Antitumor Impact of p14^ARF on Gefitinib-Resistant Non–Small Cell Lung Cancers Free

This article is featured in Highlights of This Issue, p. 1379

Molecular-targeting agents may exert a gene-specific effect on key regulators of tumor growth in various intractable malignancies (1). Epidermal growth factor receptor (EGFR)-targeting agents such as gefitinib are a first-line therapy for EGFR mutation-positive cells of non–small cell lung cancer (NSCLC; refs. 2–4). The sensitivity to gefitinib is enhanced by the point mutation of exon 21 (L858R) or the in-frame deletion of exon 19 in the EGFR tyrosine kinase domain (3, 5) and gefitinib inhibits signal transduction of the phosphoinositide 3-kinase-Akt pathway and the mitogen-activated protein kinase pathway, thereby inhibiting the growth of tumors (6–8). However, the recurrence of these tumors with drug resistance in patients treated with such agents can limit the overall utility of these agents (9, 10). The main cause of resistance to EGFR inhibitors in lung cancers is genetic alterations in the threonine residue at position 790 in exon 20 (T790M) of EGFR, a gatekeeper mutation (11, 12). This has led to an effort to identify such genetic alterations and attempt to develop more powerful agents to overcome resistance (13).

However, identification of the effector molecules that determine cellular fate within the EGFR signaling pathway is another important goal and may help facilitate the development of novel strategies to overcome drug resistance. Previous studies reported that gefitinib induces cell-cycle arrest and triggers apoptosis in some cancer cells (14, 15), although the effector molecules that crucially mediate the antiproliferation and apoptotic effect through gefitinib treatment remain unclear. Interestingly, it has been suggested that EGFR mutation correlates with low CDKN2A expression (16). Therefore, identification of gefitinib-driven cell-cycle regulators may contribute to a better understanding of NSCLCs harboring resistant EGFR mutations.

The CDKN2A (INK4a/ARF) locus is encoded by alternative splicing variants including p16^INK4a, p14^ARF, and p12 (17, 18). Recently, exon array analysis by PCR revealed the existence of p16 variant 2 in lung cancer cell lines (19). Among these CDKN2A variants, p16^INK4a and p14^ARF, which share exons 2 and 3 and use a unique exon 1, are well studied. They have distinct functions through independent signaling pathways and the amino acids sequences between p16^INK4a and p14^ARF are produced from different promoter regions (18). Moreover, methylation of their promoter is also reported in many cancer cells (20). The p16^INK4a inhibits the phosphorylation of Rb through binding to CDK4/6 (21). The role of p14^ARF antagonizes the actions of MDM2 in the MDM2-p53 pathway, resulting in the stabilization of p53 (22, 23). Thus, p16^INK4a and p14^ARF function as modulators of cell cycle and apoptosis leading to tumor suppression.

Here we report the distinct molecular response to gefitinib of drug-resistant human lung adenocarcinoma cells in comparison with drug-sensitive cells to highlight the biological significance of p14^ARF in the growth regulation of these tumors.

Non–small cell lung adenocarcinoma cell lines PC-9 and RPC-9 were used as previously described (24). PC-9 cells have a deletion in EGFR exon 19 (del E746–A750) as an activating EGFR mutation and are sensitive to gefitinib. Gefitinib-resistant RPC-9 cells were derived from parental PC-9 cells and have both an exon 19 deletion and an exon 20 mutation (T790M). EGFR mutations were confirmed by a sequence analysis of the genomic DNAs. HCC827 and H1975 (2 NSCLC cell lines), NuLi-1 (an immortalized normal human bronchus cell line), and NHDF (normal human dermal fibroblast) cell lines were obtained from American Type Culture Collection. These cell lines were passaged for less than 6 months and then replaced with those of early passages. HCC827 (exon 19 deletion of EGFR) is sensitive to gefitinib and H1975 (EGFR T790M and L858R mutations) is resistant to gefitinib (Supplementary Fig. S1A; ref. 25). MMNK-1, an immortalized line from normal cholangiocytes, and TMNK-1 from normal endothelial cells were provided by Dr. N. Kobayashi (Okayama Saidaiji Hospital; refs. 26 and 27). Cell lines used in this study were not authenticated. Cells were maintained in RPMI 1640 medium (Invitrogen Life Technologies Corp.) containing 10% FBS with 100 U/mL penicillin (Invitrogen Life Technologies Corp.) and 100 mg/mL streptomycin (Invitrogen Life Technologies Corp.) at 37°C with 5% CO₂.

RNA was extracted using an RNeasy Plus mini kit (Invitrogen Life Technologies Corp.) according to the manufacturer's protocol. The RNA was quantified by absorbance at 260 nm and 500 ng of total RNA was reverse transcribed by 6-mer oligos and the Super Script III kit in 20 μL of reaction solution. Reverse transcriptase (RT)-PCR products were diluted in 1:20 and then subjected to quantitative real-time PCR (qPCR) using a TaqMan gene expression assays and arrays kit (Invitrogen Life Technologies Corp.) in 20 μL of reaction solution. Data analysis was followed by ΔΔCt methods using the step one plus instruction (Applied Biosystems, Ltd.).

A total of 15 μL of Cell Count Reagent SF (Nacalai Tesque, Inc.) was mixed with 500 μL of RPMI 1640 medium. The mixture was added to cells, incubated for 30 minutes, and the optical density of the formazan products was measured at an absorbance at 490 nm with a microplate reader (Tecan Group, Ltd.).

Cells were seeded at a density of 5 × 10⁴ cells/0.5 mL in 24-well dishes. The next day, the medium was replaced with fresh RPMI 1640 medium with or without gefitinib (1 μmol/L). After 20-hour treatment with gefitinib, cells were lysed with the ProteoExtract Subcellular Proteome Extraction Kit (Calbiochem Corp.) and the subcellular fraction in each sample was concentrated with the Amicon Ultra 3K device (Millipore Corp.) according to the manufacturer's instructions. The protein concentration in each fraction was determined using a bicinchoninic acid protein assay kit (Pierce Corp.).

PC-9 and RPC-9 cells were treated either with gefitinib (1 μmol/L) for 24 hours or with the functional peptides (20 μmol/L) for 14 hours. The cells were changed to medium containing 10 μg/mL JC-1 dye (Invitrogen Life Technologies Corp.) and cultured at 37°C for 10 minutes. After washing with fresh medium, red or green fluorescence was detected by inverted fluorescence microscopy (Olympus IX-71; Olympus Japan Inc.).

For MitoTracker-based staining, cells were incubated with medium containing 0.2 μmol/L of MitoTracker Red (Lonza Corp.) for 15 minutes at 37°C. Then, they were placed in fresh medium for evaluation by fluorescence microscopy.

Cells were fixed with PBS containing 4% paraformaldehyde for 10 minutes, then incubated with 3% bovine serum albumin/PBS-Tween 20 blocking solution for 15 minutes at room temperature. After washing, cells were incubated with 0.1% Triton/PBS containing the primary antibodies for 1 hour. Cells were washed again with 0.1% Triton/PBS, incubated for 1 hour with the secondary antibodies, and then washed 3 times with 0.1% Triton/PBS. The nuclei were counterstained with Hoechst 33258 dye (Sigma-Aldrich Corp.). Subcellular localization of the endogenous p14^ARF was visualized with fluorescein isothiocyanate-conjugated anti-mouse immunoglobulin G (IgG) antibody as the secondary antibody. HSP60 was visualized with Cy3-conjugated goat anti-rabbit antibody.

For the fluorescent immunostaining of paraffin-embedded tumor tissues, primary antibodies were used for p14^ARF (1:500) and HSP60 (1:800), and Cy3-conjugated rabbit anti-goat IgG (1:500) or Alexa Fluor 488-conjugated rabbit anti-mouse IgG (1:250) were used as the secondary antibodies. The nuclei were stained with Hoechst 33258 dye (1:1,000). The fluorescent images were detected by confocal microscopy (Carl Zeiss, Inc.).

Antigen retrieval of paraffin-embedded sections was conducted by Pascal S2800 systems (DakoCytomation, Ltd.) in sodium citrate buffer and endogenous peroxide activity was quenched by incubation in 3% hydrogen peroxide (Kanto Chemical, Inc.) for 10 minutes at room temperature. Slides were then washed with 0.05% Tween-20 in PBS and incubated with the primary antibody [p14^ARF (1:500) or p53 (1:200)] diluted in CanGet signal solution (Toyobo, Ltd.) overnight at 4°C. The signals were developed with Envision+System-HRP (DAB) kit (DakoCytomation, Ltd.). Nuclei were counterstained with hematoxylin.

Formalin-fixed, paraffin-embedded tumor tissues that were diagnosed by pathologists as a lung adenocarcinoma were obtained surgically or via core biopsies from patients. All patients provided informed consent for the scientific use of their tissues. Tissues subjected to the analysis in this study were selected according to the mutational analysis of the EGFR gene, which was conducted by Okayama University Hospital and Toyama University Hospital. In total, 3 tumor tissues from patients with an activating EGFR mutation and 3 tumor tissues from patients with a resistant EGFR mutation were examined. The peptide nucleic acid–locked nucleic acid PCR clamp-based detection test (Mitsubishi Chemical Medience Corp.) was used to detect EGFR mutations (28, 29).

The use of these tissues for this study was approved by the Institutional Review Boards (ethics committees) at Okayama University Graduate School of Medicine, Dentistry and Pharmaceutical Sciences, University of Toyama Graduate School of Medicine and Pharmaceutical Science, and Aichi Cancer Center.

Statistical differences were analyzed by paired Student t test (MS Excel) and a value of P < 0.05 was regarded as statistically significant.

As a preliminary analysis for identifying gefitinib-specific molecular response, reactivity to gefitinib was first assessed among 4 NSCLC cell lines, 2 pairs of drug-sensitive and drug-resistant clones. The chemical structure of gefitinib used in this study is shown in Fig. 1A. The growth of PC-9 and HCC827 cells, both of which bear activating mutations in EGFR (exon 19 of EGFR), was suppressed by gefitinib in a dose-dependent manner (Supplementary Fig. S1A). For example, approximately 60% of PC-9 cells were inhibited with 1 μmol/L gefitinib and 90% of were inhibited with 10 μmol/L. However, the growth of RPC-9 and H1975 cells, both of which harbored the T790M mutation as the resistant EGFR mutation, was not affected by 1 μmol/L of gefitinib. However, 10 μmol/L of gefitinib comparably suppressed their growth (50% inhibition in RPC-9), which suggested that the T790M mutation does not completely abolish the response but still inhibits it (11). When the effect of gefitinib on cell-cycle progression was investigated in PC-9 and RPC-9 cells using the FUCCI (fluorescent ubiquitination-based cell-cycle indicator) system (30) to monitor the cell-cycle progression, the number of G₁-arrested living cells markedly increased in response to gefitinib treatment among PC-9 cells but not among RPC-9 cells. A similar effect was observed by fluorescence-activated cell sorting analysis (Supplementary Fig. S1B and S1C). These results suggest a distinct effect of gefitinib in cell-cycle regulation as a critical point of the drug action.

Based on this finding, the dynamics of CDKN1A and CDKN2A gene expression were further examined by qPCR in PC-9 and RPC-9 cells with or without gefitinib treatment. The mRNA level of CDKN2A but not that of CDKN1A in PC-9 cells showed a time-dependent increase in response to gefitinib, whereas RPC-9 showed no increase in both mRNA levels (Supplementary Fig. S1D). We found specific amplification of p14^ARF expression but not the expressions of other CDKN2A variants (p16^INK4a, p16 variant 2, p12) upon gefitinib treatment by conventional RT-PCR and qPCR only in gefitinib-sensitive NSCLC lines (PC-9 and HCC827; Supplementary Fig. S2A–S2C). This phenomenon was not observed in gefitinib-resistant lines (RPC-9 and H1975). In PC-9 cells, p14^ARF mRNA was augmented up to fourfold after 48 hours of incubation with gefitinib, which was consistent with prominent induction of the p14^ARF protein (Supplementary Fig. S2B and S2D). p16^INK4a and other CDKN2A variants were not significantly amplified in response to gefitinib (Supplementary Fig. S2C and S2D).

Based on these data, we examined the subcellular localization of endogenously induced p14^ARF in PC-9 cells in comparison with RPC-9 cells with or without treatment of gefitinib. Immunofluorescence using anti-p14^ARF antibody showed that endogenous p14^ARF was distributed to the nucleoli, nucleus, and mitochondria in gefitinib-untreated PC-9 and HCC827 cells, whereas only mitochondrial localization of p14^ARF was recognized in the untreated RPC-9 and H1975 cells (Fig. 1B and Supplementary Fig. S2E). Fraction analysis of cell lysates from gefitinib-treated PC-9 cells revealed the prominent increase of p14^ARF in the mitochondria rather than in the nucleus (Fig. 1C). In contrast to PC-9 cells, no significant accumulation of p14^ARF was observed in the mitochondrial or nuclear fraction in RPC-9 cells (Fig. 1D). When knocking down p14^ARF mRNA by the RNA interference, activation of caspase-3 was affected (Fig. 1E). Moreover, partial rescue of the cells from apoptosis was observed in the gefitinib-treated PC-9 cells when they were treated with the p14^ARF-specific siRNA (Fig. 1F). These data suggested that p14^ARF may play a crucial role in apoptotic induction by gefitinib.

However, we examined p14^ARF mRNA levels in PC-9 and RPC-9 cells treated with conventional chemotherapeutic agents such as cisplatin and etoposide to examine whether the induction of p14^ARF expression is a gefitinib-specific response. Standard doses of these agents induced the same degree of growth suppression as was observed in gefitinib treatment in PC-9 and RPC-9 cells (Supplementary Fig. S3A) and the qPCR showed no significant induction of p14^ARF mRNA by cisplatin or etoposide in contrast to gefitinib (Supplementary Fig. S3B). Furthermore, knockdown of EGFR using specific siRNA increased p14^ARF mRNA both in PC-9 and RPC-9 cells suggested that inactivation of EGFR signaling by gefitinib resulted in specific induction of p14^ARF among genes encoded by the CDKN2A locus on these lung adenocarcinoma cells (Supplementary Fig. S3C and S3D).

Genomic sequence analysis of p14^ARF in PC-9 and RPC-9 revealed that p14^ARF seemed functional because these cell lines are carrying only silent mutations (Supplementary Fig. S4A). However, both PC-9 and RPC-9 cells harbor a missense mutation (R248Q) in the DNA-binding domain of p53 (31) and nuclear localization of p53 was not altered in response to gefitinib treatment (Supplementary Fig. S4B and S4C). These findings suggest that p14^ARF may function in a p53-independent manner in these cells. H1975 had a point mutation of the glycine residue at position 83 (G83V), but its biological role has not yet been elucidated (Supplementary Fig. S4A).

To identify the specific function of p14^ARF and its functional intracellular localization that efficiently mediates growth inhibition in the NSCLC cells with EGFR mutations, we generated several expressors of the truncated p14^ARF (Fig. 2A): the d1 expressor, which lacked the MDM2 binding site; the d2 expressor, which lacked the nucleophosmin/B23 binding site; and the d3 expressor, which lacked the entire exon 1β region. These truncated p14^ARF expressors were fused with either the mitochondrial-targeting sequence (MTS) or nuclear localization signal (NLS) to regulate the subcellular localization of each protein. Although the MTS-targeted d2 and d3 mutants showed lower expression than the wild-type (WT) and d1 expressors (Fig. 2A), the MTS-p14^ARF d2 most efficiently inhibited cell growth of both PC-9 and RPC-9 (Fig. 2B). For a detailed analysis, we further generated the GFP-tagged MTS-p14^ARF expressors to determine the essential region for growth inhibitory function within p14^ARF (Fig. 2C). Among the p14^ARF expressors covering the amino acid residues 38-65 (MTS-GFP-p14^ARF 38-65), MTS-GFP-p14^ARF 65-132, and MTS-GFP-p14^ARF WT, the protein products of which were localized in the mitochondria in both PC-9 and RPC-9 cells (Supplementary Fig. S5), the MTS-GFP-p14^ARF 38-65 showed the most potent growth inhibition. The inhibitory ratio was 60% in the transfected PC-9 cells and 70% in the RPC-9 cells, which was even more potent than in MTS-GFP-p14^ARF WT-transfected cells (Fig. 2D). In the MTS-GFP-p14^ARF 38-65-transfected cells, prominent mitochondrial aggregation was observed, which was similar to the cells treated with gefitinib (Fig. 3A and B). Staining these cells with JC-1 dye revealed that the aggregation triggered mitochondrial dysfunction (Fig. 3B; ref. 32) and that eventually led to caspase-3 activation in MTS-GFP-p14^ARF 38-65-expressing PC-9 and RPC-9 cells as well as in gefitinib-treated PC-9 cells (Fig. 3C). These findings suggested that the region encoding residues 38-65 within p14^ARF plays an important role for gefitinib-induced growth inhibition in NSCLC cells with EGFR mutations.

Taking these results together, we designed a synthetic peptide with an antitumor property against the gefitinib-sensitive and gefitinib-resistant NSCLC cells carrying functional EGFR mutations. The peptide covering the amino acid residue from the 38 to 65 positions of the p14^ARF protein was generated by fusion of nona-d-arginine (r9) to create a cell-penetrating form. The p14-1C peptide encoded by the p14^ARF MDM2 binding site, which suppresses the growth of p14^ARF-negative glioblastoma cells (33), was used for a comparison to evaluate the antitumor efficacy (Fig. 4A). MTT cell-proliferation assays revealed growth suppression in all 4 NSCLC lines after 48 hours of incubation with these peptides (Fig. 4B); however, the p14 38-65 peptide showed the maximum growth inhibition on all 4 lines in comparison with the p14-1C peptide (90% inhibition vs. 60% inhibition; Fig. 4B). The p14 38-65 peptide penetrated more than 99% of the NSCLC cells and it successfully localized in the mitochondria after its incorporation into the cells as was shown by MitoTracker staining (Fig. 4C and Supplementary Fig. S6B). The peptide functioned to trigger the reduction of mitochondrial membrane potential, corroborated by JC-1 dye (Fig. 4D), and it eventually caused apoptosis via the activation of caspase-9 in both PC-9 and RPC-9 cells (Fig. 4D and E). Moreover, the antitumor effect (growth suppression) by the p14 38-65 peptide was the strongest among the 4 different functional peptides targeting specific molecules other than p14^ARF, which can restore the impaired p16, p14, and p21 functions (33) or inhibit Akt function (Supplementary Fig. S6A, S6C, and S6D). Thus, the peptide encoded by the amino acid position 38-65 of p14^ARF was a powerful antitumor tool against gefitinib-sensitive and gefitinib-resistant NSCLC cells, irrespective of the type of EGFR mutations.

Using the p14 38-65 peptide, we attempted to restore the gefitinib sensitivity in the resistant clones. As shown in the graphs, pretreatment of the resistant cells with a small amount of p14 38-65 peptide (2–10 μmol/L conc.) sensitized the response against the same dose of gefitinib (1 μmol/L conc.) as in the case of gefitinib-sensitive clones such as PC-9. For example, treatment with 1 μmol/L of gefitinib enhanced growth suppression of RPC-9 cells from 40% to 80% when the cells were pretreated with 5 μmol/L of the p14 peptide for 12 hours. However, 5 μmol/L of p14 peptide alone was not sufficient to eliminate the tumor cells. Similarly, a 20% increase in growth inhibition was obtained by 1 μmol/L of gefitinib treatment in the peptide-pretreated H1975 cells in comparison with the peptide-untreated cells (Fig. 5A and B). Unlike the case of the p14 peptide, pretreatment with the same dose of the p21-S153A peptide did not augment the response to gefitinib in these resistant clones. In this case, 5 μmol/L of the p14 peptide alone did not seriously affect the mitochondrial membrane potential, whereas the potential was dramatically reduced in these peptide-pretreated cells in the presence of 1 μmol/L of gefitinib (Fig. 5C). We also found that the p14 38-65 peptide (at 20 μmol/L conc.) has little cytotoxic effect on the cell growth of the human immortalized lines derived from normal cell origins that retain expression of p14^ARF (Fig. 5D and E).

As a next step, we examined the expression of p14^ARF in human NSCLC tissues to determine whether they reflected the subcellular localization of p14^ARF in the corresponding cell lines in vitro. Double immunofluorescence studies using anti-p14^ARF antibody and anti-HSP60 antibody as mitochondrial markers revealed that lung adenocarcinoma tissues with the deletion of EGFR exon 19 (ΔEx19) or those with the secondary mutation in T790M in addition to the deletion of EGFR exon 19 (ΔEx19+T790M) expressed colocalized p14^ARF with HSP60 (Fig. 6A). Immunohistochemistry with anti-p14^ARF antibody in these cases showed that abundant expression of p14^ARF was observed in both the nuclei and mitochondria of lung cancer cells in the cases carrying the activating EGFR mutations (ΔEx19 or L858R in exon 21), whereas endogenous expression of p14^ARF on tumor cells was comparably weaker in the cases with the resistant EGFR mutations (T790M and D761Y; Fig. 6B). Expression of the p14^ARF seemed to be independent of p53 expression in these cancerous tissues (Supplementary Fig. S7A and S7B). In normal bronchiolar epithelial cells and alveolar epithelial cells, p14^ARF was expressed in the nucleus with or without mitochondrial expression and not all of the epithelial cells expressed it (Fig. 6B). Thus, the expression pattern of endogenous p14^ARF in cellular models represented by PC-9 and RPC-9 seemed to be coincident to the genetically corresponding tumor tissues from patients.

Utility of the EGFR tyrosine kinase inhibitor “gefitinib” is restricted to NSCLCs carrying activating EGFR mutations at the present moment. Furthermore, the antagonist actions are not durable, because the tumor cells inevitably cause the development of secondary EGFR mutations that lead to gefitinib resistance. Considering this situation, characterization of the cellular and molecular pathways that underlie the development of the resistance is critical for insight into novel therapeutic countermeasures. This study showed the altered expression of CDK inhibitors as the key to characterize the mechanisms mediating gefitinib sensitivity and its resistance in NSCLCs. The CDKN2A locus is encoded by the alternative splicing variants that serve as tumor suppressor genes, including p16^INK4a, p14^ARF, and p12 (17–19). Although little is known about p12, both p16^INK4a and p14^ARF play crucial roles in regulating cell proliferation and their expression is frequently impaired by epigenetic causes in biologically aggressive tumors (20). Here we show that the specific induction of p14^ARF in response to gefitinib was distinct between the gefitinib-sensitive and resistant clones, although they retained its basal expression. In the sensitive NSCLC cells represented by PC-9, approximately fourfold enhancement of p14^ARF mRNA was observed in response to the drug, whereas there was no significant change in p14^ARF mRNA levels against gefitinib. Furthermore, induction of p14^ARF was gefitinib specific in these tumor cells, because this phenomenon was not detected by treatment with any other chemotherapeutic agents. Its induction was specifically mediated by the EGFR pathway, which was shown by the siRNA-introduced EGFR-knockdown model.

We show that mitochondrial accumulation of p14^ARF is an important response to gefitinib, although many previous studies have reported that the inhibitory mechanism of p14^ARF is in the conventional MDM2-p53 pathway, triggering cell-cycle arrest and apoptosis. Indeed, p14^ARF function still remains controversial in various cells (22, 23); however, we corroborated that mitochondrial p14^ARF efficiently inhibited the growth of NSCLC cells and triggered apoptosis through caspase-9 activation, which was more effective than nuclear-translocated p14^ARF. Because we detected a missense mutation in p53 in these cells (31), p14^ARF might trigger apoptosis in a p53-independent manner. To explain this, we tried to find a proapoptotic molecule as a specific target of the augmented p14^ARF in the gefitinib-treated NSCLC cells, but p14^ARF did not alter the expression of the mitochondrial proapoptotic proteins Bax, Bad, Bcl-2, and Bcl-xL. Although Bcl-2–interacting mediator of cell death induction was detected by gefitinib treatment as reported in previous studies, it seemed to be p14^ARF independent (data not shown). Thus, the details of the apoptotic mechanism via mitochondrial p14^ARF still remain to be elucidated.

In this study, the region spanning the 38-65 amino acid residues within p14^ARF was found to be a functional core for mitochondrial targeting and apoptosis in the NSCLC cells. Based on this, we generated the novel antitumor peptide p14 38-65 as a synthetic cell-permeable form, which may enable in vivo application. This peptide contains the amino acids positioned the 42-45 (AVAL) of p14^ARF, which was reported to be a putative motif of mitochondrial localization, although it lacks a mdm2/B23 binding site (34). This newly designed peptide is powerful for antitumor potency because its antiproliferative effect was much stronger than that of the p14 38-50 peptide (fourfold more; Supplementary Fig. S6). Other functional peptides including the p21 S153A peptide (35), the Smac peptide, and the Akt inhibitory peptide (36) showed growth inhibition to the NSCLC cells to some extent; however, the p14 38-65 peptide was superior to them. Although its precise molecular action is not fully clear, it affects the mitochondrial membrane potential and proapoptotically functions on the NSCLC cells examined. Previous studies have reported that nucleolar expression of p14^ARF in cells lacking p53 affects cell growth by inhibiting ribosome biogenesis through association with nucleophosmin (37, 38). The p16-MIS peptide that restores the p16^INK4a function comparably inhibited cell growth because the tumor cell lines examined bore a missense mutation in p16, but p16 was not the target for gefitinib action as was shown in this study (Supplementary Fig. S4A and S6; ref. 39). As shown in Fig. 5, one of the utilities of the p14 38-65 peptide may be as a cooperative agent in the treatment of the resistant clones with gefitinib, to recover their drug sensitivity up to a level of the sensitive NSCLC cells.

Histologic examination of lung adenocarcinoma specimens from patients with either ΔEx19 of EGFR or L858R in Ex21 of EGFR (both are activating EGFR mutations) coincidently reflected the expression pattern of endogenous p14^ARF and the subcellular localization that were observed in the in vitro studies. Expression of p14^ARF seemed weaker in the cases with either the T790M or D761Y resistance mutation. Furthermore, the mutant p53 seemed to be independently expressed from p14^ARF in all the cases examined (Supplementary Fig. S7).

In summary, here we show that p14^ARF is induced by gefitinib through inactivation of the EGFR signaling pathway in the EGFR-mutated NSCLC cells and it is one of the essential molecules for the drug action. Moreover, mitochondrial localization is important for p14^ARF to exert efficient cell growth suppression, which seems to function in a p53-independent manner. The mechanism of p14^ARF induction by gefitinib and the detailed interaction of p14^ARF at mitochondria still remain unclear; however, mitochondrial p14^ARF is one of the critical regulators of gefitinib action (Fig. 6C). We identified its functional core sequence and showed one of the utilities of the designed antitumor peptide against gefitinib-resistant NSCLC cells in this study. Our study may contribute to the development of therapeutics for these intractable cancers.

N. Takigawa and K. Kiura have honoraria from Speakers Bureau of AstraZeneca. No potential conflicts of interest were disclosed by the other authors.

Conception and design: K. Saito, R. Ueda, E. Kondo

Development of methodology: K. Saito, E. Kondo

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K. Saito, N. Takigawa, Y. Tomita, J. Fukuoka, E. Ichihara, K. Kiura, E. Kondo

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K. Saito, N. Ohtani, J. Fukuoka, K. Kuwahara, E. Kondo

Writing, review, and/or revision of the manuscript: K. Saito, N. Takigawa, N. Ohtani, R. Ueda, J. Fukuoka, E. Kondo

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K. Saito, H. Iioka, K. Kiura

Study supervision: K. Kiura, E. Kondo

The authors thank the members of the Kondo Lab for suggestions about experimental design, for critical evaluation of this manuscript, and for technical assistance.

E. Kondo was supported by a Grant-in-Aid for Scientific Research (Kiban-C; grant no. 23590442; E. Kondo) from the Japanese Ministry of Education, Culture, Sports, Science, and Technology (MEXT).

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.