were tumor cell-scant. When the two groups were compared, there was no significant difference in the average amount of extracted cfDNA, but the tumor cell-rich group showed a significantly higher EGFR mutation ratio (median copy number ratio) than the tumor cell-scant group. Interestingly, EGFR mutation was still detectable even in the tumor cell-scant group.

Table 3.

Mutation ratio of EGFR mutations in relation to the amounts of tumor cells obtained

	DNA conc. (ug/mL)	DNA conc. (ug/mL) median	Average of TC No. (/1HPF)	Average of NTC No. (/1HPF)	TC/(TC + NTC) (%)	TC/(TC + NTC) (%) median	EGFR mutation ratio (%)	EGFR mutation ratio (%) median
	5.1		67.4	103.8	39.4		23.2
Tumor cell rich group	1.2		52.7	17.1	75.5		9.4
	3.1	5.1	26.9	36.1	42.7	43	8.9	8.9
	14.8		47.1	20.5	69.7		1
	14.4		28.9	70.8	29		1
	4.8		2.5	151.1	1.6		2.2
Tumor cell scant group	3.8		10.4	76.5	12		2
	4.4	4.8	1.5	53.9	2.7	2.7	0.3	0.3
	15.8		1.3	65	2		0.1
	38.8		4.8	160	2.9		0.07

TC, tumor cells; NTC, non-tumor cells; EGFR mutation ratio, mutation copy number/(mutation copy number + wild type copy number).

Figure 2.

Example of whole-slide imaging (WSI). Tumor cells and normal cells were annotated using WSI, and the numbers of each were counted in 10 fields of view.

Comparison of EGFR mutation detection rate between sBWF and blood serum samples

Blood serum from the same patients as those who had also provided sBWF was examined. Using eight blood serum samples, ddPCR analysis was also performed under similar conditions and the rate of detection of EGFR mutation was compared with that based on sBWF. The cases examined included three with ex21 L858R mutation, three with ex19 E746-A750 del, and two that were negative for EGFR mutation on the basis of tissue samples. On the basis of sBWF samples, EGFR mutations were detected in all cases of both ex21 L858R and ex19 E746-A750 del (6/6, 100%) but on the basis of blood serum samples, only one case of L858R could be detected (1/6, 17%). In this case, EGFR mutation was also detected in sBWF. There were no false-positive results among the EGFR mutation-negative cases.

Discussion

In this study, we found that EGFR mutation was detectable in a high proportion (80%) of sBWF samples, particularly in cases where tumor cells could not be identified from either biopsy or cytology specimens.

Bronchoscopy is an effective procedure for diagnosis of lung carcinoma. Recently, histological and cytological specimens have become essential for not only morphological diagnosis but also molecular testing of various driver oncogene abnormalities. If sufficient tissue materials are available, not only histological diagnosis, but also molecular testing can be performed satisfactorily. For histological diagnosis of peripheral lung cancer by BLB, the sensitivity is 57%, and that for washing and brushing cytology is 43 and 54%, respectively (19). Recently, biopsy by EBUS-GS has improved the sensitivity (20), but the tumor cell volume is often lower than that obtained by BLB. Therefore, molecular testing using NGS may be impossible due to the insufficient tumor cell volume. Cytological specimens with/without tumor cells would be very useful in situations where biopsy material cannot be obtained or tissue degeneration has occurred. Cytological sampling, such as bronchial brushing/curettage, can be performed more easily than histological biopsy. Cytology specimens have also received attention in the context of molecular testing (21).

‘The IASLC Atlas of Molecular Testing For Targeting Therapy in Lung Cancer’ and ‘Guideline for Diagnosis and Treatment of the Lung Cancer/Malignant Pleural Mesothelioma/Thymic Tumors’ recommend NGS-based companion diagnostics to detect multiple driver mutation (22,23). Although we completely agree that treatment decision making with NGS-based companion diagnostics is valuable, ddPCR analysis using sBWF is very sensitive and might support such decision making if available specimens are inappropriate for NGS analysis or a second clinical biopsy is impracticable. Therefore, we focused on sBWF collected after biopsy and/or brushing cytology as a practically supportive method. Although sBWF is generally discarded after routine clinical examinations, the cfDNA it contains might be potentially useful for molecular testing (5–7). Recently, clinical application of liquid biopsy using blood plasma samples has become practical for molecular testing. As the amount of ctDNA obtained from blood plasma is very small, liquid biopsy has been employed for detection of EGFR mutations in patients with suspected tumor recurrence, for example, and not specifically for initial diagnosis of low-cStage tumors (24,25). On the other hand, BWF is a specimen obtained by washing and collection from the area containing the tumor periphery after biopsy or brushing cytology. sBWF may therefore contain relatively high amounts of tumor-derived cfDNA as well as RNA and protein (26).

We found that a sufficient amount of cfDNA could be extracted from sBWF (as shown in Table 1a, the mean DNA yield was 3.95 mg/mL). We also performed ddPCR using cfDNA extracted from sBWF collected from patients who had shown positivity for EGFR mutations in 50 tissue samples, and detected EGFR mutations in 40 (80%) of those sBWF samples (Fig. 1b). As ddPCR is much more sensitive than the PCR used for general gene mutation detection (sensitivity about 0.01%), it allows detection of EGFR mutation from very small amounts of mutant DNA (Table 4) (7,16,17,27–30).

Table 4.

Comparison of various EGFR analysis methods using BWF

First author	No. of patients	Stage	Historogical type	Methods	Mutations	Sensitivity	Specificity
Jae Young Hur et al.	137	I-IV	Adenocarcinoma, NSCLC, SCC, large cell carcinoma, Sarcomatoid carcinoma	PNA-mediated real time PCR	Exon 18, 19, 21	75.9	86.7
Kawahara et al.	74	NR	Adenocarcinoma	real-time PCR	Exon 19, 20, 21	88	100
In Ae Kim et al.	224	III-IV	Adenocarcinoma, NSCLC	PNA-mediated real time PCR	Exon 18, 19, 20, 21	97.8	97.7
Fiamma Buttitta et al.	33	NR	Adenocarcinoma	NGS	Exon 19, 21	79	NR
Sang Hoon Lee et al.	73	I-IV	Adenocarcinoma, SCC, sarcomatoid carcinoma	ddPCR	Exon 19, 21	68.42 (ex19 del), 89.47 (L858R)	98.15 (ex19 del), 96.30 (L858R)
Murata Y et al.	80	i-IV	Adenocarcinoma, NSCLC, Adenoid cystic carcinoma	ddPCR	Exon 19, 21	83.3 (ex19 del), 76.9 (L859R)	100 (ex19 del), 96.7 (L859R)

NSCLC, non-small cell lung cancer; SCC, squamous cell carcinoma; NR, not reported; ddPCR, droplet digital PCR.

As an empirical study, this time we examined two sites of EGFR abnormality (ex21 L858R mutation and ex19 E746-A750 del), since these account for almost 70% of all EGFR abnormalities, making it easy to find cases applicable to this kind of study (31). In order to cover all EGFR abnormality sites, it would be necessary to analyze more primer sets.

We divided the cytology specimens into two groups: one containing a large number of tumor cells and the other a small number (Fig. 2). As shown in Table 3, the high group had a higher EGFR mutation copy number ratio than the scant group. However, ddPCR detected EGFR mutation even in the scant group, indicating the applicability of ddPCR for molecular testing using sBWF. The rate of detection of EGFR mutation was 96% in cases that had shown positivity in biopsy or cytology specimens, and interestingly, EGFR mutation was detectable in 44% (4/9 cases) of cases where no tumor cells were present in either biopsy or cytology specimens (Table 2). Even though the detection rate was only 10% (4/39 EGFR-positive cases), it was possible to detect EGFR mutation in specimens without malignant cells. For most materials obtained by bronchoscopy including tumor cells additional ddPCR analysis is not required, but preservation of sBWF would be beneficial if the specimens included no tumor cells. Therefore, ddPCR analysis using sBWF is a useful and sensitive method for screening of overlooked patients who qualify for EGFR-TKI therapy.

Detection of EGFR mutation showed sensitivity of 83.3 and 76.9% and specificity of 100 and 96.7% for ex19 E746-A750 del and ex21 L858R, respectively (Fig. 2). This level of sensitivity is thought to be satisfactory and potentially applicable to daily clinical practice. As this was a retrospective study, we used 400 μL of each of the stored sBWFs as a trial. Therefore, the sensitivity would likely have increased if 1 or 2 mL of sBWF had been used. It will be necessary to clarify the optimum amount of sBWF required for future testing.

Table 4 lists reports of EGFR mutation analysis using sBWF. Although the analytical methods differed, the reported sensitivity and specificity rates were similar to those in our present study. Hur et al. and Kim et al. used EV DNA. cfDNA is usually present in smaller fragments of 200–400 bp, whereas EV DNA is usually dsDNA longer than 1 kb (32–34). Kim et al. considered EV DNA to give very good results. However, DNA extraction is complicated and we think that this might be difficult to perform in routine clinical situations.

Hannigan et al. used FNA specimens that had been washed with RPMI 1640 and analyzed the supernatant (9). In FNA, tumor cells are easily obtained, the amount of cfDNA is high (median 407 ng), and long cfDNA is recovered. In that study, genetic mutations were found in 85 (82%) of 104 cases. Although the sensitivity of detection was similar to that in the present study, we were able to demonstrate EGFR mutations even when no tumor cells had been detectable by biopsy or cytology. Although care should be taken to exclude false negative cases, we believe that sBWF is useful and might be an effective sample source for EGFR mutation testing.

Although the number of samples employed was limited, we were able to compare the use of sBWF and blood serum samples for EGFR molecular testing in eight cases. The median amount of cfDNA extracted from sBWF was 3.95 μg/mL, while almost the same amount, 4.40 μg/mL, was extracted from blood serum. Although we were able to detect EGFR mutations in all six sBWF samples, only one case showed serum positivity for EGFR mutation. While further validation is needed, we consider that sBWF might be superior to serum for EGFR mutation testing. Recently, EGFR mutation analysis using extracellular vesicles contained in sBWF has also been investigated (29,30). If the residual fluid from sBWF is stored, it may serve as a liquid sample to replace blood serum when tissue samples cannot be obtained from BLB specimens.

Interestingly, there was one case in which EGFR mutation (ex21 L858R) was detected in sBWF, even though the biopsy tumor specimen was negative (Fig. 1b). Both the cytology and biopsy specimens showed similar morphological characteristics. We performed ddPCR analysis several times, and obtained positive results each time (data not shown). Although the ddPCR result obtained using sBWF might have been false positive, we considered the possibility of tumor heterogeneity or EGFR mutation in non-tumorous cells. It is possible that pneumocytes with apparently normal characteristics might harbor EGFR mutation, since Hill et al. recently reported detection of EGFR mutations in 18% of histologically normal lung tissue samples from 295 individuals across 3 clinical cohorts by ultradeep mutational profiling (35).

Finally, it should be noted that several benign tumors, such as ciliated mucinous nodular papillary tumor (CMPT), also harbor EGFR mutation. Indeed, it has been reported that ~20–30% of CMPTs have such mutations (36,37). Therefore, a target tumor should not be diagnosed as ‘malignant’ without morphological diagnosis based on positivity for EGFR mutation alone.

In this study, we performed ddPCR using sBWF and demonstrated that EGFR mutation was detectable at high probability, even when no tumor cells were identifiable in cytology and biopsy specimens. These results suggest that sBWF might be an effective sample material for EGFR mutation analysis.

Conflict of interest statement

Yoshihiko Murata reports receipt of grants from Precision System Science Co, Ltd during the course of the study. None of the other authors have any disclosures.

Funding

This work was supported by a JSPS Grant-in-Aid for Encouragement of Scientists [grant number JP21H04263].

Data availability

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.