Advertisement

Defining the sensitivity landscape of EGFR variants to tyrosine kinase inhibitors

Open AccessPublished:November 05, 2022DOI:https://doi.org/10.1016/j.trsl.2022.11.002

      Abstract

      Tyrosine kinase inhibitor (TKI) is a standard treatment for patients with NSCLC harboring constitutively active epidermal growth factor receptor (EGFR) mutations. However, most rare EGFR mutations lack treatment regimens except for the well-studied ones. We constructed two EGFR variant libraries containing substitutions, deletions, or insertions using the saturation mutagenesis method. All the variants were located in the EGFR mutation hotspot (exons 18–21). The sensitivity of these variants to afatinib, erlotinib, gefitinib, icotinib, and osimertinib was systematically studied by determining their enrichment in massively parallel cytotoxicity assays using an endogenous EGFR-depleted cell line. A total of 3914 and 70,475 variants were detected in the constructed EGFR Substitution-Deletion (Sub-Del) and exon 20 Insertion (Ins) libraries. Of the 3914 Sub-Del variants, 221 proliferated fast in the control assay and were sensitive to EGFR-TKIs. For the 70,475 Ins variants, insertions at amino acid positions 770–774 were highly enriched in all 5 TKI cytotoxicity assays. Moreover, the top 5% of the enriched insertion variants included a glycine or serine insertion at high frequency. We present a comprehensive reference for the sensitivity of EGFR variants to five commonly used TKIs. The approach used here should be applicable to other genes and targeted drugs.

      Background

      Tyrosine kinase inhibitors (TKIs) therapy is a standard treatment for patients with advanced non–small-cell lung carcinoma (NSCLC) when activating epidermal growth factor receptor (EGFR) mutations are detected. However, except for the well-studied EGFR mutations, most EGFR mutations lack treatment regimens.

      Translational Significance

      The results demonstrated that patients with rare EGFR mutations were most likely to benefit from osimertinib therapy compared to afatinib, erlotinib, gefitinib, or icotinib therapy. This study provides a case of deep mutational scanning that simultaneously assayed substitution, deletion, and insertion variants. This approach is applicable for other oncogenes and targeted drugs.

      Keywords

      Abbreviations:

      AA (Amino acid), Afa (Afatinib), ATCC (American Type Culture Collection), CIViC (Clinical Interpretation of Variants in Cancer), COSMIC (Catalogue of Somatic Mutations in Cancer), DMS (Deep mutational scanning), DMSO (Dimethyl sulfoxide), EGFR (Epidermal growth factor receptor), EGFR-TKI (EGFR-tyrosine kinase inhibitor), Erl (Erlotinib), ES (Enrichment Score), FACS (Fluorescent-activated cells sorting), FDR (False discovery rate), gDNA (Genomic DNA), Gef (Gefitinib), Ico (Icotinib), Ins (Insertion), MANO (Mixed-all-nominated-mutants-in-one), MITE (Mutagenesis by Integrated TilEs), NEB (New England Biolabs), NSCLC (Non–small-cell lung carcinoma), Osi (Osimertinib), sgRNA (Single guide RNA), STR (Short tandem repeat), Sub-Del (Substitution-Deletion), TKI (Tyrosine kinase inhibitor), VUS (Variants of unknown significance)

      INTRODUCTION

      The incidence of epidermal growth factor receptor (EGFR) mutations is reported to be 28.6% among populations with non–small-cell lung carcinoma (NSCLC). East Asian and Southeast Asian populations have a higher incidence (41.3%–57.2%) than the European population (8.0%–20.2%).
      • Soraas L
      • Stebbing J.
      Geographic variation in EGFR mutation frequency in lung adenocarcinoma may be explained by interethnic genetic variation.
      Clinical trials have shown that EGFR-tyrosine kinase inhibitor (EGFR-TKI) treatment of patients with NSCLC is superior to chemotherapy in terms of progression-free survival and serious adverse effects.
      • Kuan FC
      • Kuo LT
      • Chen MC
      • et al.
      Overall survival benefits of first-line EGFR tyrosine kinase inhibitors in EGFR-mutated non-small-cell lung cancers: a systematic review and meta-analysis.
      ,
      • Wang S
      • Cang S
      • Liu D.
      Third-generation inhibitors targeting EGFR T790M mutation in advanced non-small cell lung cancer.
      Several EGFR-TKIs (eg, afatinib, erlotinib, gefitinib, and osimertinib) have been approved for first-line treatment of patients with advanced NSCLC having activating EGFR mutations.
      • Zhao Y
      • Liu J
      • Cai X
      • et al.
      Efficacy and safety of first line treatments for patients with advanced epidermal growth factor receptor mutated, non-small cell lung cancer: systematic review and network meta-analysis.
      Usually, an EGFR mutation testing is performed to check the pattern of mutations in patients, which enables the physicians to determine whether, and which, EGFR-TKI could be used. Statistical analysis of EGFR mutation records in the Catalogue of Somatic Mutations in Cancer (COSMIC) database shows that, apart from the secondary mutations (eg, p.Thr790Met, p.Cys797Ser), p.Leu858Arg (41%) and exon 19 deletions (47%) together account for 88% of the mutations in all the records, whereas rare mutations account for about 12% of all mutations.
      • Harrison PT
      • Vyse S
      • Huang PH.
      Rare epidermal growth factor receptor (EGFR) mutations in non-small cell lung cancer.
      The rare EGFR mutations mainly comprise missense variants (>63%) and exon 20 insertions (∼17%).
      • Harrison PT
      • Vyse S
      • Huang PH.
      Rare epidermal growth factor receptor (EGFR) mutations in non-small cell lung cancer.
      Although rare EGFR mutations are reportedly less prevalent, the high incidence of NSCLC has increasingly led to their detection in the clinic,
      • Bray F
      • Ferlay J
      • Soerjomataram I
      • et al.
      Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries.
      making it impossible to ignore them. Unlike a limited number of high-prevalence EGFR mutations, TKI sensitivity for rare mutations and a large number of potential mutations has not been systematically studied.
      Several groups have tried to evaluate the clinical relevance of EGFR variants on a large scale. One group developed a mixed-all-nominated-mutants-in-one (MANO) method and applied it to 101 nonsynonymous EGFR variants.
      • Kohsaka S
      • Nagano M
      • Ueno T
      • et al.
      A method of high-throughput functional evaluation of EGFR gene variants of unknown significance in cancer.
      The process of the MANO method is labor intensive, which makes it challenging to evaluate the functions of variants on a large scale, for example, for several thousands of variants. Another group reported screening for activating EGFR mutations using a library of 7216 randomly mutated single-nucleotide variants.
      • Chakroborty D
      • Kurppa KJ
      • Paatero I
      • et al.
      An unbiased in vitro screen for activating epidermal growth factor receptor mutations.
      This variant library was generated by the error-prone PCR method, which can barely cover all the interested variants.
      In view of the lack of economical and efficient methods, it is highly challenging to systematically study the TKI sensitivity of rare EGFR mutations on a large scale. Deep mutational scanning (DMS) is a cutting-edge technology that enables the functional analysis of large numbers of variants simultaneously.
      • Araya CL
      • Fowler DM.
      Deep mutational scanning: assessing protein function on a massive scale.
      ,
      • Fowler DM
      • Stephany JJ
      • Fields S.
      Measuring the activity of protein variants on a large scale using deep mutational scanning.
      Usually, a mutant library is first constructed employing synthetic biology or genome editing methods,
      • Melnikov A
      • Rogov P
      • Wang L
      • et al.
      Comprehensive mutational scanning of a kinase in vivo reveals substrate-dependent fitness landscapes.
      ,
      • Findlay GM
      • Daza RM
      • Martin B
      • et al.
      Accurate classification of BRCA1 variants with saturation genome editing.
      and functional impacts of variants are determined through parallel functional assays. To date, variants of a few clinically actionable genes, such as BRCA1, PPARG, TP53, PTEN, TPMT, NUDT15, SCN5A, CYP2C9, CYP2C19, CXCR4, CCR5, ADRB2, and MSH2,
      • Findlay GM
      • Daza RM
      • Martin B
      • et al.
      Accurate classification of BRCA1 variants with saturation genome editing.
      • Starita LM
      • Young DL
      • Islam M
      • et al.
      Massively parallel functional analysis of BRCA1 RING domain variants.
      • Majithia AR
      • Tsuda B
      • Agostini M
      • et al.
      Prospective functional classification of all possible missense variants in PPARG.
      • Giacomelli AO
      • Yang X
      • Lintner RE
      • et al.
      Mutational processes shape the landscape of TP53 mutations in human cancer.
      • Kotler E
      • Shani O
      • Goldfeld G
      • et al.
      A systematic p53 mutation library links differential functional impact to cancer mutation pattern and evolutionary conservation.
      • Matreyek KA
      • Starita LM
      • Stephany JJ
      • et al.
      Multiplex assessment of protein variant abundance by massively parallel sequencing.
      • Mighell TL
      • Thacker S
      • Fombonne E
      • et al.
      An integrated deep-mutational-scanning approach provides clinical insights on PTEN genotype-phenotype relationships.
      • Suiter CC
      • Moriyama T
      • Matreyek KA
      • et al.
      Massively parallel variant characterization identifies NUDT15 alleles associated with thiopurine toxicity.
      • Glazer AM
      • Kroncke BM
      • Matreyek KA
      • et al.
      Deep mutational scan of an SCN5A voltage sensor.
      • Zhang L
      • Sarangi V
      • Moon I
      • et al.
      CYP2C9 and CYP2C19: deep mutational scanning and functional characterization of genomic missense variants.
      • Heredia JD
      • Park J
      • Brubaker RJ
      • et al.
      Mapping interaction sites on human chemokine receptors by deep mutational scanning.
      • Jones EM
      • Lubock NB
      • Venkatakrishnan AJ
      • et al.
      Structural and functional characterization of G protein-coupled receptors with deep mutational scanning.
      • Jia X
      • Burugula BB
      • Chen V
      • et al.
      Massively parallel functional testing of MSH2 missense variants conferring Lynch syndrome risk.
      have been extensively studied using the DMS method, highlighting its potential for assessing the TKI sensitivity of rare and potential EGFR mutations. In this study, based on the DMS method, we developed massively parallel cytotoxicity assays and systematically determined the sensitivity landscape of the variants to five EGFR-TKIs (Fig 1).
      Fig 1
      Fig 1Overview of the experimental design and data analysis. Parallelly synthesized oligonucleotide pools and degenerate primers were separately used to construct EGFR substitution-deletion (Sub-Del) and insertion (Ins) variant libraries. Variants in the Sub-Del library were located within EGFR exons 18–21 (amino acid positions, 688–875). Variants in the Ins library were located within EGFR exon 20 (amino acid positions, 762–776). Each variant had only one designed mutation within a full codon-optimized EGFR-coding sequence. Two single-guide RNA (sgRNA) expression cassettes driven by human U6 (hU6) promoter, targeting the endogenous EGFR sequence, were located upstream of the EGFR variant expression module. A blasticidin S resistance gene (BSD) was fused with the EGFR variant cassette through a P2A peptide. The Sub-Del and Ins libraries were separately introduced into Cas9-expressing PC9 cells. The cells were treated with five EGFR tyrosine kinase inhibitors (TKIs) or dimethyl sulfoxide (DMSO). The number of cells in each assay was counted manually once, every two days. Two weeks later, cells were harvested and subjected to genomic DNA extraction. Mutation regions were amplified by polymerase chain reaction and subjected to next-generation sequencing. Relative drug sensitivity was determined by variant enrichment in cell numbers before and after drug screening.

      MATERIALS AND METHODS

      Cell lines and reagents

      HEK 293 and A549 cell lines cryopreserved in our laboratory were used. PC9 cells were obtained from Beijing Cancer Hospital and authenticated via short tandem repeat (STR) profiling. All cells were tested to be mycoplasma negative. The genetic background of PC9 cells was further checked by next-generation sequencing using the Illumina TruSight Cancer panel. Cells were maintained according to the instructions prescribed by American Type Culture Collection (ATCC). Unless otherwise noted, all cell culture reagents were purchased from Thermo Fisher Scientific (USA), and all molecular cloning reagents were purchased from New England Biolabs (NEB, USA). Afatinib (S1011), erlotinib (S1023), icotinib (S2922), and osimertinib (S7297) were purchased from Selleck Chemicals and gefitinib (SML1657) was purchased from Sigma-Aldrich. Total EGF Receptor (10001-R021) and anti-EGFR-PE (352904) antibodies were purchased from Sino Biological (China) and BioLegend (USA), respectively.

      CRISPR/Cas9 single guide RNA (sgRNA) for endogenous EGFR deletion

      Ten EGFR targeting sgRNAs (Supplemental Table 1) were designed using the online tool, CHOPCHOP (https://chopchop.cbu.uib.no/). The sgRNAs cloning was performed according to the reported protocol.
      • Shalem O
      • Sanjana NE
      • Hartenian E
      • et al.
      Genome-scale CRISPR-Cas9 knockout screening in human cells.
      CRISPR/Cas9 lentiviral particles were separately produced and then introduced into A549 cells. The gene-editing efficiency of the sgRNAs was determined using fluorescent-activated cell sorting (FACS) analysis and immunoblotting.

      Construction of saturation mutagenesis libraries

      EGFR saturation mutagenesis libraries were constructed based on the Mutagenesis by Integrated TilEs (MITE) method reported previously,
      • Melnikov A
      • Rogov P
      • Wang L
      • et al.
      Comprehensive mutational scanning of a kinase in vivo reveals substrate-dependent fitness landscapes.
      with necessary modifications. The complete Substitution-Deletion (Sub-Del) Library was obtained by mixing 6 sub-libraries at equal mass ratios. The complete Insertion (Ins) Library was obtained by mixing 3 sub-libraries (Ins-1, Ins-2, Ins-3) in the mass ratio of 1:5:5. The completeness and uniformity of the variant libraries were verified through next-generation sequencing using a HiSeq sequencer (Illumina).

      Lentivirus production and transduction

      Lentivirus was produced and titered as reported previously,
      • Giacomelli AO
      • Yang X
      • Lintner RE
      • et al.
      Mutational processes shape the landscape of TP53 mutations in human cancer.
      with minor modifications. For lentivirus production, 5 million 293T cells were preseeded on 10 cm dishes, the first day evening. Cells were cotransfected with the variant library plasmid and two helper plasmids (psPAX2 and pMD2.G) at a molar ratio of 1:1:1 using a cationic polymer transfection reagent, EZ Trans (LIFE iLab, China), according to the manufacturer's instructions. The medium was replaced 24 h later and the virus particles were collected and filtered through 0.45 μm filters (Millipore), 24 or 48 h later.
      PC9 cells were transduced with a Cas9 expression lentivirus and selected on puromycin for 2 weeks. PC9 cells stably expressing Cas9 (PC9-Cas9) were cultured in large quantities and seeded in twenty 15 cm dishes (∼15 million cells per dish), the first day evening. Subsequently, these cells were transduced with a pooled virus of the Sub-Del/Ins library (10 dishes for each library) at a multiplicity of infection of ∼0.25. Transduced cells were selected on a medium with puromycin and blasticidin for 2 weeks.

      Cell proliferation assay

      About 15 million cells (PC9-Cas9 cells transduced with the Sub-Del variants) were seeded in a 15 cm dish. Another 5 million cells were collected as the week 0 sample. These cells were trypsinized, manually counted, and reseeded every 2/3 days for 8 weeks. Samples (∼5 million cells) were taken once a week (weeks 1, 2, 3, 4, 5, 6, 7, 8) and stored in a −80 freezer. Subsequently, samples of all time points were subjected to genomic DNA (gDNA) extraction.

      TKI cytotoxicity assay

      For the TKI sensitivity screening, cells transduced with the Sub-Del/Ins library were separately seeded in eight 15 cm dishes (15 million cells per dish) and treated with different TKIs or dimethyl sulfoxide (DMSO): two for afatinib (50 nM/600 nM), one for erlotinib (600 nM), one for gefitinib (600 nM), one for icotinib (600 nM), two for osimertinib (200 nM/600 nM), and one for DMSO (control). The remaining cells (>5 million cells for each library) were collected as the time 0 samples. Cells were trypsinized, manually counted, and reseeded every 2 days for 14 days. Dishes of proper size were used each time to ensure that the cell density was maintained within a proper range. The TKI assays were independently repeated once (Supplemental Fig 1). After 14 days of TKI treatment (time 1), cells were harvested for genomic DNA extraction.

      Amplification and sequencing of mutational regions

      The first-round PCR was performed to amplify a ∼1.1 kb DNA fragment covering the entire EGFR mutational region from the gDNA. Multiple 50 μL PCRs were performed for each sample. PCR products from each sample were separately pooled and purified with a Universal DNA Purification Kit (Tiangen, China). In the second-round PCR, the Sub-Del library mutational region was amplified with corresponding tagged primers using the 1.1 kb fragments as templates. In contrast, the Ins library mutational region was amplified from the corresponding templates. PCR primers can be found in Supplemental Table 2. All PCRs were performed using the Q5 High-Fidelity DNA Polymerase (NEB). The amplicons were purified using a Universal DNA Purification Kit. Next-generation sequencing libraries were constructed according to the manufacturer's protocol (VAHTS, China). All samples were sequenced on the HiSeq platform (Illumina) using a 2 × 150 paired-end configuration.

      Data processing

      Specifically, raw overlapping paired reads (read1 and read2) were merged into single reads using the bbmerge tool with default parameters.
      • Bushnell B
      • Rood J
      • Singer E.
      BBMerge - Accurate paired shotgun read merging via overlap.
      For Sub-Del libraries, each read was assigned to one reference sequence using the USEARCH tool with parameters of -usearch_global -strand both -id 0.95.
      • Edgar RC.
      Search and clustering orders of magnitude faster than BLAST.
      For Ins libraries, each read was assigned to one reference sequence with the same USEARCH tool but with a different value of id parameter (0.90). The successfully assigned reads in Sub-Del and Ins libraries were translated into amino acid (AA) and were subjected to diamond blastp analysis against the reference protein sequences to call variants.
      • Buchfink B
      • Xie C
      • Huson DH.
      Fast and sensitive protein alignment using DIAMOND.
      The frequency of each mutation was calculated as the ratio of the variant count to the count of valid variants. Thereafter, the frequency of each variant was weighted with the ratio of cell count change relative to time 0. To assess the enrichment or depletion of each variant, we calculated the log2-fold change in variant frequency relative to the time 0 samples. Heatmap and violin plots were created using the R pheatmap and ggplot2 packages, respectively.

      RESULTS

      Construction of 74,389 EGFR variants

      EGFR pathogenic variants, including substitutions, deletions, and exon 20 insertions,
      • Graham RP
      • Treece AL
      • Lindeman NI
      • et al.
      Worldwide frequency of commonly detected EGFR mutations.
      are mainly located in exons 18–21 (residues 688–875). To cover all the major types of mutations, two EGFR libraries were constructed using the MITE method.
      • Melnikov A
      • Rogov P
      • Wang L
      • et al.
      Comprehensive mutational scanning of a kinase in vivo reveals substrate-dependent fitness landscapes.
      Parallel synthesized oligonucleotides were used to construct a saturation EGFR Sub-Del library containing 3572 (theoretical number) single amino acid (AA) substitution (Sub) variants, 188 (theoretical number) single AA deletion (Del-1) variants, and 182 (theoretical number, for cloning reason, 6 deletion variants were not covered) 2 consecutive AA deletion (Del-2) variants. All Sub-Del library variants were located at AA positions 688–875. Similarly, multiple degenerate primers were used to construct a saturation EGFR Ins library containing 126,300 (theoretical number) in-frame insertion variants. All Ins library variants were located in the EGFR insertion hotspot (AA positions, 762-776) and consisted of 1, 2, or 3 tandem NNS codons (N = adenine, cytosine, guanine, or thymine; S = cytosine or guanine) inserted between adjacent AAs, with each variant containing one insertion in each of the AA positions. The Ins library was composed of three subtype libraries: Ins-1 library (300 variants, each with a single-AA inserted), Ins-2 library (6000 variants, each with a double-AA inserted), and Ins-3 library (120,000 variants, each with a triple-AA inserted). Variants of Sub-Del and Ins libraries were cloned into lentiviral vectors, and the uniformity and completeness of each library were determined by high-throughput sequencing of the plasmids. The Sub-Del library showed decent uniformity (Supplemental Fig 2A), and 99.3% of the designed variants were detected more than 20 times (>20 reads) (Supplemental Fig 3A). Similarly, variants of the Ins library also showed high uniformity (Supplemental Fig 2B). Because the Ins library was derived from mixing the Ins-1, Ins-2, and Ins-3 libraries at a mass ratio of 1:5:5, the ratio of the average proportion of each variant from Ins-1, Ins-2, and Ins-3 libraries was 80:20:1. Not surprisingly, most of the designed variants in the Ins-1 (99.7%) and Ins-2 (93.6%) libraries were detected (>20 reads for each variant). However, only 53.8% of the designed Ins-3 library variants were detected (>20 reads for each variant) (Supplemental Fig 3B). Overall, the coverage and uniformity of Sub-Del and Ins libraries were as expected, and the total number of detected variants was 74,389.

      Deletion of endogenous EGFR in the PC-9 cells

      The human lung adenocarcinoma cell line, PC9 (contains EGFR exon 19 deletions), was chosen as the model. The cell line identity was confirmed by STR profiling and genomic mutations were further reviewed through next-generation sequencing. A Glu746–Ala750 deletion was confirmed in EGFR exon 19 and a p.Arg248Gln mutation was detected in TP53. The p.Arg248Gln variant is a gain-of-function mutation that promotes tumorigenesis.
      • Yoshikawa K
      • Hamada J
      • Tada M
      • et al.
      Mutant p53 R248Q but not R248W enhances in vitro invasiveness of human lung cancer NCI-H1299 cells.
      These results indicate that the PC9 cell line is a genetically clear cell model for evaluating the sensitivity of EGFR variants to TKIs.
      To eliminate the endogenous EGFR, we separately introduced Cas9 and the sgRNA-EGFR variant (codon-optimized) into PC9 cells using 2 lentiviral vectors. First, Cas9 was introduced into PC9 cells to obtain the PC9-Cas9 cell line, stably expressing Cas9. Subsequently, the sgRNA-EGFR variant was introduced into the PC9-Cas9 cells. The endogenous EGFR was knocked out using the CRISPR/Cas9 and replaced with the exogenous EGFR variant in approximately 10 days. Because direct knockout of EGFR may seriously affect the viability of PC9 cells,
      • Okamoto K
      • Okamoto I
      • Okamoto W
      • et al.
      Role of survivin in EGFR inhibitor-induced apoptosis in non-small cell lung cancers positive for EGFR mutations.
      we tested the CRISPR/Cas9 gene-editing efficiency in A549 cells (with a wildtype EGFR and activating KRAS mutation [p.Gly12Ser] background). Considering that KRAS locates downstream of the EGFR signaling pathway, EGFR can be knocked out without seriously affecting the viability of A549 cells. FACS and western blotting analyses confirmed that the endogenous EGFR had been efficiently knocked out by CRISPR/Cas9 (Supplemental Figs 4 and 5). We predicted that these validated CRISPR/Cas9 sgRNAs would also efficiently knock out the endogenous EGFR in PC9-Cas9 cells.

      Sensitivity landscape of EGFR variants to 5 TKIs

      To systematically evaluate the sensitivity of EGFR variants to different TKIs, we separately introduced Sub-Del and Ins variants into the PC9-Cas9 cells and treated them with five TKIs. The expression of exogenous EGFR variants was checked through FACS analysis and found to be slightly higher than that of EGFR in PC9 cells (Supplemental Fig 6). The concentrations of EGFR-TKIs were set mainly referring to the clinical plasma concentration of each TKI (Supplemental Table 3). Specifically, the concentrations of reversible TKIs (erlotinib, gefitinib, and icotinib) were set to 600 nM, whereas those of irreversible TKIs (afatinib and osimertinib) were set at 2 levels. The afatinib clinical plasma concentration is much lower than 600 nM
      • Wind S
      • Schmid M
      • Erhardt J
      • et al.
      Pharmacokinetics of afatinib, a selective irreversible ErbB family blocker, in patients with advanced solid tumours.
      ; we designed 2 assays for 50 and 600 nM afatinib. Because incubation with 600 nM osimertinib showed strong cytotoxicity to cells with Sub-Del variants in vitro, we designed two assays for 200 and 600 nM osimertinib. The TKI treatments lasted for two weeks and cells were counted every 2 days. The number of cells containing the Sub-Del variants in the erlotinib, gefitinib, icotinib, afatinib (50 nM), and osimertinib (200 nM) assays decreased substantially on days 5 and 7, reached the minimum on day 9, and then began to increase. The cell numbers in the afatinib (600 nM) and osimertinib (600 nM) assays reached the minimum on days 11 and 15, respectively (Fig 2A). For cells containing the Ins variants, the minimum cell number in the afatinib (600 nM) and osimertinib (600 nM) assays was reached on days 9 and 13, respectively, whereas the cell number in other TKI assays was not decreased during two-week of TKI treatment (Fig 2B).
      Fig 2
      Fig 2EGFR-tyrosine kinase inhibitors (EGFR-TKIs) showed different inhibition on Sub-Del and Ins library variants. Cell growth kinetics (log10) of the Sub-Del (A) and Ins (B) libraries treated with five TKIs and dimethyl sulfoxide (DMSO) (mechanical replicates, n = 4; error bars indicate the standard deviation). Violin plots denoting enrichment score (log2) distributions of variants in the Sub-Del (C) and Ins (D) libraries treated with the five TKIs or DMSO. (E) The enrichment scores (log2) of the three subtypes of variants (Sub: single amino acid (AA) substitution; Del-1: single AA deletion; and Del-2: dual AA deletion) in the Sub-Del library are separately displayed by treatment type in the violin plot. (F) The enrichment scores (log2) of the three subtypes of variants (Ins-1: single AA insertion; Ins-2: dual AA insertion; and Ins-3: triple AA insertion) in the Ins library are separately displayed by treatment type in the violin plot. Student's t-test, ns: P > 0.05, *: P ≤ 0.05, **: P ≤ 0.01, ***: P ≤ 0.001, ****: P ≤ 0.0001.

      Osimertinib inhibited all types of variants

      To investigate the enrichment of variants after incubation with TKIs, we harvested cells from each TKI assay for gDNA extraction, mutation region amplification, and high-throughput sequencing. The relative drug sensitivity for each variant was defined as the fold change of a variant's theoretical cell number post-TKI screening (time 1) to that at the early time point (time 0). The theoretical cell number for each variant was calculated using the read proportion of a variant among the total valid reads multiplied by the final cell number in each assay. Compared with reversible TKIs (erlotinib, gefitinib, and icotinib), irreversible TKIs (afatinib 50 nM and osimertinib 200 nM) showed stronger inhibition for Sub-Del variants (enrichment score mean: afatinib, 0.331; erlotinib, 0.793; gefitinib, 1.73; icotinib, 4.44; osimertinib, 0.388; false discovery rate (FDR)-adjusted P values <0.0001) (Fig 2C). For the Ins variants, osimertinib (200 nM) had a greater inhibition than the other TKIs (enrichment score mean: afatinib, 113; erlotinib, 89.3; gefitinib, 111; icotinib, 176; osimertinib, 4.46; FDR-adjusted P values <0.0001; Fig 2D). Overall, these results indicated that the Ins library variants were more resistant to TKI treatment than the Sub-Del library variants, and osimertinib (200 nM) showed stronger inhibition for variants in both the libraries.
      To investigate whether distinct EGFR variant subtypes present different sensitivity against the same EGFR-TKI, we separately analyzed the relative drug sensitivity of the Sub, Del-1, and Del-2 variants in the Sub-Del library and the Ins-1, Ins-2, and Ins-3 variants in the Ins library. No significant difference in the inhibition of Sub, Del-1, and Del-2 variants to any of the 5 TKIs was observed (Fig 2E; FDR-adjusted P values >0.05). In contrast, the sensitivity of Ins-1, Ins-2, and Ins-3 variants to each of the five TKIs was significantly different (Fig 2F; FDR-adjusted P values < 0.05). However, the difference in enrichment scores (log2) cannot be entirely attributed to TKI treatment because differences were also observed for Ins-1, Ins-2, and Ins-3 variants in the TKI-free assay (control).

      Cytotoxicity screening results were consistent with known clinical annotations

      Next, we evaluated the consistency of TKI screening results with known clinical annotations. For this, we compiled a list of EGFR mutations from the Clinical Interpretation of Variants in Cancer (CIViC) knowledgebase and compared the clinical annotations (resistant or sensitive) of mutations with the enrichment scores (log2) obtained in cytotoxicity assays by plotting those mutations on a scale bar comprising all the results (Supplemental Fig 7). TKI-sensitive variants (such as p.Leu858Arg and p.Gly719Ala/Ser) were located in the middle or lower half of the bar, whereas TKI-resistant variants (such as p.Thr790Met) were located in the middle or upper half of the bar (Supplemental Fig 7A). Similarly, the annotated Ins variants were distributed at expected positions in the bars (Supplemental Fig 7B). However, some annotated variants, such as p.Cys797Ser (0.53) and p.Ser768Ile (1.19), had contradictory enrichment scores in the osimertinib assay. These discrepancies could be attributed to the presence of additional mutations. For example, EGFR mutation combinations in cis (on the same allele) or trans (on different alleles) might lead to altered drug response outcomes compared with those for independent mutations.
      • Arulananda S
      • Do H
      • Musafer A
      • et al.
      Combination Osimertinib and Gefitinib in C797S and T790M EGFR-mutated non-small cell lung cancer.
      • Wang Z
      • Yang JJ
      • Huang J
      • et al.
      Lung adenocarcinoma harboring EGFR T790M and in trans C797S responds to combination therapy of first- and third-generation EGFR TKIs and shifts allelic configuration at resistance.
      • Hidaka N
      • Iwama E
      • Kubo N
      • et al.
      Most T790M mutations are present on the same EGFR allele as activating mutations in patients with non-small cell lung cancer.
      • Brown BP
      • Zhang YK
      • Westover D
      • et al.
      On-target resistance to the mutant-selective EGFR inhibitor Osimertinib can develop in an allele-specific manner dependent on the original EGFR-activating mutation.
      Indeed, p.Cys797Ser has been detected as a secondary mutation coexisting with p.Thr790Met in cis and results in resistance to osimertinib.
      • Arulananda S
      • Do H
      • Musafer A
      • et al.
      Combination Osimertinib and Gefitinib in C797S and T790M EGFR-mutated non-small cell lung cancer.
      • Wang Z
      • Yang JJ
      • Huang J
      • et al.
      Lung adenocarcinoma harboring EGFR T790M and in trans C797S responds to combination therapy of first- and third-generation EGFR TKIs and shifts allelic configuration at resistance.
      • Hidaka N
      • Iwama E
      • Kubo N
      • et al.
      Most T790M mutations are present on the same EGFR allele as activating mutations in patients with non-small cell lung cancer.
      In addition, another variant, p.Gly724Ser, has been associated with increased sensitivity to osimertinib when combined in trans with p.Leu858Arg than with exon 19 deletions.
      • Brown BP
      • Zhang YK
      • Westover D
      • et al.
      On-target resistance to the mutant-selective EGFR inhibitor Osimertinib can develop in an allele-specific manner dependent on the original EGFR-activating mutation.
      In general, our TKI sensitivity evaluation results are consistent with the known annotation.

      Cytotoxicity screening identifies potentially drug resistant variants

      The variant enrichment scores (log2) in each TKI assay for each variant were presented as heatmaps (Fig 3, Fig 4, Supplemental Figs 8 and 9). For Sub-Del variants, 280 (7.2%) variants in afatinib assay (50 nM), 841 (21.5%) variants in erlotinib assay, 1802 (46.2%) variants in gefitinib assay, 3440 (88.0%) variants in icotinib assay, and 271 (7.2%) variants in osimertinib assay (200 nM) of all variants had a positive enrichment score (log2), indicating that the number of those variants were increased after 2 weeks of TKI incubation. Enriched variants from all three reversible TKI assays have 813 variants (complex variants = 571, simple variants = 242) in common (Fig 5, A, C, and E). The variants enriched in the two irreversible TKI assays have 51 variants (complex variants = 36, simple variants =15) in common (Fig 5, B, D, and F). Here, we divided all substitution variants into 2 categories, the complex and simple variants. The complex variants can only be achieved through 2 or 3 point mutations in the corresponding codon, whereas the simple variants can be obtained through one point mutation. Under normal circumstances, the probability of multiple independent point mutations occurring in the same codon is relatively low, and therefore, simple variants are more likely to be encountered in the clinic. Overall, in our systematic cytotoxicity screening, we found 813 and 51 substitution variants that potentially mediate resistance to EGFR reversible and irreversible TKIs, respectively.
      Fig 3
      Fig 3Heatmaps depicting enrichment of EGFR Sub-Del library variants on the cytotoxicity of five tyrosine kinase inhibitors (TKIs). The enrichment scores are shown as the ratio (log2) of the cell number post-TKI treatment to that pre-TKI treatment. The top 20 rows represent all 20 possible amino acid (AA) substitutions, and the 2 rows below represent one or two AA deletion mutations (Sub: single AA substitution; Del-1: single AA deletion; and Del-2: dual AA deletion). Each column indicates one AA residue in EGFR (688–875). Blue to red represents high to low abundance of variants, respectively. Variants that were not covered are colored in white. Codon-level average enrichment scores are plotted below each heatmap. Cells were treated with dimethyl sulfoxide (DMSO) (A), 50 nM afatinib (B), 600 nM erlotinib (C), 600 nM gefitinib (D), 600 nM icotinib (E), or 200 nM osimertinib (F).
      Fig 4
      Fig 4Heatmaps depicting the enrichment results for EGFR Ins-1 library variants on the cytotoxicity of five tyrosine kinase inhibitors (TKIs). The enrichment scores are shown as the ratio (log2) of the cell number post-TKI treatment to that pre-TKI treatment. The 20 rows represent all 20 possible single AA insertion mutations. Each column indicates one AA residue in EGFR (762–776). Blue to red represents high to low abundance of variants, respectively. Variants that were not covered are colored in white. Codon-level average enrichment scores are plotted below each heatmap. Cells were treated with dimethyl sulfoxide (DMSO) (A), 50 nM afatinib (B), 600 nM erlotinib (C), 600 nM gefitinib (D), 600 nM icotinib (E), or 200 nM osimertinib (F).
      Fig 5
      Fig 5Potentially drug-resistant substitutions identified in the cytotoxicity screening. Number and proportion of enriched (red bar) and depleted (green bar) variants in the reversible (A) and irreversible (B) tyrosine kinase inhibitor (TKI) assays. (C) The enriched variants in the three reversible TKI assays have an intersection of 813 variants (571 complex and 242 simple variants). (D) The enriched variants in the two irreversible TKI assays have an intersection of 51 variants (36 complex and 15 simple variants). (E) The enrichment score (log2) of the top 5% of the selected simple variants in “C” is presented as a heatmap. (F) The enrichment score (log2) of all the 15 selected simple variants in “D” was presented as a heatmap. Afa, Erl, Gef, Ico, Osi, and ES are abbreviations for afatinib, erlotinib, gefitinib, icotinib, osimertinib, and enrichment score, respectively.

      Cytotoxicity screening identifies 221 potentially drug sensitive variants

      Through the cell proliferation assay, we identified 292 substitutions/deletions that have enrichment scores (log2) larger than 1. This means the proportion of these variants doubled after eight weeks of competing proliferation. Among them, the top 13 extremely enriched variants (p.Phe712Arg, p.Arg841Phe, p.Ser784Leu, p.Val742Trp, p.Tyr827Ile, p.Leu861Trp, p.Ala839Ile, p.Gly719Lys, p.Leu833His, p.Leu862Lys, p.Thr751Pro, p.Ile759Gln, and p.Tyr727Thr) had enrichment scores (log2) larger than 3 (Supplemental Table 4). These extremely enriched variants were very likely to be constitutively active EGFR variants. Drug-sensitive variant was subjectively defined when its enrichment score (log2) was less than zero. Of the 292 fast-proliferated variants, 221 were identified to be sensitive to 1st (erlotinib, gefitinib, or icotinib), 2nd (afatinib), and 3rd (osimertinib) generation EGFR-TKIs simultaneously (Supplemental Table 5). Variants being sensitive to any 1st generation EGFR-TKIs (erlotinib, gefitinib, or icotinib) were covered in the list. Among them, 25 selected variants proliferated faster than other variants and were more sensitive to the five TKIs in general (Fig 6). Finally, we identified 221 fast proliferated variants that were also potentially EGFR-TKIs sensitive variants.
      Fig 6
      Fig 6Cytotoxicity screening identified potentially TKI sensitive variants. Cell proliferation assay identified 35 highly enriched (enrichment scores [log2] ≥ 2.32, means enrichment scores increased for five times) substitution/deletion variants. Of the 35 variants, 25 were highly depleted in four TKIs (afatinib, 50 nM; erlotinib, 600 nM; gefitinib, 600 nM; and osimertinib, 200 nM) cytotoxicity assays, except for the icotinib (600 nM) assay. ES, enrichment score; 869delY, p.Y869del; 801del YV, p.Y801_V802del.

      The top 5% enriched insertions showed insertional position and AA type preference

      For Ins variants, the proportion of variants with positive enrichment scores (log2) was much larger than of their Sub-Del counterparts: 95% for afatinib (50 nM) assay, 93% for erlotinib assay, 94% for gefitinib assay, 97% for icotinib assay, and 57% for osimertinib (200 nM) assay. To study whether the insertional position and the AA type affect the sensitivity of exon 20 insertions to different TKIs, we selected the top 5% enriched insertions from each TKI assay and determined the cumulative frequency of the inserted AA at each AA position. The top enriched exon 20 insertions shared many common features (Fig 7A–F): (1) the highly enriched insertion variants showed a preference for insertion at AA positions 770–774; (2) the highly enriched insertion variants showed a preference of glycine and serine. In addition, we also determined the cumulative frequency of insertional AA in variants from the Chinese population
      • Qin Y
      • Jian H
      • Tong X
      • et al.
      Variability of EGFR exon 20 insertions in 24 468 Chinese lung cancer patients and their divergent responses to EGFR inhibitors.
      and COSMIC database (Fig 7G–I). The detected rare exon 20 insertions were mainly concentrated within AA positions 770–774. Our results explain the resistance of these clinically identified exon 20 insertions to TKIs. Together, these results reveal that insertion in the EGFR AA positions 770–774 might induce a stronger resistance to TKIs, especially when the insertions contain glycine or serine.
      Fig 7
      Fig 7Variants in EGFR exon 20 insertional hotspot induce stronger tyrosine kinase (TKI) resistance. (A-F) The inserted amino acids (AA) in the top 5% enriched exon 20 insertions were counted and the cumulative frequency numbers were plotted as black dots at the corresponding AA positions. Dots with cumulative frequency less than 15 were filtered. (A-E) show summary results for afatinib (50 nM), erlotinib, gefitinib, icotinib, and osimertinib (200 nM) assays, respectively. (F) shows the summary result for untreated control. (G-I) The inserted AA in exon 20 insertions from the Chinese population
      • Qin Y
      • Jian H
      • Tong X
      • et al.
      Variability of EGFR exon 20 insertions in 24 468 Chinese lung cancer patients and their divergent responses to EGFR inhibitors.
      and COSMIC database were counted and the cumulative frequency numbers were plotted as black dots at the corresponding AA positions. All the dots are shown, regardless of the cumulative frequency numbers. (G) shows the summary result for the Chinese population. (H) shows the summary result for the COSMIC database. I shows combined results of “G” and “H.” The X-axis shows AA positions.

      DISCUSSION

      In the post-genome era, a massive number of variants of unknown significance (VUS) has been identified by high-throughput sequencing; however, it is still a big challenge for the functional interpretation of VUS in the proper context. To study the TKI sensitivity of rare EGFR mutations, we designed cytotoxicity assays and systematically evaluated the sensitivity of 74,389 EGFR variants to 5 commonly used EGFR-TKIs. Afatinib and osimertinib showed intense and relatively persistent inhibition of all types of EGFR variants (substitutions, deletions, and exon 20 insertions) among the 5 tested EGFR-TKIs. Considering the tolerance to the treatment, the clinical plasma drug concentration of afatinib cannot reach 600 nM,
      • Hayashi H
      • Iihara H
      • Hirose C
      • et al.
      Effects of pharmacokinetics-related genetic polymorphisms on the side effect profile of afatinib in patients with non-small cell lung cancer.
      whereas for osimertinib, it can be more than 600 nM.
      • Brown K
      • Comisar C
      • Witjes H
      • et al.
      Population pharmacokinetics and exposure-response of osimertinib in patients with non-small cell lung cancer.
      Accordingly, we speculate that otherwise for additional evidence, patients with NSCLC having rare EGFR mutations (including exon 20 insertions) are more likely to benefit from osimertinib than from the other four TKIs. Compared with Sub-Del mutations, EGFR exon 20 insertion mutations were more difficult to deal with. A recent clinical study showed that patients with EGFR exon 20 insertions could benefit from high-dose (160 mg daily) osimertinib treatment.
      • YW Zofia Piotrowska
      • Sequist Lecia V.
      • Ramalingam Suresh S.
      ECOG-ACRIN 5162: a phase II study of osimertinib 160 mg in NSCLC with EGFR exon 20 insertions.
      With the combined use of synthetic biology and CRISPR/Cas9 methods, we can economically generate a large number of cell models with designed mutations of a specific gene. These cell models can partially simulate clinical samples and enable us to obtain rapid screening results through appropriate functional assays.
      The systematic cytotoxicity screening of Sub-Del variants has led us to identify potentially drug-resistant and drug-sensitive variants. We were curious whether these variants occurred with equal frequency in the clinic. Further analysis revealed that the substitution variants could be divided into complex and simple variants. The majority of EGFR substitution mutations detected in the clinic are simple variants and most of the complex variants are rarely detected. The difficulty in the mutation process partially explains why the highly enriched complex variants are rarely detected in the clinic.
      • Giacomelli AO
      • Yang X
      • Lintner RE
      • et al.
      Mutational processes shape the landscape of TP53 mutations in human cancer.
      To the best of our knowledge, we have designated substitution variants into complex and simple variants for the first time. More efforts should be invested on the functional interpretation of the simple variants in the clinical context.
      The systematic screening revealed that EGFR exon 20 insertions at AA positions 770–774 would induce stronger TKI resistance, especially when the insertions contain glycine or serine. Clinically, excluding the high prevalence exon 20 insertions of p.Ala767_Val769dup and p.Ser768_Asp770dup, a large number of low prevalence exon 20 insertions are mostly concentrated at AA positions 770–774.
      • Qin Y
      • Jian H
      • Tong X
      • et al.
      Variability of EGFR exon 20 insertions in 24 468 Chinese lung cancer patients and their divergent responses to EGFR inhibitors.
      ,
      • Vyse S
      • Huang PH.
      Targeting EGFR exon 20 insertion mutations in non-small cell lung cancer.
      According to the theory of Vyse et al., EGFR exon 20 insertions in the loop region (AA positions 767–775) have altered 3-dimensional structures and will stabilize the mutated EGFR in the active state even without ligand binding
      • Vyse S
      • Huang PH.
      Targeting EGFR exon 20 insertion mutations in non-small cell lung cancer.
      . Our systematic screening results provided important experimental evidence supporting this theory.
      The cytotoxicity assay-based method for functional interpretation of variants used in this study can also be applied to other genes and targeted drugs. At least 52 small molecule kinase inhibitors have been approved by US food and drug administration for targeted therapy.
      • Roskoski Jr, R
      Properties of FDA-approved small molecule protein kinase inhibitors: a 2020 update.
      An extensive functional interpretation of variants will enable us to clarify the targeted inhibitors applicable for the corresponding mutations. Moreover, high throughput sequencing of circulating tumor DNA allows us to forecast the potential drug-resistance events.
      • Pines G
      • Fankhauser RG
      • Eckert CA.
      Predicting drug resistance using deep mutational scanning.
      • Bulbul A
      • Leal A
      • Husain H.
      Applications of cell-free circulating tumor DNA detection in EGFR mutant lung cancer.
      • Zhou Q
      • Yang JJ
      • Chen ZH
      • et al.
      Serial cfDNA assessment of response and resistance to EGFR-TKI for patients with EGFR-L858R mutant lung cancer from a prospective clinical trial.
      This will dramatically promote the development of precision medicine. Moreover, for newly developed small molecule targeted inhibitors, a predrug sensitivity screening at the laboratory stage will help forecast the suitability of drug candidates for patients with specific mutations, thereby drastically reducing the cost incurred at clinical trial stages.
      Despite the significance of this work, there are several important caveats: (1) We only used PC9 cells for cytotoxicity screening. The genetic background of this cell line might have had an unwanted influence on the functional interpretation of variants.
      • Findlay GM
      • Daza RM
      • Martin B
      • et al.
      Accurate classification of BRCA1 variants with saturation genome editing.
      ,
      • Suiter CC
      • Moriyama T
      • Matreyek KA
      • et al.
      Massively parallel variant characterization identifies NUDT15 alleles associated with thiopurine toxicity.
      The use of multiple NSCLC cell lines (with different genetic background) for parallel functional screening can significantly reduce such bias. (2) The expression of the exogenous EGFR variants was higher than that of endogenous EGFR, which may lead to an overestimation of the TKI resistance of specific variants. (3) As the endogenous EGFR of PC9 cells was deleted by CRISPR/Cas9, the engineered PC9 cells only has one copy of the exogenous EGFR variant. Previously, several studies have demonstrated that EGFR variant combinations can significantly influence sensitivity to a specific TKI.
      • Arulananda S
      • Do H
      • Musafer A
      • et al.
      Combination Osimertinib and Gefitinib in C797S and T790M EGFR-mutated non-small cell lung cancer.
      • Wang Z
      • Yang JJ
      • Huang J
      • et al.
      Lung adenocarcinoma harboring EGFR T790M and in trans C797S responds to combination therapy of first- and third-generation EGFR TKIs and shifts allelic configuration at resistance.
      • Hidaka N
      • Iwama E
      • Kubo N
      • et al.
      Most T790M mutations are present on the same EGFR allele as activating mutations in patients with non-small cell lung cancer.
      • Brown BP
      • Zhang YK
      • Westover D
      • et al.
      On-target resistance to the mutant-selective EGFR inhibitor Osimertinib can develop in an allele-specific manner dependent on the original EGFR-activating mutation.
      Further studies are needed to uncover the TKI inhibition against combinations of EGFR variants. In addition, knocking out of EGFR gene may not completely deplete the endogenous EGFR protein in every cell. It is possible for very few mutated endogenous EGFR gene to produce functional EGFR protein and form complex variant combinations with the introduced EGFR variant. This may bring additional bias to the screening results. (4) In this study, no EGFR activation ligand (like epidermal growth factor, etc) was used in the cytotoxicity screening assays. Many of the introduced EGFR variants were likely to maintain only weak activity under the activation of growth factors from the fetal bovine serum. Using an EGFR activation ligand to strongly activate the EGFR variants will further enrich the TKI-resistant variant more robustly in the cytotoxicity assays. The current study has the risk of losing some drug-resistant variants because the EGFR variants were not properly activated.
      In conclusion, systematic cytotoxicity screening revealed that patients with NSCLC having rare EGFR mutations are most likely to benefit from osimertinib treatment. We expect variants of other genes to be classified to assess their sensitivity to corresponding targeted inhibitors.

      Data sharing statement

      All primary data in this study will be made available upon request to the corresponding authors for individuals with appropriate data sharing agreements in place.

      Acknowledgments

      We thank Dr. Kang Bin from Cancer Hospital of Peking University for PC9 cell line and Luo Yuqi for experimental assistance. We thank Dr. Ren Yanfang of the University of Rochester for his comments on the manuscript. The data analysis was supported by the Supercomputing Center at Zhengzhou University (Zhengzhou). This work was supported by Henan Province Key Research and Promotion Project (202102310403 and 192102310034) and the National Natural Science Foundation of China (82101963).
      Conflicts of interest: All authors have read the journal's policy on disclosure of potential conflicts of interest and agreed to the journal's authorship statement. The authors declare no conflicts of interest.
      Author contribution: L.A., Y.W. H.X. and S.C. designed research. L.A., Y.W., S.C., G.W., C.L., Z.W., C.W., and Z.S. performed research. C.N., S.D., X.L., M.Y. K.W. and W.T. contributed new reagents/analytic tools. H.X., and Y.W. analyzed data. Y.W., H.X., L.A., and S.C. wrote the paper.

      Appendix. Supplementary materials

      REFERENCES

        • Soraas L
        • Stebbing J.
        Geographic variation in EGFR mutation frequency in lung adenocarcinoma may be explained by interethnic genetic variation.
        J Thorac Oncol. 2018; 13: 454-458
        • Kuan FC
        • Kuo LT
        • Chen MC
        • et al.
        Overall survival benefits of first-line EGFR tyrosine kinase inhibitors in EGFR-mutated non-small-cell lung cancers: a systematic review and meta-analysis.
        Br J Cancer. 2015; 113: 1519-1528
        • Wang S
        • Cang S
        • Liu D.
        Third-generation inhibitors targeting EGFR T790M mutation in advanced non-small cell lung cancer.
        J Hematol Oncol. 2016; 9: 34
        • Zhao Y
        • Liu J
        • Cai X
        • et al.
        Efficacy and safety of first line treatments for patients with advanced epidermal growth factor receptor mutated, non-small cell lung cancer: systematic review and network meta-analysis.
        BMJ. 2019; 367: l5460
        • Harrison PT
        • Vyse S
        • Huang PH.
        Rare epidermal growth factor receptor (EGFR) mutations in non-small cell lung cancer.
        Semin Cancer Biol. 2020; 61: 167-179
        • Bray F
        • Ferlay J
        • Soerjomataram I
        • et al.
        Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries.
        CA Cancer J Clin. 2018; 68: 394-424
        • Kohsaka S
        • Nagano M
        • Ueno T
        • et al.
        A method of high-throughput functional evaluation of EGFR gene variants of unknown significance in cancer.
        Sci Transl Med. 2017; 9: 2265
        • Chakroborty D
        • Kurppa KJ
        • Paatero I
        • et al.
        An unbiased in vitro screen for activating epidermal growth factor receptor mutations.
        J Biol Chem. 2019; 294: 9377-9389
        • Araya CL
        • Fowler DM.
        Deep mutational scanning: assessing protein function on a massive scale.
        Trends Biotechnol. 2011; 29: 435-442
        • Fowler DM
        • Stephany JJ
        • Fields S.
        Measuring the activity of protein variants on a large scale using deep mutational scanning.
        Nat Protoc. 2014; 9: 2267-2284
        • Melnikov A
        • Rogov P
        • Wang L
        • et al.
        Comprehensive mutational scanning of a kinase in vivo reveals substrate-dependent fitness landscapes.
        Nucleic Acids Res. 2014; 42: e112
        • Findlay GM
        • Daza RM
        • Martin B
        • et al.
        Accurate classification of BRCA1 variants with saturation genome editing.
        Nature. 2018; 562: 217-222
        • Starita LM
        • Young DL
        • Islam M
        • et al.
        Massively parallel functional analysis of BRCA1 RING domain variants.
        Genetics. 2015; 200: 413-422
        • Majithia AR
        • Tsuda B
        • Agostini M
        • et al.
        Prospective functional classification of all possible missense variants in PPARG.
        Nat Genet. 2016; 48: 1570-1575
        • Giacomelli AO
        • Yang X
        • Lintner RE
        • et al.
        Mutational processes shape the landscape of TP53 mutations in human cancer.
        Nat Genet. 2018; 50: 1381-1387
        • Kotler E
        • Shani O
        • Goldfeld G
        • et al.
        A systematic p53 mutation library links differential functional impact to cancer mutation pattern and evolutionary conservation.
        Mol Cell. 2018; 71: 873
        • Matreyek KA
        • Starita LM
        • Stephany JJ
        • et al.
        Multiplex assessment of protein variant abundance by massively parallel sequencing.
        Nat Genet. 2018; 50: 874-882
        • Mighell TL
        • Thacker S
        • Fombonne E
        • et al.
        An integrated deep-mutational-scanning approach provides clinical insights on PTEN genotype-phenotype relationships.
        Am J Hum Genet. 2020; 106: 818-829
        • Suiter CC
        • Moriyama T
        • Matreyek KA
        • et al.
        Massively parallel variant characterization identifies NUDT15 alleles associated with thiopurine toxicity.
        Proc Natl Acad Sci U S A. 2020; 117: 5394-5401
        • Glazer AM
        • Kroncke BM
        • Matreyek KA
        • et al.
        Deep mutational scan of an SCN5A voltage sensor.
        Circ Genom Precis Med. 2020; 13e002786
        • Zhang L
        • Sarangi V
        • Moon I
        • et al.
        CYP2C9 and CYP2C19: deep mutational scanning and functional characterization of genomic missense variants.
        Clin Transl Sci. 2020; 13: 727-742
        • Heredia JD
        • Park J
        • Brubaker RJ
        • et al.
        Mapping interaction sites on human chemokine receptors by deep mutational scanning.
        J Immunol. 2018; 200: 3825-3839
        • Jones EM
        • Lubock NB
        • Venkatakrishnan AJ
        • et al.
        Structural and functional characterization of G protein-coupled receptors with deep mutational scanning.
        Elife. 2020; 9: 2265
        • Jia X
        • Burugula BB
        • Chen V
        • et al.
        Massively parallel functional testing of MSH2 missense variants conferring Lynch syndrome risk.
        Am J Hum Genet. 2021; 108: 163-175
        • Shalem O
        • Sanjana NE
        • Hartenian E
        • et al.
        Genome-scale CRISPR-Cas9 knockout screening in human cells.
        Science. 2014; 343: 84-87
        • Bushnell B
        • Rood J
        • Singer E.
        BBMerge - Accurate paired shotgun read merging via overlap.
        PLoS One. 2017; 12e0185056
        • Edgar RC.
        Search and clustering orders of magnitude faster than BLAST.
        Bioinformatics. 2010; 26: 2460-2461
        • Buchfink B
        • Xie C
        • Huson DH.
        Fast and sensitive protein alignment using DIAMOND.
        Nat Methods. 2015; 12: 59-60
        • Graham RP
        • Treece AL
        • Lindeman NI
        • et al.
        Worldwide frequency of commonly detected EGFR mutations.
        Arch Pathol Lab Med. 2018; 142: 163-167
        • Yoshikawa K
        • Hamada J
        • Tada M
        • et al.
        Mutant p53 R248Q but not R248W enhances in vitro invasiveness of human lung cancer NCI-H1299 cells.
        Biomed Res. 2010; 31: 401-411
        • Okamoto K
        • Okamoto I
        • Okamoto W
        • et al.
        Role of survivin in EGFR inhibitor-induced apoptosis in non-small cell lung cancers positive for EGFR mutations.
        Cancer Res. 2010; 70: 10402-10410
        • Wind S
        • Schmid M
        • Erhardt J
        • et al.
        Pharmacokinetics of afatinib, a selective irreversible ErbB family blocker, in patients with advanced solid tumours.
        Clin Pharmacokinet. 2013; 52: 1101-1109
        • Arulananda S
        • Do H
        • Musafer A
        • et al.
        Combination Osimertinib and Gefitinib in C797S and T790M EGFR-mutated non-small cell lung cancer.
        J Thorac Oncol. 2017; 12: 1728-1732
        • Wang Z
        • Yang JJ
        • Huang J
        • et al.
        Lung adenocarcinoma harboring EGFR T790M and in trans C797S responds to combination therapy of first- and third-generation EGFR TKIs and shifts allelic configuration at resistance.
        J Thorac Oncol. 2017; 12: 1723-1727
        • Hidaka N
        • Iwama E
        • Kubo N
        • et al.
        Most T790M mutations are present on the same EGFR allele as activating mutations in patients with non-small cell lung cancer.
        Lung Cancer. 2017; 108: 75-82
        • Brown BP
        • Zhang YK
        • Westover D
        • et al.
        On-target resistance to the mutant-selective EGFR inhibitor Osimertinib can develop in an allele-specific manner dependent on the original EGFR-activating mutation.
        Clin Cancer Res. 2019; 25: 3341-3351
        • Qin Y
        • Jian H
        • Tong X
        • et al.
        Variability of EGFR exon 20 insertions in 24 468 Chinese lung cancer patients and their divergent responses to EGFR inhibitors.
        Mol Oncol. 2020; 14: 1695-1704
        • Hayashi H
        • Iihara H
        • Hirose C
        • et al.
        Effects of pharmacokinetics-related genetic polymorphisms on the side effect profile of afatinib in patients with non-small cell lung cancer.
        Lung Cancer. 2019; 134: 1-6
        • Brown K
        • Comisar C
        • Witjes H
        • et al.
        Population pharmacokinetics and exposure-response of osimertinib in patients with non-small cell lung cancer.
        Br J Clin Pharmacol. 2017; 83: 1216-1226
        • YW Zofia Piotrowska
        • Sequist Lecia V.
        • Ramalingam Suresh S.
        ECOG-ACRIN 5162: a phase II study of osimertinib 160 mg in NSCLC with EGFR exon 20 insertions.
        J Clin Oncol. 2020; 38: 9513
        • Vyse S
        • Huang PH.
        Targeting EGFR exon 20 insertion mutations in non-small cell lung cancer.
        Signal Transduct Target Ther. 2019; 4: 5
        • Roskoski Jr, R
        Properties of FDA-approved small molecule protein kinase inhibitors: a 2020 update.
        Pharmacol Res. 2020; 152104609
        • Pines G
        • Fankhauser RG
        • Eckert CA.
        Predicting drug resistance using deep mutational scanning.
        Molecules. 2020; 25: 2265
        • Bulbul A
        • Leal A
        • Husain H.
        Applications of cell-free circulating tumor DNA detection in EGFR mutant lung cancer.
        J Thorac Dis. 2020; 12: 2877-2882
        • Zhou Q
        • Yang JJ
        • Chen ZH
        • et al.
        Serial cfDNA assessment of response and resistance to EGFR-TKI for patients with EGFR-L858R mutant lung cancer from a prospective clinical trial.
        J Hematol Oncol. 2016; 9: 86