Update on molecular pathology and role of liquid biopsy in nonsmall cell lung cancer

Abstract

Personalised medicine, an essential component of modern thoracic oncology, has been evolving continuously ever since the discovery of the epidermal growth factor receptor and its tyrosine kinase inhibitors. Today, screening for driver alterations in patients with advanced lung adenocarcinoma as well as those with squamous cell carcinoma and no/little history of smoking is mandatory. Multiplex molecular platforms are preferred to sequential molecular testing since they are less time- and tissue-consuming. In this review, we present the latest updates on the nine most common actionable driver alterations in nonsmall cell lung cancer. Liquid biopsy, a simple noninvasive technique that uses different analytes, mostly circulating tumour DNA, is an appealing tool that is used in thoracic oncology to identify driver alterations including resistance mutations. Additional roles are being evaluated in clinical trials and include monitoring the response to treatment, screening for lung cancer in high-risk patients and early detection of relapse in the adjuvant setting. In addition, liquid biopsy is being tested in immune-oncology as a prognostic, predictive and pharmacodynamic tool. The major limitation of plasma-based assays remains their low sensitivity when compared to tissue-based assays. Ensuring the clinical validity and utility of liquid biopsy will definitely optimise cancer care.

Short abstract

Screening for driver alterations in patients with advanced lung adenocarcinoma and those with squamous cell carcinoma and no/little smoking history is mandatory. Liquid biopsy is evolving constantly and will definitely improve outcomes in thoracic oncology. https://bit.ly/2XjuQrD

Nonsmall cell lung cancer in the era of precision medicine

Lung carcinogenesis is the outcome of an imbalance between oncogenes and tumour suppressor genes leading to a malignant mass of cells that acquire the phenotypic “hallmarks of cancer” [1]. The discovery of oncogene addiction led to the development of targeted therapies against driver mutations that yielded impressive response rates and survival outcomes in advanced nonsmall cell lung cancer (NSCLC), a disease with a dismal prognosis until a few years ago [2]. Our understanding of the molecular mechanisms of this disease has improved significantly with the latest advances in molecular testing, mainly with the advent of next-generation sequencing (NGS). Our ultimate goals are to classify lung tumours into groups that integrate clinical, genomic, epigenomic, transcriptomic, proteomic and metabolomic data and to determine the best diagnostic and treatment methods for each group [3]. Individualised genomic-driven management decisions have become a standard of care in NSCLC ever since the discovery of the epidermal growth factor receptor (EGFR) driver oncogene and its tyrosine kinase inhibitors (TKIs) [4]. Today, there are more than nine identified actionable driver alterations and molecular screening is mandatory in patients with nonsquamous cell carcinoma and those with squamous cell carcinoma (SCC) who are never- or light smokers. Simultaneous molecular testing (multiplex panels) is recommended rather than sequential testing with an objective to save both time and tissue. Our understanding of the molecular “code” of NSCLC at the time of initial diagnosis, progression and resistance to therapy is evolving rapidly. We have more diagnostic, prognostic and theragnostic molecular biomarkers than ever before and the time when every cancer patient undergoes cost-effective comprehensive molecular screening that dictates treatment is imminent [5, 6]. Moreover, the histological phenotype of NSCLC should not be underestimated, as it can predict the presence/absence of driver alterations and thus guide molecular testing, especially in the case of limited resources. In addition to having the worst prognosis among the adenocarcinoma histological subtypes, micropapillary and solid adenocarcinoma subtypes have the lowest frequency of targetable driver alterations, except for BRAF-V600E mutations, whereas sensitising EGFR mutations are most frequently found in the lepidic subtype [7]. In this review, we provide the latest updates on common oncogenic drivers in NSCLC and we focus on the latest applications of liquid biopsy in thoracic oncology.

Oncogenic driver alterations in NSCLC: where do we stand?

KRAS

KRAS mutations are the most common driver alterations in NSCLC and are almost exclusive to adenocarcinoma (ADC) [8]. They affect 25% of patients and are associated with smoking and shorter survival [9, 10]. Since they do not primarily overlap with EGFR, ALK, ROS1 or BRAF, they can be used to select patients who might not benefit from further molecular testing. Not all KRAS mutations are oncogenic drivers. The most common oncogenic alterations involve exons 2 and 3, with subsequent substitutions in codons 12, 13 or 61 [11]. KRAS p.G12C, the most frequent substitution detected in 41% of KRAS-mutant NSCLC, is associated with smoking, while the KRAS p.G12D mutation is usually seen in non/former light smokers [12]. KRAS-addicted tumours have a well-differentiated epithelial gene expression phenotype, whereas KRAS-independent tumours with KRAS mutation have an epithelial–mesenchymal transition (EMT) phenotype. Transforming growth factor-β1 is a cytokine that promotes EMT, thus decreasing KRAS oncogene addiction [13, 14]. Three co-mutational clusters of KRAS-mutant tumours with distinct biological behaviours were identified. TP53-mutant tumours (39% of co-mutated tumours) have an immune-rich microenvironment with high programmed death ligand 1 (PD-L1) expression, increased tumour infiltrating lymphocytes (TILs) and immunoediting. In contrast, STK11-inactivated tumours (20%) have a pauci-immune microenvironment with low PD-L1 expression and few TILs as well as KEAP1 mutational inactivation. Finally, the CDKN2A/B-inactivated cluster is associated with a higher incidence of TTF1-negative ADC, mucinous histology and decreased mTORC1 signalling [15, 16]. In addition, KRAS mutations are detected in other oncogene-addicted tumours (MET, EGFR, BRAF and ALK) when they develop resistance to TKIs [17]. Pre-clinical and clinical studies yielded conflicting results regarding the predictive role of KRAS mutations in chemotherapy outcomes [18, 19]. Alternatively, according to retrospective data from ImmunoTarget, a global multicentre registry, KRAS mutations were associated with an increased response to immunotherapy in comparison to other drivers [20]. KRAS-mutant NSCLC usually harbours a higher tumour mutational burden (TMB), thus explaining the increased response to immunotherapy [21]. In addition, as mentioned previously, TP53 co-mutated tumours are immunogenic and show a remarkable sensitivity to immune checkpoint inhibitors (ICIs), whereas STK11 co-mutated “cold” tumours have impaired immunosurveillance and primary resistance to these agents [21]. Despite the efforts deployed, KRAS remains a frustrating therapeutic target due to the absence of a binding pocket in the KRAS protein and its high affinity towards the abundance of cellular guanosine triphosphate [22]. Selective novel covalent direct inhibitors of the KRAS G12C mutation such as sotorasib (AMG510) and adagrasib (MRTX849) are currently being investigated and results are encouraging [22]. In contrast to former nonspecific inhibitors of KRAS, selective KRAS G12C inhibitors seem to be well tolerated, since the G12C substitution does not exist in normal tissues and thus there is no risk of off-target toxicity [12]. Results from the phase II cohort of the CodeBreak 100 study of sotorasib (AMG510, 960 mg orally, once daily) in 126 pre-treated NSCLC patients with KRAS G12C mutations showed a confirmed objective response rate (ORR) of 37.1%, a disease-control rate (DCR) of 80.6%, a median duration of response (mDOR) of 10.0 months and a median progression-free survival (mPFS) of 6.8 months. Most treatment-related adverse events were mild to moderate (grade 1 or 2) including diarrhoea (31%), nausea (19%) and increased alanine and aspartate aminotransferases (15.1% each) [23]. Sotorasib was granted breakthrough therapy designation by the US Food and Drug Administration (FDA) and China for patients with pre-treated KRAS G12C-mutated NSCLC. CodeBreak 200, a phase III randomised study comparing sotorasib to docetaxel in patients with KRAS G12C-mutated NSCLC is currently underway (NCT04303780) [24]. Adagrasib achieved an ORR of 45.1%, in 51 pre-treated KRAS G12C-mutant lung cancer patients in the pooled phase I/Ib and II KRYSTAL-1 study [25]. JNJ-746999157 is another direct G12C inhibitor that is currently being tested in a phase I trial (NCT04006301) [26]. Unlike immune checkpoint inhibitors, KRAS inhibitors are active in patients with STK11 co-mutated tumours [23, 25]. In addition, they may “heat” KRAS-mutated cold tumours by increasing T-cell infiltration, antigen presentation and interferon response, while decreasing monocyte and neutrophil recruitment [27]. Adaptive resistance to KRAS inhibitors seems to result from the increased synthesis, via EGFR and aurora kinase A signalling, of a new KRAS G12C protein that is less susceptible to inhibition [28]. Novel combinations with other agents may help overcome resistance to KRAS inhibitors.

EGFR

Activating somatic mutations in the tyrosine-kinase (TK) domain of EGFR were first described in 2004 [29]. They mostly affect exons 18 to 21 of the TK domain and consist mainly of short in-frame deletions of exon 19 clustered around the amino-acid residues 747–750 (45% of cases) in addition to a specific exon 21 missense mutation (L848R) (40% of cases) [29, 30]. Other less-common sensitising EGFR mutations (10% of cases) include exon 19 insertions, p.L861Q, p.G719X and p.S768I [31]. Sensitising EGFR mutations occur in 30–40% and 10–12% of Asian and non-Asian patients, respectively [32, 33]. They are frequently associated with ADC, female sex and never-smoker status [33]. They can occur in adenosquamous carcinoma, which is difficult to distinguish from pure SCC on small biopsies. Their occurrence in pure SCC is rare [34]. Sensitising EGFR mutations, ALK or ROS1 fusions, RET rearrangements, BRAF mutations and MET exon 14 skipping mutations are usually mutually exclusive [35]. First-generation TKIs gefitinib and erlotinib and second-generation afatinib achieved improved response rates and PFS time compared to standard chemotherapy in first-line treatment of advanced NSCLC with sensitising EGFR mutations [36–40]. None of the aforementioned randomised phase III studies demonstrated a benefit in overall survival with first- or second-generation TKIs, probably due to high crossover rates. Dacomitinib represents another alluring treatment option following the results of the ARCHER 1050 phase III study comparing dacomitinib to gefitinib (mPFS 14.7 versus 9.2 months, median overall survival (mOS) 34.1 versus 27.0 months) [41]. Osimertinib, a third-generation EGFR TKI with potent central nervous system (CNS) activity is currently the preferred first-line treatment in EGFR mutation-positive patients with advanced NSCLC following the results of the FLAURA trial (mPFS 18.9 versus 10.2 months with gefitinib or erlotinib, mOS 38.6 versus 31.8 months) [42]. Moreover, impressive results from the first interim analysis of the phase III ADAURA study of adjuvant osimertinib in early-stage (IB to IIIA) EGFR-mutant nonsquamous NSCLC were reported at the American Society of Clinical Oncology virtual meeting in 2020. Osimertinib versus placebo improved disease-free survival by 79% and 83% in the overall study population and in stage II to IIIA patients, respectively [43]. Other TKIs, such as poziotinib (NCT03318939) and AP32788 (NCT02716116), are being tested in patients with exon 20 insertions in EGFR and HER2, which confer natural resistance to existing EGFR TKIs. The combination of EGFR TKIs with antiangiogenic agents, namely bevacizumab and ramucirumab, improved PFS when compared to first-generation EGFR TKI monotherapy. While the addition of bevacizumab did not result in a significant increase in overall survival times in the final analysis of the NEJ026 trial, overall-survival data regarding the addition of ramucirumab to erlotinib remain immature [44–46]. Recently, results of the ETOP BOOSTER phase II randomised trial of second-line osimertinib and bevacizumab versus osimertinib monotherapy in EGFR T790M-mutated NSCLC did not show any PFS advantage of the combination therapy over osimertinib alone [47]. In contrast, the addition of pemetrexed and carboplatin to gefitinib versus gefitinib monotherapy significantly prolonged PFS and OS, but increased toxicity in a phase III randomised study in treatment-naïve patients with EGFR-mutated advanced NSCLC [48]. FLAURA2 is a currently recruiting phase III study of osimertinib with or without chemotherapy as first-line treatment in EGFR-mutated NSCLC (NCT04035486) [49]. Despite the impressive outcomes mentioned so far, patients eventually progress after front-line EGFR TKI treatment. EGFR p.Thr790Met (T790M) mutation was found in 60% of the patients who progressed after erlotinib, gefinitib or afatinib [50, 51]. The AURA3 phase III study showed improved ORR (71% versus 31%), mPFS time (10.2 versus 4.4 months) and numerical mOS time (26.8 versus 22.5 months, mostly due to patient crossover) in patients with acquired EGFR T790M mutation after first-line EGFR TKI therapy who received osimertinib in comparison to pemetrexed-platinum chemotherapy [52]. Primary T790M mutation is rare and possibly constitutional. TKI-naïve patients with T790M mutation should have genetic counselling in search for a familial predisposition to lung cancer [53]. Other, less common mechanisms of acquired resistance to first-line erlotinib, gefitinib and afatinib include the following.

• 1) Histological transformation to small-cell lung cancer (5%) [54]. In a retrospective analysis by Marcoux et al. [55], the transformation occurred at an average of 17.8 months after diagnosis and genomic profiling showed mutations in Rb1, TP53 and PIK3CA.

• 2) Increased activity of additional kinases due to MET (5%), ERBB2 (10%) or extracellular signal-regulated kinase (ERK) amplification, mutation of PIK3CA (which encodes the PI3 K p110α subunit) (2%) or overexpression of AXL kinase [56–58].

• 3) Modifications in splicing and deletion of the pro-apoptotic BCL-2-homology domain (BH3) due to a common BIM (also known as BCL2L11) polymorphism that probably affects intrinsic resistance to EGFR inhibitors, thus highlighting the complexity of possible resistance mechanisms [59, 60].

• 4) Other alterations such as PTEN downregulation, CRKL amplification, high-level HGF expression, FAS-NFκB pathway activation and EMT [61]. Of interest, EGFR mutation-positive patients with increased expression of IκBα (also known as NFKBIA), an inhibitor of NF-κB, had better response and survival outcomes [62].

Using paired plasma samples analysed using NGS, Ramalingam et al. [63] described the main resistance mechanisms to osimertinib in 91 patients with detectable plasma EGFR mutations in the FLAURA trial. Acquired EGFR T790M mutation was not found in the osimertinib arm, and the most common acquired resistance mechanisms were off-target MET amplification (15% of patients), on-target EGFR C797S mutation (on-target, 7%), downstream PIK3CA, BRAF V600E and KRAS G12D/C and A146 T mutations (7%, 3% and 3%, respectively) as well as off-target HER2 amplification (2%). A retrospective tissue analysis using NGS in 62 patients treated with osimertinib in the first- or later-line settings identified the following resistance mechanisms that were influenced by time on osimertinib treatment, initial EGFR mutation and line of therapy.

• 1) Histological transformation to SCC (most frequent), small cell carcinoma or pleomorphic carcinoma (15% and 14% of patients treated in the first- and later-line settings, respectively) [64].

• 2) Off-target genetic resistance (19% and 24% of patients in first- and later-line settings). MET amplification (7%), KRAS mutation (4%), BRAF (4%) and RET fusions (4%) were the most commonly identified off-target resistance mechanisms in the first-line setting [64].

• 3) On-target EGFR G724S mutation (4%) and EGFR amplification in the first-line setting (4%). On-target EGFR C797X mutation (29%), EGFR amplification (6%) and EGFR SV758IL compound mutation (3%) in the later-line setting [64].

Amivantamab (JNJ-61186372) is a fully human bispecific antibody that targets EGFR and MET. It demonstrated clinical activity across diverse EGFR-mutated NSCLC (exon 21 L858R, exon 19 deletion, exon 18 G719A, exon 20 insertion, T790M, C797S, coexistent MET amplification) [65]. It was granted FDA breakthrough therapy designation for EGFR-mutant exon 20 insertion NSCLC post-chemotherapy. Moreover, results from the CHRYSALIS phase I study of the combination of amivantamab with lazertinib, a third-generation EGFR TKI in osimertinib-resistant, chemo-naïve patients showed an ORR of 36% and a clinical benefit rate of 60% [66]. In the TATTON multi-cohort study, the addition of savolitinib, a potent selective MET inhibitor, to osimertinib, in TKI pre-treated patients with EGFR-mutant and MET-amplified/overexpressed (by fluorescent in situ hybridisation, immunohistochemistry (IHC) or NGS) advanced NSCLC, showed impressive ORRs of 33%, 65% and 62% in cohorts B1 (prior treatment with third-generation EGFR TKI), B2 (savolitinib 600/300 mg per weight-based dosing, no prior treatment with third-generation EGFR TKI and absence of T790M mutation) and D (savolitinib 300 mg, no prior treatment with third-generation EGFR TKI and absence of T790M mutation), respectively. Of note, exposure to prior TKIs was heterogenous and it was unclear whether the response to treatment was consistent across all “MET-positive” patients (FISH versus IHC versus NGS) [67]. Finally, patritumab deruxtecan (U3-1402), a novel antibody drug conjugate designed to target HER3 expression in EGFR-mutant NSCLC, demonstrated antitumour activity with a response rate of 25% a DCR of 70% and a mDOR of 6.9 months. Efficacy was observed in patients with various mechanisms of EGFR TKI resistance, including EGFR C797S, MET amplification, HER2 mutation, BRAF fusion and PIK3CA mutation [68].

ALK

Anaplastic lymphoma kinase (ALK) gene rearrangements occur in 5% of patients with NSCLC, predominantly in ADC and in never-/light smokers [69, 70]. Compared with EGFR-mutant patients, those with ALK rearrangements are younger and more likely to be male [70]. Echinoderm microtubule-associated protein-like 4 (EML4)-ALK fusion gene resulting from an inversion of the short strand of chromosome 2 was first observed in lung cancer in 2007, and many other ALK fusions, differing by the N-terminal gene fusion partner, have since been identified in lung and other cancers [71]. EMLA4-ALK is an abnormal fusion gene that encodes a cytoplasmic chimeric protein with constitutive kinase activity. Crizotinib achieved improved outcomes in front-line treatment of advanced NSCLC with ALK fusions and became standard of care (mPFS 10.9 months versus 7 months with chemotherapy; mOS not reached versus 47.5 months with chemotherapy) [72, 73]. Ceritinib, alectinib and brigatinib may be used after first-line crizotinib and ceritinib or alectinib may also be used in the first-line setting [74, 75]. The updated results of the ALEX study which compared alectinib to crizotinib in TKI-naïve patients with ALK-rearranged NSCLC showed a higher 5-year overall survival rate (62.5% versus 45.5%; mOS data are still immature) in addition to a longer mPFS time (34.8 versus 10.9 months), a higher CNS ORR (59% versus 26%) and a longer time to CNS progression [74, 76]. Similarly, in the ALESIA randomised phase III trial which included 187 treatment-naive Asian patients with advanced ALK-rearranged NSCLC, alectinib was significantly superior to crizotinib in terms of mPFS time [77]. Moreover, brigatinib (ALTA-1L trial), as well as ensartinib (EXALT-3 trial), a potent newer-generation ALK inhibitor, achieved better mPFS times in comparison with crizotinib in first-line treatment of advanced NSCLC with ALK fusions (mPFS 24 versus 11 months and 25.8 versus 12.7 months, respectively) [78, 79]. The CROWN phase III study demonstrated higher ORR (76% versus 58%), higher intracranial response rate (82% versus 23%), and longer mPFS (not reached versus 9.3 months) with lorlatinib, a third-generation brain-penetrant ALK and ROS1 inhibitor compared with crizotinib in TKI-naïve patients with ALK-positive advanced NSCLC [80]. As with other oncogene-addicted lung tumours, patients eventually develop resistance to ALK TKIs through the following mechanisms.

• 1) Secondary ALK mutations, found in one third of crizotinib-resistant tumours [35]. The most frequent mutations in this case are L1196M and G1269A [81, 82]. Other less frequent mutations include L1152R, 1151 T-ins, C1156Y, I1171 T, F1174 L, G1202R and S1206Y [83]. Alectinib, ceritinib and brigatinib are able to overcome the majority of ALK-resistant mutations (L1196M, G1269A and F1174L) acquired with crizotinib. I1171 T/N/S and G1202R are observed in tumours resistant to alectinib and F1174C/V and G1202R are found in those resistant to ceritinib [83]. Lorlatinib is particularly active against the solvent front mutation G1202R [84]. A phase II study of lorlatinib in ALK-positive NSCLC resulted in an ORR of 72.90% in patients who had only received previous crizotinib, 42.9% in patients treated with one previous non-crizotinib ALK TKI (EXP3B) and 39.6% in patients treated with two or more previous ALK TKIs [85]. TPX-0131 is a next-generation ALK inhibitor that was shown in cell proliferation assays to be active against the G1202R mutation (more than 100-fold more potent than lorlatinib) and against compound mutations of both gatekeeper and solvent front mutations (L1196M/G1202R) that are refractory to lorlatinib [86].

• 2) Activation of “bypass” pathways, present in 40% of crizotinib-resistant tumours. Activated pathways include EGFR, cKIT, IGF1R and Src. Alectinib resistance mechanisms include c-MET amplification while ceritinib-resistant tumours harbour MEK-activating mutations [83].

• 3) Amplification of the ALK fusion gene, found in 9% of crizotinib-resistant tumours [35, 87].

• 4) Other mechanisms including EMT, apoptosis regulation and epigenetic modifications, as well as transformation to small-cell lung carcinoma [83, 87–89].

ROS1

ROS proto-oncogene 1 (ROS1) rearrangements were first described in NSCLC in 2007 [90]. They lead to a fusion of a ROS1 TK domain with a partner gene (several options, with CD74-ROS1 fusion being the most prevalent) with subsequent activation of the ROS1 kinase leading to tumour proliferation and survival through downstream signalling via SHP-1/SHP-2, JAK/STAT, PI3 K/AKT/MTOR and mitogen-activated protein kinase (MAPK)/ERK pathways [90, 91]. ROS1 fusions affect 1–2% of patients with NSCLC [92]. There is a significant homology between ALK and ROS1 TK domains including the binding site for crizotinib [93]. Similar to ALK, ROS1 fusions occur mostly in patients with ADC histology who are nonsmokers or former light smokers. However, in contrast to ALK, ROS1 fusions are associated with lower rates of extrathoracic and intracranial metastases at the time of diagnosis, as well as lower cumulative incidence of intracranial metastases [94]. Crizotinib was granted FDA approval in 2016 for treatment of advanced NSCLC with ROS1 rearrangement following the results of the PROFILE 1001 phase I study [95]. Updated results showed an ORR of 72%, a mDOR of 24.7 months, a mPFS of 19.3 months and a mOS of 51.4 months [96]. In a Korean phase 2 study, Lim et al. [97] administered ceritinib to 32 patients with ROS1 rearrangement, 30 of whom were crizotinib-naïve. ORR was 62% (the two patients who had previously received crizotinib had no clinical response), mDOR was 21.0 months, median PFS was 9.3 months for all patients and 19.3 months for crizotinib-naïve patients and mOS was 24 months. Resistance to crizotinib is inevitable and occurs through two main mechanisms, as follows.

• 1) On-target mutations of the TK domain of ROS1 which affect crizotinib binding [94, 98]. The most common mutation is ROS1-G2032R (41%). Other mutations include D2033N (6%), S1986Y/F (6%), gatekeeper L2026M and L1951R [94]. Lorlatinib has in vitro activity against most known ROS1 resistance mutations such as ROS1-D2033N, S1986Y/F and L2026M mutations. Its efficacy against ROS1-G2032R remains limited [90]. In contrast, repotrectinib, a next-generation ALK/ROS1/TRKA-C inhibitor with higher potency against ROS1 than crizotinib, has shown efficacy against the latter mutation [99].

• 2) Off-target resistance mechanisms that include activation of bypass signalling pathways such as EGFR, RAS and KIT, and phenotypic changes such as EMT [94, 98].

In an open-label phase I/II trial, 69 patients with ROS1-positive NSCLC were treated with lorlatinib. TKI-naïve patients (n=13; 30%) had an ORR of 62%, an intracranial response rate of 64%, mDOR of 19.6 months and mPFS of 21 months. Among patients who had previously received crizotinib, ORR was 35%, intracranial response rate was 53% and mPFS was 8.5 months [100]. In an integrated analysis of 3 phase I/II studies (ALKA-372-001, STARTRK-1 and STARTRK-2) with a total of 53 ROS1-positive TKI-naïve patients, entrectinib, a ROS1, TRK and ALK TKI, achieved an ORR of 77.4%, a mPFS of 19 months, a mDOR of 24.6 months and an intracranial ORR of 55% [101]. TRIDENT-1 is an open-label multicentre phase I/II study of repotrectinib (TPX-0005) in patients with ROS1 or NTRK fusion-positive advanced NSCLC. Preliminary results from the phase 2 expansion 1 (EXP-1) cohort of 55 TKI-naïve ROS1-positive patients (20% had received prior chemotherapy) showed an ORR of 93% (in 15 evaluable patients at data cut-off on 31 December 2020). When pooled with data from seven patients treated in phase 1, the confirmed ORR was 91%. Based on these results, repotrectinib was granted FDA breakthrough therapy designation as treatment for TKI-naive ROS1-positive advanced NSCLC. In addition, repotrectinib achieved an ORR of 40% in five patients who were pre-treated with one TKI and chemotherapy (EXP-2 cohort) and in five other patients who were pre-treated with two TKIs and no chemotherapy (EXP-3). Finally, ORR was 67% in six patients who were previously treated with one TKI and no chemotherapy (EXP-4). All patients with ROS1 G2032R mutation had tumour regression (confirmed ORR of 40%) [102, 103]. This trial is still recruiting [104]. An updated pooled analysis of US and Japan phase 1 studies of taletrectinib, an oral, potent selective ROS1 and pan-NTRK TKI with pre-clinical activity against ROS1 G2032R solvent-front mutation, reported an ORR of 66.7%, 33.3% and 33.3% among ROS1 TKI-naïve patients (n=9), crizotinib-pre-treated patients (n=6) and patients pre-treated with two previous TKIs (n=3), respectively [105]. In the ImmunoTarget cohort, ROS1-positive tumours had the highest median PD-L1 expression among other oncogene-addicted NSCLC tumours; however, no tumour response to ICIs was reported in patients from the ALK/ROS1/RET group despite PD-L1 positivity [20].

BRAF

BRAF is a serine/threonine protein kinase whose mutations favour tumorigenesis by activating the MAPK signalling pathway. BRAF V600E is the most common BRAF mutation across all tumour types; however, in lung ADC it occurs at the same rate as other BRAF mutations (1–2% of patients) [106–108]. Unlike patients with EGFR or ALK alterations, those with BRAF mutations are more likely to be current or former smokers [107]. BRAF mutations act as an alternative oncogenic driver in NSCLC and do not overlap with EGFR mutations, ALK/ROS1/RET rearrangements, or MET exon 14 skipping mutations [108, 109]. They are almost restricted to non-mucinous ADC histological subtype [110]. In a basket trial, vemurafenib, a selective BRAF V600 inhibitor achieved a response rate of 42%, a mPFS of 7.3 months and an unreached mOS in pre-treated NSCLC patients (20 patients) [111]. Secondary resistance mechanisms to BRAF inhibitors include 1) reactivation of ERK signalling through the MAPK pathway (most common resistance mechanism) which can take place upstream or downstream of BRAF kinase. Mechanisms include BRAF splice variants (16%), BRAF gene amplification (13%), NRAS/KRAS (20%) or MEK1/2 mutations (7%) that induce a BRAF-independent reactivation of ERK signalling [112, 113]; and 2) bypass of the MAPK pathway through activation of other pathways such as PI3K-AKT (through activating mutations in AKT and PTEN loss of function) [113].

The combination of dabrafenib (another BRAF V600E inhibitor) and trametinib (a MEK1-2 inhibitor) results in dual inhibition of the MAPK pathway thus overcoming the ERK escape mechanism. In the second-line setting in advanced BRAF-mutated NSCLC, dabrafenib and trametinib achieved an ORR of 66.7%, mPFS of 10.9 months and mOS of 18.2 months [108]. As front-line treatment, the same combination yielded an ORR of 64%, mPFS of 10.2 months, mOS of 24.6 months and mDOR of 10.4 months according to a phase II study by Planchard et al. [114]. Interestingly, data from the ImmunoTarget registry show that patients with BRAF mutations who received ICIs had higher response rates when compared to patients with other driver mutations (except for KRAS). Smoking had a positive impact on PFS. More specific studies are needed [20].

MET

C-MET is a hepatocyte growth factor TK receptor that promotes cell survival and proliferation. MET exon 14 skipping mutations, MET gene copy number gain or amplification and MET protein overexpression represent the main oncogenic driver alterations of MET and do not overlap with other oncogenic drivers in NSCLC [115, 116]. They occur in 3–4% of patients with lung ADC and 1–2% of patients with other NSCLC subtypes. MET exon 14 skipping mutations and MET amplification may coexist [117, 118]. MET amplification occurs as a primary driver event or as a resistance mechanism to EGFR TKI therapy. MET exon 14 skipping mutations result in impaired MET ubiquitination, decreased MET turnover, and increased downstream signalling [119]. Patients with MET exon 14 skipping mutations (3–4% of NSCLC patients) were found to be significantly older than patients with EGFR or KRAS mutations and were more likely to be female and nonsmokers [115, 120]. In addition, these mutations are more frequent in adenocarcinoma and sarcomatoid carcinoma histological subtypes. Crizotinib and cabozantinib are effective in patients with MET abnormalities [118, 121]. Newer therapeutic candidates include tepotinib, capmatinib and savolitinib, three selective c-MET TKIs with encouraging phase Ib/II results that are now being tested in combination with EGFR TKIs in EGFR TKI-resistant NSCLC [122]. The VISION phase II study of tepotinib in 152 patients with advanced NSCLC (first- and later lines) with MET exon 14 skipping mutations yielded a response rate of 46% by independent review with a mDOR of 11.1 months (in the combined tissue and liquid biopsy group) regardless of the number of previous treatment lines [119]. In the phase II GEOMETRY mono-1 study of capmatinib in patients with MET exon 14 mutations, the response rate was 41% and 68% among 69 previously treated and 28 treatment-naïve patients, respectively. mPFS was 5.42 and 9.13 months, respectively [123]. In addition, the authors reported an ORR of 40% and 29% in TKI-naïve and TKI pre-treated patients with MET amplification who received capmatinib [124].

Recondo et al. [125] studied mechanisms of acquired resistance to MET TKIs using plasma and tissue NGS. On-target acquired mechanisms of resistance (35% of cases) included single and polyclonal MET kinase domain mutations in codons H1094, G1163, L1195, D1228 and Y1230, and high levels of amplification of the MET exon 14-mutant allele [125]. Glesatinib, a spectrum-selective MET inhibitor with a distinct mechanism of target inhibition can overcome resistance to type I MET inhibitors [126]. Off-target mechanisms of resistance (45% of cases) included KRAS mutations and amplifications in KRAS, EGFR, HER3 and BRAF [125].

RET

Rearrangements between the RET gene and other domains (mainly KIF5B and CCDC6) lead to an overexpression of the RET protein, a TK receptor that influences cell proliferation and differentiation. RET rearrangements occur in 1–2% of patients with NSCLC, mainly in ADC, and are independent of smoking status [127–129]. They are usually mutually exclusive with other oncogenic drivers, although some studies describe a possible overlap of RET rearrangements with EGFR or KRAS mutations [130]. Prior to the development of selective RET-inhibitors, response rates and survival outcomes of patients with RET-rearranged NSCLC who received multikinase inhibitors such as cabozantinib, vandetanib, sunitinib, sorafenib, alectinib, lenvatinib, nindetanib, ponatinib and regorafenib were heterogenous (small biased studies) and mostly disappointing in comparison to those achieved with selective TKIs in other oncogene-driven NSCLC tumours (ORR 16–47% and mPFS 2.3–7.3 months) [128, 131, 132]. In contrast, selpercatinib (formerly LOXO-292) and pralsetinib (formerly BLU-667) selectively block RET, thus avoiding other targets and limiting side-effects. Based on data from the LIBRETTO-001 phase I/II trial that resulted in ORRs of 70% and 90% in 105 patients previously treated with platinum-based chemotherapy and 39 treatment-naïve patients, respectively, selpercatinib obtained FDA breakthrough designation for the treatment of RET fusion-positive NSCLC in August 2018 [133]. In the phase I/II ARROW trial, pralsetinib yielded an ORR of 65% in 116 patients (first- or later line settings) and 73% in 26 treatment-naïve patients [134]. AcceleRET is an on-going phase III study comparing pralsetinib to a platinum-based chemotherapy in treatment naïve patients with RET fusion-positive metastatic NSCLC (NCT04222972) [135]. Similarly, LIBRETTO-431 is another phase III study comparing selpercatinib to platinum-based and pemetrexed chemotherapy with or without pembrolizumab in the first-line treatment of advanced RET fusion-positive NSCLC (NCT04194944) [136]. On-target RET G810 solvent front mutations detected in plasma and/or tissue samples, conferred resistance to selpercatinib in patients with KIF5B-RET and CCDC6-RET fusion-positive NSCLC [137]. NCT04161391 is an ongoing phase I/II study of TPX-0046, a RET/SRC inhibitor in advanced/metastatic solid tumours with RET fusions or mutations [138]. Off-target mechanisms of resistance include MET amplification and activation of other bypass pathways which could be targeted by combination strategies [139].

NTRK

NTRK gene fusions result in tropomyosin receptor kinase (TRK) fusion proteins that drive tumorigenesis in many solid tumours including lung, salivary gland, thyroid and sarcoma. NTRK fusions are found in <1% of patients with NSCLC and do not overlap with other driver oncogenes [140–142]. They are independent of age, sex, smoking history and NSCLC histological subtype [141]. NRTK1-3 gene fusions encode oncogenic TRKA-C fusion proteins, respectively, that can be targeted by therapies in development, including larotrectinib (LOXO-101) and entrectinib (RXDX-101). Larotrectinib and entrectinib were both shown to have marked and durable efficacy in NSCLC with NTRK fusions [140, 143]. A pooled analysis of three phase I/II trials of larotrectinib in 159 adult and paediatric patients with advanced solid tumours including 12 patients with lung cancer (8%) who had received prior therapy, resulted in an ORR of 79% (16% complete responses) and a good safety profile. In the overall population, the mDOR was 35.2 months and mPFS was 28.3 months [144]. Larotrectinib is the first tumour-agnostic drug to receive a European Medicines Agency (EMA) approval for the treatment of advanced solid tumours with NTRK gene fusion in adult and paediatric patients who have no satisfactory alternative therapeutic options. Similarly, in an integrated analysis of three phase I/II trials (ALKA-372-001, STARTRK-1 and STARTRK-2) of patients with metastatic or locally advanced solid tumours harbouring oncogenic NTRK1, NTRK2 and NTRK3 gene fusions, entrectinib achieved an ORR of 57% (7% complete responses), an mDOR of 10 months, mPFS of 11 months and a good tolerance profile. 10 of the 54 patients who were included in the study had advanced NSCLC; ORR in this subpopulation was 70% [145]. In June 2020, the EMA granted a conditional marketing authorisation to entrectinib for the treatment of solid tumours expressing NTRK gene fusions. On-target resistance mechanisms to first-generation TRK inhibitors (larotrectinib and entrectinib) include solvent-front mutations (G595R, G639R, G623R), gatekeeper mutations (F589L, F633L, F627L) and xDFG mutations (G667C, G709C, G696A) [146]. These mutations could be sensitive to second-generation TRK inhibitors, such as repotrectinib (ongoing TRIDENT-1 trial, ORR 50% in six TKI-pre-treated patients in the EXP-6 cohort) and selitrectinib [103, 147]. Off-target resistance mechanisms include KRAS mutation, MET amplification, BRAF mutation and IGF1R activation [146].

HER2

HER2/ERBB2 alterations in NSCLC can serve as primary or secondary oncogenic drivers after treatment with targeted therapies. They include mutations, amplifications and protein overexpression [148].

HER2 overexpression without mutation or amplification is able to induce de novo NSCLC in mouse models [149]. It is observed in 10–30% of NSCLCs and is associated with papillary histology and poor disease prognosis [150–154]. High protein overexpression (IHC score 3) is often the consequence of HER2 amplification [148]. In a study by de Langen et al. [155], HER2 protein overexpression increased after EGFR TKI treatment, thus resulting in a selected HER2-driven resistant cell population. The benefit from the use of trastuzumab in patients with HER2 overexpression or amplification is inconsistent; the monoclonal antibody might be more efficient in HER2-mutant tumours. Moreover, ado-trastuzumab emtansine, a HER2-specific antibody–drug conjugate, only showed a modest benefit (ORR 20%) in patients with IHC3⁺ overexpression of HER2 (no responses in the IHC2⁺ cohort) [156].

HER2 amplification, classically detected using fluorescent in situ hybridisation, is present in 2–5% of lung ADCs, particularly in females, never-smokers and patients with pleural effusions [154, 157, 158]. Ado-trastuzumab emtansine achieved an ORR of 55% in 11 patients with HER2-amplified NSCLC [159].

ERBB2 kinase domain mutations were reported in 1.5% of NSCLC cases and had a higher prevalence in females and nonsmokers [152]. Genetically, they are less diverse than EGFR alterations, with 96% consisting of exon 20 insertions [160]. HER2 mutations are associated with a poorer prognosis in lung ADC [161, 162]. With the implementation of NGS, HER2 mutations are routinely identified and are considered an oncogenic driver in NSCLC as well as a mechanism of acquired resistance after targeted therapy [148]. HER2 kinase domain mutations such as Y772_A775dup activates HER2 and EGFR, thus inducing resistance to EGFR TKI treatment [163]. Y772_A775dup occurred in one out of 60 patients treated with osimertinib in the AURA study [164]. HER2 protein expression is an insufficient surrogate marker for HER2 mutations [165]. The latter can be detected using liquid biopsies as demonstrated in a study by Lee et al. [166] and are not strictly associated with HER2 overexpression or amplification [165]. To date, there are no approved anti-HER2 targeted agents in HER2-mutant NSCLC [167]. Pan-HER TKIs such as dacomitinib, afatinib, neratinib, poziotinib and pyrotinib were evaluated in small-sized studies with variable response rates (between 3.8% and 5.8%) and PFS rates ranging from 3.7 to 6.4 months. The efficacy of these agents is moderate in comparison to EGFR, ALK or ROS1 TKIs [148]. Of these drugs, poziotinib seems to have the most promising results, especially in targeting HER2 exon 20 mutations [168]. Osimertinib is active in vitro against the relatively rare L775P and L775S mutations, but is less effective than afatinib or neratinib in patients with Y772_A775dup mutation [169]. The combination of trastuzumab with pertuzumab, a recombinant humanised monoclonal antibody that specifically targets the HER2 dimerisation domain, thus blocking ligand-dependent heterodimerisation of HER2 with other HER family members, has shown limited activity in the phase IIa MyPathway basket trial which included 30 patients with HER2-mutant or HER2-positive refractory NSCLC, with an ORR of 21% and 13%, respectively [170]. Li et al. [171] treated 18 HER2-mutant advanced NSCLC patients with ado-trastuzumab emtansine as part of a phase 2 basket trial. Partial responses were achieved in 44% of the patients, mainly those with HER2 exon 20 insertions and point mutations in the kinase, transmembrane and extracellular domains. The median PFS was 5 months. HER2 protein expression measured using IHC or mass spectrometry was not predictive of response to treatment. Recently, trastuzumab deruxtecan (DS8201-a), a novel antibody–drug conjugate, achieved an ORR of 61.9%, a mPFS of 14 months and an unreached mDOR in 42 patients with HER2-mutant NSCLC (cohort 2 of the DESTINY-Lung01 phase II trial) [172]. DESTINY-Lung01 also included 49 patients with HER2-overexpressing NSCLC (cohort 1). Results were promising albeit less spectacular in comparison with cohort 2 (ORR 24.5%, mDOR 6 months and mPFS 5.4 months). Trastuzumab deruxtecan was granted FDA breakthrough therapy designation for patients with advanced HER2-mutant NSCLC and disease progression on or after platinum-based therapy.

Table 1 lists the currently approved targeted therapies in oncogene-driven NSCLC.

TABLE 1.

Approved targeted therapies in oncogene-driven nonsmall cell lung carcinoma

	ASCO [ 173 ]	ESMO# [174] + updated version published 15 September 2020	NCCN¶ [ 167 ] + updated version April 2021
EGFR-sensitising mutations	Afatinib, erlotinib or gefitinib Evidence quality: high Strength of recommendation: strong for each	Erlotinib, gefitinib or afatinib (I, A), or dacomitinib (I, B; MCBS score: 3) None of the four EGFR TKIs is consensually considered as a preferred option (IV, C) Osimertinib (I, A; MCBS score: 4) Erlotinib and bevacizumab (II, B; MCBS score: 3) Erlotinib and ramucirumab (I, B; MCBS score 3) Gefitinib+carboplatin and pemetrexed is a first-line option (I, B; not EMA-approved) Atezolizumab and bevacizumab with carboplatin and paclitaxel may be used after targeted therapies (III, A; not EMA-approved)	Afatinib (other recommended) Erlotinib (other recommended) Dacomitinib (other recommended) Gefitinib (other recommended) Osimertinib (preferred) Erlotinib+ramucirumab Erlotinib+bevacizumab (nonsquamous)
EGFR T790M mutation	Osimertinib Evidence quality: high Strength of recommendation: strong	Osimertinib (I, A; MCBS score: 4)	Osimertinib (category 1)
ALK fusions first-line	Crizotinib Evidence quality: strong Strength of recommendation: high	Crizotinib (I, A; MBCS score: 4), ceritinib (I, B; MCBS score: 4), alectinib (I, A; MCBS: 4), brigatinib (I, B; not EMA-approved), ensartinib (I, A; not EMA-approved) CNS involvement: alectinib (III, A), brigatinib (III, B) or ceritinib (IV, B)	Alectinib (preferred) Ceritinib (other recommended) Brigatinib (other recommended) Crizotinib (useful in certain circumstances) Lorlatinib
ALK fusions subsequent therapy		Ceritinib and alectinib after crizotinib (I, A; MBCS score: 4) Brigatinib after crizotinib (III, A; MCBS score: 3) Lorlatinib in patients who progress after a second-generation ALK TKI (III, A; MCBS score: 3)	Alectinib, brigatinib, lorlatinib, ceritinib
ROS1 rearrangement first-line	Crizotinib (informal consensus) Evidence quality: low Strength of recommendation: weak	Crizotinib (III, A; MBCS score: 3) Ceritinib (III, C; not EMA approved)	Crizotinib (category 2A; preferred) Entrectinib (category 2A; preferred) Ceritinib (category 2A)
ROS1 rearrangement subsequent lines		Lorlatinib (III, B), repotrectinib (III, B), entrectinib (III, B; MCBS score: 3)	Lorlatinib (category 2A) Entrectinib
BRAF V600E mutation positive first-line and subsequent therapy	In patients who have received prior immune checkpoint therapy, dabrafenib alone or in combination with trametinib in third-line is an option (informal consensus) Evidence quality: insufficient Strength of recommendation: moderate	Dabrafenib and trametinib in first- or second-line (III, A; MBCS score: 2)	Dabrafenib and trametinib (category 2A)
NTRK fusion first-line/subsequent therapy		Larotrectinib (III, A; MCBS score: 3), entrectinib (III, B; MCBS score: 3)	Entrectinib (category 2A) Larotrectinib (category 2A)
ERBB2		Trastuzumab deruxtecan (III, B)
MET exon 14 skipping mutation first-line/subsequent therapy		Crizotinib (III, B) Capmatinib (III, B), tepotinib (III, B)	Crizotinib Capmatinib Tepotinib
RET rearrangement positive first-line/subsequent		Not currently routinely recommended and recruitment into open trials is encouraged (II, C) Selpercatinib (III, B), pralsetinib (III, B)	Selpercatinib Pralsetinib Cabozantinib Vandetanib

ASCO: American Society of Clinical Oncology; ESMO: European Society for Medical Oncology; NCCN: National Comprehensive Cancer Network; EGFR: epidermal growth factor receptor; ALK: anaplastic lymphoma kinase; ROS proto-oncogene 1; MCBS: magnitude of clinical benefit scale; TKI: tyrosine kinase inhibitor; EMA: European Medicines Agency; CNS: central nervous system. ^#: ECMO-MCBS v1.1 for new therapy/indication approved by the EMA since 1 January 2016. The score has been calculated by the ESMO-MCBS working group and validated by the ESMO guidelines committee. Levels of evidence (I to V) and grades of recommendation (A to E) for the ESMO guidelines are adapted from the Infectious Diseases Society of America–United States Public Health Service Grading System [175]; ^¶: NCCN categories of evidence and consensus include category 1 (based upon high-level evidence, there is uniform NCCN consensus that the intervention is appropriate), category 2A (based upon lower-level evidence, there is uniform NCCN consensus that the intervention is appropriate), category 2B (based upon lower-level evidence, there is NCCN consensus that the intervention is appropriate) and category 3 (based upon any level of evidence, there is major NCCN disagreement that the intervention is appropriate).

Liquid biopsy

Molecular testing has become an inevitable standard in thoracic oncology. Oncologists are now confronted with another challenge on how to select tissue biopsy versus liquid biopsy to identify driver alterations and guide treatment decisions in NSCLC [176]. The term liquid biopsy refers to the analysis of plasma circulating tumour cells (CTC) or circulating tumour DNA (ctDNA). Other analytes in the peripheral blood include circulating cell-free DNA (cfDNA), circulating cell-free RNA (cfRNA), circulating extracellular vesicles and tumour educated platelets, as well as proteins and metabolites [177]. In addition, molecular analysis could be conducted on urine, pleural and sputum samples. While ctDNA might derive from apoptotic cells, the other analytes may be the product of viable cells, thus suggesting a possible difference in the acquired information. Hence, comparative studies between the different analytes are needed [177]. In addition, combination strategies of the different analytes that result in an optimal characterisation of a tumour are still lacking [177]. Compared to tissue biopsy, liquid biopsy is a simple, noninvasive method with an almost nonexistent complication rate, which is therefore much more appealing to patients, particularly those who are unfit or elderly. It is mostly useful and therefore recommended when tumour tissue is scarce or when there is bone-only metastatic disease, as the bone decalcification solutions can damage tumour DNA. It provides a global perspective since the collected tumour cells/DNA may be derived from several tumour sites and therefore partially resolves the problem of tumour heterogeneity. However, the accuracy in reflecting the heterogeneity of a tumour and its subclones or all tumour lesions in metastatic patients is yet to be determined [177]. Serial assessments can be performed with phylogenetic ctDNA construction, thus facilitating response monitoring and allowing for an earlier detection of secondary resistance mutations and eventually, earlier switching of targeted therapies. Moreover, liquid biopsy may be used to determine TMB, an important predictive biomarker for ICI response [176, 178, 179]. In addition, ctDNA variant allele frequency (VAF) and pre-treatment ctDNA concentration are a reflection of tumour burden and metabolic tumour volume, respectively [177]. Furthermore, liquid biopsy is being tested as a screening tool for early detection of cancer [180]. The major limitation of this technique remains in its lower sensitivity compared to tissue biopsy (∼85% in advanced-stage disease using the most sensitive assays) [179]. The isolated DNA is highly fragmented, has a short half-life and a low overall yield. The VAF is often low in the range of background noise of analytical methods. In addition, standard operating procedures are still lacking [177]. Tumour burden and anatomical site of disease have a clear impact on the sensitivity of liquid biopsies within and across cancer types [181]. Among the methods developed to improve the sequencing sensitivity are unique molecular identifiers, i.e. molecular barcodes that facilitate bioinformatic alignment of sequences derived from the same DNA fragment, thus increasing the accuracy (number of errors per base pair sequenced) [177]. Other solutions to increase sensitivity include enriching for ctDNA fragments on the basis of their length, since they are shorter than normal cfDNA [177]. The International Association for the Study of Lung Cancer (IASLC) guidelines provide a comprehensive review of the pros and cons of liquid biopsy in NSCLC [179]. Clinicians should have a thorough understanding of the test's characteristics and its validation against a reference standard such as tissue genotyping. They should recognise the differences between the available assays and be able to select the correct patients for liquid biopsy versus tissue biopsy and to correctly interpret the results of the assay [182].

cfDNA/ctDNA

Upon cell death or apoptosis, cfDNA, composed mostly of germline DNA, is released into the bloodstream. The concept of cfDNA “shed”, meaning the release of tumour DNA into the circulation, is central to the development and interpretation of plasma genotyping assays. The concentration of cfDNA in the plasma is directly proportional to the rate of cfDNA shedding by the tumour and indirectly proportional to the rate of DNA elimination by renal clearance. It is also related to the overall burden of metastatic disease, the presence of particular metastatic sites such as the liver or the bone, the mitotic rate as well as the amount of tumour necrosis and vascularisation [183–185]. In contrast, the analysis of plasma cfDNA testing in patients with isolated brain metastases has limited sensitivity [186].

Droplet digital (dd)PCR and beads, emulsions, amplification and magnetics (BEAMing) PCR are highly sensitive and specific currently used molecular platforms that employ emulsion technology to quantify the amount of mutant DNA in the plasma of patients with advanced NSCLC [185, 187]. The main limitation of these assays is their inability to detect complex genomic alterations and perform multiplex testing. They can be used to detect and monitor levels of common EGFR and KRAS mutations [187, 188]. NGS assesses larger panels and detects complex alterations including rearrangements, but is slower than ddPCR. Moreover, its sensitivity to detect copy number alterations remains low. Multiple NGS-based commercial plasma genotyping assays are available (Guardant360, FoundationACT). Both PCR and NGS are valid approaches for plasma genotyping. Rigorous validation of the available assays against a reference standard within a given tumour type, stage and treatment time point is crucial to ensure reliability of the obtained results [182].

Applications of liquid biopsy in NSCLC

Identification of driver alterations

The best-validated clinical context for use of plasma cfDNA is in newly diagnosed NSCLC to guide treatment selection, because tumour shedding is considerably reduced in patients already receiving treatment. Given the high specificity of validated plasma-based assays, the identification of an actionable driver alteration should be considered as sufficient evidence to start treatment with the corresponding targeted therapy without the need for additional tissue biopsy.

The cobas EGFR Mutation Test v2 (Roche Molecular Diagnostics) is an FDA-approved cfDNA-based test that detects activating EGFR mutations in NSCLC patients. These mutations are a strong predictive biomarker of the efficacy of EGFR TKIs in this setting [185]. Other methods such as ddPCR, BEAMing and NGS have a higher sensitivity and their use is expanding. According to the IASLC recommendations, the detection of a sensitising EGFR mutation using the cobas EGFR mutation test v2 or ddPCR on ctDNA is sufficient to start treatment with EGFR TKIs. A negative result mandates further evaluation with either a NGS-based test of ctDNA or tissue biopsy. Plasma genotyping may be particularly useful in situations where a tissue biopsy is not feasible or where there is an urgent need to rapidly identify a driver alteration [182]. Circulating tumour DNA studies were successful in detecting somatic mutations that had been missed in the corresponding tissue samples, thus highlighting the concept of intratumour heterogeneity and subclonal mutations [177]. Liquid biopsy may also be used to exclude the presence of other potential drivers, particularly when a KRAS mutation (non-overlapping with EGFR and ALK alterations) is identified [182]. In addition, mutant allele fraction quantified by NGS, ddPCR or BEAMing should be carefully considered in the interpretation of plasma cfDNA results. In the AURA3 study of osimertinib versus platinum-pemetrexed for T790M mutation-positive advanced NSCLC, the plasma T790M positive percentage agreement was 51% by cobas testing, 58% by ddPCR and 66% by NGS, suggesting that all three test platforms are suitable for routine clinical practice [189]. Caution is advised when interpreting mutations that are positive at low levels (allelic fraction <1% using the latter quantitative assays) as they might represent false positives (assay artefacts). In a recent noninferiority prospective study of treatment-naïve patients with NSCLC, Guardant360, a validated cfDNA test identified guideline-recommended biomarkers at a rate at least as high as standard-of-care tissue genotyping with high tissue concordance and shorter median turnaround time [190]. Despite these promising results, negative cfDNA sequencing is not sufficient to rule out the presence of driver mutations and tissue genotyping remains the standard diagnostic procedure. B-FAST (NCT03178552) is an on-going phase II/III, global, multicentre, open-label, multi-cohort study designed to evaluate the safety and efficacy of targeted therapies or immunotherapy as single agents or in combination in patients with advanced NSCLC and driver mutations or high TMB as identified by two blood-based NGS ctDNA assays [191].

Compared to EGFR mutations, data on ALK rearrangement assessment using ctDNA in treatment-naïve patients are less convincing with NGS reaching acceptable levels of sensitivity. The sensitivity of ddPCR is much higher than that of reverse transcriptase (RT)-PCR in detecting ALK rearrangements; however, prospective studies are awaited [179]. The IASLC recommends against the use of PCR for ALK or ROS1 rearrangement detection from ctDNA [179].

Multiplex panels using NGS platforms on ctDNA detect fewer common mutations and have acceptable levels of sensitivity and optimal levels of specificity to initiate first-line therapy in patients with EGFR, ALK, ROS1 and BRAF alterations. In addition, cfRNA can be used for the measurement of RNA transcripts of fusion genes (NTRK, ALK, ROS1, RET) and MET-14 splicing variant by RT-PCR [179]. Again, a negative plasma result mandates a tissue biopsy to formally exclude the presence of a driver alteration [179].

Identification of resistance mechanisms

Identifying resistance mechanisms and guiding subsequent therapy decisions is another useful application of liquid biopsy in NSCLC, particularly in EGFR-mutant disease where identification of a T790M mutation is an indication to switch from first-/second-generation EGFR TKIs to osimertinib therapy [185]. Furthermore, due to tumour heterogeneity, T790M mutation might not be identified in all tumour sites which, again, highlights the utility of plasma biopsy [182]. T790M mutation could also be detected in urine and pleural fluid samples [192, 193]. The concordance rate of T790M testing in plasma and tissue is up to 74% [194]. Moreover, the profile of resistance alterations after treatment with osimertinib is becoming clinically relevant with the advent of fourth-generation EGFR TKIs and with strategies that impair parallel pathway activation [195, 196]. The MELROSE phase II trial is an ongoing French multicentric study whose primary objective is to determine the genetic tumour profile at progression after first-line therapy with osimertinib in patients with EGFR-sensitising mutations, using both tissue biopsy and ctDNA [197]. In patients progressing after first-line EGFR TKIs, the IASLC recommends an NGS multiplex panel over PCR methods, as it could also detect fewer common alterations. Negative results need to be confirmed by NGS testing on tissue biopsies. Of note, liquid biopsy cannot detect small cell lung cancer transformation; the co-detection of TP53/RB1 loss (associated with a ×43 higher risk of transformation to small cell lung cancer) on a liquid biopsy does not indicate that a transformation has occurred [198]. In this setting, tissue biopsies are critical upon disease progression.

The use of liquid biopsy, and particularly NGS-based techniques, is currently emerging as a means for detecting ALK resistance alterations in patients progressing during ALK TKIs. This approach has not yet been implemented in daily practice; however, the detection of certain mutations can help select the subsequent ALK TKI to treat patients. For instance, I1171 T and the V1180 L mutations confer resistance to alectinib and crizotinib but not to ceritinib; L1196M or S1206Y mutations confer resistance to crizotinib but not ceritinib; G1202R, G1123S or F1174 mutations confer resistance to crizotinib [199, 200]. Pailler et al. [201] explored resistance mutations in CTCs isolated at the single-cell level from 17 patients with ALK-rearranged advanced NSCLC who progressed on crizotinib and lorlatinib. They identified an array of co-occurring mutations (mainly the RTK-KRAS and TP53 pathways) in ALK-independent pathways in nine out of 14 crizotinib-resistant patients. Additionally, in one lorlatinib-resistant patient, they found two compound ALK mutations G1202R/F1174C and G1202R/T1151M, a previously described “on-target” resistance mechanism.

The CHRONOS trial is currently testing the efficacy of a second anti-EGFR treatment in the third-line setting, in patients with RAS-wild metastatic colorectal carcinoma who received anti-EGFR agents in the first-line setting and who had a >50% drop in ctDNA RAS levels (RAS mutation being the most common resistance mechanism to anti-EGFR therapy) [202]. This approach would be particularly interesting in thoracic oncology where the decay of resistance mutations over time following the withdrawal of targeted therapy might allow for an effective rechallenge with a previously active targeted therapy [181].

Ortiz-Cuaran et al. [203] analysed ctDNA before treatment with anti-BRAF agents, upon response and progression. BRAF V600E mutation was detected in ctDNA in 72.7% of samples at baseline and in 54.3% of samples at disease progression during treatment with BRAF inhibitors. Longitudinal monitoring of ctDNA reflected the observed overall clinical and radiological tumour response. At baseline, BRAF mutations were associated with liver and adrenal metastases. The presence of concomitant activating mutations of FGFR2 and CTNNB1 was associated with a shorter time on BRAF TKIs. While reactivation of the MAPK pathway remained the most frequent resistance mechanism, other alterations in the PI3 K pathway as well as in IDH1, FGFR2 and CTNNB1 were identified and should be explored in upcoming studies.

Monitoring of treatment response

Besides its established use in newly diagnosed advanced NSCLC and in identifying secondary resistance to TKIs, plasma cfDNA could also be used to monitor treatment response. In a retrospective analysis of plasma samples in patients treated within the FASTACT-2 study, patients with undetectable mutant cfDNA after two cycles of therapy had a significantly longer PFS time compared to those with residual disease (12 months versus 7.2 months, hazard ratio 0.32; p<0.0001) [204]. Conversely, the persistence of T790M mutation in ctDNA 6 weeks after starting treatment with osimertinib in the AURA trial predicted a shorter mPFS (5.5 versus 10.9 months) and a decreased ORR (35% versus 70%) [205]. APPLE is an ongoing randomised, open-label, multicentre, phase II trial in advanced, EGFR-mutant, EGFR TKI-naïve NSCLC to evaluate the best strategy for sequencing gefitinib and osimertinib treatment. A cfDNA T790M test is used as a predictive marker for making treatment decisions in one of the three study arms. Positive results would prospectively validate the role of liquid biopsies in assessing tumour progression compared with standard radiological guidelines [206]. In addition, data from the AURA and FLAURA studies show that 15–32% of patients with NSCLC do not shed ctDNA into the circulation and have a better prognosis than those with detectable ctDNA, the latter usually reflecting a higher tumour burden [207].

In the previously mentioned VISION study of tepotinib in patients with MET exon 14 skipping mutations, a high concordance was found between the molecular cfDNA response and clinical response on the basis of Response Evaluation Criteria in Solid Tumours measurement [119].

Lung cancer screening

Through a tumour-specific phylogenetic approach, the TRACERx (Tracking Non-Small-Cell Lung Cancer Evolution Through Therapy) study revealed the presence of lymphovascular invasion, the non-ADC histology and a high Ki67 proliferation index as three independent predictors of ctDNA detection. ¹⁸F-2-fluoro-2-deoxy-d-glucose avidity on positron emission tomography predicted ctDNA detection and tumour volume from computed tomography (CT) analyses correlated with mean clonal plasma VAF. It was predicted that a primary tumour burden of 10 cm³ would result in a mean clonal plasma VAF of 0.1% (95% CI 0.06–0.18%). The minimal size and volume of a lung nodule detected by low-dose CT screening exam are 4 mm and 0.034 cm³, respectively, which would equate to a plasma VAF of 1.8×10⁻⁴% (95% CI 9.8×10⁻⁶–0.0033%). The latter value being the extreme of detection limits of current ctDNA assays, the sensitivity of ctDNA-directed early-NSCLC screening remains low [208].

NCT03774758 is an ongoing prospective observational study of ctDNA dynamics using the Guardant Health ct-DNA LUNAR assay in patients who undergo lung cancer screening at the San Francisco VA Medical Center, UCSF Medical Center and the San Francisco General Hospital (San Fransisco, CA, USA) [209]. The methylation of the promoter region of septin 9 (SEPT9), a tumour suppressor gene, is commonly observed in colorectal cancer and can be detected using ctDNA. Powrózek et al. [210] found a SEPT9 promoter methylation in 44.3% of lung cancer patients with different histological subtypes and disease stages. Similar to colorectal cancer, identification of SEPT9 promoter methylation using ctDNA might be a useful biomarker in early diagnosis of lung cancer [181, 210].

Liquid biopsy may also be used to identify other tissue- and cancer-specific epigenetic alterations such as DNA hydroxymethylation, inferred histone occupancy and DNA fragment length [181].

Micro-RNAs extracted from peripheral blood mononucleated cells have also been used to diagnose NSCLC with a sensitivity and specificity rates nearing 71.43% and 82.61%, respectively [211]. Furthermore, a panel of four miRNAs extracted from exosomes could detect lung cancer with 96% sensitivity and 60% specificity in a retrospective cohort of 30 people of whom 10 had lung cancer [212].

Combining ctDNA with low-dose CT of the chest could possibly improve the positive predictive value of the latter exam in lung cancer screening and reduce the need for unnecessary and potentially harmful follow-up procedures. In addition, improving the sensitivity of ctDNA testing to the point that it replaces imaging exams as a screening tool would reduce the risk of secondary radiation-induced cancers [181].

Early detection of relapse in the adjuvant setting

Liquid biopsy may serve as a future tool to identify earlier stage NSCLC patients at high risk of relapse, and to eventually personalise adjuvant therapy [182]. In the TRACERx study, patients with early NSCLC were followed-up regularly with serial liquid biopsies in addition to clinical and imaging assessment. Circulating tumour DNA detection in the plasma was defined as the detection of two out of 18 tumour-specific single-nucleotide variants (SNV) using multiplex-PCR assay panels. 13 (93%) out of 14 relapsing patients had positive ctDNA before or at clinical relapse. The median lead time defined as the median interval between ctDNA detection and CT-confirmed NSCLC relapse was 70 days (range 10–346 days). Detection of SNVs in ctDNA after surgery predicted relapse and resistance to adjuvant chemotherapy [208].

Liquid biopsy and immunotherapy

The integration of liquid biopsies into immune-oncology is challenging, especially since tissue biopsy provides a more comprehensive overview of the tumour microenvironment. Copy number instability of cfDNA could be used as a predictive biomarker of response to immunotherapy [213]. PD-L1 expression, a prognostic and predictive biomarker of response to ICIs usually assessed on tissue samples, was detected in CTCs with a PD-L1 specific antibody and in extracellular vesicles by Western blotting [214, 215]. Assessment of TMB, another predictive biomarker of efficacy ICIs, was also performed on ctDNA samples using panels of 54–70 genes [216]. In a retrospective analysis of two large randomised trials, Gandara et al. [217] showed that TMB measured in plasma (bTMB) was predictive of PFS (cut off ≥16 derived from sequencing 1.1 Mb of coding region) in patients treated with atezolizumab in second-line and higher NSCLC. In another cohort study of patients with NSCLC, a cancer gene panel named NCC-GP150 was used to estimate bTMB and correlated well with matched tumour TMB calculated by whole-exome sequencing. A bTMB ≥6 was associated with a better PFS and ORR in patients treated with anti-PD1 or anti-PDL1 agents [218]. In a single-institution phase II study of pembrolizumab in patients with advanced solid tumours, changes in ctDNA levels from baseline to the third cycle of treatment were predictive of benefit to pembrolizumab across cancer types and clearance of ctDNA during treatment yielded superior clinical outcomes independent of tumour type, TMB or PD-L1 status [219]. Changes in variant allele frequencies of all mutations were correlated with tumour response to durvalumab in a phase I/II trial in 304 patients with advanced NSCLC [220]. Other studies are needed to establish the clinical validity of serial ctDNA assays before and during treatment with ICIs, as a prognostic, predictive and pharmacodynamic tool [219]. In addition, ctDNA liquid biopsy might aid in the selection of patients who can benefit from de-escalation of treatment (after clearance of detectable ctDNA) especially in the field of immune-oncology where the optimal duration of treatment with ICIs remains unclear.

Conclusion

Modern thoracic oncology and precision medicine are inseparable. With the recent genomic sequencing of NSCLC tumours, we now possess huge amounts of genomic data that will markedly improve our understanding of lung carcinogenesis in order to develop appropriate, less toxic and more effective treatment strategies. Liquid biopsy is a precious tool that facilitates tumour sampling and has thus become quite endearing to both patients and physicians. However, caution is advised due to the lower sensitivity of plasma-based assays that cannot thus far replace tissue genotyping. Ensuring clinical validity and clinical utility will allow liquid biopsies to reach their full potential and significantly impact precision oncology with subsequent optimisation of cancer care.