MLN2480

Acquired BRAF Rearrangements Induce Secondary Resistance to EGFR therapy in EGFR-Mutated Lung Cancers

Morana Vojnic 1, Daisuke Kubota 1, Christopher Kurzatkowski 1, Michael Offin 2, Ken Suzawa 1, Ryma Benayed 1, Adam J Schoenfeld 2, Andrew J Plodkowski 3, John T Poirier 4, Charles M Rudin 5, Mark G Kris 5, Neal X Rosen 6, Helena A Yu 5, Gregory J Riely 5, Maria E Arcila 1, Romel Somwar 7, Marc Ladanyi 8

Abstract
Introduction
Multiple genetic mechanisms have been identified in EGFR-mutant lung cancers as mediators of acquired resistance (AR) to EGFR tyrosine kinase inhibitors (TKIs), but many cases still lack a known mechanism.

Methods
To identify novel mechanisms of AR, we performed targeted large panel sequencing of samples from 374 consecutive patients with metastatic EGFR-mutant lung cancer, including 174 post-TKI samples, of which 38 also had a matched pre-TKI sample. Alterations hypothesized to confer AR were introduced into drug-sensitive EGFR-mutant lung cancer cell lines (H1975, HCC827, and PC9) by using clustered regularly interspaced short palindromic repeats/Cas9 genome editing. MSK-LX138cl, a cell line with EGFR exon 19 deletion (ex19del) and praja ring finger ubiquitin ligase 2 gene (PJA2)/BRAF fusion, was generated from an EGFR TKI–resistant patient sample.

Results
We identified four patients (2.3%) with a BRAF fusion (three with acylglycerol kinase gene (AGK)/BRAF and one with PJA2/BRAF) in samples obtained at AR to EGFR TKI therapy (two posterlotinib samples and two posterlotinib and postosimertinib samples). Pre-TKI samples were available for two of four patients and both were negative for BRAF fusion. Induction of AGK/BRAF fusion in H1975 (L858R + T790M), PC9 (ex19del) and HCC827 (ex19del) cells increased phosphorylation of BRAF, MEK1/2, ERK1/2, and signal transducer and activator of transcription 3 and conferred resistance to growth inhibition by osimertinib. MEK inhibition with trametinib synergized with osimertinib to block growth. Alternately, a pan-RAF inhibitor as a single agent blocked growth of all cell lines with mutant EGFR and BRAF fusion.

Conclusion
BRAF fusion is a mechanism of AR to EGFR TKI therapy in approximately 2% of patients. Combined inhibition of EGFR and MEK (with osimertinib and trametinib) or BRAF (with a pan-RAF inhibitor) are potential therapeutic strategies that should be explored.

Introduction
Long-term clinical benefits of EGFR tyrosine kinase inhibitors (EGFR TKIs) in EGFR-mutant NSCLCs remain limited by the development of drug resistance.1 The most common mechanism of resistance to first- and second-generation EGFR TKIs, the EGFR T790M mutation (which accounts for 60% of cases), can be targeted successfully with third-generation drugs such as osimertinib.2 Recently, osimertinib has been approved as first-line therapy for EGFR-mutant lung cancer.3 As expected, disease progression due to acquired resistance to osimertinib has emerged, with the most common alteration described to date being the EGFR C797S mutation.

Activation of parallel signaling pathways (amplification of erb-b2 receptor tyrosine kinase 2 gene [HER2] and gene MNNG HOS Transforming gene [MET]) or downstream signaling (mutations in KRAS, phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha gene [PIK3CA], and BRAF) has also been shown to induce resistance to osimertinib and other EGFR TKIs.5 Several acquired gene rearrangements, including fusions involving ALK receptor tyrosine kinase gene (ALK), BRAF, fibroblast growth factor receptor 3 gene (FGFR3), neurotrophic receptor tyrosine kinase 1 gene (NTRK1), and ret proto-oncogene gene (RET), have recently been reported as possible mechanisms of resistance to EGFR TKI therapy.6–11 Clinically, a combination of an EGFR TKI with a drug targeting either ALK or RET, has been shown to be effective in patients with lung cancer with an EGFR mutation and the respective gene fusion. However, only RET or FGFR3 fusions have been experimentally proved to confer resistance to EGFR inhibitors in EGFR-mutant lung cancers.

Activating BRAF fusions typically occur at a very low frequency across a wide variety of cancers, with the exception of pilocytic astrocytomas.9,14 BRAF rearrangements are divided into N-terminal deletions, kinase domain duplications, and BRAF fusions. Overall, BRAF alterations are present in 4.4% of NSCLCs, and rearrangements represent 4.3% of all alterations.10 Rearrangements retain the BRAF kinase domain at the 3′ end while lacking the N-terminal inhibitory domain. Because of the lack of the N-terminal BRAF-inhibitory domain, rearrangements result in constitutive dimerization of RAF proteins independently of RAS activation,15 thereby activating downstream MAP kinase signaling.

Here, we report four cases of EGFR-mutant NSCLC with concurrent expression of a BRAF fusion and provide functional data supporting BRAF fusions as a mechanism of acquired resistance to EGFR therapy in EGFR-mutant lung adenocarcinomas. In addition, we present data supporting treatment strategies focusing on targeting RAF with pan-RAF inhibitors.

Methods
Generation of AGK/BRAF Fusion by Using CRISPR/Cas9 in EGFR-Mutant Lung Cancer Cell Lines
gRNA Design and Cloning
To faithfully model a BRAF fusion, guide RNAs (gRNAs) were designed to generate a fusion linking acylglycerol kinase gene (AGK) exons 1 and 2 with BRAF exons 8 to 18. Specifically, four gRNAs were designed (Supplementary Table 1) to target the intron between AGK exons 2 and 3 or BRAF exons 7 and 8 and selected by using the tool at http://crispr.mit.edu/. The placement of these gRNAs aimed to minimize splicing interference by being positioned at least 50 to 250 bp from the splice site. gRNA cloning was performed according to previously published protocols.16 A quantity of 1 μg of px458 (containing Cas9 and green fluorescent protein complementary DNAs [cDNAs]) was digested with 1 μL of FastDigest Bpil, 1 μL of fast AP, and 2 μL of 10× FastDigest buffer diluted in 6 μL of water at a final volume of 20 μL at 37oC for 45 minutes. The resulting DNA was cleaned up with a polymerase chain reaction (PCR) clean-up kit to a final concentration of 15 to 30 ng/μL.

Next, 1 μL of each pair of gRNAs was phosphorylated and annealed with 1 μL of 10× T4 ligation buffer, 0.5 μL of T4 polynucleotide kinase, and 6.5 μL of RNAse/DNAse-free water to a final volume of 10 μL in a Bio-Rad (Hercules, CA) C1000 Touch thermocycler for 30 minutes at 37oC, then in decrements in temperature from 95oC to 25oC in 5-minute intervals. Annealed gRNAs were ligated into px458 by using 50 ng of vector, 1.5 μL of annealed gRNAs (1:100 dilution), 5 μL of quick ligase buffer, and 1 μL of quick ligase with the addition of RNAse/DNAse-free water to a final volume of 10 μL at room temperature for 30 minutes. The annealed and ligated plasmids were then transformed by using TOP10 chemically competent Escherichia coli (ThermoFisher). The plasmid DNA was extracted and insertion of gRNAs into px458 was confirmed by Sanger sequencing.

Validation of gRNAs
To test the efficacy of fusion generation, human bronchoepithelial cells (plated at a density of 500,000 cells/well in 6-well plates) were transfected with each possible pair of gRNAs with 1 μg of each clustered regularly interspaced short palindromic repeat (CRISPR)/Cas9 construct. After 48 hours, green fluorescent protein was readily visible, and after 72 hours each pool of cells was harvested for RNA extraction, cDNA synthesis, and reverse-transcriptase to detect the fusion mRNA. The pool of cells containing BRAF gRNA 4 and AGK gRNA 4 generated detectable fusion mRNA. This pairing (BRAF gRNA 4 and AGK gRNA 4) was transfected into H1975, HCC827, and PC9 cells plated at a density of 1,500,000 cells per 10-cm dish with 7.5 μg of each plasmid and fusion mRNA detected as already described.

Clonal Selection
Transfected cells were maintained in growth media supplemented with 0.5 μM osimertinib to select for a drug-resistant population. The resistant cells were then replated at a density of 500 cells per 10-cm dish to isolate single-cell clones. Colonies were then picked with cloning disks and replated in a 96-well plate for expansion in Roswell Park Memorial Institute 1640 medium containing 10% fetal bovine serum, 1% antibiotic/antimycotic, and 0.5 μM osimertinib. Once the clones began to proliferate readily, cells were serially moved to larger containers and then subjected to reverse-transcriptase PCR for fusion detection.

Generation of BRAF Fusion Constructs
The cDNA encoding praja ring finger ubiquitin ligase 2 gene (PJA2)/BRAF fusion was amplified from MSK-LX138cl cells by PCR and cloned into a pENTR/TOPO vector. The cDNA encoding AGK/BRAF fusion was generated by first amplifying the AGK and BRAF fragments from HEK-293T cells and then assembling the two pieces into a KpnⅠ-EcoRⅤ–digested pENTR1A vector by Gibson assembly. These BRAF fusion cDNAs were then subcloned into a pLX303 lentiviral transfer vector. The primer sequences used for plasmid construction are shown in Supplementary Table 2. All constructs were confirmed by DNA sequencing. Lentiviruses harboring the two fusions were generated using HEK-293T cells and then used to infect cells of the lung cancer lines. Cells stably expressing the cDNAs were selected with blasticidin (20 μg/mL) for 10 days. The primers used for confirmation of genomic fusions and expression of fusion transcripts are provided in Supplementary Table 3.

shRNA Infection and BRAF Knockdown
Short hairpin RNAs (shRNAs) (TRCN0000195066 [Sh2], TRCN0000195609 [Sh3], and TRCN0000196844 [Sh1]) were obtained from Sigma-Aldrich (St. Louis, MO). Lentiviruses were generated by using HEK-293T cells and used to infect cells. The cells were plated in six-well plates at a density of 150,000 to 500,000 cells/well and then infected 24 hours later with viral supernatant (multiplicity of infection = 7) expressing BRAF shRNAs or nontargeting sequence, mixed with polybrene (10 μg/mL). Infected cells were selected with puromycin (5 to 10 μg/m) and then replated for viability testing with osimertinib or used to detect BRAF by RT quantitative PCR.

Results
Coexpression of EGFR Mutation and BRAF Fusions in Lung Cancer Samples Refractory to EGFR TKI Therapy
We performed targeted large panel sequencing using the Memorial Sloan Kettering–Integrated Mutation Profiling of Actionable Cancer Targets (MSK-IMPACT) assay14,17 on 374 consecutive patients with metastatic EGFR-mutant lung cancers, including 200 patients who were tested before receiving an EGFR TKI, 136 patients who were tested after progression while taking an EGFR TKI, and 38 patients with both types of samples. We identified four patients (four of 174 [2.3%]) with disease progression while taking an EGFR TKI (two who were taking erlotinib and two who were taking erlotinib followed by osimertinib) with an EGFR mutation and a BRAF fusion (three with AGK/BRAF fusion and one with PJA2/BRAF fusion). In one patient for whom paired pre-TKI and post-TKI samples were available, an AGK/BRAF fusion was present only after progression while the patient was taking erlotinib, supporting the hypothesis that BRAF fusions are possible acquired mechanisms of drug resistance. The other three EGFR TKI–resistant patients lacked pre–EGFR TKI therapy data from a same-site sample; however, one patient had pre–EGFR TKI therapy MSK-IMPACT testing on a lung biopsy specimen and a BRAF fusion on a pleural effusion specimen (Supplementary Table 4). Importantly, none of the 200 patients whose tumor sample was obtained before EGFR TKI treatment showed a concurrent BRAF fusion.
Clinical Course of Two Patients with Matched Pre-TKI and Post-TKI Samples

Case 1
A 48-year-old male never-smoker presented with a right 12th cranial nerve palsy and was found to have metastatic lung adenocarcinoma to the brain, liver, and bone and lymphangitic spread in the lungs (see patient 2 in Supplementary Table 5) (Fig. 1). Evaluation of the right upper lobe biopsy sample by PCR revealed a 15-bp EGFR exon 19 deletion (ex19del). The patient received whole-brain radiation therapy followed by single-agent erlotinib, 150 mg daily. A complete depiction of the treatment course and diagnosis is shown in Fig. 1A. The patient initially had a robust clinical and radiologic response, but after 7.7 months of treatment he was noted to have progressive disease in the bone and subsequently in the liver. A biopsy was performed on the liver, and MSK-IMPACT17 testing revealed a newly acquired EGFR T790M mutation. The patient was subsequently given osimertinib, 80 mg daily, with an initial response. After 9.5 months the patient had radiologic and clinical progression with right-sided chest pain and a pleural effusion. A thoracentesis was done, with MSK-IMPACT testing revealing a BRAF structural rearrangement. Targeted RNA sequencing using MSK-Fusion Solid, a panel assay based on anchored multiplex PCR,18 was performed and showed the AGK-BRAF fusion (see Supplementary Table 4). The patient’s treatment was changed to nivolumab, but he showed further clinical deterioration and died 2 weeks later (see Fig. 1).

Figure 1 Clinical information for patients with resistance to EGFR tyrosine kinase inhibitor and BRAF fusions. (A) Disease course of patient 2, who was found to have acquired resistance to osimertinib and acylglycerol kinase gene (AGK)/BRAF fusion. (B) Disease course of patient 4, who was found to have acquired resistance to erlotinib and progressive mediastinal disease with AGK/BRAF fusion.

Case 2
A 69-year-old female former smoker (22 pack-years) had her cancer diagnosed as stage IIa (pT2aN1M0) lung adenocarcinoma after a left upper lobe lobectomy and mediastinal node dissection. She was treated with adjuvant cisplatin and pemetrexed for four cycles (see patient 4 in see Supplementary Table 5) (see Fig. 1). She was maintained on active surveillance for 2.9 years, at which time she was noted to have a metastatic recurrence in the brain. She underwent a metastasectomy with pathologic examination of a right frontal lobe metastasis showing a 1.8- cm tumor with predictive immunohistochemistry negative for EGFR L858R and ALK. MSK-IMPACT analysis revealed an EGFR ex19del. A time line of the clinical course is shown in Figure 1B. An interval computed tomography scan found enlarging mediastinal lymph nodes, and the patient was given erlotinib, 150 mg daily. She had a clinical and radiologic response to erlotinib and continued taking it for 1.5 years, at which point she was noted to have progression in her mediastinal lymph nodes. MSK-IMPACT analysis of the posterlotinib treatment sample identified the original EGFR ex19del as well as a new AGK/BRAF fusion (see Supplementary Table 4). The patient was placed on multiple lines of cytotoxic chemotherapy; however, she had continued disease progression and died.

Generation of Isogenic EGFR-Mutant Cell Lines with BRAF Fusions
To examine the influence that acquired BRAF rearrangements have on sensitivity to EGFR TKIs, we inserted an AGK/BRAF fusion into the EGFR-mutant lung cancer cell lines H1975 (T790M and L858R), HCC827 (EGFR ex19del), and PC9 (EGFR ex19del) by using CRISPR/Cas9 genome editing.16,19 An outline of the experimental scheme is illustrated in Figure 2A. Guides were designed to target intron 1 of AGK and intron 7 of BRAF to generate a fusion of AGK exon 2 with exon 8 of BRAF (see Fig. 2A) and cloned into pX458 expression plasmid. Plasmids were introduced and cells selected with osimertinib as described in Methods. All osimertinib-resistant clones were positive for the AGK/BRAF fusion by PCR (22 H1975, 16 HCC827, and 20 PC9). Two clones of each cell line that was positive for AGK/BRAF fusion at the mRNA and genomic DNA level were selected for further analysis (Fig. 2B). The fusion junction of the two genes was confirmed by Sanger sequencing (Fig. 2C).

Figure 2 Generation and characterization of CRISPR/Cas9 acylglycerol kinase gene (AGK)/BRAF cell lines. (A) clustered regularly interspaced short palindromic repeat (CRISPR)/Cas9 genome engineering protocol scheme. (B) Expression of AGK/BRAF. (C). Sanger sequencing of polymerase chain reaction (PCR) product confirming fusion in cell line of AGK exon 2 with exon 8 of BRAF. (D) H1975 and PC9 cells expressing AGK/BRAF were treated with 0.5 osimertinib for 1 hour, whole-cell extracts were prepared, and then lysates were subjected to immunoblotting. The results represent three independent experiments. (E) The indicated cell lines were plated at a density 7500/well in 96-well plates and treated with osimertinib for 96 hours; the concentration that inhibits 50% (IC50) values for growth inhibition are shown in the Table (F). (G and H) Cells were plated at a density of 150,000 cells/well in six-well plates and then infected 24 hours later with lentiviruses harboring the indicated short hairpin RNAs (shRNAs) and then selected with puromycin 48 hours later.

BRAF mRNA levels was determined by quantitative reverse-transcriptase PCR (G) or cells were replated in osimertinib and growth assessed 96 hours later (H). The IC50 values for growth inhibition were determined by nonlinear regression analysis by using Graphpad Prism software (GraphPad Software). The results represent the mean plus or minus SD of two experiments. GAPDH, glyceraldehyde-3-phosphate dehydrogenase; GFP, green fluorescent protein; gDNA, guide DNA; p, phosphorylated; STAT3, signal transducer and activator of transcription 3; PJA2, praja ring finger ubiquitin ligase 2; -RT, control reaction without reverse transcriptase.

Expression of AGK/BRAF Fusion in EGFR-Mutant Cell Lines Induces Resistance to EGFR TKIs
Western blot analysis with phospho-specific BRAF antisera confirmed the presence of the AGK/BRAF chimeric protein at the predicted molecular mass of 50 kDa. All cells expressing the AGK/BRAF fusion showed enhanced phosphorylation of MEK1/2 and ERK1/2 (Fig. 2D) compared with the parental control cells. Importantly, treatment with osimertinib (0.5 μM) blocked activation of MEK1/2, ERK1/2, and signal transducer and activator of transcription 3 in parental cells but not in the AGK/BRAF-expressing cells, probably on account of EGFR-independent activation of the pathways by the BRAF fusion (see Fig. 2D).

To confirm that expression of BRAF fusions can impede osimertinib efficacy, we assessed the viability of EGFR-mutant cell lines expressing AGK/BRAF fusion in dose-response studies. Osimertinib inhibited growth of parental HCC827, H1975, and PC9 with concentration that inhibits 50% (IC50) values of 0.001, 0.006, and 0.009 μM, respectively. In contrast, clones expressing AGK/BRAF were approximately 1000-fold less sensitive to osimertinib than the parental lines were (Figs. 2E and F). To provide orthogonal validation of these results obtained with the CRISPR-generated AGK/BRAF isogenic lines, we also expressed AGK/BRAF and PJA2/BRAF fusions by using lentiviral plasmids harboring the respective fusion cDNAs in H1975 and PC9 cells (Supplementary Fig. 1A). As shown in Supplementary Figure 1B, expression of the two BRAF fusions in H1975 and PC9 cells resulted in lower sensitivity to osimertinib.

To ensure that the resistance to EGFR TKIs was due to expression of BRAF fusions, we examined viability in two independent clones (H1975-AGK/BRAF-C8 and HCC827-AGK/BRAF-C1) in which BRAF was knocked down with two independent shRNAs targeting the kinase domain of BRAF. Expression of BRAF mRNA was reduced by 98.2% to 100% by the two shRNAs compared with in cells expressing a nontargeting control shRNA (Fig. 2G). HCC827 and H1975 clones expressing the AGK/BRAF fusion were resensitized to osimertinib upon treatment with these shRNAs targeting the portion of BRAF included in the fusion, confirming that the resistance we observed was due to expression of the fusion (a >100 fold decrease in IC50 versus that in the nontargeting control) (Fig. 2H). Similar results were obtained in cells expressing AGK/BRAF and PJA2/BRAF cDNAs (Supplementary Figs. 1C and D). Taken together, these results confirm BRAF fusions as a resistance mechanism to EGFR TKIs.
An EGFR TKI-Resistant Patient-Derived Cell Line with EGFR ex19del and PJA2/BRAF

We generated a patient-derived xenograft (MSK-LX138) and paired cell line (referred to as MSK-LX138cl) from tissue isolated from patient 1 after resistance to EGFR TKI emerged. MSK-IMPACT analysis of this tumor sample showed a PJA2-BRAF fusion, and we confirmed the presence and expression of the fusion at the genomic DNA, mRNA and protein levels (Figs. 3A–C). In agreement with the MSK-IMPACT data, Sanger sequencing of the PCR amplicon indicated a fusion between exon 7 of PJA2 and exon 11 of BRAF (see Fig. 3B), and EGFR ex19del (Supplementary Fig. 2A). Similar to the observations in the isogenic lung cancer cell lines with AGK/BRAF fusion that we generated, MSK-LX138cl cells exhibited high levels of MEK and ERK phosphorylation, which persisted despite osimertinib treatment. Interestingly, this cell line exhibited no AKT phosphorylation (Fig. 3C), suggesting that BRAF fusions can drive growth and survival signals independently of the PI3K-AKT pathway. Growth of MSK-LX138cl cells was insensitive to EGFR TKIs (afatinib, erlotinib, osimertinib, and gefitinib) and several BRAF inhibitors (vemurafenib, dabrafenib, sorafenib, and RXDX-105) (Fig. 3D and E). Similar results with BRAF inhibitors were obtained in the established CRISPR fusion clones (Fig. 4A). As BRAF fusions function as dimers, the lack of activity of vemurafenib and dabrafenib was not unexpected. In agreement with the reduced sensitivity of these cells to EGFR and BRAF inhibitors, there was no apoptosis-related induction in caspase 3/7 activity in response to treatment with several EGFR and RAS-MAPK pathway inhibitors as single agents (osimertinib, erlotinib, trametinib, RXDX-105, sorafenib, and dabrafenib), further supporting the resistant nature of this cell line (Fig. 3F).

Figure 3 Characterization of an EGFR-mutant cell line with praja ring finger ubiquitin ligase 2 gene (PJA2)/BRAF fusion. (A) Expression of PJA2/BRAF in a patient-derived cell line, MSK-LX138cl, and xenograft tissue. (B) Sanger sequencing of the polymerase chain reaction product, confirming a fusion between PJA2 exon 7 and BRAF exon 11. (C) MSK-LX138cl cells were treated with 0.5 μM osimertinib for 1 hour, whole-cell extracts were prepared, and then lysates were subjected to immunoblotting. The results represent 3 independent experiments. (D–E) MSK-LX138cl cells were plated at a density 7500/well in 96-well plates and treated with EGFR (D) or BRAF (E) inhibitors. The concentration that inhibits 50% (IC50) values for growth inhibition were determined by nonlinear regression analysis using Graphpad Prism software (GraphPad Software). (F) A quantity of 50,000 MSK-LX138cl cells/well were plated in 96-well plates, and the next day they were treated with 0.1 or 1 μM of the indicated agents or a combination of osimertinib and trametinib. Caspase 3/7 activity was determined 48 hours later. The results represent the mean plus or minus SD of two experiments. GAPDH, glyceraldehyde-3-phosphate dehydrogenase; cDNA, complementary DNA.

Figure 4 Combined inhibition of MEK and EGFR is effective in reducing growth of cells with EGFR mutation. (A) Cells were plated at a density of 7500 cells/well in 96-well plate and treated with vemurafenib or dabrafenib for 96 hours, after which growth was determined. The concentration that inhibits 50% (IC50) values for growth inhibition are shown (right). (B) MSK-LX138cl cells were plated at a density of 7500 cells/well in 96-well plates and treated with trametinib (0–1 μM) and osimertinib (0–1μM) for 96 hours. Growth inhibition was then determined and the combination index was calculated by using CompuSyn software. Data represent the mean value of growth inhibition ratio at each concentration of the drugs in two independent experiments. Dot plot indicates the combination index and fraction affected (inhibition ratio) of various drug concentrations. AGK, acylglycerol kinase.

Combined Inhibition of MEK and EGFR as Potential Therapy
Given the presence of two actionable oncogenes in our cell lines and patients, we tested the hypothesis that coinhibition of MEK and EGFR could be an efficient therapeutic strategy. Treatment with trametinib alone decreased the growth of all cell lines with EGFR mutation alone or with a BRAF fusion; cell lines with the BRAF fusion were more sensitive, which is consistent with published data showing only limited activity of MEK inhibitors in EGFR-mutated cell lines.20 Indeed, we found that a maximum inhibition of cell growth of only 50% was achieved by trametinib (results not shown). This may indicate that a subpopulation of cells from lines expressing a BRAF fusion can still survive after MEK inhibition. Taken together, these results support the notion that MEK inhibition alone is not sufficient to fully inhibit growth of BRAF fusion–positive cell lines. To determine the combinatorial effect of osimertinib and trametinib we used the method of Chou-Talalay that was developed to identify whether two or more agents act in a synergistic manner.21 We calculated the combination index (CI) of multiple combinations of different drug concentrations, where a CI value less than 1 represents synergy, a CI value of 1 represents an additive effect, and a CI value greater than 1 represents an antagonistic effect, to determine whether the two agents can act synergistically. The experiment was performed in MSK-LX138cl (Fig. 4B), as well as in two of our isogenic EGFR-mutant cell lines with AGK/BRAF fusion (H1975-C1, HCC827-C1). Our results show that trametinib and osimertinib acted in a synergistic manner to inhibit growth up to a concentration of 100 nM.

LY3009120, a Pan-RAF Inhibitor Is Effective as a Single Agent in BRAF Fusion–Positive Cell Lines
BRAF rearrangements are grossly insensitive to BRAF inhibitors (e.g., vemurafenib, dabrafenib) owing to homodimerization and heterodimerization with other RAF proteins.22,23 As we have shown here and as has been previously described,24 BRAF inhibitors are not effective in cell lines with BRAF fusions. However, RAF inhibitors such as BGB659 and LY3009120 are expected to be effective against the fusions because they can bind and inhibit all RAF isoforms.25–29 Treatment of MSK-LX138cl with LY3009120 resulted in complete inhibition of growth with an IC50 of 0.006 μM (Fig. 5A). The melanoma cell line A375, which harbors a BRAF V600E mutation,30 was used as a positive control in these experiments. We next examined the effect of LY3009120 on caspase 3/7 activity. LY3009120 caused a small but significant increase in caspase 3/7 activity, similar to that observed with trametinib plus osimertinib in MSK-LX138cl (Fig. 4F) and HCC827-AGK/BRAF-C1 cells but not in the parental HCC827 cell line (Fig. 5C). Finally, to better understand how LY3009120 inhibits growth, we examined its effect on phosphorylation of BRAF fusion and other downstream signaling proteins. LY3009120 at 0.1 μM caused a profound reduction in phosphorylation of MEK, ERK, and signal transducer and activator of transcription 3 (Fig. 5B).

Figure 5 The pan-RAF inhibitor LY3009120 effectively inhibits growth of EGFR-mutant cell lines with BRAF fusions. (A) Cells were plated in 96-well plates and treated with LY3009120, after which growth was determined (left). The concentration that inhibits 50% (IC50) values for growth inhibition were determined by nonlinear regression analysis by using GraphPad Prism software (GraphPad Software) (right). (B) MSK-LX138cl cells were treated with the indicated concentration of LY3009120 or vemurafenib for 1 hour, and then whole-cell extracts were prepared and immunoblotted for the indicated proteins. The results are representative of two independent experiments. (C) Cells were treated with 0.1 or 1 μM of the indicated compounds or osimertinib/trametinib combinations, and caspase 3/7 activity was determined 48 hours later.

Discussion
The present study establishes BRAF fusions as a mechanism of resistance to EGFR TKI in EGFR-mutant NSCLC and offers two potential therapeutic strategies to circumvent the drug resistance, namely, combined inhibition of MEK and EGFR and inhibition of BRAF fusion. Of seven patients with BRAF fusion present on MSK-IMPACT (n = 374), four had concomitant EGFR driver mutations, suggesting that the co-occurrence of EGFR mutations and BRAF fusion is a rare but recurrent event. Importantly, all four patients were resistant to EGFR TKIs (erlotinib or osimertinib), and in two of four cases the BRAF fusions were confirmed to have been acquired because pre-TKI samples were available and negative for BRAF fusion. AGK, which like BRAF resides at 7q34, encodes a mitochondrial membrane protein involved in lipid and glycerolipid metabolism, whereas PJA2 protein (encoded by a gene at 5q21) has E2-dependent E3 ubiquitin-protein ligase activity. AGK/BRAF fusions have been described in melanomas; gliomas; and cancers of the lung, thyroid, and breast.14 To the best of our knowledge, this is the first report of a PJA2/BRAF fusion. Multiple cellular models, including one patient-derived cell line and isogenic EGFR-mutated lung cancer cell line models, demonstrated that expression of BRAF fusion mediated resistance to EGFR TKIs. In these models, we observed activation of the MEK-ERK pathway independently of EGFR. We further found that loss of BRAF fusion expression restores sensitivity to EGFR TKIs. Taken together, these results indicate that BRAF fusions drive resistance to EGFR TKIs in a subset of EGFR-mutant NSCLC and provides a pathway for developing a viable therapeutic strategy.

It is customary to express a cDNA for a given gene either by lipid-based transfection or by viral transduction to determine whether expression of the gene alters biological properties such as response to small molecules. These methods of gene expression typically result in supraphysiologic levels of the protein because multiple copies of the cDNA are usually introduced and expression is driven from a strong, constitutively active viral promoter. Therefore, there is often a concern that the observed results may be an artifact of overexpression of the protein. To avoid this pitfall, we engineered the AGK/BRAF fusion by using CRISPR-Cas9–mediated genome editing in lung cancer cells harboring drug-sensitive EGFR mutations. The advantage of this method over conventional exogenous cDNA expression is that the BRAF fusion is expressed from the endogenous, nonamplified allele, resulting in a physiologic level of protein expression. Using this approach, we demonstrated that a physiologic level of AGK/BRAF fusion was able to induce resistance to EGFR inhibitors in EGFR-mutant lung cancer cell lines. These results were concordant with data obtained by overexpression of BRAF fusions with use of lentiviral vectors harboring the respective cDNAs. Our approach represents a methodological advance that can be used to model other fusion genes as mechanisms of resistance.

Although BRAF fusion may seem to be an obvious therapeutic target, U.S. Food and Drug Administration (FDA)-approved BRAF inhibitors have not been effective against BRAF fusions24,31–33 and have not always been effective against BRAF mutants. As previously described, BRAF V600E–positive colorectal cancers have been notoriously insensitive to selective BRAF inhibitors (vemurafenib or dabrafenib) owing to feedback activation of EGFR and subsequent RAF dimerization.34 A similar resistance mechanism has been noted in BRAF V600E melanomas, where RAF inhibition abrogates ERK activation, which then terminates the high ERK negative feedback on RTK-dependent activation of RAS, leading to consequent RAF dimerization and ERK reactivation.

This ineffectiveness of FDA-approved BRAF inhibitors against BRAF fusions is highlighted by the fact that an AGAP3/BRAF fusion has been proposed as a mechanism of acquired resistance to vemurafenib in BRAF V600E melanoma.24 When compounds such as vemurafenib and dabrafenib bind to the first site within the BRAF dimer, affinity for the second dimer site is significantly lower, explaining why traditional BRAF inhibitors do not abrogate ERK phosphorylation in BRAF fusion-positive tumors.36 BRAF V600E mutants can become resistant to BRAF inhibitors through formation of BRAF dimers by the mechanism already described,23,37 which prompted development of pan-RAF inhibitors.26,38 Yao et al. have reported superiority of a dual RAF inhibitor BGB659 over BRAF inhibitors.

Similarly, Peng et al. reported the development of LY3009120, a pan-RAF inhibitor that binds simultaneously to two sites of BRAF dimers and equally inhibits all RAF isoforms (ARAF, BRAF, and CRAF) with similar affinity.25–29 We found LY3009120 to be very active against cell lines with BRAF fusions as a single agent. Taken together, this indicates that pan-RAF inhibition is a superior single-agent strategy for targeting these acquired BRAF fusions. Notably, we (N. X. R.) have recently shown that another promising RAF inhibitor, PLX8394, selectively disrupts BRAF dimers, sparing cRAF homodimer-dependent RAF function in normal cells, which may lead to a better safety profile clinically.39 Clinical studies will be needed to determine whether RAF inhibitors are effective against tumors with BRAF fusions as a single agent or whether they should be combined with an EGFR TKI to maximize the effectiveness.

To identify potential therapeutic strategies for patients with coexpression of EGFR mutation and BRAF fusion that can be utilized immediately, we investigated the effectiveness of a combination of trametinib and osimertinib. We found that combined inhibition of MEK and EGFR with these drugs inhibited growth of isogenic cell line models and growth of a patient-derived cell line with a BRAF fusion and EGFR mutation (established from a patient who did not respond to EGFR drugs) in a synergistic manner. The maximum inhibition of growth that we observed was 76%, which may have been due to reactivation of erb-b2 receptor tyrosine kinase 2 and ERK after the prolonged treatment of cells with osimertinib and trametinib, respectively (data not shown). Nevertheless, MLN2480 these results suggest that combined MEK and EGFR inhibition is a possible treatment strategy in the absence of more effective FDA-approved RAF inhibitors that are effective against BRAF fusions.

Acknowledgments
This work was supported by National Institutes of Health grants P01 CA129243, P30 CA008748, and U54 OD020355. The authors are grateful to Igor Odintsov and Dr. Zebing Liu for critical reading of this article and helpful suggestions during the course of this study, respectively.