Background
Lung cancer is the second most frequent cancer and the leading cause of cancer-related death worldwide [
1]. Non-small cell lung cancer (NSCLC) is the major type of lung cancer, and around 14–38% of NSCLC patients harbor genetic alterations in epidermal growth factor receptor (
EGFR) [
2], with the incidence of
EGFR mutations higher in East Asian patients than in Caucasian patients [
3,
4]. Short in-frame deletions in exon 19 (19-Del) and point mutations in
EGFR exon 21 p.L858R are the most common activating mutations in
EGFR, accounting for approximately 90% of all
EGFR mutations in NSCLC [
5,
6]. EGFR tyrosine kinase inhibitors (TKIs) have shown profound clinical benefits and are thus used as the first-line treatment in
EGFR-mutated NSCLC patients [
7‐
12]. Besides 19-Del and
EGFR exon 21 p.L858R, extensive research has uncovered a wide array of rare
EGFR activating or resistant mutations in NSCLC, including
EGFR exon 18 p.G719X,
EGFR exon 20 p.S768I,
EGFR exon 21 p.L861Q,
EGFR exon 20 p.T790M, and
EGFR exon 20 insertions (20ins). Qin et al. found that
EGFR 20ins had at least 80 different insertion patterns, and lung cancer patients with
EGFR 20ins showed different clinical responses to various EGFR TKIs [
13]. The
EGFR exon 20 p.T790M mutation confers drug resistance to first-generation EGFR TKIs, and it has been shown to occur in 1–2% of treatment-naïve
EGFR-mutated NSCLC patients [
14,
15]. In addition to these well-studied common and rare
EGFR mutations,
EGFR variants of uncertain significance (VUS) were observed in lung cancer patients, but the clinical relevance and TKI sensitivity of these VUSs are largely unknown [
16,
17].
Although the majority of
EGFR-positive NSCLC patients harbor a single
EGFR mutation, recent advances in next-generation sequencing (NGS) technologies have revealed that around 10% of patients harbor compound
EGFR mutations, defined by the presence of double or multiple distinct
EGFR genetic alterations at baseline [
18‐
20]. Several groups reported that patients with compound
EGFR mutations tended to be less responsive to TKI therapies than those with a single
EGFR mutation [
21‐
24]. Furthermore, researchers found that the different types of
EGFR compound mutations might be associated with distinct treatment efficacies [
18,
19]. Despite the potential clinical implications of
EGFR compound mutations, most of the previous studies were based on limited patient cohorts, so it is imperative to perform large-scale analyses to gain a deeper insight into the complexity and diversity of compound
EGFR mutations in NSCLC. In the present study, we retrospectively studied the NGS data of treatment-naïve tumor samples from 8485
EGFR-mutated NSCLC patients, of whom 1025 had compound
EGFR mutations. We explored the clinical characteristics and genetic architecture of different types of compound
EGFR mutations, as well as their responses to EGFR TKIs and the associated drug-resistant mechanisms.
Discussion
We performed a large-scale retrospective study of 1025 NSCLC patients who harbored baseline compound EGFR mutations. Intriguingly, compound EGFR mutations had a significantly higher frequency of EGFR exon 21 p.L858R and rare EGFR mutations and a dramatically lower rate of EGFR 19-Del mutation than single EGFR mutation. Different types of compound EGFR mutations demonstrated distinct subtypes of mutated genes, aberrant signaling pathways, mutational signatures, and chromosomal instability. Notably, the rare EGFR mutation-dominant subtypes were associated with significantly shorter FPS. In addition, VUSs in the rare + VUSs subtype were more likely to locate at the EGFR kinase domain, and patients with rare + VUSs (KD +) had worse PFS than those with other VUS-containing subtypes. In terms of TKI-resistant mechanism, the common + VUSs subtype was highly enriched for EGFR exon 20 p.T790M and/or other RAS/RAF/MEK pathway-related mutations. Therefore, different compound EGFR mutation subtypes had distinct clinical/genetic characteristics and therapeutic responses.
The first-generation EGFR TKIs (e.g., gefitinib and erlotinib) are ATP‑competitive small molecules that reversibly target the EGFR tyrosine kinase domain. Despite its significant clinical benefits when compared with chemotherapies in NSCLC patients, drug resistance inevitably developed [
43]. To overcome the resistance to first-generation TKIs, the second-generation TKIs (e.g., afatinib and dacomitinib), which are irreversible inhibitors, were designed. Although second-generation TKIs generally showed improved EGFR inhibition, they also exhibited high potency against wild-type EGFR, leading to lower maximum dose tolerance, more adverse events, and limited clinical utilities [
44,
45]. One of the most common resistance mechanisms against both the first- and second-generation TKIs is
EGFR exon 20 p.T790M mutation [
46‐
48]. The gatekeeper hypothesis suggests that the steric hindrance between the methionine residue on the gatekeeper side chain of
EGFR exon 20 p.T790M and the aniline moiety of first-generation TKIs is the underlying mechanism of the drug resistance, although other putative mechanisms have been proposed, including elevated ATP-binding affinity for
EGFR exon 20 p.T790M, changes in the catalytic domain, and variations in the conformational dynamics [
49,
50]. In our study, we found that a significant proportion of patients with common + VUSs subtype (44%) acquired
EGFR exon 20 p.T790M mutation after EGFR TKI treatments, but not for other
EGFR subtypes. Because the percentage of acquiring
EGFR exon 20 p.T790M is similar between the common + VUSs subtype in our study and other studies using patients with a single
EGFR common mutation [
51], we speculate that the common + VUSs subtype might resemble the function of a single
EGFR common mutation. In particular, the
EGFR VUSs in the common + VUSs subtype might be passenger mutations and did not contribute to the oncogenic activation of EGFR. In contrast, some
EGFR compound mutation subtypes (e.g., rare + rare and rare + VUSs) are less likely to acquire
EGFR exon 20 p.T790M, implying that these subtypes might rewire the signaling network to make them prone to utilize other resistance mechanisms to bypass first- and second-generation TKIs. The third-generation TKIs, especially osimertinib, demonstrated satisfactory efficacy against
EGFR exon 20 p.T790M. Osimertinib formed irreversible covalent bonds with the cysteine 797 residue in the ATP-binding site, and it exhibited selective potency against the mutant EGFR rather than wild-type EGFR, resulting in its accelerated approval by US Food and Drug Administration to treat
EGFR-mutated NSCLC [
52]. One patient in our cohort gained
EGFR exon 20 p.C797S mutation after first-line TKIs and became resistant to osimertinib. This patient might be treated with TKI combinations or next-generation TKIs to overcome this resistance mutation [
53].
Around 12.1% of
EGFR-positive NSCLC patients in our cohort harbored compound
EGFR mutations, which is consistent with previous studies [
18‐
20]. Only around 2% of all compound
EGFR mutation-positive patients had more than 2 baseline
EGFR mutations, and these patients generally had high tumor mutation loads. Kauffmann-Guerrero et al. reported that compound
EGFR mutations were more often observed in patients with a smoking history [
22]. Although our patient cohort did not have complete records of the patient’s smoking status, the mutational signature results suggested that not all subtypes of compound
EGFR mutations had the same level of smoking-related signatures, with common + VUSs and rare + rare subtypes being more likely to occur in smokers than other subtypes. Additionally, Kim et al. found that compound
EGFR mutations were frequently co-mutated with some actionable genes, such as
ALK rearrangement,
KRAS mutation, and
PIK3CA mutations [
23]. We also detected multiple co-mutated genes, which exhibited distinct subtypes according to the specific type of compound
EGFR mutations. Particularly, unlike other compound
EGFR mutations, the rare + rare subtype had a significantly low frequency of mutations in the PI3K and RAS/RAF/MEK signaling pathways, implying that tumors harboring double rare
EGFR mutations might less rely on these oncogenic pathways. On the other hand, the VUSs + VUSs subtype had the highest mutational frequency in almost all the tested oncogenic pathways. This indicates that many of the detected
EGFR VUSs might have little or very mild oncogenic activities, and tumors harboring the VUSs + VUSs subtype had to heavily depend on other oncogenic mutations for tumorigenesis and maintenance.
Another interesting observation of our study is that compound
EGFR mutations had a much lower frequency of
EGFR 19-Del and a significantly higher frequency of
EGFR exon 21 p.L858R than the single
EGFR mutation. The two types of common
EGFR mutations also had different preferences in the co-existed
EGFR mutations. Furthermore, the
EGFR 19-Del + X subtype and
EGFR exon 21 p.L858R + X subtype had opposite trends in the therapeutic response to second-generation TKIs. Multiple previous studies on single
EGFR mutation have found that
EGFR 19-Del and
EGFR exon 21 p.L858R demonstrated different clinical features and treatment outcomes. Hong’s group reported that patients with a single
EGFR 19‑Del mutation had significantly improved clinical outcomes than patients with a single
EGFR exon 21 p.L858R mutation following first‑line TKI, but not first‑line chemotherapy or second‑line TKI [
54]. NSCLC patients with
EGFR 19-Del also had a higher risk of lymph node metastasis than those with
EGFR exon 21 p.L858R [
55]. Despite the clinical difference between
EGFR exon 21 p.L858R and
EGFR 19-Del, the underlying mechanism is still elusive. Sordella et al. discovered that
EGFR exon 21 p.L858R and
EGFR 19-Del had differential levels of EGFR autophosphorylation on some specific sites, which may affect their drug sensitivity to TKIs [
56]. Nevertheless, future studies are needed to elucidate the distinguishing preference of
EGFR exon 21 p.L858R and
EGFR 19-Del in compound
EGFR mutations.
We found that patients with compound
EGFR mutations tended to be less responsive to EGFR TKIs than those with single
EGFR mutation, especially the patients with single
EGFR 19-Del, which is consistent with previous studies [
21‐
24]. Additionally, we discovered that different subtypes of compound
EGFR mutations were also significantly associated with patient’s prognosis to first-line TKIs. Specifically, the presence of a common mutation in compound
EGFR mutations can sufficiently predict prognosis, regardless of the type and location of the other
EGFR mutation. However, for rare
EGFR mutation-containing patients, their prognosis is likely to highly rely on the type of mutation combinations. In particular, rare + common was associated with good PFS, rare + VUSs (KD −) might be related to good to intermediate PFS, while rare + rare and rare + VUSs (KD +) are likely to associate with short PFS. Therefore, both the type of
EGFR mutations (common vs rare vs VUSs) and the specific combination of compound mutations might contribute to the overall prognosis of NSCLC patients.
The common + common subtype was extremely rare, accounting for only 2.3% of patients in our cohort. Given that common
EGFR mutations could efficiently activate EGFR kinase activity and promote tumorigenesis, it is highly unlikely that a single tumor would acquire two
EGFR common mutations simultaneously. As a result, we suspect that the two different
EGFR common mutations might mainly reside in different tumor cells. In other words, we think those patients might have two subclones of cancer cells, one is driven by
EGFR exon 21 p.L858R and the other is driven by
EGFR 19-Del, and both of them are likely to be sensitive to EGFR TKIs. For the common + rare and common + VUSs subtypes, the two
EGFR mutations could be either in the same or in different tumor cells. However, if some common and rare
EGFR mutations are in the same cancer cells, they might interfere with the response to certain EGFR TKIs. For example, Yu et al. found that if lung cancer patients had co-occurred baseline common
EGFR mutation and baseline
EGFR exon 20 p.T790M, they had poor responses to first-generation TKIs [
57]. Indeed, several previous studies reported that common
EGFR mutations and
EGFR exon 20 p.T790M were almost always in cis configurations in order to confer resistance to first-generation EGFR TKIs [
58]. Additionally, we found that rare
EGFR mutations were specifically enriched for
EGFR VUS (KD +) mutations. We speculate that
EGFR VUSs (KD +) and rare
EGFR mutations are within the same cancer cell or even on the same allele, and the additional KD aberrations from the VUSs might help reinforce the oncogenic activities of rare
EGFR mutations. Strikingly, we found that patients with the rare + VUSs (KD +) subtype are generally associated with a poorer prognosis than those with other subtypes, which further implies that they might reside in the same cancer cells to drive tumorigenesis and/or tumor progression. Nevertheless, our NGS results were not ideal to elucidate whether the compound
EGFR mutations were from the same cancer cell/DNA allele or not. Among the 1025 patients in our cohort, the compound
EGFR mutations of 282 patients were on the same exon. We then analyzed whether the mutations were on the same sequencing read (i.e., the same allele) or not. Strikingly, in 98.9% (279/282) of cases, the compound
EGFR mutations were located on the same allele, which also infers that they were in the same cancer cell (Additional file
1: Table S8). Future studies using more appropriate approaches (e.g., NGS on multi-site sampling tissues, single-cell sequencing, sequencing complementary DNAs, long-read sequencing, or fluorescent in situ hybridization) are necessary to further check the cis/trans configuration and cellular distribution of compound mutations.
There were several limitations of our study. Firstly, a large proportion of patients had missing clinical information, including the PD-L1 expression and disease stages, which can potentially impede thorough analyses of the correlation between the clinical characteristics and compound EGFR mutation subtypes. Secondly, because the tumor samples were collected by different hospitals spanning the past 4.5 years, the samples were generically profiled by 3 different targeted sequencing panels. Fortunately, all 3 targeted sequencing panels were designed and performed by the same sequencing institute. Specifically, all the assay validations were performed using a method-based validation approach to detect a specific type of mutation at a specific sequencing depth under the entire NGS system, and all three sequencing panels showed a similar capacity to detect mutations (cross-panel accuracy > 97%). Therefore, the result of overlapping genes from the three sequencing panels is comparable. Lastly, only 95 patients who had paired baseline and PD samples were available for drug resistance analyses, and future studies with larger patient sizes are necessary to fully elucidate the differential resistant mechanisms for various compound EGFR mutations.