Histone Deacetylase 3 Inhibition Overcomes BIM Deletion Polymorphism–Mediated Osimertinib Resistance in EGFR-Mutant Lung Cancer Free

This article is featured in Highlights of This Issue, p. 2919

The majority of patients with non–small cell lung cancer (NSCLC) with EGFR-activating mutations, such as exon 19 deletion and L858R point mutation, show marked responses to the first-generation reversible EGFR tyrosine kinase inhibitors (EGFR-TKI), gefitinib and erlotinib (1, 2). However, the acquisition of TKI resistance is almost inevitable, and is commonly associated with the so-called EGFR-T790M gatekeeper mutation, which substitutes a threonine with a methionine at the amino acid position 790 of exon 20. Accordingly, the T790M mutation is detected in 50%–60% of patients who develop clinical resistance to the first-generation EGFR-TKIs, gefitinib or erlotinib (3, 4).

Osimertinib, a mono-anilino-pyrimidine compound, is a third-generation irreversible EGFR-TKI, which has activity against EGFR with sensitizing mutations, such as the exon 19 deletion, L858R mutation, and T790M resistance mutation, but spares wild-type EGFR (5). Although the second-generation irreversible EGFR-TKI, afatinib, can inhibit the T790M mutation in vitro, it also has high activity against wild-type EGFR and has failed to demonstrate an objective response rate (ORR; less than 10%) in NSCLC patients with the EGFR-T790M mutation (6). In contrast, osimertinib exhibited a prominent anticancer effect (confirmed ORR, 61%) among an equivalent cohort of patients (7), and was thus approved for the treatment of patients harboring the T790M mutation in Europe, the United States, and Japan.

BIM, also called Bcl-2-like protein 11, is a proapoptotic molecule that belongs to the Bcl-2 family. BIM upregulation is essential for the induction of apoptosis in lung cancer cells with EGFR mutations treated with first-generation EGFR-TKIs, and low BIM protein level is associated with resistance to EGFR-TKIs (8, 9). In East Asians, but not Caucasians or Africans, a 2,903-bp deletion polymorphism in the BIM gene was found to be present at incidences of around 13% and 0.5% for heterozygous and homozygous carriers, respectively (10). Another study has recently reported that 15.7% of hispanic patients with NSCLC carried the deletion allele (11). Importantly, the BIM deletion polymorphism results in the preferential splicing of exon 3 over the BH3-encoding exon 4 in the BIM pre-mRNA, and leads to the production of inactive BIM isoforms lacking the BH3 domain. This in turn reduces expression of proapoptotic BIM protein isoforms in EGFR-mutated lung cancer cell lines following TKI exposure, and is sufficient to confer TKI resistance (10). The polymorphic fragment includes multiple and redundant splicing silencers that repress exon 3 inclusion (12). Since its initial discovery, several meta-analyses have reported the association between BIM deletion polymorphism and shorter progression-free survival (PFS) of patients with NSCLC harboring EGFR mutations, who received gefitinib or erlotinib treatment (13). However, it is unknown if the BIM deletion polymorphism affects the antitumor efficacy of third-generation EGFR-TKIs including osimertinib.

We previously reported that the combined use of gefitinib and the histone deacetylase (HDAC) inhibitor, vorinostat, was able to preferentially upregulate the expression of proapoptotic BIM isoforms in EGFR-mutated NSCLC cell lines heterozygous for the BIM deletion, and overcome EGFR-TKI resistance in vitro and in vivo (14). However, it remained unclear how vorinostat corrected the splicing defect conferred by the BIM deletion, and whether vorinostat could overcome resistance in the setting of cells with homozygous BIM deletions.

In this study, we examined the ability of osimertinib, in comparison with afatinib and gefitinib, to induce apoptosis in EGFR-mutated lung cancer cell lines with either heterozygous or homozygous configurations of the BIM deletion polymorphism. We also determined the effect of vorinostat on BIM deletion polymorphism-mediated resistance to osimertinib both in vitro and in vivo. We further identified the target HDAC molecule whose inhibition can overcome EGFR-TKI resistance in EGFR-mutated lung cancer cells bearing the BIM deletion polymorphism.

NSCLC cell lines, PC-9 and PC-3, which have an exon 19 deletion in the EGFR, were obtained from Immuno-Biological Laboratories Co., Ltd., in May 2015, and Human Science Research Resource Bank (Osaka, Japan) in March 2013, respectively (14). PC-9 cells with a homozygous BIM deletion polymorphism (PC-9BIM^i2−/−) were established by editing with zinc finger nuclease, as reported previously (10). All three cell lines were subcultured in RPMI1640 medium supplemented with 10% FBS and antibiotics within 3 months of thawing the frozen stock. Mycoplasma infection in the cells was regularly checked using a MycoAlert Mycoplasma Detection Kit (Lonza). The cell line authentication was performed by short tandem repeat analysis at the laboratory of the National Institute of Biomedical Innovation (Osaka, Japan) in May 2015. Gefitinib, afatinib, osimertinib, vorinostat, belinostat, droxinostat, and RGFP966 were obtained from Selleck Chemicals. All drugs were dissolved in DMSO and preserved at 30°C.

Cellular DNAs were extracted from the cells using a DNeasy Blood and Tissue Kit (Qiagen). To recognize the presence of the wild-type and deletion alleles, we conducted PCR reactions using the discriminating primers for the wild-type alleles (forward: 5′-CCACCAATGGAAAAGGTTCA-3′; reverse: 5′-CTGTCATTTCTCCCCACCAC-3′) and deletion alleles (forward: 5′-CCACCAATGGAAAAGGTTCA-3′; reverse: 5′-GGCACAGCCTCTATGGAGAA-3′). The genomic DNAs were amplified using a Veriti Thermal Cycler (Applied Biosystems) with GoTaq Hot Start Polymerase (Promega). The PCR amplicons for the wild-type (362 bp) and the deletion (284 bp) alleles were separated by agarose gel electrophoresis.

Cellular apoptosis induced by the drugs was determined through the use of a FACSCalibur flow cytometer (BD Biosciences) with a PE Annexin V Apoptosis Detection Kit I (BD Biosciences), which detects and quantifies apoptotic cells with phycoerythrin (PE) Annexin V and 7-amino-actinomycin (7-AAD) staining.

The proteins harvested were separated via SDS-PAGE. The proteins were transferred onto polyvinylidene fluoride membranes (Bio-Rad), which were immersed in StartingBlock T20 (TBS) Blocking Buffer (Thermo Fisher Scientific) for 1 hour at about 20°C, followed by incubation above 8 hours at 4°C with antibodies against phospho-EGFR (Tyr1068), Akt, phospho-Akt (Ser473), cleaved PARP, cleaved caspase-3, histone H3, acetylated histone H3 (Lys27), BIM, and β-actin (Cell Signaling Technology); and against phospho-Erk1/2 (Thr202/Tyr204), Erk1/2, and EGFR (R&D Systems). After washing three times in the Tris-buffered saline with the polyoxyethylene sorbitan monolaurate (TBST), the membranes were incubated for 1 hour at room temperature with horseradish peroxidase–conjugated secondary antibodies. The proteins labeled with secondary antibodies were visualized using SuperSignal West Dura Extended Duration Substrate Enhanced Chemiluminescent Substrate (Thermo Fisher Scientific). Each experiment was independently carried out at least three times.

The cells (1 × 10⁵) cultured in medium containing 10% FBS (antibiotic free) for 24 hours were treated with Stealth RNAi siRNA against BIM and HDAC 1, 2, 3, 6, and Stealth RNAi siRNA Negative Control Lo GC (Invitrogen) using Lipofectamine RNAiMAX (Invitrogen) for 48 hours.

Total cellular RNAs were extracted from the cells using RNeasy PLUS Mini kit (Qiagen). Reverse transcription of the collected RNAs was performed using SuperScript VILO cDNA synthesis Kit and Master Mix (Invitrogen). Expression of BIM mRNA was quantitatively measured by ViiA 7 Real-Time PCR System (Applied Biosystems) using the following primers: BIM exon 2A (forward: 5′-ATGGCAAAGCAACCTTCTGATG-3′; reverse: 5′-GGCTCTGTCTGTAGGGAGGT-3′), BIM exon 3 (forward: 5′-CAATGGTAGTCATCCTAGAGG-3′; reverse: 5′-GACAAAATGCTCAAGGAAGAGG-3′), BIM exon 4 (forward: 5′-TTCCATGAGGCAGGCTGAAC-3′; reverse: 5′-CCTCCTTGCATAGTAAGCGTT-3′), and β-actin (forward: 5′-GGACTTCGAGCAAGAGATGG-3′; reverse: 5′-AGCACTGTGTTGGCGTACAG-3′).

Male 5- to 6-week-old BALB/c-nu/nu mice were injected subcutaneously into both flanks with cultured tumor cells (5 × 10⁶ cells/0.1 mL/mouse). After tumor volumes reached 80 to 100 mm³, the mice were randomized and treated once daily by oral gavage with osimertinib and/or vorinostat. Each tumor was monitored using an electronic caliper. Tumor volume was measured in two dimensions, and calculated using the following formula: tumor volume (mm³) = 1/2 × length (mm) × width (mm)². After the mice were treated for 4 days, two tumors in each control and treatment group were excised, lysed, and subjected to Western blot analysis. All animal experiments complied with the Guidelines for the Institute for Experimental Animals, Kanazawa University Advanced Science Research Center (approval no. AP-081088).

Differences between groups were analyzed by one-way ANOVA. All statistical analyses were conducted using Graph-Pad Prism Ver. 6.05 (GraphPad Software Inc.). The threshold for significance was P < 0.05.

We first carried out PCR to confirm the presence of the BIM deletion polymorphism in EGFR-mutated NSCLC cell lines. PC-9 cells did not harbor the BIM deletion, but as expected, PC-3 and PC-9BIM^i2−/− cells were heterozygous and homozygous for the deletion, respectively (Fig. 1A; refs. 10, 14). Next, we assessed the degree of apoptosis induction by second- and third-generation EGFR-TKIs in these cell lines. Using a FACS-based assay, we found that both afatinib and osimertinib induced significant apoptosis in PC-9 cells, but not in PC-3 or PC-9BIM^i2−/− cells (Fig. 1B). Western blot analyses for cleaved PARP and caspase-3 confirmed low apoptosis induction in PC-3 and PC-9BIM^i2−/− cells, in contrast to PC-9 cells, when treated with gefitinib, afatinib, and osimertinib. All three EGFR-TKIs inhibited EGFR phosphorylation, as well as its downstream kinases AKT and ERK, in PC-9, PC-3, and PC-9BIM^i2−/− cells. Although the EGFR-TKIs induced active BIM (BIM_EL) protein in PC-9 cells, the level of BIM_EL protein in EGFR-TKI–treated PC-3 and PC-9BIM^i2−/− cells was much lower (Fig. 1C). Moreover, knockdown of BIM protein by BIM-specific siRNA resulted in the abrogation of apoptosis, as shown by the absence of cleaved PARP and cleaved caspase-3 in EGFR-TKI–treated PC-9 cells (Fig. 1D). These results clearly indicate that the presence of the BIM deletion polymorphism, in one or both alleles, is sufficient to mediate resistance to all three generations of EGFR-TKIs in EGFR-mutated NSCLC cells.

We previously reported that the combined use of an HDAC inhibitor, vorinostat, with gefitinib induces apoptosis in EGFR-mutated NSCLC cells harboring a single allele of the BIM deletion polymorphism. We therefore investigated the effect of vorinostat combined with second- or third-generation EGFR-TKIs on apoptosis induction in PC-3 and PC-9BIM^i2−/− cells. Consistent with the results shown in Fig. 1C, afatinib or osimertinib alone induced very low levels of BIM_EL, cleaved PARP, and cleaved caspase-3 in PC-3 and PC-9BIM^i2−/− cells, while at the same time, completely inhibited EGFR phosphorylation and its downstream targets, AKT and ERK (Fig. 2A). Meanwhile, combined use of vorinostat with either afatinib or osimertinib markedly upregulated BIM_EL expression and thus induced cleaved PARP and cleaved caspase-3 (Fig. 2A). Moreover, knockdown of the BIM protein by BIM-specific siRNA resulted in the inhibition of apoptosis induced by vorinostat, combined with afatinib or osimertinib, in PC-9BIM^i2−/− cells (Fig. 2B). These results indicate that vorinostat overcame apoptosis resistance to second- and third-generation EGFR-TKIs by inducing active BIM protein (BIM_EL) expression in EGFR-mutated NSCLC cells which are either heterozygous or homozygous for the BIM deletion polymorphism.

Vorinostat is a nonspecific HDAC inhibitor, and targets both class I HDACs (HDAC 1, 2, 3, and 8) and class II HDACs (HDAC 6 and 10; ref. 15). Considering the various side effects of vorinostat (16), a selective HDAC inhibitor may bring about less toxicity in combination with osimertinib. To elucidate which HDAC plays a role in the induction of active BIM protein expression and cell death, we treated PC-9BIM^i2−/− cells with specific siRNA for HDAC 1, 2, 3, or 6. Interestingly, HDAC3 knockdown induced upregulation of BIM exon 4-containing transcripts that encode proapoptotic BH3-containing (active) BIM isoforms (Fig. 3A and B; Supplementary Fig. S1). Moreover, knockdown of HDAC3 restored active BIM protein expression and induced apoptosis by osimertinib in PC-3 and PC-9BIM^i2−/− cells (Fig. 3C). These results strongly suggest that HDAC3 inhibition is important for the transcription of BIM exon 4–containing isoforms, and is sufficient for the induction of active BIM protein isoforms, including BIM_EL. This was further supported by the results of other HDAC inhibitors with different HDAC-inhibitory profiles. Droxinostat, an HDAC inhibitor under development, has much weaker inhibitory activity against HDAC3 compared with vorinostat (IC₅₀ values of HDAC3 inhibition by droxinostat and vorinostat are 16.9 ± 5.0 μmol/L and 20 nmol/L, respectively; refs. 17, 18). At equivalent concentrations, droxinostat, in contrast to vorinostat, failed to significantly upregulate active BIM protein, and was unable to induce apoptosis in PC-9BIM^i2−/− cells even in combination with osimertinib (Fig. 3D). We also tested belinostat, a pan-HDAC inhibitor (IC₅₀ value of HDAC3 inhibition is 27 nmol/L), whose inhibitory profile is similar to that of vorinostat (19), and which is approved for the treatment of relapsed or refractory peripheral T-cell lymphoma (PTCL) by the FDA. As is the case with vorinostat, combined use of belinostat and EGFR-TKIs (gefitinib, afatinib, and osimertinib) enhanced the expression of active BIM, and induced apoptosis in PC-3 and PC-9BIM^i2−/− cells (Supplementary Fig. S2A). Furthermore, a selective HDAC3 inhibitor, RGFP966 (IC₅₀ value of HDAC3 inhibition is 80 nmol/L; ref. 20), clearly induced apoptosis in combination with osimertinib together with upregulating active BIM in PC-9BIM^i2−/− cells (Supplementary Fig. S2B). These findings implicate HDAC3 inhibition as an important target in the ability of vorinostat to induce apopotosis in EGFR-mutated NSCLC cells with the BIM deletion polymorphism.

In our previous study, which used heterozygous BIM deletion–positive NSCLC (PC-3) cells with EGFR mutations, vorinostat preferentially induced the expression of the exon 4–containing isoform (encoding BH3-domain containing BIM), although its exact mechanisms of action remain unclear. There are two possibilities to explain this observation: (i) vorinostat upregulated the transcription of the exon 4–containing isoform from either or both BIM alleles, or (ii) vorinostat affected BIM splicing, resulting in the production of exon 4 rather than exon 3–containing transcripts from the deletion polymorphism allele. To elucidate the mechanism, we evaluated the ratio of exon 3 to exon 4 transcripts in PC-9BIM^i2−/− cells (homozygous BIM deletion), compared with those in PC-3 (heterozygous BIM deletion) and PC-9 (with only full-length BIM alleles) cells. As reported previously, the ratio of exon 3 to exon 4 transcripts in PC-3 cells was higher than that in PC-9 cells. As expected, the ratio of exon 3 to exon 4 transcripts in PC-9BIM^i2−/− cells was also higher than in PC-3 cells (Supplementary Fig. S3A). In PC-9BIM^i2−/−cells, vorinostat upregulated the transcription of the exon 4–containing isoform (Fig. 4A), which was further enhanced in combination with the EGFR-TKI gefitinib or osimertinib (Fig. 4A; Supplementary Fig. S3B). Vorinostat markedly decreased the ratio of exon 3 to exon 4 transcripts compared with the control (Fig. 4B). These results indicate that vorinostat affected the splicing process in the BIM deletion allele, rather than the full-length allele.

We next sought to elucidate whether HDAC3 inhibition prominently affects the alternative splicing of BIM pre-mRNA. Knockdown of HDAC3 elevated the mRNA expression of exon 2A and exon 4, and thus strongly decreased the ratio of exon 3 to exon 4 transcripts (Fig. 4C). Although droxinostat (which weakly inhibits HDAC3) did not significantly decrease the ratio of exon 3 to exon 4 transcripts, RGFP966 with selective inhibitory activity to HDAC3 did decrease it (Supplementary Fig. S4A and S4B). These findings suggest that HDAC3 inhibition contributes to the alternative splicing of BIM by promoting exon 4 inclusion.

We next examined the effect of combination therapy with osimertinib and vorinostat on EGFR-mutated NSCLC cells homozygous for the BIM deletion polymorphism in vivo. PC-9BIM^i2−/−tumor-bearing mice were treated with osimertinib, vorinostat, or a combination of both. In contrast to PC-9 xenograft tumors, which decreased in volume following osimertinib treatment, PC-9BIM^i2−/− xenograft tumors continued to grow, albeit more slowly (Fig. 5A; Supplementary Fig. S5). However, the combination of osimertinib with vorinostat led to a significant reduction in the size of PC-9BIM^i2−/− xenograft tumors, without causing weight loss in treated mice (data not shown). In PC-9BIM^i2−/− tumors, Western blot analysis revealed that combined treatment with osimertinib and vorinostat markedly induced active BIM protein expression, and induced apoptosis as represented by increased cleaved caspase-3 (Fig. 5B).

Studies have demonstrated that most patients with EGFR-mutated NSCLC and the BIM deletion polymorphism showed shorter PFS compared with those without the BIM deletion polymorphism (13). Because the BIM deletion polymorphism induces apoptosis resistance, but not growth impairment, to EGFR-TKI exposure, tumors with the BIM deletion polymorphism may be permissive for the acquisition of additional resistance mechanisms, as recently demonstrated in chronic myeloid leukemia (CML; ref. 21). To assess this question, we established gefitinib-resistant cells from PC-9BIM^i2−/− cells by continuous exposure to increasing concentrations of gefitinib and cloning by limiting dilution in vitro, which was designated as PC-9BIM^i2−/−GR. Interestingly, PC-9BIM^i2−/−GR cells acquired the EGFR-T790M mutation over time. Although PC-9BIM^i2−/−GR cells were highly resistant to gefitinib in terms of cell viability, they had the same sensitivity to osimertinib compared with PC-9BIM^i2−/− cells (Fig. 6A). Notably, while neither gefitinib, osimertinib, nor vorinostat alone markedly induced apoptosis in PC-9BIM^i2−/− and PC-9BIM^i2−/−GR cells, combined use of osimertinib and vorinostat markedly induced cell apoptosis in both PC-9BIM^i2−/− and PC-9BIM^i2−/−GR cells (Fig. 6B). Western blotting demonstrated that although gefitinib did not reduce the expression of phosphorylated EGFR in PC-9BIM^i2−/−GR cells, osimertinib suppressed the phosphorylation of EGFR and its downstream kinases, AKT and ERK (Fig. 6C). Moreover, the combination of osimertinib and vorinostat upregulated both PARP and caspase-3 cleavage in PC-9BIM^i2−/−GR cells, indicating that apoptosis was induced in these tumor cells.

The BIM deletion polymorphism is found in a significant proportion (∼20%) of normal individuals of East Asian ethnicity (10), with the majority carrying one allele of the deletion polymorphism and only a minority who are homozygous (∼0.5%; ref. 10). Accordingly, native EGFR-mutated NSCLC cell lines that are homozygous for the BIM deletion polymorphism have yet to be described. In this study, we utilized the genetically edited PC-9BIM^i2−/− cells as a homozygous BIM deletion polymorphism-positive EGFR-mutant NSCLC cell line with which to explore the mechanism of HDAC activity (10). In PC-9BIM^i2−/− cells, vorinostat decreased the ratio of exon 3 to exon 4 transcripts, increased active BIM protein expression, and resensitized cells to EGFR-TKI–induced apoptosis. These findings clearly indicate that vorinostat predominantly affects the deletion allele of BIM to overcome EGFR-TKI resistance, and also provides a rationale of combined treatment with vorinostat and EGFR-TKIs for patients with EGFR-mutated NSCLC who are homozygous for the BIM deletion polymorphism. In addition to the BIM deletion polymorphism in intron 2, several single-nucleotide polymorphisms (SNP) within the BIM locus were recently discovered (22, 23). A silent SNP (the T allele in the c465C>T) in exon 4 of BIM is reported to exist in approximately 30% of French individuals, and is associated with a delay in major molecular responses to imatinib in CML (22). Moreover, the BIM C29201T variant, located within the BH3-domain coding region, is reported to be associated with lower overall survival in children with acute lymphoblastic leukemia (23). Thus, further investigations are warranted to examine the effect of HDAC inhibitors on the target drug sensitivity of tumors with BIM SNPs.

HDACs can affect alternative mRNA splicing (24–26). For instance, the Hu proteins are thought to regulate pre-mRNA splicing through HDAC2 inhibition, which in turn modulates chromatin structures to alter splicing of NF1 in HeLa cells (27). The inhibition of HDAC1 but not HDAC2 is also important for the alternative splicing of fibronectin in HeLa cells (28). A recent study revealed that HDAC1 and HDAC2 copurified with the U2 small nuclear ribonucleoprotein splicing factor, and that knockdown of these two deacetylases but not that of HDAC3 modified the splicing patterns of CD44 (29). These reports suggest that the different HDACs differentially regulate alternative splicing of specific pre-mRNAs. We here report that HDAC3 inhibition plays pivotal role on the alternative splicing of BIM caused by vorinostat. This is consistent with the results of Hnilicova and colleagues (28), who showed that HDAC inhibition by NaB did not change the splicing pattern in BIM, and which we reason is because NaB has much weaker HDAC3-inhibitory activity than vorinostat (30). Although HDAC3 is reported to have multiple functions in stem cell differentiation, embryonic cardiovascular development, and endothelial cell differentiation and integrity maintenance (31), the role of HDAC3 in alternative splicing is virtually unknown, as opposed to HDAC1 or HDAC2. Previous reports have documented that HDACs play a role in the expression of miRNAs (32–34), and that miRNAs regulate alternative splicing (35–37), suggesting that the influence of HDAC3 in splicing might be indirect. Specifically, Chen and colleagues reported that knockdown of HDAC3 upregulated the expression of miR15a and miR16-1, which are important suppressors that modulate BCL-2 and other molecules (38). Although we have established HDAC3 inhibition as important in the alternative splicing of BIM in deletion-positive EGFR-mutated NSCLC, the precise mechanism by which it does so needs to be further elucidated.

We observed that tumors of BIM deletion polymorphism–positive EGFR-mutated NSCLC cells were stable in size during EGFR-TKI treatment, indicating that the acquisition of additional resistance mechanisms is necessary for tumor enlargement during EGFR-TKI treatment. EGFR-T790M is the most frequent resistance mechanism detected in patients with EGFR-mutated NSCLC who are refractory to reversible EGFR-TKI treatment (4). In PC-9BIM^i2−/−GR cells, we detected the EGFR-T790M mutation which allows tumor cell growth even in the presence of gefitinib in vitro. Therefore, we speculate that T790M may be detected in a certain population of patients with BIM deletion–positive NSCLC harboring EGFR mutations who fail to respond to reversible EGFR-TKI therapy. EGFR-T790M–positive resistant tumors can occur either by selection of preexisting EGFR-T790M–positive clones or via genetic evolution of initially EGFR-T790M–negative drug-tolerant cells (39, 40). In the latter case, drug-tolerant cells become a base for de novo evolution of the EGFR-T790M mutation (40). Because the cells with BIM deletion polymorphism are resistant to EGFR-TKI–induced apoptosis, these cells may become a base for de novo evolution of the EGFR-T790M mutation. Therefore, therapeutic strategies with vorinostat and EGFR-TKI to eradicate BIM deletion polymorphism–positive apoptosis-resistant cells may be useful for preventing the acquisition of the EGFR-T790M mutation in NSCLC. We are now conducting a phase I trial (NCT02296125) to assess the feasibility of combined treatment with vorinostat and gefitinib in BIM deletion polymorphism–positive EGFR-mutated NSCLC.

Third-generation EGFR-TKIs, including osimertinib, show high activity against T790M-positive EGFR-mutated NSCLC, but most patients subsequently develop resistance to this class of inhibitors (7, 41). Recent studies demonstrated that the C797S mutation in EGFR exon 20 is acquired in approximately 20% of osimertinib-resistant cases (42). In nearly half of cases who acquired resistance to third-generation EGFR-TKIs, tumors lose the T790M mutation and acquired other resistance mechanisms (42), including bypass track activation by MET amplification (43, 44) or HER2 amplification (43), and small-cell transformation (45). We here report that BIM deletion polymorphism is one of the mechanisms that cause apoptosis resistance to osimertinib in EGFR-mutated NSCLC cells with or without T790M. Because the BIM deletion polymorphism is associated with shorter PFS in patients with EGFR-mutated NSCLC who were treated with first-generation EGFR-TKIs, it may also be associated with worse outcomes for osimertinib-treated EGFR-T790M–positive patients. However, Lee and colleagues reported that the BIM polymorphism was not a predictive biomarker of EGFR-TKI resistance (46). Prospective studies with a larger number of cases will be necessary in the future. An ongoing phase III trial is currently comparing osimertinib and gefitinib as the first-line treatment for EGFR-mutated NSCLC (NCT02296125). The results of this trial may indicate that osimertinib might be used as a first-line treatment for EGFR-mutated NSCLC in the future. If it is the case, elucidation of the resistance mechanisms to first-line osimertinib treatment in T790M-negative EGFR-mutated NSCLC would become more clinically important.

In summary, this study demonstrated that the combination of vorinostat and osimertinib can be used to overcome osimertinib resistance in EGFR-mutated NSCLC, which are either heterozygous or homozygous for the BIM deletion polymorphism, both in vitro and in vivo. Notably, HDAC3 inhibition by vorinostat plays a crucial role in apoptosis induction via promoting transcription and modulating alternative splicing to upregulate active BIM protein in BIM deletion polymorphism–positive EGFR-mutated NSCLC cells. Furthermore, acquisition of the EGFR-T790M mutation allows BIM deletion polymorphism–positive EGFR-mutated NSCLC cells to grow in the presence of gefitinib, and combined use of vorinostat with osimertinib could induce apoptosis even when BIM deletion polymorphism–positive EGFR-mutated NSCLC cells acquire the T790M mutation. These findings illustrate the importance of developing HDAC3-selective inhibitors, and provide a rationale for their combined use with osimertinib to treat lung cancer with EGFR mutations and the BIM deletion polymorphism.

O. Tiong reports receiving speakers bureau honoraria from Chipscreen Biosciences. S. Yano reports receiving speakers bureau honoraria from Boehringer-Ingelheim and Chugai and other commercial research support from AstraZenica, Boehringer-Ingelheim, and Chugai. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A. Tanimoto, S. Takeuchi, S. Yano

Development of methodology: A. Tanimoto, S. Takeuchi, S. Yano

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Tanimoto, S. Takeuchi, S. Arai, S. Yano

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K. Fukuda, S. Yano

Writing, review, and/or revision of the manuscript: A. Tanimoto, S. Takeuchi, X. Roca, S.T. Ong, S. Yano

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Tanimoto, S. Takeuchi, K. Fukuda, S. Yano

Study supervision: A. Tanimoto, S. Takeuchi, T. Yamada

This work was supported by JSPS KAKENHI grant number JP16H05308 (to S. Yano), the Project for Cancer Research and Therapeutic Evolution (P-CREATE) grant number 16cm0106513h0001 (to S. Yano), and grants from the Japan Agency for Medical Research and Development, AMED, grant numbers 15Aak0101016h0003 and 15Ack0106113h0002 (to S. Yano). X. Roca and S.T. Ong were supported by the Singapore Ministry of Health's National Medical Research Council under its Clinician Scientists Individual Research Grant (NMRC/CIRG/1330/2012), and S.T. Ong by the Clinician Scientist Award (NMRC/CSA/0051/2013), administered by the Singapore Ministry of Health's National Medical Research Council.

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.