Expression of Neuroendocrine Factor VGF in Lung Cancer Cells Confers Resistance to EGFR Kinase Inhibitors and Triggers Epithelial-to-Mesenchymal Transition

Activating mutations in EGFR constitute one of the major subsets among those molecular aberrations that occur preferentially in patients with clinicopathologic characteristics of lung adenocarcinoma (1–4). EGFR-tyrosine kinase inhibitors (TKI), such as gefitinib, erlotinib, and afatinib, have been shown to produce profound therapeutic responses in lung adenocarcinoma harboring EGFR mutations (exon 19 deletions or the L858R mutation; refs. 5–10). Despite this initial response, patients with EGFR-mutated lung adenocarcinoma ultimately developed resistance to EGFR-TKIs.

A secondary mutation in EGFR (T790M), which abrogates the inhibitory activity of the TKIs, is reported to be the major contributing factor for the development of acquired resistance to EGFR-TKIs (11–13). In contrast, the mutation of T790M infers better survival outcomes and negatively correlates with distant metastasis, thereby predicting a favorable prognosis in lung cancer patients (14–17). New EGFR-TKIs, such as AZD9291 and rociletinib, have been designed to overcome T790M-mediated resistance (18–22). However, treatment of T790M-positive patients with AZD9291 or rociletinib will still eventually develop resistance and in part convert tumors to T790M-negative ones (23, 24). Thus, other non-T790M factors may affect cancer progression and predict worse survival to EGFR-TKI treatment. Pathologic transformation from adenocarcinoma toward the neuroendocrine phenotype similar to small-cell lung cancer or large-cell neuroendocrine carcinoma, has been detected in specimens during EGFR-TKI treatments (13, 25–27). Several studies also revealed that epithelial-to-mesenchymal transition (EMT), a proinvasive phenotype, can endow EGFR-mutated lung cancer cells with EGFR-TKI resistance (28, 29). Nonetheless, the biological mechanism underlying EMT-mediated resistance to EGFR-TKIs remains elusive.

The VGF (nerve growth factor inducible) gene encodes a neuronal and neuroendocrine polypeptide that is secreted by normal neurons and neuroendocrine cells and is responsible for energy balance and metabolism (30, 31). VGF expression enhances neuronal growth and prevents apoptosis (32, 33). VGF has been detected in several neuroendocrine cells and cancers (34–36); however, the role of VGF in tumor progression is not known. In this study, we discovered that VGF is highly expressed in EGFR-TKI–resistant lung adenocarcinoma cells and the expression of VGF induces EMT in EGFR-mutated cells. Therefore, we further characterized the role of VGF in EGFR-TKI resistance and cancer cell dissemination in lung adenocarcinoma.

Gefitinib, erlotinib, afatinib, AZD9291, and rociletinib were purchased from Cayman Chemical. Doxycycline was ordered from Sigma-Aldrich Inc. RPMI1640 and FBS were purchased from GIBCO-BRL.

HCC827 cells were kindly provided by Dr. Jeff Wang (Development Center for Biotechnology, Taiwan) and further certified via short tandem repeat-PCR DNA profiling in 2013 (37). H1975 cells were obtained from Dr. Wayne Chang (National Health Research Institutes, Taiwan) and further authenticated via EGFR sequencing analysis in 2015 (38). HCC827GR cells were established in our laboratory, as described in Supplementary Materials and Methods, and tested positive for human origin carrying specific EGFR mutations by EGFR sequencing analysis (Supplementary Fig. S1) in 2015. All lung cancer cells were cultured in RPMI1640 medium containing 4 mmol/L l-glutamine, 1 mmol/L sodium pyruvate, 10 mmol/L HEPES, and 10% FBS. Cells were checked routinely and found to be free of contamination by Mycoplasma.

The VGF was amplified by PCR from cDNA library of HCC827GR cells and then subcloned into HR′-puro plasmid. pLOK.1-shVGF#1 (TRCN0000429075) and pLKO.1-shVGF#2 (TRCN0000180034) were obtained from the National RNAi Core Facility, Academia Sinica (Taipei, Taiwan). pLKO.1-Scrambled shRNA and Tet-PLKO-Puro were obtained from Addgene. Tet-On-shVGF was constructed by inserting a short hairpin double-stranded oligonucleotide targeting VGF into Tet-PLKO-Puro vector and the target sequence is 5′-CTCAGCTCTGAGCATAAAGAG-3′.

Quantitative real-time PCR (qPCR) was performed with specific primers and TaqMan probes or probes from the Universal Probe Library (Roche Applied Science) in the StepOne Real-Time PCR system (Applied Biosystems Inc.). 18S rRNA was used as a reference transcript. Primer sequences designed to detect specific genes and probes are listed in Supplementary Table S1. For quantification of gene copy numbers, total 50 ng genomic DNA extracts from HCC827 and HCC827GR cells were subjected to qPCR analysis with primer sets specific to EGFR and MET genes. Gene copy numbers from each sample were normalized to β-actin (ACTB) as a reference gene. Primer sequences and probes were listed in Supplementary Table S2.

Cells were harvested in RIPA lysis buffer supplemented with a protease inhibitor cocktail. The antibodies against specific proteins and the titers for immunoblotting and immunofluorescence used in this study are listed in Supplementary Tables S3 and S4, respectively.

Cells (1.0 × 10⁵ per well) were cultured in 400 μL of RPMI with 10% FBS on 8W1E ECIS Cultureware arrays (Applied Biophysics). The arrays were kept in the incubator with 5% CO₂ at 37°C and measured on an ECIS Models Z Theta at multifrequency in 5-minute intervals for 48 hours. Data were analyzed on electrical cell substrate impedance sensing (ECIS) software and normalized to the starting values.

IHC was conducted as described previously (1, 39). The primary VGF antibody used in the study was VGF antibody (1:300 dilution, GeneTex, catalog no. GTX87831). The immunoreactivity pattern and histologic appearance of all clinical specimens and tissue microarray slides were examined and scored by the pathologist. The Allred scoring system was used to give the staining scores for the expression of VGF based on the sum of the percentage of cells that were stained by IHC (on a scale of 0–5) and the intensity of the staining (on a scale of 0–3).

Tumor growth assays were conducted as described previously (39). Detailed procedures for the xenograft assays are described in the Supplementary Materials and Methods. All animal experiments were performed in accordance with the animal guidelines of the National Tsing Hua University Institutional Animal Care and approved by Animal Care Committee (permit number: 10260).

The specimens used in IHC were obtained from a high-density tissue array (catalog no. LC2085A, US BIOMAX) of lung tumors or from surgery at Taipei Medical University Hospital. Approval for this study was granted by the Institutional Review Board protocol number CRC-04-11-05. The public gene expression profiling datasets used in this study were analyzed as described previously (38). The sources of these gene expression profiling datasets are listed in Supplementary Table S5. The median levels of specific biomarkers (VGF, CEACAM6, SYP, and CHGA) were used as cut-off points for the overall survival analysis.

Differences in the IHC staining of human lung cancer specimens were assessed using a Pearson χ² test. Associations among the various markers and levels were calculated using Spearman ρ correlation coefficient. Overall survival (OS) curves were generated with Prism 5.0 (GraphPad software Inc.) and estimated by the Kaplan–Meier method. The differences of OS between the high- and low gene–expressing groups were compared by log-rank test. All statistical analyses were performed using SPSS software, version 16 (SPSS, Inc.). A P value less than 0.05 was considered to indicate a significant difference.

To investigate the mechanism of resistance to EGFR-TKIs in lung cancer, HCC827 cells, which harbor the EGFR delE746_A750 mutation, were treated with an incremental concentration of gefitinib. Surviving cells were pooled together, propagated, and named HCC827GR cells. The cell viability assay was performed to determine the IC₅₀ values of EGFR-TKIs showing that HCC827GR cells were resistant to not only gefitinib but also erlotinib and afatinib (Fig. 1A). Moreover, HCC827GR also exhibited resistance to AZD9291 and rociletinib, the third generation of EGFR-TKIs (Fig. 1A; Supplementary Fig. S2). Consistently, the clonogenic assay demonstrated that HCC827GR cells survived better with EGFR-TKI treatment than HCC827 cells, indicating that HCC827GR cells are resistant to EGFR-TKIs (Fig. 1B). As activation of AKT and ERK (both downstream molecules of EGFR signaling) by phosphorylation is responsible for cellular survival and proliferation, respectively, we examined EGFR, AKT, and ERK phosphorylation in HCC827GR cells as compared with that in HCC827 cells. The immunoblotting assay showed that upon gefitinib treatment, EGFR, AKT, and ERK phosphorylation was attenuated in HCC827, whereas AKT phosphorylation was not diminished by gefitinib in HCC827GR cells as compared with that in HCC827 cells, suggesting the involvement of AKT signaling in EGFR-TKI resistance (Fig. 1C, top). Western blot analysis revealed that gefitinib treatment induced two apoptosis markers, activated caspase-3 and PARP expressions, in HCC827 but not HC827GR cells. These results demonstrated that HCC827GR cells are resistant to gefitinib-induced apoptosis (Fig. 1C, bottom).

HCC827GR cells, though resistant to EGFR-TKIs, neither acquire the mutation of EGFR T790M nor the aberrant expression of MET or HER2 (Supplementary Fig. S1). Phase-contrast image analysis showed that HCC827GR cells exhibited a spindle-like shape in contrast to the epithelial morphology in HCC827 cells (Fig. 1D). The qPCR assay revealed that E-cadherin and EpCAM, two epithelial markers, were highly expressed in HCC827 but not HCC827GR cells, whereas levels of vimentin and TWIST1, two mesenchymal markers, were much higher in HCC827GR than those in HCC827 cells (Fig. 1E). In agreement, immunofluorescence staining confirmed the differential expression of cell type–specific markers, E-cadherin and vimentin, between HCC827 and HCC827GR cells (Fig. 1F). These data indicate that induction of EGFR-TKI resistance promotes EMT.

The loss of barrier function is the key cellular event in EMT. Cellular electric resistance (impedance) is determined by cell–cell contact–mediated barrier function (Rb) and cell–matrix interaction (alpha). Electric cell–substrate impedance sensing (ECIS) analysis revealed that after seeding, the level of impedance surged in HCC827, but not HCC827GR cells (Fig. 2A, top left). Moreover, huge elevations in the barrier resistance (Rb) occurred in HCC827, but not HCC827GR cells (Fig. 2A, top right and bottom). These findings indicate the loss of barrier function in HCC827GR cells. Cell tracking and wound-healing assays displayed that HCC827GR cells exhibited better migration and wound-healing abilities than HCC827 cells (Fig. 2B and C). In addition, transwell migration and invasion assays revealed that HCC827GR cells were more migratory and invasive than HCC827 cells (Fig. 2D and E). Our findings indicate the EGFR-TKI–induced EMT facilitates cancer cell dissemination in lung cancer cells.

To identify genes involved in EGFR-TKI resistance and EMT processes in lung adenocarcinoma, a gene expression profiling assay followed by qPCR analysis was performed. We found that a gene encoding a neurosecretory polypeptide, VGF, is highly elevated in gefitinib-resistant HCC827GR cells as compared with that in the parental HCC827 cells (Fig. 3A, left). The RNA in situ hybridization assay confirmed that VGF transcript was highly expressed in a primary lung neuroendocrine carcinoma and HCC827GR cells, but not HCC827 cells (Fig. 3A, top right; Supplementary Fig. S3A). Consistent with these findings, VGF expression was elevated in the independently isolated EGFR-TKI–resistant HCC827 pool and single-cell clones (Fig. 3B; Supplementary Figs. S4A and S5A S5B). The immunoblotting assay confirmed that more VGF protein was generated in HCC827GR than in HCC827 cells (Fig. 3A, bottom right). In agreement, the immunomagnetic reduction assay displayed that HCC827GR cells secreted more VGF than HCC827 cells in the conditioned medium (Supplementary Fig. S6A). We further examined VGF expression in primary tumors and cell lines derived from different pathologic types of lung cancer. We observed that VGF levels were higher in pulmonary neuroendocrine tumors, such as small-cell lung cancer and large-cell neuroendocrine carcinoma, than adenocarcinoma and squamous cell carcinoma, while a few of adenocarcinoma cells exhibited high levels of VGF expression (Fig. 3C; Supplementary Figs. S7 and S8). To examine whether VGF levels are associated with EGFR-TKI–resistant status in lung cancer cells, the IC₅₀ values of gefitinib in different EGFR-mutated lung adenocarcinoma cell lines were determined (Fig. 3D, left). qPCR analysis revealed that VGF levels were low in EGFR-TKI–sensitive cells but high in resistant cells (Fig. 3D, right; Supplementary Fig. S9). Moreover, in single-cell clones selected from gefitinib-treated HCC827 cells, VGF levels positively correlated with IC₅₀ values of gefitinib and EMT marker expression (Supplementary Fig. S5). These data suggest a possible association between VGF expression and EGFR-TKI resistance in lung adenocarcinoma cells. Intriguingly, VGF transcript was detected in H1975, a cell line harboring EGFR L858R/T790M double mutations (Supplementary Fig. S3B). We found that H1975 was more resistant to afatinib than its VGF-low counterpart (Supplementary Fig. S10). Moreover, we observed that VGF was expressed in not only treatment-naïve lung neuroendocrine carcinoma (case 1) but also acquired EGFR-TKI resistant lung adenocarcinomas (case 2 and case 3), suggesting the involvement of VGF expression in EGFR-TKI resistance (Fig. 3E). To study the role of VGF in EGFR-TKI resistance, VGF was knocked down in HCC827GR cells, followed by gefitinib-mediated cell viability analysis (Fig. 3F). The cell viability assay revealed that VGF silencing rendered HCC827GR cells sensitive to gefitinib (Fig. 3F, lower). These data indicate that VGF is critical for the maintenance of EGFR-TKI resistance.

To further elucidate the role of VGF in EGFR-TKI resistance, we ectopically expressed VGF in HCC827 cells (Fig. 4A). The cell viability assay demonstrated that VGF expression increased the IC₅₀ values of various EGFR-TKIs in the HCC827 cells (Fig. 4B). Consistently, clonogenic analysis showed that VGF expression endowed the HCC827 cells with enhanced EGFR-TKI resistance, indicating that VGF expression enhances EGFR-TKI resistance in lung cancer cells (Fig. 4C; Supplementary Fig. S2). We further investigated the effect of EGFR-TKI treatment on the phosphorylation of EGFR, ERK, or AKT in VGF-expressing HCC827 cells. Immunoblotting revealed that gefitinib treatment attenuated the phosphorylation of AKT in the control HCC827 cells, but not in VGF-expressing HCC827 cells (Fig. 4D, top). Moreover, gefitinib treatment induced the expression of cleaved PARP and activated caspase-3 in control but not VGF-expressing HCC827 cells (Fig. 4D, bottom). As EMT regulators, such as TWIST1, SLUG, and SNAIL, have been reported to regulate cell survival, we further examined their expression in HCC827 and HCC827GR cells (40). qPCR analysis revealed that TWIST1, but not SLUG and SNAIL, was highly expressed in HCC827GR cells (Fig. 4E; Supplementary Fig. S11A). Correlation analysis showed that VGF expression was closely associated with TWIST1 levels in primary lung adenocarcinoma (Supplementary Fig. S11B). Moreover, we confirmed that VGF expression induced TWIST1 in HCC827 cells (Fig. 4E). We further examined the effect of TWIST1 on EGFR-TKI resistance. Clonogenic analysis displayed that TWIST1 expression endowed HCC827 cells with EGFR-TKI resistance (Fig. 4F; Supplementary Fig. S11C). These data indicate that VGF-TWIST1 signaling encourages EGFR-TKI resistance in EGFR-mutated lung cancer cells.

Because the above data suggest that VGF expression enhances EGFR-TKI resistance, which induces EMT, we further examined the effect of VGF expression on EMT and cancer cell dissemination. Phase-contrast image analysis showed that the VGF expression induced a morphologic change from an epithelial to a spindle-like phenotype (Fig. 5A). qPCR and immunoblotting assays revealed that ectopic expression of VGF attenuated the expression of E-cadherin and EpCAM but elevated the levels of vimentin and TWIST1 (Fig. 5B and D). Immunofluorescence staining demonstrated that the VGF expression induced a switch from E-cadherin to vimentin in the HCC827 cells, supporting that VGF induces EMT (Fig. 5C). ECIS analysis displayed that VGF expression diminished impedance (Fig. 5E, top left) and attenuated Rb levels in HCC827, indicating a loss of barrier function in VGF-expressing cells (Fig. 5E, top right and bottom). The transwell assay revealed that VGF expression promoted migration and invasion in lung cancer cells (Fig. 5F and G). As VGF induces TWIST1, we further examined the effect of TWIST1 on cancer cell dissemination. Ectopic expression of TWIST1 induced a spindle-like morphologic change in HCC827 cells accompanied with the increased expression of mesenchymal markers, such as vimentin and N-cadherin (Supplementary Fig. S11D and S11E). Moreover, cell-tracking analysis revealed that TWIST1 expression encouraged cell migration (Supplementary Fig. S11F). Our findings support that VGF-TWIST1 signaling induces EMT and lung cancer cell dissemination.

To evaluate the biological significance of elevated endogenous VGF in EGFR-TKI resistant cells, we knocked down the VGF expression and measured its effect on lung cancer cell growth. Clonogenic assays showed that the knockdown of VGF attenuated cell growth in the EGFR-TKI–resistant cell lines HCC827GR and H1975 (Fig. 6A). Treatment of VGF-silenced HCC827GR cells with the condition medium from VGF-transfected HEK293T cells rescued cells from growth arrest (Supplementary Fig. S6B). Clonogenic analysis revealed that VGF-positive lung cancer cell lines, including HCC827GR, H1975, and H1299, were more resistant to the low serum (1% FBS) condition than VGF-negative HCC827 (Supplementary Fig. S12A and S12B). Moreover, overexpression of VGF in HEK293T endowed cells with an increased cell survival capacity under the low serum condition (Supplementary Fig. S12C). We observed that conditioned media from HCC827GR or VGF-expressing cells enhanced growth of HCC827 under low serum stress (Supplementary Fig. S12D). These findings suggest that VGF is essential for cell growth and survival in vitro. To evaluate the importance of VGF in maintaining cell growth in vivo, VGF was knocked down in HCC827GR cells prior to subcutaneous injection into a xenograft nude mouse tumor model. The HCC827GR cells formed tumors readily in this xenograft animal model; however, knockdown of VGF diminished tumor growth, indicating that the VGF expression enhances tumor growth in vivo (Fig. 6B). We further generated HCC827GR/tet-on-shVGF cells in which doxycycline induces shVGF expression to silence VGF (Fig. 6C; Supplementary Fig. S13A). The clonogenic assays showed that doxycycline treatment attenuated cell growth in HCC827GR/tet-on-shVGF but not in control HCC827GR/tet-on cells (Fig. 6C; Supplementary Fig. S13A). To further test whether VGF could function as a therapeutic target, HCC827GR/tet-on-shVGF cells were subjected to a xenograft animal assay. When palpable tumor bulges were observed in the host mice, shVGF was induced by doxycycline treatment in the xenograft tumors. We found that doxycycline-induced knockdown of endogenous VGF attenuated tumor growth in HCC827GR/tet-on-shVGF cells, causing decrease of tumor weight, and doxycycline alone had no effect on tumor growth in HCC827GR/tet-on control cells (Fig. 6D; Supplementary Fig. S13B). Our findings support the notion that VGF is essential for oncogenic tumor growth and is a potential therapeutic target in lung cancer.

To characterize the role of VGF in the tumor progression of lung cancer, we measured VGF expression in lung adenocarcinoma by IHC analysis of a panel of 70 specimens. IHC staining revealed that VGF was preferentially expressed in poorly differentiated (75% grade III) lung adenocarcinomas as compared with well (25% grade I) and moderately differentiated (30% grade II) ones, and a χ² analysis indicated that the association of VGF expression with pathologic grades was significant (P = 0.001; Fig. 7A). The aforementioned data indicated that VGF induced EMT in lung cancer cells. In primary lung adenocarcinoma, we further validated the correlation of VGF with EMT markers. Correlation analysis demonstrated the presence of a positive correlation of VGF with VIM, N-cadherin, and TWIST1 (Fig. 7B; Supplementary Fig. S14A and S14B). Because CEACAM6 is currently used as a biomarker for diagnosis and prognosis in lung adenocarcinoma (41), we further examined its expression in EGFR-TKI–resistant versus sensitive cells. The qPCR assay revealed that the expression of CEACAM6 was lost in EGFR-TKI resistant cells, such as HCC827GR and the independently selected cells (ER3, T15-2), as compared with that in the parental HCC827 cells (Supplementary Fig. S14C). Correlation analysis revealed that VGF negatively correlated with CEACAM6 (Supplementary Fig. S14D). Kaplan–Meier survival analysis was then conducted for lung adenocarcinoma to determine the prognostic significance of the expression of VGF versus CEACAM6. In lung adenocarcinoma, we found that VGF expression correlated with poor survival; in contrast, the expression of CEACAM6 and neuroendocrine markers, such as Chromogranin A, Chromogranin B, and Synaptophysin, did not correlate with survival outcome (Fig. 7C; Supplementary Fig. S15). Moreover, tumors harboring a VGF^high/TWIST1^high signature were associated with worse survival as compared with those carrying a VGF^low/TWIST1^low signature (Fig. 7C, right). Consistently, in EGFR-mutated lung adenocarcinoma, we observed that VGF expression predicted poor survival, whereas CEACAM6 expression did not correlate with the survival outcome (Fig. 7D, left and middle; and Supplementary Fig. S16). In EGFR-mutated lung adenocarcinoma, we observed that a VGF^high/TWIST1^high signature predicted a worse prognosis than a VGF^low/TWIST1^low signature (Fig. 7D, right; Supplementary Fig. S16). These data indicate the potential of VGF and TWSIT1 as prognostic biomarkers for lung adenocarcinoma.

Although EMT has been linked to EGFR-TKI resistance, the mechanism is not clear. In this study, we found that VGF, a neuroendocrine factor, is highly expressed in EGFR-TKI–resistant lung adenocarcinoma cells, the growth of which is dependent on VGF expression. The ectopic expression of VGF supports cell growth with EGFR-TKI resistance and induces EMT. Our findings revealed for the first time that VGF functions as an emerging factor in EGFR-TKI resistance and EMT in lung adenocarcinoma.

VGF was originally identified in the neurons and neuroendocrine cells and is responsible for normal metabolism as well as cell survival and proliferation in the hippocampus (30, 42). Moreover, VGF was reported to protect neuron cells from ER stress–induced cell death, suggesting its involvement in stress-induced cell survival (33). In lung cancer, VGF was first detected in neuroendocrine lung carcinoma cell lines by proteomic analysis, while the biological and clinical significance of VGF in tumors have not been known (35). In this study, we found that VGF is highly expressed in EGFR-TKI–resistant cells. We discovered that VGF expression activates AKT survival signaling, thus preventing the cells from undergoing EGFR-TKI–induced apoptosis in lung adenocarcinoma cells. We found that VGF-containing conditioned media can promote cell growth in the low serum culture (Supplementary Figs. S6 and S12); moreover, VGF silencing in HCC827GR cells attenuates tumor cell growth both in vitro and in vivo (Fig. 6). These data support a critical role of VGF in growth and survival in VGF-positive lung cancer cells. Thus, we propose that VGF-induced survival signaling is essential for the development of acquired EGF-TKI resistance in both T790M-dependent and independent mechanisms. In support of this hypothesis, we observed that VGF is highly expressed in the case 2 and case 3 of T790M-positive lung adenocarcinomas, which were treated with EGFR-TKIs before (Fig. 3E). Consistently, we found that H1975, harboring EGFR L858R/T790M mutations, exhibits a high level of VGF, and H1975 is more resistant to afatinib and expresses lower E-cadherin than its VGF^low counterpart (Supplementary Fig. S10). Moreover, the knockdown of VGF in H1975 attenuates cell growth (Fig. 6A), supporting the notion that VGF expression promotes EGFR-TKI resistance, EMT, and cell growth. As described in the ATCC database, H1975 lung adenocarcinoma cell line was established by Adi Gazdar and John Milnor from a female nonsmoker in 1988, prior to the clinical use of TKIs. Thus, high VGF expression in EGFR-mutated lung tumors cannot all be attributed to the acquired resistance of EGFR-TKI treatment, but may, in part, naturally exist before the selection. In support of this view, VGF is highly expressed in the case 1 of lung neuroendocrine carcinoma, which harbors wild-type EGFR and was never treated with EGFR-TKIs (Fig. 3E). We observed that VGF-positive lung cancer cells, such as HCC827GR, H1975, and H1299, are more resistant to low serum stress and EGFR-TKI treatment than VGF-negative HCC827, whereas H1975 and H1299 were derived from patients never treated with EGFR-TKIs (Supplementary Fig. S12A and S12B). Our in vitro and in vivo animal models indicated that VGF silencing inhibits cancer cell growth and tumor formation, suggesting that VGF contains oncogenic properties with the potential to serve as a therapeutic target (Fig. 6). New modalities targeting VGF, such as neutralizing anti-VGF mAbs, may offer solutions for patients with EGFR-TKI–resistant lung tumor.

EMT has been linked to EGFR-TKI resistance; however, the mechanism is not known (28, 29, 43). Here we observed that during EGFR-TKI selection, the EMT phenotypic conversion occurs in HCC827GR cells, which contain high levels of VGF and TWIST1, and HCC827GR cells develop EGFR-TKI resistance in a non-T790M–dependent manner. We found that ectopic expression of VGF in HCC827 cells not only confers resistance to EGFR-TKIs but also induces the EMT phenotypic alteration accompanied with TWIST1 upregulation, whereas other EMT regulators, such as SNAIL and SLUG, are not induced by VGF expression. TWIST1 has been reported to determine cell fates and regulate normal cell differentiation and EMT (44). In addition, TWIST1 encourages cancer cell survival and dissemination (40, 45). Here, we observed that VGF levels correlate with TWIST1 expression in primary lung adenocarcinoma. Ectopic expression of VGF induces TWIST1 expression. Moreover, TWIST1 expression facilitates EGFR-TKI resistance and induces the expression of mesenchymal markers. These data indicate that VGF-TWIST1 signaling plays a critical role in EGFR-TKI resistance and EMT.

The human carcinoembryonic antigen (CEA), mainly referred to as CEACAM5 and CEACAM6 with shared antigenic determinants, has been wildly used as a tumor marker in colorectal cancer and lung cancer. CEACAM6 expression is higher than that of CEACAM5 in lung adenocarcinoma (41). However, the use of CEA as a prognostic and predictive marker in lung cancer patients is controversial (46). We observed that CEACAM6 expression is lost in HCC827GR cells and other independently selected EGFR-TKI–resistant cells, all of which develop EGFR-TKI resistance by a non-T790M pathway (Supplementary Fig. S14C). These data suggest that CEACAM6 expression may not be a predictive biomarker for the prognosis of EGFR-TKI–resistant tumors driven by a non-T790M pathway. Moreover, VGF is negatively correlated with CEACAM6 expression in the primary lung adenocarcinoma (Supplementary Fig. S14D). We found that CEACAM6 expression does not correlate with overall survival while VGF expression predicted a poor survival not only in lung adenocarcinoma patients but also in the EGFR-mutated subpopulation. Recently, VGF is detected in Synaptophysin- and/or Chromogranin A-positive neuroendocrine breast cancer and in a subgroup of basal-like breast cancer, in which Chromogranin A and Synaptophysin are absent (36). In this study, we found that VGF is expressed in lung neuroendocrine carcinoma and EGFR-TKI–resistant lung adenocarcinoma. We observed that although VGF Synaptophysin, Chromogranin A, and Chromogranin B are highly expressed in neuroendocrine tumors, such as small-cell lung cancer and large-cell neuroendocrine carcinoma, only VGF levels correlates with EGFR-TKI resistance in EGFR-mutated lung adenocarcinoma (Fig. 3D; Supplementary Figs. S7–S9). While the transformation toward small-cell lung cancer in resistant EGFR-mutated lung adenocarcinoma has been attributed to the RB loss accompanied with EGFR downregulation (47), none of these alterations occur in HCC827GR and H1975 cells. Moreover, VGF, but not Chromogranin A or Synaptophysin, correlated with a poor survival in patients with lung adenocarcinoma (Fig. 7C and D; Supplementary Figs. S15 and S16). These findings suggest that VGF expression is not only limited to neuroendocrine cancer cells but also could serve as a prognostic biomarker in patients with lung adenocarcinoma.

Taken together, we found that VGF is highly expressed in a subgroup of lung adenocarcinoma and favors EGFR-TKI resistance and EMT, thereby predicting a poor survival. These findings provide new insights for the role of VGF, a neurotrophic factor, in lung cancer oncogenesis with the potential to serve as a biomarker and therapeutic target for lung cancer intervention.

No potential conflicts of interest were disclosed.

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): W. Hwang, Y.-F. Chiu, J.-L. Chang, S.-H. Hsiao, S.-E. Lin

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): W. Hwang, Y.-F. Chiu, M.-H. Kuo, K.-L. Lee, J.-L. Chang, W.-C. Huang, S.-E. Lin

Writing, review, and/or revision of the manuscript: W. Hwang, Y.-F. Chiu, A.-C. Lee, M.-H. Kuo, Y.-T. Chou

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y.-F. Chiu, A.-C. Lee, C.-C. Yu, J.-L. Chang, W.-C. Huang

Study supervision: Y.-T. Chou

We thank Dr. Michael S. Lan (Louisiana State University) for critical reading of this manuscript and insightful comments. We also thank staff at the Eighth Core Lab, Department of Medical Research, National Taiwan University Hospital, for technical support.

This work was supported by National Tsing Hua University and SMOBiO Inc.

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.