Longitudinal multi-gene panel assessment of circulating tumor DNA revealed tumor burden and molecular characteristics along treatment course of non-small cell lung cancer.

BACKGROUND: Most studies associating circulating tumor DNA (ctDNA) with outcome in lung cancer treatment were either cross-sectional or, if longitudinal, only analyzed a limited number of genes. This study evaluated the potential of utilizing ctDNA profiled by a panel of common cancer genes to monitor tumor burden and to reveal molecular characteristics of tumor along treatment course.
METHODS: Twenty Chinese non-small cell lung cancer (NSCLC) patients with serial plasma samples collected (I) before starting on either first- or second-line treatment, (II) at stable disease on treatment, and (III) upon disease progression, were analyzed for mutations in ctDNA using the PGDx 64-gene panel. Paired statistics compared mutation profiles between any two of the three time points.
RESULTS: Proportions with detectable ctDNA decreased from 65% at baseline to 35% at stable disease and rose to 80% at progression (P=0.012, between stable disease and progression); median ctDNA levels (mutated fragments per mL) were 7.8, 0, and 24.7 at the three time points, respectively (P=0.013 between baseline and progression; P=0.007 between stable disease and progression). Although plasma epidermal growth factor receptor (EGFR) mutations were commonly detected, 15% of patients had mutations other than EGFR detected during progression, such as various types of TP53 mutations.
CONCLUSIONS: ctDNA profiling in serial blood samples reflected tumor burden over time, and a multi-gene panel was more sensitive in indicating lung cancer progression on treatment than a single gene approach. The detection of additional oncogenic mutations or their disappearance suggested evolution of tumor heterogeneity along treatment course. 2020 Translational Lung Cancer Research. All rights reserved.

Introduction

Lung cancer is the leading cause of cancer death worldwide (1). For patients with advanced stage lung cancer, treatment options include conventional systemic chemotherapy or molecular-targeted therapy, and the latter is indicated with identification of therapeutic tumor targets (2). Actionable tumor targets, such as epidermal growth factor receptor (EGFR) mutations and anaplastic lymphoma kinase (ALK) translocation, are present more commonly in lung adenocarcinomas in Asians, and the presence of these actionable targets indicate that the tumour would be sensitive to specific tyrosine kinase inhibitors (TKIs) (2). Therefore, the identification of actionable mutations in tumors can guide treatment choice. Nevertheless, it is often impractical to obtain tumor tissue biopsy for molecular profiling, let alone to repeat tissue samplings over time to evaluate acquired resistance due to further genetic aberrations. The liquid biopsy approach, which offers the option of detecting molecular characteristics of circulating tumor cells or circulating tumor DNA (ctDNA) in blood (3), has the potential to revolutionize clinical care in cancer patients (2). Longitudinal follow up with liquid biopsy may be a new way to inform clinicians of subtle changes in underlying tumor characteristics or early knowledge of disease progression. This may allow for discovery of acquisition of oncogenic mutations that could be biomarkers for the next line of anti-cancer treatment or for prognostication (4,5).

ctDNA is cell-free DNA released from tumor cells into the circulation (6). Our group had previously shown the feasibility of detecting concordant EGFR mutations in plasma and tumor tissue of advanced stage lung cancer patients, and the presence or increasing plasma levels of ctDNA was associated with a worse prognosis (7). Similarly, it has been reported in other studies that somatic mutations in ctDNA of patients with non-small cell lung cancer (NSCLC) reflect molecular characteristics in tumor tissue, and ctDNA levels correlated with tumor stage and tumor burden (8-10). These observations have prompted subsequent studies to examine if ctDNA testing is useful for monitoring tumor status and tumor response to therapy over time. However, most of these previous studies tested for either a single gene (e.g., EGFR) or a few genes in ctDNA (11-13) or when a larger panel of genes was used, ctDNA was only examined at one single time point (14,15). In this proof-of-concept study, we employed a longitudinal design and a panel of 64 genes to examine whether (I) serial testing of ctDNA over time in the same patient was feasible for monitoring tumor burden, and (II) sequencing a panel of genes, as compared to the single gene approach, provided additional advantage in revealing tumor status and acquisition of new ctDNA mutations along treatment course. We present the following article in accordance with the AME publishing reporting checklist.

The authors present the following article in accordance with the MDAR reporting checklist (available at http://dx.doi.org/10.21037/tlcr-20-675).

Methods

Study population

Consecutive patients with advanced stage lung cancer attending the clinics of the Department of Medicine, Queen Mary Hospital, were recruited. The inclusion criteria were: (I) patients had diagnosis of advanced stage NSCLC and would undergo anti-cancer treatment, with either first- or second-line EGFR-TKI for patients with EGFR mutations, or first line chemotherapy for patients with EGFR wildtype tumors; and (II) patients gave informed written consents and agreed to have study follow up with plasma samples taken at three time points along their course of treatment: enrollment/baseline, at stable disease, and upon clinical progression of disease according to the RECIST 1.1 criteria. Blood samples of at least 30 mL were collected at each clinical visit time-point along the course of treatment. Plasma and serum samples were processed within 1 hour of collection and stored at –80 °C. Clinical information of recruited subjects was recorded. At the time of enrollment, each participant also completed a questionnaire pertaining to their demographic information, lifestyle factors, and medical history. The study was conducted in accordance with the Declaration of Helsinki (as revised in 2013). The study protocol was approved by the Ethics Committee of the University of Hong Kong and Hong Kong Hospital Authority Hong Kong West Cluster Institutional Review Board (IRB Reference Number UW 16-104). Informed consent was obtained from all the participants.

The first patient was recruited in April, 2016 and the last patient who completed all the follow-up with disease progression had the last blood sample collected in October 2018. Stable disease was confirmed with review of all the records for the 20 recruited subjects, with neither recent sufficient shrinkage to qualify for partial or complete response nor sufficient increase to qualify for progressive disease.

ctDNA measurements

For each patient, a plasma sample of at least 1 (median: 2, range: 1 to 4) mL, processed from EDTA tubes from each time point was shipped frozen overnight to Personal Genome Diagnostics (PGDx, Baltimore, MD, USA) for ctDNA analysis (16,17). Samples were labeled with a unique sample number, and serial samples of the same participant could not be identified by laboratory personnel. Circulating cell-free DNA was extracted and analyzed for a panel of 64 well-characterized cancer genes (PlasmaSELECTTM 64) by next generation sequencing with about 30,000× coverage (16,17). PlasmaSELECTTM 64 allows the identification of single nucleotide variants (SNVs), indels, amplifications, translocations, and microsatellite instability with sensitivity of 99.4% for SNVs and indels and a per base specificity of >99.9% (16,17). Somatic mutations in ctDNA (point mutations, insertion, deletion and substitution) were identified by the proprietary VariantDx bioinformatics pipeline that incorporates information from public databases, such as dbSNP, the 1000 Genome Project, and COSMIC, and excludes potential germline as well as hematopoietic variants (18-21). The total level of ctDNA in a sample was assessed as the total number of mutant fragments per mL of plasma. The ctDNA level of a particular gene (e.g., EGFR) was reported as the total number of mutant fragments detected in that gene per mL of plasma.

Statistical analysis

Our goal was to determine if mutation profiles in ctDNA reflected tumor burden over time and hence varied with tumor status at enrollment/baseline, stable disease, and progression. Statistical analysis focused on within-individual pairwise comparison of ctDNA profiles at any two of the three time points. Wilcoxon signed-rank test and exact binomial McNemar’s test were used for comparison of continuous variable (ctDNA quantitative level) and categorical variable (ctDNA detection), respectively. Reported P values are two-sided.

Results

Twenty patients who fulfilled inclusion criteria were recruited and their data analyzed. Their demographics, disease staging, driver mutations, treatment, tumor measurements by the longest diameter, sites of metastasis and concentration of ctDNA at baseline, stable disease and progressive disease were all listed in . A summary of these patients’ baseline characteristics are shown in . All patients were of Chinese descent and had stage IV NSCLC. The majority of patients were females, never-smokers, diagnosed with adenocarcinoma of lung, and 16 of them (16/20, 80%) were tested positive for EGFR mutations in tumor tissue. The median time interval between ctDNA samples were 84 days [inter-quartile range (IQR): 53–193] between baseline and stable disease, 145 days (IQR: 80–252) between stable disease and progression, and 259 days (IQR: 171–376) between baseline and progression. Median cell-free DNA yield per mL of plasma sample was 20.5 ng (IQR: 12.4–26.2), and the median total yield was 37.3 ng (IQR: 28.7–51).

Table 1

A summary of the baseline demographics of recruited patients and their tumor characteristics and treatment received

Patient characteristics	N (%)
Average age (± SD), years	61.3±7.9
Gender
Male	7 (35)
Female	13 (65)
Smoking status
Never-smoker	14 (70)
Former smoker	5 (25)
Missing information	1 (5)
Histology
Adenocarcinoma	18 (90)
NSCLC-NOS	2 (10)
EGFR mutation in tumor tissue at baseline
Wildtype	4 (20)
Mutant	16 (80)
Treatment
First line EGFR-TKI	14 (70)
Second line EGFR-TKI	2 (10)
Chemotherapy	4 (20)
Average number of days between blood draws (± SD)
Baseline and stable disease	114.8±78.2
Stable disease and progression	197.6±169.8

SD, standard deviation; NSCLC, non-small cell lung cancer; NOS, not otherwise specified; EGFR, epidermal growth factor receptor; TKI, tyrosine kinase inhibitor.

Table 2

A list of recruited patients with demographics, disease staging, targets, treatment received, primary tumor site and measurements (the longest diameter), metastatic sites and ctDNA concentration at baseline, stable disease and at disease progression

Patients	Gender	Age	Smoking	Stage	Target	Baseline					Stable disease				Disease progression
						Primary tumor site & size	Metastatic sites	ctDNA conc (ng/mL)	Treatment		Primary tumor site & size	Metastatic sites	ctDNA conc (ng/mL)		Primary tumor size	Metastatic sites	ctDNA conc (ng/mL)
1	M	57	Ex	IV	Del 19	RLL 2.1 cm	L SCF LN, R pleura, R pleural effusion	10.78	Gefitinib		RLL 2.0 cm	All metastatic sites improved	9.37		RLL 24 cm	Worsening of all metastatic sites	10.2
2	F	57	NS	IV	Del 19	LLL 8.5 cm	R lung, L pleura, L effusion, LN, liver, bone	10.45	Afatinib		LLL 3.9 cm	Resolution of L pleural effusion. Same for other site metastasis	15.83		LLL 13.0 cm	Worsening of all metastatic sites	21.07
3	F	64	NS	IV	Del 19	RUL 2.6 cm	R pleura, LN, liver	47.6	Afatinib		RUL 0.8 cm	All metastatic sites improved	11.63		RUL 1.2 cm	New intrapulmonary nodules and increased R pleural metastasis, same LN and liver metastasis	33.63
4	F	65	NS	IV	L858R	RLL 7.7 cm	RML, LN, bone	18.95	Afatinib		RLL tumor 1.2 cm scarring	Similar to baseline	21.9		RLL 3.8 cm	Worsening of all metastatic sites	74.45
5	F	50	NS	IV	L858R	Numerous bilateral lung nodules 1–1.5 cm	R cervical LN, bone	68.4	Gefitinib		LUL 1.3 cm, all other bilateral lung nodules reduced to <1 cm	Disappearance of R cervical LN. Same bone metastasis	14.6		LUL 2 cm	Worsening of all metastatic sites	44.3
6	F	66	NS	IV	Del 19	RUL 6 cm	R pleura, R pleural effusion, LN, bilateral adrenal, pancreatic tail, bone	25.5	Gefitinib		RUL 3.5 cm	All metastatic sites improved	13.1		RUL 5 cm	Worsening of all metastatic sites	64.5
7	F	59	NS	IV	Del 19	RLL 7.3 cm	R SCF LN, liver, bone	34.2	Afatinib		RLL 2.6 cm	All metastatic sites improved	27.3		RLL 2.7 cm	Worsening of all metastatic sites	44.2
8	F	62	NS	IV	Del 19	RLL 5.2 cm	Intrapulmonary, LN	23.3	Gefitinib		RLL 3.6 cm	Same as baseline	23.6		RLL 4.5 cm	Worsening of all metastatic sites	14.35
9	F	60	NS	IV	L858R	LUL 5.4 cm	Bilateral lung, LN, L pleura, L pleural effusion, pericardial effusion, L adrenal	6.27	Gefitinib		LUL 3.5 cm	All metastatic sites improved	26.17		LUL 3.5 cm	No lung met, increased L pleural effusion and pericardial effusion, same L adrenal met	12.33
10	M	55	NS	IV	Del 19	Multiple lung nodules both lungs	LN, bone	21.4	Gefitinib		All lung nodules decreased in sizes	All metastatic sites improved	13.2		All lung nodules increased in size	All lung and bone metastasis increased in size. New right adrenal met	38.3
11	F	64	NS	IV	Del 19	R hilar mass 1.7 cm	L pleura, L effusion	28.8	Gefitinib		Not measurable	All metastatic sites improved	30.8		Not measurable	New LLL lesions, L pleural metastasis	19.45
12	M	77	Ex	IV	Del 19	RUL 4.1 cm	LN, bone	25.1	Afatinib		RUL 3 cm	All metastatic sites improved	11.05		RUL 3.3 cm	Worsening of all metastatic sites	21.8
13	M	55	NS	IV	Del 19	LUL 2.3 cm	Multiple LN, pleural deposits, massive L pleural effusion	29.93	Erlotinib		LUL 2.1 cm	All metastatic sites improved	12.45		LUL 3 cm	Worsening of all metastatic sites	12.37
14	F	65	NS	IV	L858R	RUL 2.9 cm	L lung nodules, R pleural effusion, rib	16.8	Gefitinib		RUL 2.9 cm	Same as baseline	16.55		RUL 4 cm	Worsening of all metastatic sites	12.7
15	F	45	EX	IV	L858R	LLL 2.7 cm	L pleural deposits, LN, large L pleural effusion	15.75	Gefitinib		LLL 1.4 cm	All metastatic sites improved	20.4		LLL 2.6 cm	Worsening of all metastatic sites	10.3
16	M	66	Ex	IV	L858R	LLL 1.7 cm	LN, bone	21.45	Erlotinib		LLL 1.7 cm	Same as baseline	25.7		LLL 3.7 cm	New L pleural metastasis, same LN and bone metastasis	39.6
17	M	66	Ex	IV	WT	RUL 12.3 cm	R intrapulmonary, bone	10.63	G/Cis		RUL 9.5 cm	Same as baseline	21.43		RUL 12.2 cm	Worsening of all metastatic sites	23.3
18	F	67	NS	IV	WT	RLL 5.3 cm	Intrapulmonary, liver, bone	12.43	P/Carb		RLL 4.4 cm	Same as baseline	16.33		RLL 6.1 cm	Worsening of all metastatic sites	83.23
19	M	51	NS	IV	WT	RUL 1.8 cm	LN, bone	25.75	P/Cis		RUL 1.4 cm	Same as baseline	20.5		RUL 3.2 cm	New LUL mass, increased LN and bone metastasis, new bilateral pleural effusion	8.97
20	F	70	NS	IV	WT	RUL 4 cm	R pleural effusion, bone	22.45	P/Carb		RUL 3 cm	All metastatic sites improved	17.35		RUL 2.5 cm	Worsening of all metastatic sites	4.83

ctDNA, circulating tumor DNA; NS, non-smoker; Ex, ex-smoker; Conc, concentration; WT, wildtype; RUL, right upper lobe; RML, right middle lobe; RLL, right lower lobe; LUL, left upper lobe; LLL, left lower lobe; LN, lymph node; G/Cis, gemcitabine/cisplatin; P/Carb, pemetrexed/carboplatin; P/Cis, pemetrexed/cisplatin.

The detection frequencies of ctDNA in these 20 patients at the three time points are summarized in . Even in this cohort of patients with a high prevalence of EGFR driver mutations, more than 30% of patients tested positive for ctDNA have alterations in non-EGFR genes (38.5%, 42.9%, and 31.3% at baseline, stable disease, and progression, respectively). None of the patients had microsatellite instability, amplification, or rearrangement. In general, the proportion of patients with detectable ctDNA decreased from baseline (65%) to stable disease (35%), and rose to the highest at the time of progression (80%) (P=0.012, pairwise comparison of ctDNA detection between stable disease and progression). Similar patterns were observed for the detection of somatic EGFR and TP53 alterations over the three time points. ctDNA quantities in terms of the number of mutant fragments per mL of plasma are shown in . Median levels of ctDNA were 7.8 at baseline, zero at time of stable disease, and 24.7 at disease progression (pairwise difference: P=0.013 between baseline and progression; P=0.007 between stable disease and progression). ctDNA quantitative levels in individual patients over three time points are plotted in .

Table 3

A summary of ctDNA profiles with mutation detection in recruited lung cancer patients at three time points at baseline, stable disease and disease progression

ctDNA mutation levels	Baseline	Stable disease	Disease progression
Any detectable mutations in ctDNA, n (%)*
Yes	13 (65)	7 (35)	16 (80)†
No	7 (35)	13 (65)	4 (20)
Any EGFR mutations in ctDNA, n (%)*
Yes	8 (40)	4 (20)	11 (55)†
No	12 (60)	16 (80)	9 (45)
Any TP53 mutations in ctDNA, n (%)*
Yes	8 (40)	4 (20)	10 (50)
No	12 (60)	16 (80)	10 (50)
ctDNA levels, median [IQR]‡
Number of mutant molecules/mL	7.8 [0, 47.6]	0 [0, 11.4]	24.7 [1.5, 449.2]†§
Number of EGFR mutant molecules/mL	0 [0, 21.9]	0 [0, 0]	1.7 [0, 254.8]†§
Number of TP53 mutant molecules/mL	0 [0, 19.3]	0 [0, 5.6]	7.0 [0, 263.8]†§

*, Paired comparison of ctDNA detection frequencies between two time points by exact binomial McNemar’s test; †, P<0.05 for comparing ctDNA at stable disease vs. progression; ‡, Paired comparison of ctDNA quantities between two time points by Wilcoxon signed rank test; §, P<0.05 for comparing ctDNA at baseline vs. progression. ctDNA, circulating tumor DNA; EGFR, epidermal growth factor receptor; IQR, inter-quartile range.

Figure 1

Quantification of mutant fragments expressed as amount of ctDNA per mL of plasma in the 20 lung cancer patients included at three time points showing the changes in the quantity of ctDNA changes in the treatment time course. ctDNA, circulating tumor DNA.

The largest difference in ctDNA detection frequency was observed between stable disease and progression, within-subject results are detailed in . Using a 64-gene panel, we found that in 50% of the patients, ctDNA detection changed from negative at the time of stable disease to positive during disease progression. Contrarily, change in EGFR detection from absence to presence occurred in 35% of the patients when disease progressed (mostly in EGFR T790M and/or E746_A750del). Therefore, if a single gene of EGFR were assessed instead of using a multi-gene panel, the change in ctDNA pattern indicative of disease progression would have been missed in 15% of the patients who had ctDNA mutations in non-EGFR genes.

Table 4

A table showing the within-subject detection of ctDNA and comparison of mutation pattern changes at the times of stable disease and disease progression, with highlights of changes in mutations detected in EGFR and TP53 genes at disease progression

Pattern of ctDNA mutation changed between stable disease and progression				Number of individuals (%) with the corresponding ctDNA pattern change at disease progression
ctDNA pattern	Stable disease	Progression		Any mutations in 64 genes	Mutations in EGFR	Mutations in TP53
A	No	No		3 (15)	9 (45)	9 (45)
B	No	Yes		10 (50)	7 (35)	7 (35)
C	Yes	No		1 (5)	0 (0)	1 (5)
D	Yes	Yes		6 (30)	4 (20)	3 (15)

ctDNA, circulating tumor DNA; EGFR, epidermal growth factor receptor.

The detection of additional ctDNA mutations or their disappearance during treatment course in the 20 patients is detailed in . Based on their tumors being EGFR mutants or wildtype, and the serial changes in ctDNA mutation patterns, recruited patients could be broadly divided into four groups:

Table 5

A table listing the serial changes of ctDNA mutations detected along treatment courses in all the 20 recruited subjects

Patients	Tissue EGFR mutation at baseline	Tissue EGFR T790M detection at re-biopsy (disease progression)	ctDNA mutation detected
			At baseline (%)	At stable disease (%)	Upon disease progression (%)
Group 1: disease progression with EGFR T790M (on first line EGFR-TKI: either gefitinib/erlotinib/afatinib)
1	EGFR exon 19 E746_A750del	Not done	EGFR exon 19 E746_A750del (2.98)	No mutation detected	EGFR exon 19 E746_A750del (3.68)
			–		EGFR T790M (2.26)
			TP53 G245V (4.25)		TP53 G245V (3.45)
2	EGFR exon 19 E746_A750del	Not done	–	EGFR exon 19 E746_A750del (22.51)	EGFR exon 19 E746_A750del (29.07)
				EGFR T790M (12.23)	EGFR T790M (14.61)
			RNF43 E380D (0.34)	–	–
			ERBB2 D1048Y (0.42)	–	–
			–	BRCA1 K1290I (0.23)	BRCA2 C1573 (0.49)
				MTOR R1987Q (0.11)	–
				TP53 R249S (0.07)	–
				–	MET Y234F (0.21)
					MET Y291F (0.25)
					MYC Y192N (0.13)
3	EGFR exon 19 E746_A750del	EGFR exon 19 E746_A750del	EGFR exon 19 E746_A750del (1.7)	No mutation detected	EGFR exon 19 E746_A750del (15.83)
		EGFR T790M	–		EGFR T790M (0.08)
		–	TP53 R248W (1.84)		TP53 R248W (11.93)
			–		MET L674F (0.11)
			–		–
4	EGFR L858R	Not done	EGFR L858R (11.19)	No mutation detected	EGFR L858R (16.05)
			–		EGFR T790M (4.58)
			TP53 M246del (10.84)		TP53 M246del (18.69)
			RET E732K (0.24)		–
			–		ALK W614 (0.12)
			–		AR E323K (0.15)
5	EGFR L858R	Not done	No mutation detected	EGFR L858R (9.2)	EGFR L858R (18.95)
				–	EGFR T790M (0.14)
Group 2: disease progression appeared independent of EGFR T790M (on first line EGFR-TKI: either gefitinib/erlotinib/afatinib)
6	EGFR exon 19 E746_A750del	Not done		No mutation detected	EGFR exon 19 E746_A750del (0.14)
			KRAS D33H (0.38)		KRAS D33H (0.29)
			PTCH1 E1242K (0.22)		PTCH1 E1242K (0.15)
7	EGFR exon 19 E746_A750del	Not done	EGFR exon 19 E746_A750del (13.2)	No mutation detected	No mutation detected
			TP53 V272L (9.38)
8	EGFR exon 19 E746_A750del	Not done	No mutation detected	RNF43 Q543* (0.43)	No mutation detected
9	EGFR exon 19 E746_A750del	Not done	No mutation detected	No mutation detected	CD274 M267V (0.49)
10, 11	EGFR exon 19 E746_A750del	Not done	No mutation detected	No mutation detected	No mutation detected
12	EGFR exon 19 E746Vfs*16	Not done	No mutation detected	No mutation detected	EGFR exon 19 deletion E746Vfs*16 (31.04)
	EGFR exon 19 insertion L747Ffs*20				EGFR exon 19 insertion L747Ffs*20 (33.28)
	–				BRCA1 D1065N (0.71)
	–				TP53 T155L (15.42)
	–				TP53 F109L (13.3)
13	EGFR exon 19 L747_E749del	Not detected	EGFR exon 19 L747_E749del (0.32)	No mutation detected	–
			EGFR exon 19 E749Q (0.38)		–
			EGFR exon 19 A750P (0.38)		–
			RB1 R251* (0.62)		–
			–		TP53 M237Cfs10 (0.58)
14	EGFR L858R	Not done	EGFR L858R (1.43)	No mutation detected	EGFR L858R (0.21)
Group 3: EGFR T790M on osimertinib [on second line EGFR-TKI (osimertinib) after first line EGFR-TKI failed]
15	EGFR L858R	EGFR L858R; EGFR T790M (re-biopsy after first line EGFR-TKI failure)	EGFR L858R (8.06)	EGFR L858R (0.89)	EGFR L858R (0.31)
			EGFR T790M (1.92)	EGFR T790M (0.17)	–
			TP53 L35* (2.23)	–	–
16	EGFR L858R	EGFR L858R; EGFR T790M (re-biopsy after first line EGFR-TKI failure)	EGFR L858R (6.38)	EGFR L858R (5.78)	EGFR L858R (4.79)
			EGFR T790M (0.66)	EGFR T790M (0.67)	EGFR T790M (0.80)
Group 4: EGFR wildtype [on first line platinum-based chemotherapy (pemetrexed-platinum)]
17	EGFR WT	Not done	CDKN2A L63P (2.71)	–	CDKN2A L63P (5.27)
			TP53 E294* (3.18)	TP53 E294* (1.37)	TP53 E294* (5.07)
			–	TP53 Y205D (0.22)	TP53 Y205D (1.04)
			–	TP53 H193R (0.25)	TP53 H193R (0.17)
			–	–	TP53 C238Y (0.12)
18	EGFR WT	Not done	TP53 R267G (1.58)	TP53 R267G (1.72)	TP53 R267G (22.96)
19	EGFR WT	Not done	No mutation detected	No mutation detected	TP53 R158H (0.24)
					TP53 R175H (0.15)
20	EGFR WT	Not done	TP53 R248W (0.12)	No mutation detected	No mutation found

The 20 subjects were divided into Group 1: subjects who received first or second generation EGFR-TKI but treatment failed with disease progression and development of EGFR T790M mutations; Group 2: subjects who received first or second generation EGFR-TKI but treatment failed with disease progression but no development of EGFR T790M mutations; Group 3: subjects recruited after they have failed primary or secondary EGFR-TKI and have their ctDNA followed during the course of osimertinib; and Group 4: EGFR wildtype subjects who received systemic chemotherapy. *, stop codon. ctDNA, circulating tumor DNA; EGFR, epidermal growth factor receptor; TKI, tyrosine kinase inhibitor; WT, wildtype.

❖ Group 1 (patients 1–5): patients had an EGFR mutation, but not EGFR T790M, in tumor tissue at diagnosis, were treated with first line EGFR-TKI (gefitinib, erlotinib or afatinib) at baseline and had mainly EGFR T790M mutation detected at disease progression;

❖ Group 2 (patients 6–14): these patients were similar to Group 1 in terms of having had a drug-sensitive EGFR mutation in tumor tissue and were treated with EGFR-TKI, but they never had EGFR T790M mutation detected along their treatment course;

❖ Group 3 (patients 15 and 16): patients had sensitizing EGFR mutations and disease progression after first line EGFR-TKI, switched to second line EGFR-TKI (osimertinib) with baseline blood taken when they started osimertinib, and followed with stable disease then further disease progression on osimertinib;

❖ Group 4 (patients 17–20): EGFR wildtype subjects on first line platinum-based chemotherapy (pemetrexed-platinum) at baseline.

Among the 14 EGFR mutant subjects on first line EGFR-TKI treatment, 5 (5/14, 35.7%) (Group 1) had detection of EGFR T790M mutation at disease progression (). In these five subjects, other oncogenic mutations were also found at the time of disease progression, such as MYC and MET mutations. Nine of the 14 patients (64.3%) progressed without EGFR T790M mutation detected, but new ctDNA mutations were also found in various genes such as BRCA1, CD274 and TP53.

Two EGFR mutant patients (Group 3) showed persistent detection of EGFR T790M at osimertinib baseline through stable disease on treatment. At further disease progression on osimertinib, one of them (patient 15) lost the EGFR T790M in ctDNA while the other one (patient 16) showed persistent EGFR T790M mutation detection at further disease progression. Neither patient had new mutations detected at time of progression.

In some patients with either EGFR mutations or wildtype EGFR in tumor tissue, different mutations in the TP53 gene were detectable in ctDNA at different time-points. Nevertheless, TP53 mutations appeared to be more prevalent and were more likely to accumulate diverse types of point mutation in the wildtype EGFR tumors when the respective patients were treated with chemotherapy (Group 4).

Discussion

In this proof-of-concept longitudinal study using a multi-gene panel, we showed that ctDNA detection and quantity reflected tumor status over time—after initiation of treatment, levels decreased during stable disease and then increased again when disease progressed. Multi-gene ctDNA assessment also revealed mutations acquired and accumulated during anti-cancer treatment and reflected the underlying heterogeneous tumor biological characteristics.

Previous cross-sectional studies had found ctDNA levels to be positively associated with tumor stage, tumor burden, or tumor volume of lung cancer (8,10,22). In a study using CAPP-Seq for 139 genes, ctDNA detection increased from 50% in stage I NSCLC tumors to 100% among stage II–IV patients, and ctDNA levels correlated with tumor volume as measured by imaging (R2=0.89) (8). Because of the cross-sectional nature of these studies, they were not able to evaluate whether ctDNA changed with the evolution of tumor characteristics over time or with course of treatment. On the other hand, longitudinal studies on the utility of ctDNA in monitoring tumor status were mostly done with testing of EGFR gene only. In general, these studies showed that increased levels of plasma-mutant EGFR and detection of specific EGFR mutations in ctDNA at baseline were associated with poorer response to EGFR-TKI and poor progression-free and overall survivals (9,12,23). In serial samples, the levels of ctDNA containing EGFR mutations usually declined after the start of treatment (11,24). Only a few longitudinal studies had used a multi-gene panel for ctDNA analysis (25,26). In a study that utilized a similar gene panel and sequencing method as our study, it was reported that ctDNA reflected tumor load over time. Among 12 patients who had radiographic response to target therapy, their mutant allele concentrations in ctDNA decreased from an average of 10.8% at baseline to 0.2% at a median time of 19 days after treatment initiation. Contrarily, among five patients who did not respond to treatment and had radiographic progressive disease, they showed a relatively high average level of mutant allele fraction at baseline (14.2%) and a modest variation after initiation of therapy (11.8%) were found (26). Our longitudinal study, nevertheless, was more advantageous to demonstrate the utility of ctDNA in monitoring tumor status than previous studies—all our 20 subjects had completed follow-up from baseline, through stable disease, to disease progression, and hence our data showed within-subject variations in ctDNA pattern over the entire course of treatment.

To date, available data support the potential utility of ctDNA as a tumor biomarker for monitoring treatment response and tumor burden over time, particularly when repeat biopsy is not feasible. One question of interest is whether testing for mutations in multiple genes is advantageous over a single gene approach. Our data showed that even in NSCLC patients with a high prevalence of EGFR mutations, testing for multiple genes in ctDNA could help to identify additional patients who had disease progression. Not all mutations detected in ctDNA are actionable or have clinical significance, but changes in ctDNA pattern and quantitative levels informed by a large gene panel are more sensitive in reflecting tumor load than a single gene analysis. Additional benefits of a multi-gene panel include the potential of identifying actionable targets when more drugs are available and revealing therapy resistance mechanisms (9). EGFR mutant patients (Group 3) showed persistent detection of EGFR T790M at Osimertinib baseline through stable disease on treatment. Neither patient had new mutations detected at the time of progression. This observation may imply that tumors bearing acquired EGFR T790M could continue to evolve and may change acquired resistance pathways, through other mechanisms like epithelial-mesenchymal transition (EMT) rather than tumor mutational changes (27).

Different types of TP53 mutations were detectable at different time-points in all the 20 subjects (overall 10/20, 50%), although TP53 mutations appeared to be more prevalent in EGFR wildtype subjects on chemotherapy than in patients treated with EGFR-TKI. TP53 point mutations of various types have been reported to result in loss of tumor suppressor function and promote tumor growth, hence disease progression while on treatment (28,29). The emergence of serial changes in the various different types of TP53 mutations as well as other non-EGFR mutations such as BRCA1, BRCA2, MYC or MET mutations along treatment course probably reflected the evolution of tumor heterogeneity. The contributions of these non-EGFR mutations towards development of acquired drug resistance, regardless of whether the tumor initially carried sensitizing EGFR mutations or not, deserves further investigation (30).

Our study is one of the few in lung cancer literature that assessed serial measurements of ctDNA for a panel of genes. The results provided support for application of serial monitoring of ctDNA mutations in lung cancer patients, especially those on EGFR targeted therapies, and such serial monitoring may allow for discovery of new acquired mutations that could be of therapeutic or prognostication importance. However, this study comprised of a small sample size and lacked a time-to-event outcome, as the level of ctDNA might have started to rise well before clinical progression based on RECIST criteria. As such, we were not able to estimate how much sooner ctDNA could detect disease progression than standard-of-care radiologic imaging, as suggested by some small-scale studies (12,25,26). Future studies need to determine the frequency of ctDNA testing, define the extent of increase in ctDNA, in terms of total quantitative levels or allele fractions of specific genes, to define disease progression, and identify ctDNA biomarkers for early detection of disease progression.

Conclusions

In summary, our results suggest that ctDNA levels in serial blood samples reflect tumor burden over time, and a multi-gene panel would be a more sensitive way of detecting lung cancer disease progression or biomarkers for drug resistance than a single gene approach. The detection of additional oncogenic mutations, some of which are cumulative, or their disappearance in plasma during treatment course, reflects underlying evolution of tumor heterogeneity.

The article’s supplementary files as