Frontiers of ctDNA, Targeted Therapies, and Immunotherapy in Non-Small-Cell Lung Cancer

Chennianci Zhu; Weihao Zhuang; Limin Chen; Wenyu Yang; Wen-Bin Ou


Transl Lung Cancer Res. 2020;9(1):111-138. 

In This Article

ctDNA Aids Diagnosis and Treatment Tracking

Tissue biopsy plays an important role in analyzing tumor properties but remains invasive and may cause harm to patients. In addition, tissue biopsy is not always feasible, and it cannot fully account for the temporal and spatial heterogeneity of cancer cells.[25] Liquid biopsy, on the other hand, is noninvasive and provides a dynamic view of tumors with overall heterogeneity.[26] Because liquid biopsy is highly sensitive to ctDNA, it can be utilized as an early detection method for cancer and remains useful during treatment procedures to evaluate treatment response.[27,28]


ctDNA molecules produced by tumors have certain properties not possessed by normal cell-free DNAs (cfDNAs).[27] A wide variation in length is the most distinctive feature of ctDNAs. ctDNAs are typically highly fragmented; thus, their size varies between about 70 and 200 bp. Some studies reported that the most frequently observed ctDNAs are 180 bp in size with a classic ladder pattern that correlates with the caspase-activated DNase digestion of chromatin in apoptotic tumor cells,[29] and ctDNA fragments larger than 10 kb might originate from necrosis.[30] Two recent studies pointed out that the modal size of ctDNA for many cancer types is about 166 bp, which is the length of DNA wrapped around the chromatosome, and the result of protection from enzymatic degradation due to histone binding to nuclear DNA.[31,32] Furthermore, an enrichment in mutated ctDNA fragments longer than 167 bp were also reported, notably around 250–320 bp.[32] However, shorter length of ctDNA fragments compared to wild-type allele was found in human hepatocellular carcinoma (134–144 vs. 167 bp), melanoma (132–145 vs. 165 bp), lung cancer, renal cell carcinoma, and colorectal cancer.[33–35] In ctDNA analysis, an additional DNA fragment length cutoff could be applied for mutation identification to increase overall sensitivity.[36] It is likely that due to different origins, various fragment sizes are generated. The mechanisms behind ctDNA fragmentomics still require further research.

Because of the low concentration of ctDNA in blood, it is difficult to detect ctDNA using nucleic acid detection methods suitable for the qualitative detection of a large amount of DNA. However, with the development of highly sensitive DNA detection techniques with a low false-positive rate, such as quantitative polymerase chain reaction (qPCR), digital polymerase chain reaction (dPCR), next-generation sequencing (NGS), NGS-based techniques and other methods, the detection of ctDNAs at low concentrations became possible.[37] Mutations in ctDNA molecules from tumor tissues can clearly reflect tumoral mutations and tumor heterogeneity,[28,38] thus enabling the detection of tumors by detecting ctDNA mutations.

qPCR has been used to detect cfDNA from blood taken from lung cancer patients.[39–41] A significant correlation between NSCLC severity and the levels of cfDNA was observed, and high cfDNA concentrations indicated more severe disease.[41] cfDNA detection is currently more often used as a supplementary measure for diagnosis and to monitor NSCLC patients.

Detecting EGFR Mutations by ctDNA

EGFR mutations can cause resistance to EGFR-TKIs in NSCLC patients.[38,42] Sensitizing EGFR mutations are present in 14% to 38% of NSCLC patients depending on the location of the mutation and the ethnicity of the patient.[43] In recent years, commonly used techniques to detect ctDNA EGFR mutations have been mainly the amplification refractory mutation system (ARMS) approach, droplet digital PCR (ddPCR) and NGS-based methods.[44–46] These techniques differ in their sensitivity, accuracy, specificity and coverage in detecting EGFR mutations.[37]

ARMS is based on qPCR that uses specific probes to identify EGFR mutant sequences. As ARMS has an overall sensitivity of 0.1–1%,[47,48] the abundance of the mutated DNA must be above this threshold to depress the false-negative rate. However, the false-negative rate in plasma samples, which is approximately 30%, is still relatively high compared with that in tumor tissue. The sensitivity and specificity of plasma EGFR mutation detection are 92% and 100%, respectively, compared with those of tumor EGFR mutation detection.[49] There are two subcategories of ARMS: ADx-ARMS and cobas-ARMS. Cobas-ARMS exhibits superior sensitivity in detecting the EGFR T790M mutation (0.1%) and sufficient sensitivity (90%).[44] The obvious advantage of ARMS is its use of simple workflows and consequent rapid turnaround time. Furthermore, ARMS is normally performed at initial tumor biopsy and therefore offers first-hand information about tumors.[44] However, flaws in this method exist as cobas-ARMS can detect only EGFR variants, and studies have shown that ARMS cannot replace tumor tissue biopsy to detect EGFR mutation status.[50] In addition, the sensitivity of ARMS is unmatched with that of ddPCR, which is approximately 0.01% and thus ten times greater than that of ARMS.[48] The increased precision and accuracy of ddPCR used to detect plasma T790M status are also advantages of ddPCR over ARMS.

ddPCR, which exhibits an unparalleled sensitivity of 0.001–0.4%, can detect EGFR, KRAS, ALK and other mutations with greater precision than other methods.[51–53] Due to the small amount of ctDNA present in the serum of lung cancer patients, a highly sensitive method such as ddPCR is essential. ddPCR costs less and requires shorter turnaround time; besides, ddPCR is highly accessible without complex bioinformatics support for data analysis. And it was shown by studies[54–56] that the sensitivity of ddPCR for the detection of EGFR or KRAS mutations in lung cancer can be as sensitive and highly concordant as NGS, sometimes even more. However, ddPCR requires specific primers, and we can detect only one locus per reaction, which limits its use in multiplex tests.

NGS-based methods are the most accepted and recognized method used to detect EGFR mutations with both high and low allele frequencies, especially those with low allele frequency. In a study performed by Xu et al.[44] cobas-ARMS, ddPCR and NGS all detected mutations with low allele frequencies well, and ddPCR and NGS also yielded excellent positive coincidence rates. Overall, the NGS platform demonstrates sensitivity equivalent to that of ddPCR, the most sensitive platform for EGFR mutational profiling. A sensitivity of 100% (30/30) for amplicon-based NGS was achieved compared to 87% (26/30) for ddPCR in the detection of EGFR activating mutations in advanced NSCLC.[57] Similar studies generated data of 100% (18/18) NGS vs. 94% (17/18) ddPCR,[54] and 83% (30/36) NGS vs. 69% (25/36) ddPCR.[58] What makes NGS-based techniques truly stand out is their ability to detect a wide range of mutations in EGFR and other clinically important genes, such as ALK and RAS, which ddPCR is unable to do because it can detect one only locus in each reaction. Researchers believe that ctDNA can be used to confirm all EGFR primary driver mutations, along with additional mutations in other disease-relevant genes. These advantages have made NGS relatively costly compared to other methods. One setback of NGS is its high false-positive rate. It is recommended that the results of both NGS and ddPCR are combined or repeatedly compared with an NGS library to validate the detected mutation. For example, the combination of NGS to discover EGFR primary mutations and other mutations with the specificity of the ddPCR assay was shown to be the optimal workflow for T790M analysis.[37] In summary, NGS analysis is an ideal method for detecting EGFR primary mutations that avoids the possible harm caused by tumor biopsy. Furthermore, ddPCR can be used as an additional reassurance during NGS analysis, especially when detecting mutations with low allele frequencies (<0.1%). When the amount of cfDNA is limited, NGS is a better option than dPCR and qPCR because it requires only 10 ng of cfDNA per analysis, while dPCR and qPCR require 15 and 40 ng, respectively, according to Bartels et al..[37]

In recent years, different branches of NGS-based methods such as amplicon-based NGS[54,57,59] and ultra-deep NGS[55,60] have shown transcendent sensitivity and specificity in the detection of mutations in NSCLC. Amplicon-based NGS demonstrated high accuracy for point mutations and indels, and it can also detect chromosomal rearrangement and fusion genes in ctDNA. This barcoded NGS in which PCR was used to enrich target genes provided excellent sensitivity by limiting sequencing artifacts compared to hybrid capture-based NGS, and further lowed the required amount of ctDNA sample.[61] A shallow sequencing depth was a problem for NGS method, however, ultra-deep NGS assays can overcome this and achieved a 75% sensitivity in detecting lung cancer oncogenic driver mutations, according to Li et al..[55]

ctDNA Assists Clinical Decision Making

EGFR-activating mutations are the most important mutations among the many driver oncogenes that play critical roles in NSCLC.[62] Initial patient responses to EGFR-TKI treatment are often very positive. As ctDNA is a convenient and precise parameter collected in real-time, it can show sudden outbreaks that appears as spikes in data that correlate with the apoptosis of tumor cells within 1 week after first-time TKI use.[63,64] Riediger et al.[65] reported an 11-fold acute increase in ctDNA levels only 26 hours after therapy started. After the initial outbreak, the amount of ctDNA continues to decrease over the following weeks. However, after a partial response (PR) or even complete response (CR) for a short time, the efficiency of TKIs decreases, which may lead to a stable disease (SD) condition and in some cases prevent TKIs from countering the effect of the mutation, finally eliminating its efficacy. The state in which specific TKIs no longer function is called a resistant state. Median progression-free survival (PFS) is 10–16 months among patients treated with EGFR-TKIs, and almost half of acquired TKI resistance is caused by secondary EGFR T790M mutation.[5,66]

Tracking ctDNA Dynamics Helps Determine Disease State. Tracking changes in ctDNA levels is informative for determining a patient's disease state and capturing dynamic changes during TKI treatment.[67] Changes in ctDNA levels also reflect the different sensitivities of heterogeneous cancer cell clones because of the different responses shown by EGFR ctDNA to EGFR-TKIs. All of the advantages of ctDNA monitoring are crucial for clinical intervention. ctDNA levels show a good correlation with radiologic CR/PR but exhibit a poor correlation with patients in SD/progressive disease (PD) states.[68] Cancer cell clones in radiologic SD/PD patients with good ctDNA responses may be a mixture of cancer cells with wild-type EGFR and EGFR with activating mutations. Poor ctDNA responses indicate that the tumor may carry de novo resistance to EGFR-TKIs.[69] Notably, early (often within one week) ctDNA responses are not a good predictor of radiologic response because the sudden increase in ctDNA results from rapid cancer cell apoptosis.[63]

Differences in ctDNA Levels Before and After Treatment. Husain et al. observed peaks in the level of ctDNA, followed by median decreases of 86% and 81% from weeks one to two, respectively, indicating EGFR L858R and exon 19 del.[64] Because patients who do not undergo therapy cannot exhibit patterns such as temporal spikes in their ctDNA profiles, instantaneous changes in ctDNA profiles before and after treatment reveal a significant quantitative rise in the number of EGFR copies after therapy, which reflects cell apoptosis within days of exposure to the drug.

In a drug-sensitive state, the continuous decline in ctDNA levels in T790M-positive patients continues for a median of 6 months.[70] Urine ctDNA levels decline to a greater extent than widely used plasma ctDNA levels. Urine ctDNA detection also has a remarkable 88% overall concordance rate with the detection of ctDNA in tissue samples. The concordance rate with plasma ctDNA detection is 98%. This shows the feasibility of replacing plasma ctDNA detection with urinal ctDNA detection, which exhibits a similar sensitivity and improved convenience.[70]