EGFR-activating mutation occurs in 47% of Asian patients, 15% of European patients, 21% of African patients and has a higher mutation frequency in female patients [16]. Exon 19 deletion (EX19del) and exon 21 L858R mutation (L858R) are the main types of EGFR mutations, accounting for 47% and 41%, respectively. There are more than ten known rare mutations, including exon 19 insertion (EX19ins), exon 20 insertion (EX20ins), and point mutations S768I, L861Q, G719X, G709X [17,25].

So far, three generations of EGFR-TKIs have been approved for first-line treatment of NSCLC (Table 1), with favorable response rate (56%–83%) and progression-free survival (8.4–18.9 months) [26,27]. The first-generation EGFR-TKIs, such as gefitinib and erlotinib, are mainly used to treat EGFR EX19del and L858R mutant NSCLC by reversible interaction with tyrosine kinase [28]. Although the treatment showed a good positive clinical response (50%–80%), secondary drug resistance occurs after about one year, mostly caused by T790M mutation that was observed in ~60% of patients with acquired resistance [29–31]. The second-generation EGFR-TKIs are mainly irreversible ErbB family blockers, such as afatinib, dacomitinib and neratinib, which produce longer-lasting inhibitory effects than first-generation TKIs and more complete blockade of EGFR signaling pathway [32]. However, 40%–50% of patients treated with second-generation TKIs still develop drug resistance due to T790M mutation; and indiscriminate blockade of the EGFR signaling pathway may result in more severe side effects [33,34]. At present, represented by osimertinib, the third-generation EGFR-TKIs are in full bloom (Table 1) and mainly applied to treat NSCLC patients with T790M mutation-related secondary resistance [35]. These drugs selectively and irreversibly target EGFR L858R/EX19del/T790M mutations with about 200 times higher potency than targeting wild-type EGFR [36]. According to the AURA phase III clinical trial, the median PFS of osimertinib treatment was significantly higher than that of platinum plus pemetrexed therapy (10.1 vs. 4.4 months; HR, 0.30; 95% CI, 0.23–0.41; p < 0.001) in patients with T790M-positive NSCLC after first-generation EGFR-TKIs treatment [37]. When applied as first-line treatment, osimertinib compared with first-generation EGFR-TKIs significantly prolonged PFS (18.9 vs. 10.2 months; HR, 0.80; 95% CI, 0.37–0.57; p < 0.001) in patients with EGFR EX19del/L858R mutations [9,38]. Moreover, third-generation EGFR-TKIs also exhibited lower epithelial toxicity, with fewer adverse events of grade 3 or higher in osimertinib group vs. the second-generation EGFR-TKI group (40% vs. 48%) [39,40].

Currently, resistance mutations to third-generation drugs, bypass activations and some rare undruggable point mutations remain challenges for treating EGFR-mutant NSCLC; fourth-generation EGFR-TKIs and new therapeutic strategies are thus developed [41]. The mechanisms of resistance to third-generation EGFR-TKI treatment are heterogeneous. C797X, a point mutation at position 797 in exon 20 that is covalently linked to osimertinib, occurs in 22%–25% of T790M-positive patients treated with osimertinib, but is found in only 2% of patients treated with another third-generation EGFR-TKI rociletinib [42,43]. C797S is the main type of C797X mutation. Among osimertinib-treated patients with C797X mutations, the prevalence is 82% for C797S/T790M mutations in cis and 10% for C797S/T790M mutations in trans [44]. Interestingly, when C797S and T790M are mutated in trans but not in cis, lung cancer cells with ternary mutations (EGFR-activating mutation/T790M/C797S) retain sensitivity to first- and second-generation EGFR-TKIs [44,45]. In addition, cells with EGFR-activating mutation/C797S are sensitive to the first- and second-generation EGFR-TKIs [44]. Therefore, different generations of EGFR-TKIs can be combined to overcome resistance [41]. Meanwhile, allosteric drug designs for C797X mutation are ongoing [46]. An allosteric small-molecule inhibitor, JBJ-04-125-02 showed high inhibitory activity and low toxicity against EGFR L858R/T790M/C797S mutation [47,48]. BI-4020, a non-covalent triple mutant EGFR targeting inhibitor, effectively induced EGFR EX19del/T790M/C797S mutant tumour regression in a mouse xenograft model [49]. A growing number of fourth-generation EGFR-TKIs against C797X are in clinical trials (Table 1). In addition to the frequent C797X mutation, other causes of third-generation EGFR-TKI resistance include EGFR mutations like L718Q/V (4%), L792H (2%) and G796S (1%), other alterations like MET amplification (20%) and HER2 amplification (10%), as well as aberrant activation of RAS-MAPK pathway, PI3K pathway, etc. [50,51]. Studies have shown that the combination of MET inhibitors (such as crizotinib) and osimertinib overcame drug resistance in osimertinib-resistant EGFR-mutant NSCLC cell lines with MET amplification, suggesting that simultaneous targeting MET and EGFR may be an effective strategy [51]. The combination of anti-VEGF (vascular endothelial growth factor) therapy and EGFR-TKIs is also proposed as a promising strategy, yet further assessment is warranted because the combination of different anti-VEGF strategies and different generations of EGFR-TKIs had different outcomes. The PFS of erlotinib plus ramucirumab (a VEGFR2 inhibitor) group was superior to that of erlotinib plus placebo group in a phase III trial (median PFS: 19.4 vs. 12.4 months; HR, 0.59; p < 0.001) [52]. In 2020, ramutuzumab combined with erlotinib was approved for first-line treatment of NSCLC [25]. However, for NSCLC patients with acquired T790M mutations after failure on previous EGFR-TKI therapy, bevacizumab (a VEGF inhibitor) combined with osimertinib did not improve median PFS compared to osimertinib alone (15.4 vs. 12.3 months, stratified log-rank P = 0.83; HR, 0.96; 95% CI, 0.68–1.37), with a similar ORR (objective response rate) of 55% in both groups [25,53].

Efforts have also been made to develop EGFR-TKIs targeting rare EGFR mutations, such as EX20ins mutation which accounts for 4%–9% of EGFR mutations [54]. A phase II trial of mobocertinib, a TKI targeting both EGFR EX20ins and HER2 EX20ins, showed a confirmed ORR of 43% and a median PFS of 7.3 months [54]. Poziotinib as an irreversible pan-HER inhibitor inhibited the growth of EGFR EX20ins mutant cells with approximately ~100 times higher potency than osimertinib; however, it showed obvious toxic side effects in a phase II trial during which 60% of patients experienced grade 3 toxicity and 45% required dose reduction [55]. The preliminary efficacy of another inhibitor CLN-081 is being evaluated in a multi-center phase I–II study among NSCLC cases harboring EGFR EX20ins (NCT04036682). Moreover, amivantamab have been approved in 2021 for the treatment of locally advanced or metastatic NSCLC adult patients with EGFR EX20ins mutation [56] (Table 1) .

KRAS (Kirsten rat sarcoma oncogene homologue) mutation as another frequent driver mutation occurs in 20%–25% of NSCLC patients, with a higher incidence in Western populations than in Asian populations (26% vs. 11%) [18,19] (Fig. 2A). KRAS mutation frequency is closely related to smoking, ranging from 25% to 35% in smokers and 5% in non-smokers [57]. Of all KRAS mutations, 83% were located at codon 12 and 14% at codon 13; G12C is the most common alteration at codon 12 (41%), followed by G12V (22%) and G12D (16%) [58–60].

KRAS is a membrane-regulatory small guanine nucleoside bound protein (G protein), which has a smooth surface and lacks binding pockets, making it difficult to target. Belonging to the guanosine triphosphatase (GTPase) family, KRAS exists in two different states (GDP binding-inactivation state, GTP binding-activation state). The high concentration of GTP in vivo and the high affinity of GTP to KRAS are also difficulties in developing competitive inhibitors for KRAS [61,62]. Therefore, indirect strategies are adopted, with focus on reducing KRAS expression, blocking the membrane position of KRAS, interfering with the signal transduction of KRAS; however, none of these efforts has achieved clinical application [63]. For example, salirasib, as a farnesyltransferase inhibitor, inhibited the modification of KRAS protein to hinder its binding to the membrane but showed insufficient therapeutic activity in phase II clinical trials [64].

The emergence of small molecules directly targeting KRAS mutants changed this situation. KRAS G12C mutant has a binding pocket near the 12th cysteine residue when binding to GDP in an inactivation state, which can be utilized for drug design to stabilize this inactivation state [65,66]. This notion led to the development of sotorasib (AMG-510) [67]. NSCLC patients treated with sotorasib showed ORR of 37.1% (95% CI, 28.6–46.2), DCR (disease control rate) of 80.6% (95% CI, 72.6–87.2), and median PFS of 6.8 months (95% CI, 5.1–8.2); moreover, the drug has a tolerable safety profile: 19.8% of patients experienced grade 3 adverse events, and 0.8% experienced grade 4 adverse events [68]. In 2021, sotorasib was approved for the treatment of NSCLC patients with KRAS G12C mutation who have undergone at least one systemic treatment [69]. Patients treated with another oral KRAS G12C inhibitor adagrasib (MRTX849) showed ORR of 42.9%, DCR of 79.5%, median DOR (duration of response) of 8.5 months and median PFS of 6.5 months in phase I/II trial (NCT03785249) [70,71]. Recently, PROTACs (proteolytic targeting chimeras) comprising of ligands targeting proteins of interest, E3 ligase recruiting elements and linkers, show game-changing potential in treating cancers driven by mutant proteins without deep binding pockets via selectively mediating ubiquitin-proteasome degradation of target proteins [72]. LC-2, the first-in-class endogenous KRAS G12C degrader, combines adagrasib warhead and a E3 ligase ligand VHL to form a ternary complex with KRAS G12C and induce proteasome degradation [73]. Another example is KRAS G12C degrader YF135, which induces VHL-mediated KRAS G12C degradation and attenuates pERK signaling in a reversible manner [74]. In addition, KRAS G12D inhibitors achieved a breakthrough recently. The first non-covalent selective KRAS G12D inhibitor MRTX1133 showed a 1000-fold higher potency against KRAS mutants than wild-type KRAS in vivo [75]. It is also demonstrated that selective targeting of KRAS G12D can be achieved by cyclic peptides in the GTP-bound state, suggesting that peptides may be an option for targeting KRAS G12D beyond small chemical molecules [76]. Currently, inhibitors of KRAS G12V and other KRAS mutations have not been reported.

In clinical practice, poor clinical efficacy occurred in 50%–60% of patients receiving KRAS inhibitors. A possible reason is that some KRAS mutant cells have low KRAS dependency, due to the activation of other pathways, such as AKT and mTORC1 pathways, that lead to intrinsic resistance; while the investigation is warranted to further interpret the mechanism [77]. Acquired resistance to KRAS inhibitors also occurs, possibly due to abnormal compensatory activation of bypass pathways [78]. The reactivation of the adaptive KRAS feedback pathway is mediated by multiple RTKs (receptor tyrosine kinases) after treatment with sotorasib and another KRAS G12C inhibitor ARS-1620, while SHP2 inhibitor can serve as a blocker of multiple RTK signal transduction to suppress KRAS feedback pathway and be applied with KRAS G12C inhibitor in combination [79]. Activation of PI3K pathway is also responsible for acquired resistance to KRAS G12C inhibitor by promoting epithelial-mesenchymal transition (EMT) [80]. Additionally, treating KRAS G12C mutant cells with sotorasib and adagrasib, respectively resulted in totally 142 resistant clones, of which 124 (87%) harbored secondary KRAS mutations. In sotorasib resistant clones, KRAS G13D was the most common secondary mutation (23%), followed by R68M (21.2%) and A59S (21.2%). In adagrasib resistant clones, KRAS Q99L was the most common secondary mutation (52.8%), followed by Y96D (15.3%) and R68S (13.9%) [69,81]. Acquired resistance to KRAS G12C inhibitors and underlying mechanisms have received great attention nowadays and has been systematically reviewed [82–84].

To overcome drug resistance and maximize the potential of KRAS inhibitors, multiple combination strategies are designed. As mentioned above, the combination of KRAS inhibitors and SHP2 inhibitors showed stronger efficacy. Co-treatment of ARS-1620 and a SHP2 inhibitor SHP-099, compared to mono treatment, led to a more significant reduction of tumour volume in vivo; the combination of adagrasib and another SHP2 inhibitor RMC-4550 also showed higher anti-tumour activity in adagrasib-resistant cells [80,85]. The therapeutic effect of combining KRAS G12C inhibitors with SHP2 inhibitors is under evaluation in ongoing clinical trials (NCT04330664 and NCT04185883). Combined use of KRAS inhibitors and inhibitors targeting SOS1 (another common downstream effector in multiple RTK signaling pathways), such as BAY293, have shown synergistic anti-tumour effects [86]. The combination of KRAS inhibitors with tumour metabolism therapy or immune therapy is also of great concern [63].

Mesenchymal-epithelial transition (MET) encoded by MET proto-oncogene is a tyrosine kinase receptor that binds to hepatocyte growth factor (HGF). MET gene amplification and exon 14 skipping (EX14ski) are main categories of MET mutations. Mutations in MET exon 14 are found in about 3% of NSCLC patients. Importantly, MET amplification is significantly increased in patients with EGFR-TKI resistance, accounting for 5%–20% of patients with first-generation TKI resistance and up to 25% of patients with third-generation TKI resistance [20,51]. MET mutations and amplification cause activation of a series of pathways, including RAS, ERK/MAPK, PI3K/AKT, Wnt/beta-catenin, JAK/STAT pathway, thereby promoting the proliferation and migration of cancer cells [87]. Currently, several MET inhibitors have been approved. Capmatinib, a highly selective oral MET inhibitor, was approved in 2020 for the treatment of metastatic NSCLC patients harboring MET EX14ski. Patients treated with capmatinib showed ORR of 41% (95% CI, 29–53) and DOR of 9.7 months (95% CI, 5.5–13.0) [88]. Another MET EX14ski inhibitor tepotinib received FDA approval in 2021, with ORR of 54.4%, median DOR of 18.5 months, median PFS of 12.1 months, and median OS of 20.4 months [89,90]. Other MET targeting drugs in clinical practice include savolitinib, which is applied for the treatment of metastatic NSCLC carrying MET EX14ski mutation who progresses after chemotherapy or cannot tolerate platinum-based chemotherapy; and amivantamab, a bispecific monoclonal antibody targeting EGFR and MET, is recently approved for treating NSCLC patients with EGFR EX20ins mutation, and its effect on those with MET amplification warrants further evaluation [91,92]. In addition, different types of MET inhibitors are under development worldwide, including chemicals, monoclonal antibodies, polyclonal antibodies, and ADCs (antibody-drug conjugates), among which glumetinib (NCT05507294), telisotuzumab vedotin (NCT03539536), Gb263T (NCT05332574), MCLA-129 (NCT04868877) have entered clinical trials (Table 1).

Human epidermal growth factor receptor 2 (HER2; eRBB2) is a member of the tyrosine kinase receptor family, which also includes EGFR, HER1, HER3 and HER4 [93]. HER2 mutations occur in about 2%–4% of NSCLC cases, the most common of which is exon 20 insertion mutation; other point mutations have also been reported, including G776C, L755S, etc. HER2 amplification, as previously described as a mechanism of EGFR-TKI resistance, occurs in approximately 3% of patients who have not been treated with EGFR-TKIs, with a significant increase of incidence (approximately 10%) among patients with EGFR-TKI resistance [21,94,95]. In 2022, enhertu became the first drug approved by the FDA for NSCLC with HER2 mutation. According to data from its phase II trial (NCT03505710), the median DOR was 9.3 months (95% CI, 5.7–14.7), median PFS was 8.2 months (95% CI, 6.0–11.9), and median OS was 17.8 months (95% CI, 13.8–22.1) [96,97]. In addition, another HER2 inhibitor XMT-1522, an auristatin-derivative molecule conjugated to a novel compound dolaflexin, was well tolerated in a phase I trial and showed early signs of anti-tumour activity with DCR of 83%, partial remission of 17% and stable disease of 67% [15]. Current clinical trials about HER2-targeting ADCs are overall encouraging, bringing hope for the NSCLC patients with HER2 mutations and those with acquired HER2-related TKI resistance [98].

Anaplastic lymphoma kinase (ALK) is a tyrosine kinase receptor, and its rearrangement is reported in 3%–7% of global NSCLC cases [22]. Echinoderm microtubule-associated protein-like 4 (EML4) is ALK’s most common fusion partner. In addition, there are at least 20 other fusion genes, such as TGF-ALK, KIF5B-ALK, and STRN-ALK [99]. Crizotinib was the first ALK-TKI drug approved as a second-line treatment for ALK-positive NSCLC in 2011 and approved by European Medicines Agency to be applied in first-line NSCLC treatment in 2015. The second-generation ALK-TKI drugs alectinib, ceritinib, brigatinib, and ensatinib achieve greater curative effects, yet acquired resistance and recurrence are inevitable [100]. ALK mutation is a common cause for ALK-TKI resistance; frequently seen resistant mutations to crizotinib include L1196M, G1269A, C1156Y, G1202R, I1171T/N/S, S1206C/Y, E1210K, L1152P/R, V11180L, I1151T, G1128A, and F1174V [101]. The most known ALK G1202R mutation occurred in 21%, 29% and 43% of patients resistant to treatment of ceritinib, alectinib, and brigatinib [102]. A noteworthy fact is that patients carrying ALK-rearranged NSCLC with a prior history of ALK inhibitors have a high incidence (45%–70%) of central nervous system (CNS) metastases, indicating that brain metastasis is a common failed form of ALK targeted therapy [103]. Compared with second-generation inhibitors, third-generation ALK-TKI lorlatinib is designed to overcome known secondary resistance mutations in the ALK tyrosine kinase domain and to penetrate the CNS [104]. Preclinical studies demonstrate that lorlatinib is effective against most known single ALK-resistant mutations, including the highly refractory ALK G1202R [104,105]. Consistently, in a phase II study, lorlatinib showed beneficial activity in patients treated with first- or second-generation ALK inhibitors [106].

The ROS1 gene belongs to the subfamilies of tyrosine-kinase insulin-receptor genes. ROS1 fusions produce defective genes that act as tumour drivers, leading to excessive proliferation of tumour cells. ROS1 fusions occur in about 1%–2% of NSCLC patients [23]. This change is most common in patients with NSCLC who have adenocarcinoma and are also negative for ALK, KRAS, and EGFR mutations. The kinase domains of ROS1 and ALK share about 70% homology. Crizotinib, which is approved for the ALK-positive NSCLC, is also an inhibitor of ROS1 and improves survival in NSCLC patients with ROS1 fusions [107]; in 2016, Crizotinib was approved for the treatment of ROS1-positive NSCLC. Ceritinib, a second-generation ALK inhibitor, also benefit patients with ROS1-positive NSCLC [108]. In 2019, entrectinib that targets ROS1 and ALK was approved for adults with metastatic ROS1-positive NSCLC [109]. Entrectinib can pass through the blood-brain barrier and is clinically proven effective against primary and metastatic brain diseases [110]. Moreover, preclinical studies suggested that the third-generation ALK-TKI lorlatinib can effectively inhibit ROS1 mutants [111]. Acquired resistance to ROS1-TKIs can be mediated by secondary mutations within the ROS1 kinase domain (E1935G, L1947R, L1951R, G1971E, L1982F, S1986F/Y, L2026M, G2032R, D2033N, C2060G, V2098I and L2155S and L2086F), or by activation of alternative signaling pathways (KRAS, NRAS, EGFR, HER2, MET, BRAF and MEK) [112]. ROS1 G2032R, the most common resistance substitution found in approximately one-third of the cases, is highly resistant to crizotinib as well as entrectinib and lorlatinib [113]. A new selective ROS1 inhibitor DS-6051b may overcome G2032R drug resistance as demonstrated in a preclinical study [114].

The RET gene is located on human chromosome 10 and encodes a single-pass transmembrane RTK. RET fusion, as an independent oncogenic driver, occurs in 1%–2% of NSCLC. Chromosomal fusions between the RET gene and its fusion partners lead to RET overexpression [115]. The most common gene fusion partners are KIF5B and CCDC6, accounting for about 70%–90% and 10%–25% of RET-positive cases, respectively. The chimeric fusion proteins can activate the ligand-independent activation of RET and promote the growth and survival of cancer cells [116]. Recently, the development of highly selective RET inhibitors, such as an oral small-molecule inhibitor selpercatinib (LOXO-292), has greatly improved the outcome of RET fusion-positive NSCLC patients [117,118]. Selpercatinib actively against not only RET fusions (KIF5B-RET, CCDC6-RET, etc.) but also some RET-activating point mutations (V804L, V804M, and M918T) [119]. The ORR of selpercatinib treatment was 64% (95% CI, 54–73) in RET fusion-positive NSCLC patients who previously received at least platinum-based chemotherapy and was 85% (95% CI, 70–94) in those untreated [120]. Moreover, another approved highly selective RET inhibitor pralsetinib (BLU-667) was able to overcome acquired resistance to EGFR-TKIs, including osimertinib, according to data from cell studies and the clinic [121].

BRAF is a cytosolic serine/threonine kinase belonging to the RAF kinase family, serving as an important step of signal transmission from the cell surface to the nucleus after EGFR activation [122]. As a part of MAPK pathway, BRAF is involved in cell growth, proliferation, survival, and differentiation [123]. BRAF mutations occur in 1.5%–3.5% of NSCLC cases. BRAF-activating mutations are divided into BRAF V600E and BRAF non-V600E mutation; the former accounts for more than 50% of BRAF mutated NSCLC cases [124] and appears more common in female patients with lung adenocarcinoma, while the latter is more common in smokers [24,125]. The PFS of NSCLC patients with BRAF V600E mutation is shorter than those without [126]. Inhibitors have been developed to specifically bind to the ATP binding pocket of mutant BRAF, especially BRAF V600E, such as vemurafenib or dabrafenib [127]. Vemurafenib was the first MAPK inhibitor tested in BRAF mutant lung cancer. however, resistance eventually develops, mostly due to MAPK pathway reactivation [128]. Combined use of MEK inhibitors such as binimetinib can maximally block MAPK pathway and delay the emergence of drug resistance. In 2017, a combination therapy of dabrafenib and a MEK inhibitor trametinib received FDA approval for the treatment of metastatic NSCLC carrying BRAF V600E mutation [129]. In addition, cancer cells carrying BRAF V600E showed resistance to osimertinib, while a BRAF V600E inhibitor encorafetnib restored osimertinib sensitivity, suggesting its potential to promote the therapeutic effect of the third-generation EGFR-TKIs [130].

DNA methylation mainly occurs in CpG dinucleotides (concentrated in high-density regions called CpG islands), which inhibits the binding of RNase to gene fragments and thereby silences related genes [172,173]. DNA methyltransferases (such as DNMT1, DNMT3A and DNMT3B) and DNA demethylases (such as TET1, TET2 and TET3) are mainly responsible for the regulation of DNA methylation [168,174]. The functions of DNMT members in NSCLC vary; DNMT1 knockdown inhibits the growth of lung cancer cells in vitro and in vivo, whereas low expression of DNMT3A is associated with poor prognosis, and knockout of DNMT3A in Kras mutant mouse models promotes tumour growth and progression [175,176]. DNMT inhibitors, including azacitidine and decitabine, have not shown obvious efficacy on NSCLC in early clinical trials. However recent preclinical studies demonstrated that they could restore cancer cell sensitivity to EGFR-TKIs. Combined treatment of azacitidine and gefitinib lead to growth inhibition and apoptosis of drug-resistant cancer cells [177,178]. Loss-of-function mutations of TET can promote cancer development, and these mutations often co-occur with oncogenic KRAS mutations. In a Kras G12D NSCLC mouse model, deletion of TET promoted tumour development by up-regulating Wnt signaling pathway [179]. In addition, gefitinib repressed TET1 through the C/EBPα transcription factor, and knockdown of TET1 resulted in resistance to gefitinib. TET1 expression was also lower in gefitinib-resistant patients than in sensitive patients [180]. More recently, a correlation between DNA methylation and EGFR-TKI response was found in 79 patients with NSCLC. Transcription factor enrichment analysis showed that the hypermethylation in the enhancer region of HOXB9 mainly occurred in patients with poor EGFR-TKI response [181]. Although some agents targeting DNA methylation are under clinical trial (Table 2), more trials are warranted to further evaluate whether inhibition of DNA methylation can be applied to fight against EGFR-TKI resistance. Moreover, challenges remain in dealing with the low specificity and high side effects of DNA methylation drugs. These drugs can be a double-edged sword, up-regulating tumour suppressor genes while also activating proto-oncogenes [182]. Preclinical studies demonstrated that combination therapy could help overcome this limitation by countering the dependency on the activated oncogene or synergistically strengthening tumour-suppressing effect. For example, upregulation of oncofetal protein SALL4 was induced in SALL4-negative cancer cells treated with azacytidine, while this also provides a vulnerability to entinostat, which can suppress SALL4; azacytidine in combination with entinostat significantly repressed tumour growth [183]. Combination of azacytidine with histone deacetylase inhibitors also exhibited a robust anti-tumour effect in NSCLC cells and mouse models by reversing tumour immune evasion [184]. Appropriate dosage is also important for using DNA methylation drugs in clinical practice since adverse events caused by high doses were observed in earlier clinical trials [11,172].

Histones are key components of chromatin, and their post-translational modification plays a crucial role in regulating gene expression and developing lung cancer [185,186]. Chromatin modification enzymes are classified into histone methyltransferases (HMTs), histone demethylasehistone (HDMs), histone acetyltransferases (HATs) and histone deacetylases (HDACs) [187,188]. HMTs play a key role in many cellular processes and are typically dysregulated in cancer, with diverse consequences. One of the most well-studied HMTs in NSCLC is EZH2. Overexpression of EZH2 promotes lung cancer progression through multiple signaling pathways, including VEGF-A, AKT, E2F/Rb and TGF-β, which are also associated with resistance to chemotherapy and poor survival [189,190]. EZH2 inhibitors such as JQEZ5 and GSK126 showed anti-tumour effects against NSCLC in preclinical studies, yet further evaluation in clinical trials is warranted [191,192]. In contrast, the absence of another HMT called SETD2 leads to accelerated tumour progression in a Kras G12D NSCLC mouse model [193]. These contrary results indicate heterogeneous roles of HMTs in NSCLC development. In addition, a histone lysine methyltransferase SMYD3 is up-regulated in Ras-driven lung cancer cells. It enhances the activation of Ras/Raf/MEK/ERK signaling module through methylating a non-histone protein MAP3K2, suggesting the potential of epigenetic regulators as a therapeutic target for NSCLC patients with KRAS inhibitor resistance [194].

Inhibition of HDACs alone lacks specific efficacy in clinical trials, but its combination with targeted drugs has great potential in overcoming TKI resistance. HDAC inhibitors help to overcome EGFR-TKI resistance by suppressing EMT [12,195] and inhibiting the self-renewal of cancer stem cells [196]. Preclinical studies have shown that vorinostat, which selectively inhibits HDAC3, increases EGFR-mutant cellular sensitivity to osimertinib in vitro and in vivo [197]. Moreover, vorinostat combined with ALK-TKI brigatinib showed more potent anti-tumour activity in lung adenocarcinoma cells carrying EGFR L858R/T790M/C797S mutations [198]. In a phase II trial, patients with high E-cadherin levels treated with erlotinib combined with an HDAC inhibitor entinostat had longer OS than those treated with erlotinib alone (9.4 vs. 5.4 months) [199]. More combination therapies, including HDAC inhibitors and EGFR-TKIs (NCT02151721, NCT02520778), HDAC inhibitors and DNMT inhibitors (NCT00387465) are under clinical trials, providing ideas for overcoming targeted resistance and enhancing TKI efficacy in NSCLC [7] (Table 2). In addition, dual inhibitors against driver mutations and HDAC are being developed. Compound 9E, a dual inhibitor based on osimertinib and vorinostat, showed superior antiproliferative activity against several tumour cell lines [200].

Similar to DNA methylation, the adverse reactions of HDAC inhibitors also raised broad concern. The use of nanocarrier technologies (such as polymeric nanoparticles, PEG-coated nanoparticles, and colloid carrier systems) can deliver HDAC inhibitors with enhanced solubility, tumour specificity and less toxicity [201,202]. Developing HDAC inhibitors with higher tumour selectivity, exploring proper timing for administration, and identifying predictive biomarkers to better select patients will also be helpful in improving HDAC-based therapy [12].

Non-coding RNAs are involved in the occurrence and development of lung cancer. They are closely related to targeted therapy resistance, among which microRNAs (miRNAs), long non-coding RNAs (lncRNAs) and circular RNAs (circRNAs) have been most widely studied [170,212,213]. Although most studies are preclinical, emerging positive results suggest non-coding RNAs as potential therapeutic targets for NSCLC (Table 3). MiRNAs typically consist of 18–25 nucleotides that target RNA to regulate gene expression. Their contribution to TKI resistance involves multiple signaling pathways (such as PI3K/AKT/mTOR pathway) and EMT process [214]. For example, miR-483-3p can induce EMT and produce resistance to gefitinib through methylation of its own promoter [215]. In both univariate and multivariate analyses, gefitinib was associated with a significant improvement in OS in NSCLC patients with reduced miR-21 expression, suggesting targeting specific miRNA may promote response to EGFR-TKI [216]. Moreover, miRNAs are also involved in chemotherapy or radiotherapy resistance. After radiotherapy, NRF2-induced up-regulation of miR-140 transcription plays an important role in obtaining radiation protection [217].

LncRNA is non-coding RNA with a length >200 bp, which is a major regulator of gene expression [212]. LncRNA can interact with miRNA and affect the proliferation, invasion and metastasis of NSCLC cells [218]. LncRNA XLOC_008466 is highly expressed in patients with NSCLC and binds to miR-874 to down-regulate its level and increase miR-874 downstream target expression to promote cancer cell proliferation and invasion [219]. A tumour suppressor lncRNA SNHG10 can significantly reduce the miRNA-21 level and is positively associated with better survival and prognosis [220]. LncRNA can also interact with various signaling pathways such as Wnt, STAT3, PTEN/PI3K/AKT pathways, as well as histone modifiers like EZH2 [212]. For example, lncRNA CBR3-AS1 promotes migration and invasion of lung adenocarcinoma cells by activating Wnt/β-catenin signaling pathway [221]. Another LncRNA, TSLNC8, significantly enhanced the anti-tumour effect of osimertinib by inhibiting the EGFR-STAT3 signaling pathway [222]. LncRNA CASC9 repressed tumour suppressor DUSP1 by recruiting EZH2, thereby promoting gefitinib resistance in vitro and in vivo [100]. Moreover, the expression of LncRNA H19 is increased in gefitinib-resistant cells, and it can be transferred to non-resistant cells through exosomes to “spread” drug resistance [223]. Numerous preclinical studies (Table 3) suggested the potential of lncRNA in suppressing cancer progression and the occurrence of drug resistance, while current clinical trials of lncRNA-based therapies for NSCLC remain rare.

CircRNAs are a kind of non-coding RNAs with a stable covalent closed loop structure, with recently reported involvement in lung cancer development [224]. The production of some circRNAs is closely related to the oncogene fusion gene. F-circEA-2a, a novel circRNA produced by the EML4-ALK fusion gene, promotes the migration and invasion of cancer cells in EML4-ALK positive NSCLC [225]. An important function of circRNAs is to act as miRNA sponges to interact with miRNAs, forming a circRNA-miRNA-mRNA regulatory axis in lung cancer that regulates related gene expression [226,227]. CiR-7 interacts with miR-7 and down-regulates miR-7 level, thus promoting lung cancer cell proliferation, migration and invasion by upregulating genes such as NF-κB, EGRF, CCNE1, PIK3CD, etc. [228,229]. In addition, a recent study has shown that circRNA is associated with drug resistance. The up-regulation of hsa_circ_0004015 significantly increased cellular resistance to gefitinib, while its down-regulation decreased gefitinib IC50 in resistant cancer cells, mechanistically via a hsa_circ_0004015/miR-1183/PDPK1 axis [230]. More agents targeting circRNAs are summarized in Table 3.