DDX3X mediates EGFR-TKI resistance through VEGFR signaling


 Although epidermal growth factor receptor (EGFR)-tyrosine kinase inhibitors (TKIs) are remarkably effective against non-small-cell lung cancer (NSCLC) with EGFR-activating mutations, lung cancer cells acquire resistance to EGFR-TKIs without exception. Several mechanisms of EGFR-TKI resistance have been reported, but there are many aspects that remain to be clarified. We previously identified DDX3X as an immunogenic protein preferentially expressed in murine melanoma with a cancer stem cell (CSC)-like phenotype. DDX3X induced epithelial-mesenchymal transition and reduced the sensitivity to EGFR-TKIs in PC9 cells, human lung cancer cells harboring EGFR exon 19 deletion. We also reported that there was a small nonadherent subpopulation of parental PC9 cells that highly expressed DDX3X and had CSC properties. In this study, we found that VEGFR2 was upregulated in lung cancer cells that strongly expressed DDX3X and that these cells were addicted to VEGFR signaling. The blockade of both EGFR and VEGFR signaling reduced the phosphorylation of downstream signals in the cells with DDX3X that acquired EGFR-TKI resistance. The addition of VEGFR-TKIs or anti-VEGF antibodies to EGFR-TKIs significantly inhibited the progression of EGFR-mutated NSCLC in a xenograft mouse model. These data suggest that the blockade of VEGFR signaling enhances the antitumor effects of EGFR-TKIs by eradicating cancer stem cells, which mediate resistance to EGFR-TKIs.


Introduction
Mutations in the epidermal growth factor receptor (EGFR) have been shown to increase the kinase activity of EGFR and over-activate the downstream pro-survival signaling pathway 1,2 . In the presence of ligandindependent activation of EGFR signaling, cancer cells become dependent on EGFR signaling for prosurvival signaling pathways. Treatments targeting this signaling dependence have led to breakthrough therapeutic outcomes in the clinical setting. The use of EGFR tyrosine kinase inhibitors (TKIs) has signi cantly improved progression-free survival (PFS) in non-small-cell lung cancer (NSCLC) patients harboring activating EGFR mutations; however, lung cancer eventually acquires resistance and recurs without exception 3,4 . Most cases of acquired resistance re ect the selection of cancer cells harboring stochastic resistance-conferring genetic alterations. The mechanisms of EGFR-TKI resistance include T790M mutation, c-Met ampli cation, PIK3CA mutation and epithelial-to-mesenchymal transition (EMT) 5 , which account for 50-70% of cases of resistance, but the remaining mechanisms remain unknown 6 .
Recent evidence has demonstrated that cancer stem cells (CSCs) are critically involved in resistance to cytotoxic therapies. Sharma et al. reported that a small subpopulation of reversibly drug-tolerant cells existed in all examined cancer cells and that the drug-tolerant cells behaved as mother cells, giving rise to drug-resistant cells harboring additional mutations 7 . This small subpopulation shares biological characteristics with stem cells, is able to survive lethal stresses such as chemotherapy and radiotherapy and plays an important role in cancer treatment.
DEAD/H (Asp-Glu-Ala-Asp/His) box polypeptide 3, X-linked (DDX3X) is a member of the DEAD-box family of ATP-dependent RNA helicases and is located on the X chromosome 8 . DEAD-box helicases have multiple functions in processes including RNA splicing, mRNA export, transcriptional and translational regulation, RNA decay, ribosome biogenesis, and miRNA regulation 9,10 . Thus, DDX3X is thought to be involved in the epigenetic regulation of gene expression. Our previous proteome analyses identi ed DDX3X as a protein preferentially expressed in puri ed CD133 + B16 melanoma cells, which possess CSClike properties 11,12 . Recently, we demonstrated that EGFR-mutated NSCLC cells with DDX3X transgenes lost EGFR signaling addiction and acquired EGFR-TKI resistance without any additional EGFR mutations. Moreover, we found that a small subpopulation of parental lung cancer cells with a mesenchymal cell phenotype strongly expressed DDX3X and exhibited EGFR-TKI resistance 13 .
In this study, we sought to investigate alternative prosurvival signaling pathways induced by DDX3X in EGFR-TKI-resistant lung cancer cells and found that vascular endothelial growth factor receptor 2 (VEGFR2) expression was increased in DDX3X-expressing lung cancer cells. Recent evidence has shown that anti-VEGF therapy signi cantly enhances the antitumor e cacy of EGFR-TKIs in EGFR-mutated NSCLC patients 14,15 . In addition, VEGF-mediated signaling has been shown to contribute to key aspects of tumorigenesis, including the self-renewal and survival of CSCs 16 . Therefore, we investigated whether signaling through VEGFR is involved in EGFR-TKI resistance in lung cancer cells expressing DDX3X.

Results
High expression of DDX3X reduces EGFR signaling in cancer cells harboring EGFR-activating mutations.
To examine the effects of DDX3X on the cellular phenotype, we transfected PC9 cells, human lung adenocarcinoma cells harboring an EGFR exon 19 deletion mutation, with cDNA encoding DDX3X and established a novel cell line, termed A-4, which overexpressed DDX3X. Immunoblotting analysis con rmed the overexpression of DDX3X in the transfectants (Fig. 1A).
The survival and proliferation of PC9 cells are highly dependent on EGFR signaling because PC9 cells possess driver mutations in the EGFR gene that enhance tyrosine kinase activity. We analyzed whether EGFR signaling was affected by the overexpression of DDX3X. As shown in Fig. 1B, EGFR phosphorylation was observed in parental PC9 cells but not in A-4 cells. The suppression of EGFR phosphorylation was also observed in A-1, A-2 and A-5 cells transfected with DDX3X (Fig. 1B).
Our previous study demonstrated that the CSC-like subpopulation of PC9 cells expressed DDX3X and exhibited scaffold-independent growth 13 . To examine whether the overexpression of DDX3X affects the scaffold-independent growth of lung cancer cells, the ratio of nonadherent to adherent cells among parental PC9, A-1 and B-2 cells (mock) was evaluated. The ratio of nonadherent cells to adherent cells signi cantly increased in A-1 cells, suggesting that DDX3X promotes adhesion-independent proliferation ( Fig. 1C). Because nonadherent cells were also present among parental PC9 cells, we assessed the expression of DDX3X in parental nonadherent PC9 cells. Similar to A-4 cells, nonadherent PC9 cells showed DDX3X overexpression (Fig. 1D). In addition, nonadherent PC9 cells lacked EGFR phosphorylation despite the presence of EGF (Fig. 1E). The subsequent phosphorylation of extracellular signal-regulated kinase (ERK) and serine/threonine kinase (AKT) in nonadherent PC9 and A-4 cells was also suppressed.

Nonadherent cells overexpressing DDX3X express VEGFR2.
Recent clinical trials showed that the addition of anti-VEGF therapy to EGFR-TKIs prolonged PFS in NSCLC patients with activating EGFR mutations 14,15 . Therefore, we hypothesized that VEGF might be involved in EGFR-TKI resistance and performed FACS analysis to determine whether VEGFR was expressed on the surface of EGFR-mutated NSCLC cells. The results showed that approximately 50% of the nonadherent population of parental PC9 cells, which overexpressed DDX3X, strongly expressed VEGFR2, while only approximately 1% of adherent cells among PC9 parental cells expressed VEGFR2 (Fig. 2).
The combination of EGFR-TKIs and VEGFR-TKIs inhibits the phosphorylation of ERK and AKT in DDX3Xoverexpressing cells.
We next performed immunoblotting analysis to investigate the in uence of VEGFR-TKIs on EGFR downstream signaling. In the parental PC9 cells, the EGFR-TKI erlotinib alone inhibited the phosphorylation of ERK and AKT, whereas the VEGFR-TKI lenvatinib alone did not interfere with the phosphorylation of ERK and AKT in PC9 cells ( Fig. 3A and B). In A-4 cells, neither erlotinib nor lenvatinib alone reduced the phosphorylation of ERK and AKT ( Fig. 3A and B). In contrast, the combination of erlotinib and lenvatinib reduced the phosphorylation of ERK and AKT in A-4 cells ( Fig. 3A and B).
The combination of EGFR-TKIs and VEGFR-TKIs or anti-VEGF antibodies had synergistic effects in DDX3X high-expressing cells in vivo.
To investigate the antitumor effect of the combination of EGFR-TKIs and VEGFR-TKIs in vivo, we generated a mouse xenograft tumor model with PC9 and A-4 cells. As shown in Fig. 4A, the progression of A-4 skin tumors was faster than that of PC9 skin tumors. Therefore, each drug treatment was started on day 21 for the PC9 cell groups and on day 14 for the A-4 cell groups, and treatments for both groups were continued for 2 weeks. In the PC9 cell group, tumor growth was signi cantly inhibited by erlotinib (Fig. 4B). On the other hand, in the A-4 cell group, erlotinib appeared to inhibit tumor growth slightly in the early phase of drug administration, but tumor growth was generally the same as that in the untreated group (Fig. 4C). However, the addition of lenvatinib to erlotinib markedly delayed the growth of tumors derived from A-4 cells (Fig. 4D). We next investigated whether bevacizumab, an anti-VEGF monoclonal antibody (mAb), could exert a synergistic antitumor effect when used in combination with erlotinib. As shown in Fig. 4E, the combination of bevacizumab and erlotinib showed signi cantly greater antitumor effects than erlotinib alone in A-4 cells. The combination of lenvatinib and erlotinib also signi cantly inhibited the progression of tumors derived from PC9 cells compared to that with erlotinib alone (Fig. 4F).

Discussion
Our previous study demonstrated that strong expression of DDX3X in lung cancer cells with EGFRactivating mutations was accompanied by CSC-like properties, EMT and resistance to EGFR-TKIs due to the loss of EGFR signaling addiction 13 . In the current study, we investigated alternative prosurvival signals induced by DDX3X in the context of EGFR-TKI resistance using human lung cancer cell lines. We found that DDX3X induced the upregulation of VEGFR2 and signal transduction by VEGFR2. An in vivo xenograft model demonstrated that dual inhibition of EGFR and VEGFR signaling had synergistic durable therapeutic effects on DDX3X-expressing cancer cells.
CSCs can survive under potentially lethal stress conditions because they have multiple mechanisms by which they can resist cell death, including altered chromatin status; overexpression of multidrug e ux transporters, antiapoptotic factors, or DNA repair gene products; and stem cell-speci c growth signaling 7,17−22 . Recent studies have demonstrated that CSCs are involved in TKI resistance 23,24 and that the suppression of CSC properties could inhibit EGFR-TKI resistance in lung adenocarcinoma 25 . The present study showed that forced expression of DDX3X in lung cancer cells resulted in the suppression of EGFR phosphorylation, suggesting that these DDX3X-expressing cells proliferate independently of EGFR signaling (Fig. 1B). Indeed, lung cancer cells with high expression of DDX3X showed resistance to erlotinib in vivo (Fig. 4C). These DDX3X-expressing A-4 cells showed anchorage-independent proliferation (Fig. 1C). Parental PC-9 cells also included nonadherent cells, which showed a high level of DDX3X expression and a reduction in EGFR signaling ( Fig. 1D and E). These ndings indicated that EGFRmutated lung cancer cells include CSC-like cells that strongly express DDX3X and that these cells play an important role in resistance to EGFR-TKIs.

Recent basic and clinical studies have demonstrated the importance of dual inhibition of EGFR and
VEGFR in EGFR-mutated lung cancer cells. Cancer cells express VEGFRs, and autocrine VEGF/VEGFR signaling promotes cancer cell growth, survival, migration, and invasion 26 . VEGF expression is driven by many factors that are characteristic of tumors, including the expression of oncogenes, (e.g., ras, src, ERBB2, and EGFR) and hypoxia, and the EGFR and VEGF/VEGFR signaling pathways are tightly connected with each other 27 . Preclinical studies have shown that dual inhibition of EGFR and VEGFRdependent signaling overcomes the resistance to EGFR-TKIs 28, 29 . Randomized phase III studies have demonstrated that combination therapy with bevacizumab or ramucirumab and erlotinib signi cantly prolongs PFS compared with erlotinib monotherapy 14,15 . VEGF has been found to be involved in the promotion of cancer stemness and self-renewal; VEGF/VEGFR2/neuroplilin-1 autocrine signaling regulates CSCs to promote glioma growth 30 , and in skin cancer, the blockade of VEGFR2 not only decreased the microvascular density but also reduced pool size and impaired CSC renewal properties, resulting in tumor regression 31 . Additionally, in 2015, Zhao et al. showed that VEGF promotes CSC selfrenewal in breast and lung cancers through VEGFR2/STAT3-mediated upregulation of Myc and Sox2 32 . In the current study, combination therapy with VEGFR-TKI or anti-VEGF antibodies and EGFR-TKI signi cantly delayed the progression of lung cancer cells with CSC-like properties (Fig. 4D and E). Previous studies showed that the mechanisms of anti-VEGF therapy were normalization of intratumor blood ow and improvement of drug delivery 33,34 . We demonstrated that in lung cancer cells that overexpressed DDX3X and acquired CSC-like characteristics, EGFR-TKIs or VEGFR-TKIs alone did not suppress downstream signaling; however, the combined use of VEGFR-TKIs with EGFR-TKIs suppressed the phosphorylation of ERK and AKT (Fig. 3A and B). Because these results were observed in in vitro experiments without stromal cells, the blockade of signaling through VEGFR2 in EGFR-TKI-resistant cells is one of the mechanisms underlying the effectiveness of anti-VEGF therapy.
In conclusion, this study demonstrates that a DDX3X-dependent increase in VEGFR2 expression may confer EGFR-TKI resistance in lung cancer cells with EGFR-activating mutations. Combination therapy with EGFR-TKIs and VEGFR-TKIs seems to be a promising approach to overcome EGFR-TKI resistance mechanisms based on the eradication of CSCs.

Materials And Methods
Tumors PC9 human lung adenocarcinoma cells harboring an EGFR exon 19 deletion were obtained from Riken BioResource Center and maintained in culture medium containing RPMI 1640 medium supplemented with 10% heat-inactivated lipopolysaccharide-quali ed fetal calf serum, 0.1 mM nonessential amino acids, 1 µM sodium pyruvate, 100 U/mL penicillin, and 100 µg/mL streptomycin sulfate (all from Life Technologies, Inc., Tokyo, Japan) 13 .

Transfection of PC9 cells with DDX3X cDNA
Transfection of PC9 cells with DDX3X cDNA was obtained using a Myc-DDK-tagged open reading frame clone of Homo sapiens DDX3X transcript variant 1 as transfection-ready DNA (Origene Technologies, Inc., Rockville, MD, USA) according to the manufacturer's protocol 13 . Experimental cells were incubated with fresh medium containing G418 (600 µg/mL, Promega, Madison, WI, USA), and the medium was replaced with fresh G418-containing medium every 3-4 days until resistant colonies were identi ed 13 .
Cell viability assessment Cells were cultured on 24-well plates for 3 days 13 . Nonadherent cells were collected by gentle rocking in culture medium 13 . The ratio of nonadherent cells to adherent cells was calculated 13 . Dead cells were excluded via the trypan blue exclusion method 13 .
Monoclonal antibodies and ow cytometry PE-conjugated anti-VEGFR2 (Avas 12α1) antibody was purchased from BD Biosciences (San Jose, CA, USA). Analysis of cell-surface phenotypes was carried out by direct immuno uorescent staining of 0.5-1×10 6 cells with conjugated monoclonal antibodies and then treatment with 1% paraformaldehyde 12 . In each sample, 10,000 cells were analyzed using a FACSCalibur ow micro uorometer (BD Biosciences) 12 . PE-conjugated subclass-matched antibodies used as isotype controls were also purchased from BD Biosciences 12 . The samples were analyzed with CellQuest software (BD Biosciences) 12 .

Animals
Female C.B-17/lcr-scid/scidJcl (SCID) mice were purchased from CLEA (Japan), maintained in a speci cpathogen-free environment, and used for experiments at the age of 10 weeks. The experimental protocols were based on the National Institutes of Health Guide for the Care and Use of Laboratory Animals and approved by the Niigata University Institutional Animal Care and Use Committee. At the end of the experiment, the mice were euthanized due to cervical dislocation. This study was reported in accordance with the ARRIVE guidelines.
Adoptive medication therapy SCID mice were injected s.c. with parental PC9 or DDX3X-transfected PC9 (PC9-DDX3X) tumor cells in 100 μl of Hanks' balanced salt solution to establish skin tumors. Two or three weeks after inoculation, each group was randomized to groups of ve mice each, and vehicle, 25 mg/kg erlotinib (Selleck Chemicals, Houston, Texas, USA), 25 mg/kg lenvatinib (Selleck Chemicals, Houston, Texas, USA) or both was given intraperitoneally to the mice every 5 days a week. Bevacizumab was administered at a dose of 5 mg/kg three times a week. The perpendicular diameter of skin tumors was measured with digital calipers.

Statistical analysis
For experiments in the animal tumor model, the signi cance of differences between groups was analyzed using Student's t-test. All other experimental data were analyzed by one-way ANOVA, and each group was compared by Bonferroni's multiple comparison test. All statistical analyses were carried out using GraphPad Prism 6 (GraphPad Software, Inc., La Jolla, CA, USA) and EZR (Saitama Medical Center, Jichi Medical University, Saitama, Japan) 35 . A 2-tailed P value of < 0.05 indicated statistical signi cance. All experiments were repeated at least twice.

Declarations Data availability
The data generated or analyzed during the current study are available from the corresponding authors on reasonable request.