Understanding the Mechanisms of Resistance in EGFR-Positive NSCLC: From Tissue to Liquid Biopsy to Guide Treatment Strategy

Liquid biopsy has emerged as an alternative source of nucleic acids for the management of Epidermal Growth Factor Receptor (EGFR)-mutant non-Small Cell Lung Cancer (NSCLC). The use of circulating cell-free DNA (cfDNA) has been recently introduced in clinical practice, resulting in the improvement of the identification of druggable EGFR mutations for the diagnosis and monitoring of response to targeted therapy. EGFR-dependent (T790M and C797S mutations) and independent (Mesenchymal Epithelial Transition [MET] gene amplification, Kirsten Rat Sarcoma [KRAS], Phosphatidyl-Inositol 4,5-bisphosphate 3-Kinase Catalytic subunit Alpha isoform [PI3KCA], and RAF murine sarcoma viral oncogene homolog B1 [BRAF] gene mutations) mechanisms of resistance to EGFR tyrosine kinase inhibitors (TKIs) have been evaluated in plasma samples from NSCLC patients using highly sensitive methods (i.e., digital droplet PCR, Next Generation Sequencing), allowing for the switch to other therapies. Therefore, liquid biopsy is a non-invasive method able to detect the molecular dynamic changes that occur under the pressure of treatment, and to capture tumor heterogeneity more efficiently than is allowed by tissue biopsy. This review addresses how liquid biopsy may be used to guide the choice of treatment strategy in EGFR-mutant NSCLC.


Introduction
The Epidermal Growth Factor Receptor mutant (EGFR mut ) is an important molecular subtype of non-small cell lung cancer (NSCLC) and is highly sensitive to anti-EGFR tyrosine kinase inhibitors (TKIs). The EGFR mut NSCLC is a good model of the "oncogene addiction" theory, in which a specific oncogenic signaling pathway drives the transformation and proliferation of cancer cells [1][2][3]. The identification of the EGFR mutations and the related targeted agents allowed an important treatment in T790M positive plasma EGFR mut NSCLC patients, might provide additional data on this issue [27].
Interestingly, the disappearance of EGFR mutations such as T790M or the LREAT747del/T790M double mutant clone at progression to osimertinib has been demonstrated, suggesting the loss of the drug target as a mechanism of resistance [28]. Similarly, an association has been demonstrated between the loss of the T790M mutation and a shorter time to treatment discontinuation (6.1 vs. 15.2 months), suggesting the emergence of pre-existing resistant clones and a range of competing resistance mechanisms [29]. Recently, the EGFR C797S mutation has been reported to be an acquired mechanism of resistance to osimertinib [30]. Moreover, it has been reported that if C797S and T790M mutations are detected in trans, a combination of first-and third-generation EGFR TKIs may be more effective [31,32].
On the contrary, if mutations are in cis, the EGFR TKIs alone or in combination are not able to suppress the EGFR activity. In addition, when osimertinib is administered in a first-line setting and the C797S mutation develops without the presence of the T790M mutation, a re-challenge with first-generation TKIs has been proposed as an effective strategy [33]. To overcome the resistance caused by the C797S mutation, the efficacy of brigatinib in combination with osimertinib has been demonstrated in an in vitro study [34]. In the last few years, a number of studies have been published that use cfDNA to identify the presence of the T790M/C797S mutations or to monitor their amount in order to predict tumor response to treatment [29,35,36]. Overall, considering the use of different platforms (e.g., ddPCR, Real Time PCR, NGS), liquid biopsy has demonstrated a good sensitivity and specificity (>80%) in both the discovery of acquired mutations and monitoring tumor dynamics [37,38].

Using Liquid Biopsy to Track EGFR-Independent Mechanisms of Primary and Acquired Resistance
Considering the EGFR-independent mechanism of resistance, multiple primary or secondary mutations have been reported in the following genes: Kirsten rat sarcoma (KRAS), phosphatidylinositol 4,5-bisphosphate 3-kinase catalytic subunit alpha isoform (PI3KCA), human epidermal growth factor receptor 2 (HER2), and RAF murine sarcoma viral oncogene homolog B1 (BRAF) [39][40][41]. It is well-known that the occurrence of mutations in the RAS gene promotes cell proliferation and drug resistance, by-passing the blockade of EGFR signaling [39]. KRAS, a member of the RAS family, is highly mutated in NSCLC (15-30%) and is associated with a lack of response to EGFR inhibitors [42]. A recent study investigated the role of concomitant driver mutations (e.g., KRAS, NRAS, BRAF, PIK3CA) on the outcome of 133 EGFR mut NSCLC patients treated with EGFR TKIs [43]. In particular, PFS was significantly shorter in patients with concomitant driver mutations than it was in patients with EGFR mut only (7 vs. 11.3 months; p = 0.04), suggesting that patients with clonal EGFR and other sub-clonal driver mutations benefit less from treatment with EGFR TKIs [43].
Interestingly, in cases carrying both EGFR and KRAS mutations, patients with a KRAS allele fraction that was higher than that of the EGFR mut had a significantly shorter PFS (2.42 vs. 11.09 months; p = 0.0081) and a lower response rate (16.7 vs. 57.1%) [43]. Similarly, a study on cfDNA showed that 48.5% of plasma samples were positive for KRAS mutation after progression to EGFR TKIs and 39.4% of those had a KRAS and EGFR co-mutation [44]. Many other studies, on both plasma and tissue samples, detected KRAS mutations in EGFR mut patients with EGFR-TKI resistance, confirming its predictive role in resistance to treatment [29,45,46]. NRAS mutations are detected in a low number of NSCLC patients (∼1%), and Q61K, E63K, G12V, and G12R mutations were found in patients with acquired resistance to EGFR TKIs in pre-clinical and clinical studies as a mechanism of resistance to first-and third-generation EGFR TKIs [29,[47][48][49][50].
Interestingly, a case report showed that a patient positive for BRAF V600E mutation at progression to osimertinib benefitted from the combination of a BRAF inhibitor (encorafenib) and osimertinib [56].
Other well-known mechanisms of resistance include the dysregulation of Mesenchymal Epithelial Transition (MET) gene signaling, which is involved in the control of cell differentiation [57], proliferation [58,59], and angiogenesis [60,61]. Of note, high MET expression or amplification are connected with poor outcomes in patients with NSCLC [62,63]. The MET signaling pathway is linked to the EGFR network through the PI3K/Protein kinase B (Akt) and Mitogen-activated Protein Kinases (MAPK) nodes, showing mutual compensation [64]. For these reasons, MET activation is one of the potential mechanisms of resistance to EGFR TKIs in NSCLC. As a matter of fact, MET amplification is frequently reported as a mechanism of loss of efficacy of EGFR TKI therapy among EGFR mut patients [65][66][67].
Several studies report the presence of MET amplification in NSCLC treated with anti-EGFR TKIs, with or without the T790M mutation [67,68]. A study of 34 NSCLC patients evaluated the amount of EGFR mutation in cfDNA as a marker of response/resistance to osimertinib. Eight patients showed early progression during treatment, and a tumor re-biopsy revealed the presence of MET amplification in one case [69]. Similarly, the FLAURA trial (AZD9291 Versus Gefitinib or Erlotinib in Patients With Locally Advanced or Metastatic Non-small Cell Lung Cancer, NCT02296125) showed MET amplification and BRAF mutation as mechanisms of resistance in patients treated with osimertinib, opening up the possibility of future combinations to overcome resistance [70].
A study showed a plasma tissue correlation using NGS in 13 NSCLC patients with acquired resistance to osimertinib. Four patients were found to be positive for MET amplification in both plasma and tissue samples. In addition, a survival analysis showed a better PFS/overall survival (OS) in patients without MET alterations, confirming MET to be a mechanism of resistance to third-generation EGFR TKIs. To overcome this resistance, an exploratory evaluation with a treatment combination of first/third-generation EGFR TKIs and the MET inhibitor crizotinib was conducted; partial responses were clinically and radiographically achieved, and a cfDNA analysis was negative for common cancer-related mutations, suggesting the efficacy of the treatment combination [71]. Similarly, a combination of full-dose osimertinib and crizotinib was administered to two patients with emergent MET amplification in a liquid biopsy after progression to erlotinib [72,73]. A partial response was achieved without experiencing serious adverse events in one patient [72], while a dose-reduction of crizotinib due to hematological toxicity was needed in the other patient [73].
These cases demonstrate that combination therapy with osimertinib and crizotinib can be effective in patients with EGFR mut and MET amplification detected by liquid biopsy [72,73]. Moreover, cfDNA may help us to understand the molecular response to pharmacological treatment and provide information on clonal heterogeneity, showing the correlation between dynamic changes in the EGFR activating mutation (L858R) and MET amplification in treatment response [74]. In the last update of the Phase III AURA3 trial (AZD9291 Versus Platinum-Based Doublet-Chemotherapy in Locally Advanced or Metastatic Non-Small Cell Lung Cancer, NCT02151981), an analysis of the ctDNA genomic profile was also carried out in patients with the T790M mutation who progressed on osimertinib during the study, and several resistance mechanisms were observed, including MET amplification [75]. In order to overcome resistance, several agents have been developed to target MET or its ligand Hepatocyte Growth Factor (HGF) [66,72,[76][77][78][79], such as small molecules (e.g., capmatinib, tepotinib, and tivantinib) [80,81] or monoclonal antibodies (e.g., onartuzumab and emibetuzumab) [82] and anti-HGF antibodies (e.g., ficlatuzumab and rilotumumab) [83,84]. Moreover, several MET inhibitors have been investigated in combination with EGFR TKIs or cytotoxic agents in NSCLC patients who acquired resistance to TKIs due to the appearance of MET amplification [66,85,86].

Small-Cell Lung Cancer (SCLC) Transformation: Still a Challenge for Liquid Biopsy?
Transformation to SCLC is reported to be one of the mechanisms of resistance to treatment and has been observed after both first-and subsequent generation of EGFR TKIs [87,88], occurring in approximately 5-14% of patient biopsies at the time of TKI resistance [21,89]. Tumor heterogeneity in EGFR mut NSCLC has been widely described, and concurrent SCLC transformation and EGFR T790M mutation have been reported [90][91][92]. Clinical cases of EGFR mut NSCLC patients who received TKIs and developed SCLC transformation at progression have been published [91,93]. In one case, SCLC developed after treatment with gefitinib; cisplatin and etoposide were used as a second-line therapy, followed by chemotherapy and immunotherapy with amrubicin, irinotecan, and nivolumab. At this point, the primary lesion that had transformed into SCLC reconverted into an adenocarcinoma with EGFR L858R and T790M mutations. Thus, the patient was treated with osimertinib, showing a clinical remission [93].
Similarly, another study presented an EGFR mut lung adenocarcinoma, which was treated with erlotinib and chemotherapy and was later found to have transformed into SCLC. The patient was treated with cisplatin and irinotecan and then developed resistance to the therapy; the cfDNA revealed the presence of the EGFR T790M mutation, allowing for treatment with osimertinib, which resulted in a good clinical response [94].
Two case reports described the transformation into SCLC as a possible mechanism of resistance to afatinib. The first report described a case of a lung adenocarcinoma harboring EGFR exon 19 deletion that, after seven months of treatment with afatinib, progressed and showed SCLC transformation at re-biopsy with concomitant EGFR exon 19 deletion. Afatinib was discontinued and chemotherapy was administered with a cisplatin and irinotecan regimen, with no disease progression after four cycles of chemotherapy [95]. Similarly, the second case report described a switch of tumor histotype to SCLC with features of a G3 neuroendocrine carcinoma and positivity for exon 19 deletion of EGFR. Interestingly, the switch occurred during hepatic progression, which was the only site not responsive to afatinib. Thus, the patient was treated with a carboplatin plus etoposide chemotherapy, showing a complete response [96]. Transformation into SCLC has also been described as a mechanism of acquired resistance to osimertinib [97,98]. Two case reports showed that the transformation into SCLC occurred after 13-18 months of treatment with osimertinib, and a molecular analysis showed the presence of the EGFR mut exon 19 deletion and L858R without the T790M mutation. Patients were treated with a chemotherapy regimen, showing a complete or partial response [97,98].
Despite the fact that transformation into SCLC is difficult to detect by liquid biopsy, a recent publication demonstrated that ctDNA may be analyzed in terms of changes in global copy number to monitor its dynamics in patients with a histological transformation into SCLC. In particular, TP53 mutation levels change in accordance with the clinical status of the patient and their response to chemotherapy. Moreover, copy number alterations in the avian myelocytomatosis viral oncogene lung carcinoma derived homolog 1 [MYCL1], Sry-related HMG box 2 [SOX2], and SOX4 genes and a gain/loss of cancer genes were associated with transformation into SCLC and were linked to genomic instability due to TP53 mutant clones [99].

The Clinical Utility of a Liquid Biopsy in Guiding Treatment with EGFR TKIs
The development of sensitive molecular diagnostic tests has increased our knowledge of the genomic landscape of NSCLC, which shows a complex pattern of molecular abnormalities [100]. While in EGFR mut NSCLC both tumor growth and response to therapies are driven by EGFR signaling, the co-occurrence of genomic alterations has been described and used to identify several biological subsets of NSCLC patients with different outcomes to TKI treatments [101]. While, on the one hand, TKIs are effective against cell clones harboring EGFR-activating mutations, on the other hand the drug treatment is able to select cell clones carrying different molecular subtypes that are often resistant to TKI treatment [102]. In fact, tumor heterogeneity promotes different mechanisms of resistance at multiple metastatic sites [103,104].
Loss of sensitivity to EGFR TKIs may be divided into three groups, based on the mechanisms of selection by treatment: 1) mutations acquired by the target (EGFR), e.g., the T790M or C797S mutations, which reduce the activity of the drug because of a steric hindrance between the target and the drug without diminishing the kinase activity of the receptor [105,106]; 2) activation of a bypass signaling pathway, e.g., RAS mutations or MET amplification, in the presence of the EGFR mut [43,69]; and 3) histologic transformation to SCLC [107] (Figure 1A,B). In this context, the ideal approach to monitor tumor dynamics and comprehensively understanding NSCLC's heterogeneity would be a non-invasive one that is able to capture the molecular events that occur at different tumor sites. For these reasons, liquid biopsy is a useful instrument for following tumor dynamics and heterogeneity [108]. of a tissue biopsy, such as its invasive nature and its inability to represent the tumor's heterogeneity [116][117][118][119].
Although a liquid biopsy includes an analysis of the circulating free and tumor nucleic acids (DNA and RNA), exosomes, and circulating tumor cells (CTCs) in body fluids, only the analysis of EGFR mut in cfDNA has been approved for NSCLC patients [15]. Due to the low amount of DNA that can be obtained from CTCs [120], it seems that the major role of CTCs may be that of a prognostic biomarker both in lung cancer and in other solid tumors [121]. However, the lack of Federal Drug Administration (FDA) approval and the cost of CTC isolation remain major issues. Exosomes are micro-vesicles released by cells and contain a wide variety of molecules, such as DNA, RNA, proteins, and lipids, and seem to be implicated in intercellular communication and tumor-host interactions [122]. Exosomes have been described in the literature as markers for the monitoring of tumor dynamics; however, despite their importance, methods to detect and analyze exosomes require further development before they can be introduced into clinical practice [123][124][125][126][127][128]. Based on the evidence from the AURA trial, in which an analysis of cfDNA was demonstrated to be comparable to a tissue biopsy for the identification of EGFR mutational events, many other studies investigated the predictive potential of liquid biopsy and its advantages for the longitudinal monitoring of tumors, providing results with great relevance to the clinical setting [69,[109][110][111][112][113][114][115]. A liquid biopsy, obtained from a routine blood draw of 6-20 mL, can overcome most of the limitations of a tissue biopsy, such as its invasive nature and its inability to represent the tumor's heterogeneity [116][117][118][119].
Although a liquid biopsy includes an analysis of the circulating free and tumor nucleic acids (DNA and RNA), exosomes, and circulating tumor cells (CTCs) in body fluids, only the analysis of EGFR mut in cfDNA has been approved for NSCLC patients [15]. Due to the low amount of DNA that can be obtained from CTCs [120], it seems that the major role of CTCs may be that of a prognostic biomarker both in lung cancer and in other solid tumors [121]. However, the lack of Federal Drug Administration (FDA) approval and the cost of CTC isolation remain major issues. Exosomes are micro-vesicles released by cells and contain a wide variety of molecules, such as DNA, RNA, proteins, and lipids, and seem to be implicated in intercellular communication and tumor-host interactions [122]. Exosomes have been described in the literature as markers for the monitoring of tumor dynamics; however, despite their importance, methods to detect and analyze exosomes require further development before they can be introduced into clinical practice [123][124][125][126][127][128].
cfDNA has clearly demonstrated advantages over other markers that make it appropriate for use in clinical practice as a predictive biomarker to monitor response to treatment. In particular, it provides a minimally invasive approach to the early detection of disease recurrence and information about the molecular profile underpinning drug resistance [35]. However, the use of cfDNA presents some limitations and challenges, especially considering the occurrence of false-negative results depending on the following reasons: 1) shedding of cfDNA differs among different tumor types [120]; 2) detection of cfDNA depends on tumor location and volume [129]; 3) detection of cfDNA is lower in patients without progression or who are responding to therapy [130]; and 4) cfDNA shedding is variable [111,131,132].
False-negative results are related to the abovementioned factors; however, the technical limits of detection may also define the level of "false-negatives", being strictly related to the analytical platform.
While it is quite difficult to obtain a technical false-negative result when using very sensitive techniques, such as ddPCR or NGS (which have a lower limit of detection that ranges from 0.001% to 1%) [133,134], technical false-negative results may occur more frequently with less-sensitive techniques, such as the Scorpion Amplified-Refractory Mutation System (SARMS) or Peptide Nucleic Acid-Locked Nucleic Acid (PNA-LNA PCR clamp) [135].
On the other hand, the detection of somatic mutations in cfDNA released from non-cancer cells should be taken into consideration. It is known that clonal hematopoiesis of an indeterminate potential (CHIP) can lead to the development of such mutations [136,137]. Finally, in order to use liquid biopsy in an overall diagnostic/therapeutic strategy for EGFR mut lung cancers, the analysis of cfDNA should be able to identify EGFR-dependent and independent mechanisms of resistance to help us understand the relationship between baseline and resistance mutations.
A liquid biopsy holds the promise to help us understand the biology and heterogeneity of the tumor and the characteristics of drug-tolerant cells, and has the considerable advantage of short laboratory turn-around time, low cost, mini-invasiveness and the potential of being repeated over time. Such a test may help directing targeted therapies against EGFR and other druggable mutations to prevent the emergence of resistant clones. Figure 2 describes the potential applications of cfDNA in diagnosis, monitoring response and progression of disease and provides suggestions on how a liquid biopsy may be implemented in clinical practice. cfDNA has clearly demonstrated advantages over other markers that make it appropriate for use in clinical practice as a predictive biomarker to monitor response to treatment. In particular, it provides a minimally invasive approach to the early detection of disease recurrence and information about the molecular profile underpinning drug resistance [35]. However, the use of cfDNA presents some limitations and challenges, especially considering the occurrence of false-negative results depending on the following reasons: 1) shedding of cfDNA differs among different tumor types [120]; 2) detection of cfDNA depends on tumor location and volume [129]; 3) detection of cfDNA is lower in patients without progression or who are responding to therapy [130]; and 4) cfDNA shedding is variable [111,131,132].
False-negative results are related to the abovementioned factors; however, the technical limits of detection may also define the level of "false-negatives", being strictly related to the analytical platform. While it is quite difficult to obtain a technical false-negative result when using very sensitive techniques, such as ddPCR or NGS (which have a lower limit of detection that ranges from 0.001% to 1%) [133,134], technical false-negative results may occur more frequently with less-sensitive techniques, such as the Scorpion Amplified-Refractory Mutation System (SARMS) or Peptide Nucleic Acid-Locked Nucleic Acid (PNA-LNA PCR clamp) [135].
On the other hand, the detection of somatic mutations in cfDNA released from non-cancer cells should be taken into consideration. It is known that clonal hematopoiesis of an indeterminate potential (CHIP) can lead to the development of such mutations [136,137]. Finally, in order to use liquid biopsy in an overall diagnostic/therapeutic strategy for EGFR mut lung cancers, the analysis of cfDNA should be able to identify EGFR-dependent and independent mechanisms of resistance to help us understand the relationship between baseline and resistance mutations.
A liquid biopsy holds the promise to help us understand the biology and heterogeneity of the tumor and the characteristics of drug-tolerant cells, and has the considerable advantage of short laboratory turn-around time, low cost, mini-invasiveness and the potential of being repeated over time. Such a test may help directing targeted therapies against EGFR and other druggable mutations to prevent the emergence of resistant clones. Figure 2 describes the potential applications of cfDNA in diagnosis, monitoring response and progression of disease and provides suggestions on how a liquid biopsy may be implemented in clinical practice.  In conclusion, a liquid biopsy may be implemented in the clinical management of patients at diagnosis, during treatment (repeated serial liquid biopsies may identify mechanisms of resistance even after transformation into SCLC [93,94]), or during disease progression in order to select an appropriate treatment according to the therapy-dependent clonal selection.
Funding: This research received no external funding.

Conflicts of Interest:
The authors declare no conflicts of interest.