Almost all patients with EGFR-driven lung cancer who are treated with EGFR tyrosine kinase inhibitors (TKI) develop resistance to treatment. A single base (c.2369C>T) transition mutation, EGFR T790M, is the most frequent resistance event after first-generation exposure to EGFR TKIs. Whether T790M mutation is acquired or is selected from a preexisting clone has been a matter of significant debate. In this study, we show that treatment with EGFR TKIs leads to activation of the NFκB pathway, which in turn induces expression of activation-induced cytidine deaminase (AICDA). In turn, AICDA causes deamination of 5-methylcytosine to thymine at position c.2369 to generate the T790M mutation. Pharmacologic inhibition of the NFκB pathway or knockout of AICDA decreased the frequency or prevented the development of T790M mutation, respectively. In addition, patients treated with first-line EGFR TKI displayed increased expression of AICDA and detection of the T790M mutation upon progression. These results identify the mechanism of T790M acquisition and present an opportunity to target the process to delay or prevent it.
These findings identify the mechanism behind acquisition of a common resistance mutation to TKI treatment in lung cancer.
Lung cancer is the leading cause of cancer-related deaths in the United States with estimated 234,000 new cases and 154,000 deaths this year (1). Multiple driving mutations have been identified in non–small cell lung cancer (NSCLC) including EGFR mutations, which occur in 10% to 15% of patients of Caucasian ethnicity with lung adenocarcinoma and in 50% of Asian population (2). EGFR mutations mainly affect the EGFR tyrosine kinase domain with the two most common mutations being deletions in exon 19 and missense substitution in exon 21, L858R, where leucine is replaced by arginine (3–5). These mutations lead to constitutive activation of EGFR and NSCLC cellular dependence on this pathway. Treatment with EGFR tyrosine kinase inhibitors (TKI) decreases the activity of downstream cell survival and proliferation of signaling pathways (6, 7). Currently, EGFR TKIs are the standard-of-care for patients with EGFR-mutant NSCLC who have a better response rate and progression-free survival (PFS) as compared with chemotherapy (8, 9). Unfortunately, almost all patients with EGFR-mutant NSCLC treated with first-generation EGFR TKIs develop resistance with mean PFS range of 9.2 to 13.1 months (9). Mechanisms of resistance include various gene amplifications and mutations as well as epithelial–mesenchymal transition and transformation into small-cell lung cancer (10–13). However, the main mechanism of developing resistance to first-generation TKIs is a single-nucleotide transition mutation in EGFR; a cytosine to thymine (C>T) at position 2369 (CAC/CAT) causing threonine to methionine amino acid change at codon 790, T790M (14, 15). This T790M mutation occurs in cis with the original driving mutation and leads to steric hindrance and increased binding affinity for ATP (16, 17).
T790M mutation is usually not detected at baseline and it has been controversial whether the lack of detection is due to assay sensitivity. Multiple reports have suggested high incidence of T790M mutation (18); however, others have refuted this as artifact due to formalin tissue fixation. T790M mutations were detected in the formalin-fixed cancer specimen, but not their fresh frozen counterparts; in addition, they were even present in formalin-fixed normal lung tissue (19). Although the presence of T790M at baseline cannot be totally ruled out, reports have shown that it can be acquired after TKI treatment (20).
Deamination of cytosine can lead to the generation of either uracil or thymine depending on the methylation status (Supplementary Fig. S1). Uracil is usually removed by the activity of mismatch repair system through uracil-DNA glycosylase (UDG). This creates an apurinic/apyrimidinic (AP) site that could lead to mutation at that position. Activation-induced cytosine deaminase (AICDA) is usually expressed in germinal center B-lymphocytes upon antigen exposure. AICDA translocates to the nucleus, deaminates cytosine (C) in single-stranded DNA, and converts it to uracil (U). The U:G mismatch lesions are converted into point mutations known as somatic hypermutation allowing for a diverse antibody repertoire (21). Furthermore, AICDA causes other genes mutations and is implicated in the development of diffuse large B-cell lymphoma (22). Overexpression of AICDA in chronic myeloid leukemia (CML) is associated with increased frequency of C>T mutation and resistance to imatinib exposure (23). In solid tumors, AICDA is expressed in human hepatocellular carcinoma (24), gastric cancer (25), colorectal cancer (26), and lung cancer (27), suggesting a potential role in these tumors too. Because the T790M mutation is a C>T mutation that is acquired upon TKIs exposure, in this report, we explored whether AICDA is activated after EGFR TKIs use in lung adenocarcinoma. We further assessed whether AICDA manipulation affects T790M mutation frequency and whether cytosine at position 2369 is methylated (5-methylcytosine) where its deamination would directly lead to the observed C>T mutation rather than C>G or C>A (Supplementary Fig. S1).
Materials and Methods
Cell line PC9 was obtained from Sigma-Aldrich. NCI-H3255 was a generous donation from Pasi Janne (Dana Farber Cancer Institute, Boston, MA). NCI-H1650, NCI- H1299, NCI-H1975, and NCI-H460 were obtained from ATCC. All cell lines were authenticated by short tandem repeat profiling and Mycoplasma testing was conducted. Cell lines were grown in RPMI1640 with 10% FBS final concentration per recommendation.
Droplet digital PCR
Allele-specific fluorescent primer/probes for wild-type (dHsaCP2000020) and T790M (dHsaCP2000019)-mutated EGFR were obtained from Bio-Rad. ddPCR was then performed on DNA extracted from cell lines. Each ddPCR run included a positive control derived from NCI-H1975 sample harboring the T790M mutation and was performed as such: a master mix containing 2X ddPCR Supermix, 20X Primer/Probe Mix, 25X Droplet Stabilizer (RainDance Technologies), and 2–5U Restriction Endonuclease (HindIII; New England BioLabs) was added to separate wells containing 5–500 ng gDNA, for a total reaction volume of 20 μL, and then incubated at room temperature for 15 minutes. Samples were then pipetted into a RainDrop Source Chip (RainDance Technologies), and 3.5 to 4.5 million droplets were generated with the RainDrop Source machine. The resulting emulsion was amplified using the following settings: 95°C for 10 minutes, and then 40 cycles at 94°C for 30 seconds, and 55°C for 60 seconds, followed by 98°C for 10 minutes. Endpoint fluorescence was then measured with a RainDrop Sense machine and analyzed with RainDrop Analyst software. The wild-type and mutant signals of each control sample were used to draw positive gates to then analyze events in samples.
To determine sensitivity of ddPCR for T790M mutation, the T790M assay was carried out using 10 ng of H1975 genomic DNA serially diluted by factor of 10 in 50,000 copies of genomic DNA. The mean lowest copy number detected at 0.01 ng was three, and thus the limit of detection was 0.012% (Supplementary Fig. S2). We also detected 1 to 2 events of T790M in normal genomic DNA; thus, using the mean number of events and 2 SDs as the cutoff, we determined the specificity of the T790M mutation to be five in 50,000 copies.
Cell line treatment
For continuous long-term gefitinib treatment, approximately 1 × 106 cells were plated in T75 flasks overnight and treated the second day with 1/10th of TKI IC50. After 48 hours, cells were harvested and split into three parts. One part was used for genomic DNA extraction and the other part for RNA extraction. The third part was replated overnight and treated with double the previous dose. This process was repeated, with a DMSO-treated control, until a final dose of 100× the IC50 was reached. For combination treatment, Bay11-7082 was added at a constant dose of 0.5 μmol/L throughout the increasing dose of gefitinib exposure. For short-term EGFR TKIs treatment, cell lines were treated with various indicated doses for 24 hours to assess for AICDA or 6 hours to assess for NFκB pathway components.
Total RNA was isolated using RNeasy Mini Kit (Qiagen) RT-PCR assays were done in a ABI PRISM 7900HT sequence detection system. TaqMan primer/probes were selected from Thermo Fisher Scientific: AICDA Hs00757808_m1 and control RPS13 Hs01945436_u1. The relative expression mRNA level of AICDA was calculated with respect to internal standard RPS13 gene to normalize for variations in RNA quality and the amount of cDNA. The results are a representation of at least two biological repeats and three technical replicates. Unpaired t test was used where P < 0.05 was considered statistically significant.
AICDA CRISPR knockout
We used an all-in-one red fluorescent protein (RFP)-labeled Cas9 and guide RNA (gRNA) expression plasmids from Sigma-Aldrich targeting AICDA exon 2. Briefly, PC9 cells were transfected with the CRISPR/Cas9 AICDA plasmid using standard transfection methods. RFP-positive cells were selected and single-cell clones were generated in 96-well plates. Each clone was expanded, the area around the expected cutting site was amplified, and Sanger sequencing performed. Using the TIDE web-tool software (28), sequencing results were assessed and AICDA-mutated clones were selected.
UDG hydrolysis assay
We have synthesized an EGFR oligomer, which allows detection of cytosine deamination through the hydrolysis activity of uracil-DNA glycosylase (UDG) in alkaline environment. This oligomer is 79 nt in length; the first 5′ end 28 nt represent EGFR sequence 2355 to 2383. The remaining 51 nt represent the 2355–2406 nucleotide sequence of EGFR. This will provide a repeat sequence for the deamination process to occur. AICDA protein from a baculovirus system was a generous gift from Dr. Myron Goodman (University of Southern California, Los Angeles, CA) and AICDA was purified by Center for Structural Biology. In summary, AICDA is incubated with 0.5 μmol/L probe in a buffer containing 50 mmol/L HEPES, pH 7.5, 10 mmol/L MgCl2, and 1 mmol/L dithiothreitol for 30 minutes at 37°C. The reaction will be terminated by an extraction with phenol/chloroform/isoamyl alcohol (25:24:1) and incubated with UDG at 37°C for 1 hour. Hydrolysis is achieved by heating at 95°C for 10 minutes in 0.2 mol/L NaOH. Electrophoresis was done using a 4% to 20% TBE gel.
Patient samples were collected through the University of Michigan Molecular Pathology Laboratory. The clinical variables are provided in Supplementary Table S1. Evaluation of patient samples was conducted using surplus pathology materials with a waiver of informed consent granted by the Institution Review Board (IRB) in accordance with recognized ethical guidelines of Declaration of Helsinki, Belmont Report, and US Common Rule. Samples used for this research were patients’ tumors that were sequenced for original EGFR-driving mutation, and upon progression, biopsies were obtained and sequenced for the T790M mutation. RNA was isolated using RNeasy Mini Kit and RT-PCR was done as mentioned above.
Genomic DNA (500 ng) was bisulfite treated and cleaned using EZ DNA Methylation Kit (Zymo Research). Then, 20 ng of the bisulfite-treated DNA was used for PCR amplification with Roche HiFidelity PCR system, in a final volume of 10 μL, and using 400 nmol/L of forward/reverse primers (aggaagagagTTTGTTGGGTATTTGTTTTATTTTT/cagtaatacgactcactatagggagaaggctACTAAAAACCAATATTATCTTTATATTCCC). The following PCR amplification settings were used: (i) 94°C, 15 minutes; (ii) 45 cycles of: 94°C, 20 seconds; 62°C, 30 seconds; 72°C, 1 minute; (iii) 72°C, 3 minutes; and (iv) hold at 4°C. A volume of 2 μL of each reaction was applied to the HiSeq D1000 TapeStation Assay (Agilent Technologies) to verify the presence of an amplicon at the expected size. The amplicons were then submitted to the University of Michigan DNA Sequencing Core for SAP, in vitro transcription/cleavage, cleanup, and spotting onto a SpectroCHIP (Agena BioScience) for mass spectra acquisition. The data were analyzed with the EpiTyper software (Agena Bioscience) and exported to an excel spreadsheet.
T790M is acquired after TKIs treatment
The absence of baseline EGFR T790M in contrast to its prevalence after first-line EGFR TKI treatment, raises an important question about being an acquired event. In fact, previous work done by Engleman's laboratory (20) showed that PC9, an EGFR del19 lung adenocarcinoma cell line, can acquire T790M mutation from T790M-negative clones. Thus, using flow cytometry, we established PC9 single-cell clones in 96-well plates. And then verified by digital droplet PCR (ddPCR) that these PC9 single-cell clones had no baseline T790M mutation (Fig. 1, panels I–VIII). However, upon treating these clones with an increasing dose of gefitinib, we detected the emergence of T790M mutation, confirming the acquired nature of this mutation (Fig. 1, panels IX–XII).
AICDA is induced upon EGFR TKI exposure
AICDA capability has been implicated in resistance to imatinib in CML where its overexpression can induce blast crisis as well as the C>T gatekeeper-resistant mutation, T315I (23). We thus explored the possibility that exposure to EGFR TKIs induced a cytosine deamination process, promoting the cytosine to thymine mutation. We screened for multiple known cytosine deaminases mRNA expression after exposure to gefitinib, a first-line EGFR TKI. We found that gefitinib mainly increased AICDA expression (Supplementary Fig. S3). To determine whether this is a unique response to gefitinib or a class effect, we used other known EGFR TKIs: first generation, erlotinib; second generation, afatinib; and third generation, osimertinib. As shown in Fig. 2A, there is significant increase in the mRNA expression of AICDA upon treatment with different TKIs in PC9 cell line. This is also seen in other EGFR-mutated lung adenocarcinoma cell lines (Supplementary Fig. S3).
AICDA is induced through noncanonical NFκB
It is intriguing that after EGFR TKI exposure and before T790M resistance development, a tolerant cell population persists. Blakely and colleagues have shown that initial cell survival occurs due to activation of NFκB pathway (29), which is a known AICDA transcription factor. Thus, we tested whether NFκB pathway was induced after EGFR treatment and whether its components bind to AICDA promoter. As shown in Fig. 2B, the expression of RelB and p52 (noncanonical) were increased after 24-hour exposure to TKI. This indicates that the NFκB noncanonical pathway is activated by TKI inhibition of EGFR. Using Bay11-7802, a known inhibitor of NFκB pathway, the expression of AICDA was abrogated after TKI treatment (Supplementary Fig. S4). We then used a chromatin immunoprecipitation assay to evaluate whether RelB interacts directly with AICDA in PC9 cells after TKI exposure. RelB and RNA polymerase II antibodies were used for immunoprecipitation with IgG as control. Then qPCR was performed with primers covering the AICDA promoter region. As seen in Fig. 2C, there is a significant increase in binding of RelB to AICDA promoter after exposure to TKI. This indicates that the noncanonical NFκB pathway induces AICDA expression after EGFR inhibition. RNA polymerase II is also recruited to AICDA promoter, further supporting the increase in AICDA transcription after EGFR inhibition.
AICDA knockdown decreases T790M frequency
Because AICDA is induced after EGFR TKI treatment, we evaluated its ability to deaminate cytosine in EGFR gene. To this end, we purified AICDA protein and tested cytosine deamination of an EGFR oligomer designed to contain the cytosine of interest, c.2369, twice. We found that AICDA does recognize EGFR sequence and induces deamination of cytosine, leading to hydrolysis of the oligomer at these nucleotides (Supplementary Fig. S5). We next asked whether inhibiting AICDA expression would have an impact on the development of T790M mutation. We used two approaches to interrogate AICDA role. First, we inhibited NFκB pathway induction by using Bay11-7082 and found that combination treatment of gefitinib and Bay11-7082, abrogated the development of T790M mutation after long-term exposure to EGFR TKI (Fig. 3A). Second, we generated CRISPR/Cas9 AICDA knockout clones verified by Sanger sequencing (Supplementary Fig. S6) and treated with an increasing dose of gefitinib. As seen in Fig. 3B, using ddPCR we did not detect T790M mutations in the AICDA knockout clones (panels IV and VI) at a gefitinib dose of 64 nmol/L in comparison with PC9 control (panel II). Thus, AICDA is necessary to deaminate cytosine at position 2369 in PC9 cells in response to TKIs. We further evaluated AICDA expression in primary lung cancer tissues from patients before and after the development of resistance to EGFR TKI treatment. As shown in Fig. 4, AICDA expression significantly increased upon progression on first-line TKI treatment and detection of the T790M-resistant mutation, supporting the induction of AICDA after TKI exposure.
EGFR c.2369 cytosine is methylated
EGFR T790M–resistant mutation is a single-nucleotide transition of cytosine to thymine, C>T. Because deamination of 5-methylcytosine leads to C to T transition (Supplementary Fig. S1), we proposed that cytosine at position c.2369 is methylated. Using methylation-specific primers and mass spectrometry, we found that c.2369 cytosine is methylated in number of lung cancer cell lines (Supplementary Table S2) including PC9. Our findings suggest that deamination of the 5-methylcytosine to thymidine at position c.2369 causes the threonine/methionine amino acid change that alters TKI-binding affinity and causes resistance.
The impact of targeted therapy in oncology is undeniable since the 2001 FDA approval of imatinib for CML. This opened the door to the discovery of multiple drugs directed at known specific driver mutations in various cancers. Drugs against EGFR mutations in NSCLC were among the first to be FDA approved after imatinib. Interestingly, the resistant mutations to EGFR TKIs were predicted on the basis of imatinib CML resistance where the discovered EGFR T790M mutation has the same “gate keeper” role of T315I mutation in BCR-ABL. Moreover, it is very intriguing that a similar mutational profile is also seen in gastrointestinal stromal tumors (GIST) and hypereosinophilic syndrome (HES) tumors after treatment with imatinib. In fact, a single-nucleotide C>T mutation is detected in BCR-ABL (CAC/CAT), c-KIT (CAC/CAT), and FIP1L1–PDGFRA (CAC/CAT) in imatinib-resistant CML (30), GIST (31), and HES (32), respectively.
In all these tumors, the resistant mutation is highly prevalent, is not detected at baseline, occurs after several months of treatment, and is always C to T point mutation. Although a random mutation event followed by clonal selection cannot be completely ruled out, a specific mechanism responsible for this event is worth interrogating.
AICDA is expressed in germinal center B lymphocytes and induced upon antigen exposure, causing cytosine deamination and point mutations in immunoglobulin genes. These acquired mutation processes, known as somatic hypermutations, provide the needed diversified antibody repertoire (33). AICDA has been implicated in cytosine deamination in nonimmunoglobulin genes as well (34) and in the development of resistance to imatinib (23). In CML, overexpression of AICDA was associated with increased frequency of C>T mutations and imatinib resistance (23).
In this report, we investigated whether AICDA plays a role in the development of T790M mutation after EGFR TKI treatment. Using ddPCR, one of the most sensitive mutation assays, we first verified that the T790M mutation is not detected at baseline in the tested single-cell PC9 clones; however, T790M mutation is identified upon exposure to TKI, suggesting that it is an acquired event. We also show that AICDA is induced after treatment with first-, second-, and third-generation EGFR TKIs. Interestingly, verifying that cytosine at position c.2369 is methylated, provides strong explanation to the deamination mechanism of C>T mutation. Finally, knocking down AICDA pharmacologically through inhibiting NFκB pathway or genetically through CRISPR precluded T790M development after TKI exposure.
This article explicitly links EGFR TKIs to the development of a specific DNA-resistant mutation through AICDA activity. Thus, targeting AICDA directly by developing specific inhibitors or indirectly by inhibiting the NFκB pathway, provides an opportunity to delay/prevent T790M mutation. Furthermore, AICDA was shown to increase genome instability in B-cells after phosphatidylinositol 3-kinase δ blockade (35). We performed whole exome sequencing at an average depth of 100× comparing mutation profile before and after treatment with gefitinib. We found an increase in frequency of variants and specifically of SNPs after gefitinib exposure (Supplementary Table S2). This finding is similar to previously reported results by Jia and colleagues, where an increase in mutation frequency was detected in cell lines after EGFR TKI treatment (36). In addition, the direct effect of EGFR TKIs on base excision repair through the degradation of HSP70 was recently described (37). The consequences of deamination might even go beyond facilitating the generation of resistance mutations to perhaps contributing to tumorigenesis. Consistent with this idea, the development of small-cell lung cancer transformation after TKI exposure is associated with APOBEC mutation profile in lung EGFR-driven tumors (38). Our findings suggest that specific DNA mutations are the sequelae of a dynamic response to an environmental stress. The activation of the NFκB pathway ensures cancer cells’ tolerance and induces AICDA-driven mutagenesis, leading to resistant mutation acquisition and eventually the selection of the best-fit clone, reminiscent of antibody affinity maturation. Targeting AICDA activity directly (developing specific inhibitors) or indirectly with NFκB pathway inhibitors as well as using DNA repair inhibitors might lead to the delay or prevention of T790M resistance.
Disclosure of Potential Conflicts of Interest
G.P. Kalemkerian reports receiving a commercial research grant from Merck, Takeda, Abbvie, GlaxoSmithKline, and Pfizer. No potential conflicts of interest were disclosed by the other authors.
Conception and design: N. El Kadi, K.A. Hassan
Development of methodology: N. El Kadi, L. Wang, H. Korkaya, B.L. Betz, K.A. Hassan
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): N. El Kadi, L. Wang, A. Davis, A. Cooke, V. Vadnala, N.A. Brown, K.A. Hassan
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): N. El Kadi, L. Wang, A. Davis, H. Korkaya, A. Cooke, M. Cascalho, G.P. Kalemkerian, K.A. Hassan
Writing, review, and/or revision of the manuscript: N. El Kadi, L. Wang, N.A. Brown, M. Cascalho, G.P. Kalemkerian, K.A. Hassan
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): N. El Kadi, A. Davis, A. Cooke, N.A. Brown, G.P. Kalemkerian, K.A. Hassan
Study supervision: K.A. Hassan
This work was supported in part by NCI 5K08 CA158425-03, ALA Lung Cancer Discovery Award, UM Thoracic Oncology Program Elizabeth A. Crary Fund, and CSHL-JAX-SWOG ITSC & HOPE Foundation (to K. A. Hassan). Research reported in this article was supported by the NCI of the NIH under award number P30CA046592 by the use of the Rogel Cancer Center Flow Cytometry, Vector, DNA sequencing, Bioinformatics, and Epigenomics core facilities.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.